A method, system, electronic equipment, and vehicle for anti-icing and de-icing of a high-voltage relay.

By constructing an icing risk prediction model and implementing full-process closed-loop control, the icing problem of high-voltage relays in extremely cold and humid environments was solved, achieving precise anti-icing and de-icing, optimizing energy consumption and reliability, and extending device life.

CN123104394APending Publication Date: 2026-11-17DEEPAL AUTOMOBILE TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611338022.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-31
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing high-voltage relays are prone to freezing in extremely cold and humid environments, which can lead to contact sticking or high-voltage circuit failure. Traditional anti-icing solutions rely on a single temperature determination, which can easily cause false triggering or missed triggering, and lack self-learning capabilities, resulting in unnecessary energy consumption and device aging.

Method used

By combining environmental parameters, relay temperature, and weather forecast data, an icing risk prediction model is constructed to achieve closed-loop control throughout the entire process, including anti-icing insulation during the dormant period and de-icing operation during the awakening period. Multi-source information fusion and adaptive heating strategies are adopted to dynamically adjust the heating power and actuator coordination, combined with de-icing effect verification and model updates.

Benefits of technology

It enables precise anti-icing and de-icing of high-voltage relays in complex environments, reduces false triggering and missed triggering, optimizes energy consumption, improves system reliability and safety, and extends device life.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123104394A_ABST
    Figure CN123104394A_ABST
Patent Text Reader

Abstract

This invention relates to the field of vehicle technology and discloses a method, system, electronic device, and vehicle for anti-icing and de-icing of a high-voltage relay. The method includes: after acquiring environmental parameters of the vehicle and temperature data of the high-voltage relay, determining the dew point temperature based on the environmental parameters and calculating the difference between the temperature data of the high-voltage relay and the dew point temperature; determining the icing risk state of the high-voltage relay based on the difference; when the vehicle is in a dormant state and the icing risk state meets a first preset condition, controlling the battery management system to output energy to drive a heating component to perform anti-icing and heat preservation operations on the high-voltage relay; and when the vehicle receives a wake-up command and the high-voltage relay meets a second preset condition, controlling the heating component to perform de-icing operations on the high-voltage relay.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of vehicle technology, specifically to a method, system, electronic device, and vehicle for preventing and de-icing a high-voltage relay. Background Technology

[0002] With the rapid development of new energy vehicle technology, high-voltage relays, as core electrical components controlling the connection and disconnection of the power battery and the vehicle's high-voltage circuit, are of paramount importance in terms of operational stability and safety. Under complex climatic conditions such as extreme cold and high humidity, the high-voltage relay and its surrounding components are prone to freezing due to the decrease in ambient temperature and condensation, which may lead to serious malfunctions such as contact adhesion, failure of the high-voltage circuit to engage or disengage properly.

[0003] Existing anti-icing or de-icing solutions for high-voltage relays typically rely solely on a single ambient temperature sensor reading for simple threshold judgment. When the detected outside ambient temperature is lower than a preset threshold, the system triggers an open-loop heating component to periodically heat the relay. However, whether ice forms on the relay surface depends not only on the outside ambient temperature but also on ambient humidity and the relay's own actual temperature state. A single temperature judgment method cannot accurately reflect the actual risk of relay icing, easily leading to inaccurate heating control. Furthermore, traditional solutions often only passively perform de-icing after the vehicle is powered on, ignoring the relay's temperature evolution during different operating phases, such as when the vehicle is in sleep mode, making it difficult to guarantee the operational reliability of high-voltage relays under various vehicle operating conditions. Summary of the Invention

[0004] In view of the shortcomings of the prior art, the purpose of this application is to provide a method, system, electronic device and vehicle for anti-icing and de-icing of high-voltage relays, which can accurately assess the actual icing risk of high-voltage relays under different operating stages of vehicles and achieve efficient and reliable anti-icing and de-icing control.

[0005] In a first aspect, embodiments of this application provide a method for anti-icing and de-icing a high-voltage relay, comprising: after acquiring environmental parameters of a vehicle and temperature data of the high-voltage relay, determining a dew point temperature based on the environmental parameters, and calculating the difference between the temperature data of the high-voltage relay and the dew point temperature; determining an icing risk state of the high-voltage relay based on the difference; when the vehicle is in a dormant state and the icing risk state meets a first preset condition, controlling the battery management system to output energy to drive a heating component to perform anti-icing and heat preservation operations on the high-voltage relay; and when the vehicle receives a wake-up command and the high-voltage relay meets a second preset condition, controlling the heating component to perform de-icing operations on the high-voltage relay.

[0006] The anti-icing and de-icing method provided in this application accurately identifies icing risks by calculating the physical temperature difference between the actual temperature of the high-voltage relay and the ambient dew point temperature. This overcomes the problems of false triggering and missed triggering that are easily caused by traditional technologies that rely solely on a single ambient temperature threshold for open-loop judgment. Simultaneously, anti-icing and de-icing control is integrated into both the vehicle's dormancy and wake-up phases. During dormancy, the battery management system coordinates with the system to perform low-power preventative insulation, while during wake-up, dynamic power adaptive de-icing is implemented. This constructs a closed-loop protection system that combines proactive prevention during dormancy with efficient de-icing during wake-up. This eliminates the risk of contact adhesion or power loss in the high-voltage relay caused by icing at the source, while also preventing energy waste in ice-free conditions, thus balancing the safety and reliability of the high-voltage system with the overall vehicle energy efficiency.

[0007] In one embodiment, determining the icing risk status of the high-voltage relay based on the difference includes: acquiring weather forecast data for a preset future time period, and combining it with the temperature of the vehicle's high-voltage battery pack to construct a time evolution trend of the difference within the future time series; and determining the icing risk status of the high-voltage relay as high-risk level when the time evolution trend indicates that the difference is less than or equal to a preset temperature difference threshold.

[0008] In this implementation, by incorporating future weather forecast trends and the evolution of the high-voltage battery pack heat source, a forward-looking prediction of the relay icing risk is achieved, providing sufficient response time for anti-icing operation.

[0009] In one embodiment, determining the icing risk status of the high-voltage relay further includes: inputting the difference, the environmental parameters, the high-voltage battery pack temperature, and a preset relay historical operation record into a preset risk prediction model; calculating a first risk probability using the heat conduction logic in the risk prediction model, and calculating a second risk probability using the statistical regression logic in the risk prediction model; and fusing the first risk probability and the second risk probability to generate a comprehensive risk level indicating the icing risk status.

[0010] In this implementation, by constructing a physical mechanism-based heat conduction logic and a data-driven statistical regression logic in parallel and fusing their prediction probabilities, the confidence and robustness of the icing risk assessment are improved.

[0011] In one possible embodiment, the step of controlling the battery management system to output energy to drive the heating component to perform anti-icing and heat preservation operation on the high-voltage relay when the vehicle is in a dormant state and the icing risk state meets a first preset condition includes: after the vehicle disconnects the high-voltage power supply, obtaining the current remaining charge of the high-voltage battery pack; and when the current remaining charge is greater than a preset safe reserve charge threshold and the icing risk state indicates the existence of an icing risk, triggering the battery management system to connect the auxiliary power supply circuit and controlling the heating component to heat at a maintenance power lower than the preset rated power.

[0012] In this embodiment, a high-voltage battery pack power safety threshold is added during the anti-icing and heat preservation process in the dormant period to ensure that the anti-icing energy dispatch will not cause the battery to be over-discharged or depleted, thus ensuring the power consumption safety of the whole vehicle during the dormant period.

[0013] In an exemplary embodiment, the step of controlling the heating component to perform de-icing on the high-voltage relay when the vehicle receives a wake-up command and the high-voltage relay meets a second preset condition includes: obtaining the vehicle's start-up urgency level and the maximum allowable output power of the high-voltage battery pack; determining a target heating power curve based on the start-up urgency level and the maximum allowable output power; and controlling the heating component to dynamically adjust the heating power of the high-voltage relay based on the target heating power curve.

[0014] In this implementation, a dynamic heating curve is generated by combining the urgency of starting and the battery's allowable output power, balancing the need for rapid vehicle wake-up and start-up with the protection against thermal stress damage to the relay caused by high-power heating.

[0015] In some embodiments, after the heating component performs de-icing on the high-voltage relay, the method further includes: acquiring at least one of the temperature recovery rate, insulation resistance, and closed contact resistance of the high-voltage relay; determining a de-icing effect evaluation result based on the acquired temperature recovery rate, insulation resistance, or closed contact resistance; and updating the configuration parameters of the risk prediction model based on the data from the current de-icing operation if the de-icing effect evaluation result indicates successful de-icing.

[0016] In this implementation, the multimodal fusion evaluation based on thermal and electrical properties can accurately verify whether de-icing is thorough, and the model configuration parameters are updated online using motion data, thus constructing an adaptive closed-loop evolution mechanism.

[0017] In one possible embodiment, the method further includes: when the de-icing effect evaluation result indicates that de-icing has failed and the number of retry heating attempts has reached a preset limit, locking the closing command of the high-voltage relay and sending a fault warning message to the vehicle's on-board terminal.

[0018] In this implementation, when de-icing fails repeatedly, a closing command is locked and a safety alarm is triggered, thus constructing a system-level degradation protection mechanism. This effectively prevents contact welding damage or high-voltage safety accidents caused by forced closure with ice.

[0019] In one possible embodiment, acquiring the vehicle's environmental parameters and the temperature data of the high-voltage relay includes: acquiring the ambient air temperature and relative humidity of the environment in which the vehicle is located, and using the ambient air temperature and relative humidity as the environmental parameters; and acquiring the reading of a temperature sensor attached to the housing of the high-voltage relay, and using the reading as the temperature data of the high-voltage relay.

[0020] In this embodiment, by combining the ambient temperature and humidity with the local real temperature of the relay housing, the physical conditions of water vapor condensation in the relay's microscopic environment can be accurately clarified, thereby improving the objectivity and accuracy of the dew point temperature difference calculation.

[0021] Secondly, embodiments of this application provide an anti-icing and de-icing system for a high-voltage relay. The anti-icing and de-icing system includes: a sensing module for acquiring environmental parameters of the vehicle and temperature data of the high-voltage relay; a controller for determining the dew point temperature based on the environmental parameters, calculating the difference between the temperature data of the high-voltage relay and the dew point temperature, and determining the icing risk state of the high-voltage relay based on the difference; and a heating component deployed in the area where the high-voltage relay is located and connected to the controller. The controller is further configured to, when the vehicle is in a dormant state and the icing risk state meets a first preset condition, control the battery management system to output energy to drive the heating component to perform anti-icing and heat preservation operations on the high-voltage relay; and when the vehicle receives a wake-up command and the high-voltage relay meets a second preset condition, control the heating component to perform de-icing operations on the high-voltage relay.

[0022] The corresponding effects of the system described in the second aspect can be found in the relevant content of the method described in the first aspect, and will not be repeated here.

[0023] In one possible embodiment, the heating component includes: a first heating unit attached to the housing surface of the high-voltage relay; a second heating unit disposed on the base of the high-voltage relay; and a flow path regulating unit configured to control the flow rate of the heat medium guided to the vicinity of the high-voltage relay.

[0024] In this embodiment, the heat transfer efficiency to the core of the high-voltage relay is improved and the de-icing response time is shortened by the coordinated linkage of multi-zone heating and heat medium flow path adjustment.

[0025] Thirdly, embodiments of this application provide a vehicle that includes an anti-icing and de-icing system for a high-voltage relay as provided in the second aspect or any possible implementation thereof.

[0026] The corresponding effects of the vehicle described in the third aspect can be found in the relevant content of the first or second aspect, and will not be repeated here.

[0027] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method provided as in the first aspect or any possible implementation thereof.

[0028] Fifthly, embodiments of this application provide a computer program product, including a computer program or instructions, which, when executed by a processor, implement the method provided as in the first aspect or any possible implementation thereof.

[0029] In a sixth aspect, embodiments of this application provide an electronic device, including: a memory having a computer program stored thereon; and a processor coupled to the memory for running the computer program to implement the method provided by the first aspect or any possible implementation thereof.

[0030] The anti-icing and de-icing methods, systems, and vehicles provided in this application can achieve the following technical effects: First, by acquiring ambient air temperature, ambient relative humidity, and high-voltage relay temperature, the difference between the relay temperature and dew point temperature is calculated, and the risk of icing is determined based on this difference. This allows for the simultaneous consideration of the effects of temperature and humidity on condensation and icing, reducing false triggering and missed triggering caused by relying solely on a single temperature threshold.

[0031] Second, by acquiring weather forecast data for a preset time period in the future and combining it with the time evolution trend of the temperature difference of the high-voltage battery pack, risks can be identified before icing occurs, and the control method can be changed from passive handling after icing to active prevention during dormancy.

[0032] Third, the probability of the first risk is calculated by heat conduction logic, and the probability of the second risk is calculated by statistical regression logic. The two are then integrated into a comprehensive risk level, which can take into account both the physical heat transfer law and the historical operating characteristics of the vehicle, thereby improving the adaptability and reliability of icing risk assessment in complex environments.

[0033] Fourth, in the vehicle's dormant state, the auxiliary power supply circuit is only activated when the current remaining charge of the high-voltage battery pack is greater than the safe reserve charge threshold and there is a risk of icing. The auxiliary power supply is used to heat the battery at a level lower than the rated power, which can reduce the probability of icing while preserving the vehicle's starting energy and reduce unnecessary dormant energy consumption.

[0034] Fifth, after the vehicle is woken up, the target heating power curve is determined based on the emergency level of starting and the maximum allowable output power of the high-voltage battery pack. This allows for coordination between starting time, heating energy consumption, battery output capacity and relay thermal load, avoiding energy waste or excessively high local temperatures caused by fixed-power heating.

[0035] Sixth, by evaluating the de-icing effect through at least one of the temperature recovery rate, insulation resistance, and closing contact resistance, the assumption of de-icing completion based solely on heating time can be avoided, thereby improving the reliability of the state verification before the high-voltage relay closes.

[0036] Seventh, after successful de-icing, the configuration parameters of the risk prediction model are updated using the data from this de-icing operation, enabling the system to continuously correct the model based on individual vehicle differences, relay aging status, and actual usage environment.

[0037] Eighth, when de-icing fails and the number of retry heating attempts reaches the limit, the high-voltage relay closing command is locked and a fault warning message is sent. This can prevent the high-voltage circuit from being forcibly connected under conditions of suspected icing, insulation abnormality, or mechanical obstruction, thereby improving the fault safety of the vehicle's high-voltage system.

[0038] Ninth, by setting up a first heating unit, a second heating unit, and a flow path adjustment unit, the relay housing, base, and surrounding heat medium can be adjusted respectively, improving the uniformity of temperature distribution and increasing anti-icing and de-icing efficiency.

[0039] Tenth, by setting an offline degradation prediction mode when weather forecast data is unavailable, the system can continue to perform icing risk prediction when there are underground parking lots, network signals are missing, or cloud services are unavailable, reducing the system's dependence on external networks. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of this application or the background art, the accompanying drawings used in the embodiments of this application will be described below.

[0041] Figure 1 This paper shows a schematic diagram of the architecture of the high-voltage relay anti-icing and de-icing system provided in an embodiment of this application; Figure 2 A schematic flowchart of an anti-icing and de-icing method for a high-voltage relay provided in an embodiment of this application is shown. Figure 3A schematic diagram of the overall process of another high-voltage relay anti-icing and de-icing method provided in an embodiment of this application is shown; Figure 4 It shows Figure 3 A flowchart illustrating the process of environmental perception and icing risk prediction in China; Figure 5 A schematic diagram of the heat conduction logic operation process is shown; Figure 6 This diagram illustrates the statistical regression logic performed by the statistical analysis model. Figure 7 A schematic diagram of the process of an empirical prediction model performing offline degradation prediction mode is shown; Figure 8 An example diagram illustrating the operation of the multi-model fusion decision module is shown. Figure 9 A schematic diagram illustrating the specific process of predictive energy scheduling and proactive prevention is shown. Figure 10 This diagram illustrates the online self-learning and continuous optimization process. Figure 11 A flowchart illustrating the real-time system health monitoring, assessment, and operation and maintenance mechanism is provided. Figure 12 A flowchart illustrating real-time situation assessment, intelligent de-icing, and effect verification is shown. Figure 13 This diagram illustrates the process of initiation decision-making, tiered security degradation, and fault report generation. Figure 14 A flowchart illustrating the vehicle-level safety defense system and emergency fault-tolerant intervention mechanism is shown. Detailed Implementation

[0042] The terms “first,” “second,” etc., are used for descriptive purposes only and have no sequential or technical meaning, nor should they be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated.

[0043] In the embodiments of this application, "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0044] Unless otherwise stated, the "preset" parameters in this application may be parameters obtained and stored in the controller during the vehicle development phase through environmental chamber testing, road testing, or simulation calibration, or they may be parameters updated by the controller based on vehicle operating data. The parameters such as temperature, humidity, power, time, remaining battery power, probability, and resistance listed in the specification are examples provided for ease of understanding and do not constitute a limitation on the scope of protection of this application.

[0045] The embodiments of this application are described below with reference to the accompanying drawings.

[0046] With the dwindling availability of fossil fuels and the increasing severity of environmental problems, the development of new energy sources has become an inevitable trend, and new energy vehicles have become an important development direction for the automotive industry. With the rapid development of the new energy vehicle industry, electric vehicles, hybrid vehicles, and other new energy models are rapidly gaining popularity worldwide. These vehicles generally use high-voltage electrical systems, among which high-voltage relays, as key safety components between the power battery and the drive system, are responsible for controlling the circuit's continuity during power-on / off, charging, and driving conditions. Their reliability directly affects the safety of the entire vehicle.

[0047] However, in cold and humid environments, condensation or even ice formation can easily occur inside and on the surface of high-voltage relays, leading to abnormally increased contact resistance and decreased insulation performance. In severe cases, this can cause the relay to fail to close properly, resulting in safety hazards such as the vehicle failing to start or sudden power loss while driving. This has become one of the technical bottlenecks restricting the promotion of new energy vehicles in cold and humid regions.

[0048] Currently, industry solutions for high-voltage relay icing mainly fall into two categories: passive protection and active de-icing. Passive protection measures, such as improving the relay sealing structure, using hydrophobic coatings, and filling with moisture-proof gel, can delay icing to some extent, but cannot fundamentally eliminate the risk of icing in extreme environments, and will increase costs and design complexity.

[0049] Regarding active de-icing, several control schemes have been proposed in related technologies. For example, one scheme involves setting the vehicle's wake-up time based on the temperature difference when the vehicle is powered off. If the outside temperature is detected to be lower than a preset freezing point threshold, and the temperature difference between the inside and outside of the vehicle is greater than a preset temperature difference threshold, then de-icing is performed upon wake-up. This scheme relies solely on a simple temperature signal for control and lacks comprehensive judgment based on multi-dimensional information such as ambient humidity, wind speed, and historical operating data. This can easily lead to false triggering of heating when there is no ice, or missed triggering when there is ice but no heating.

[0050] Another approach focuses primarily on relay fault diagnosis, determining whether the heating relay is stuck by acquiring sampling voltages and pre-charge times from multiple locations such as the heating relay and the main positive relay. However, this approach mainly focuses on hardware-level fault diagnosis and fails to provide effective systemic operational guidance for safe control and degraded operation after relay sticking or de-icing failure.

[0051] In general, most existing active de-icing technologies employ simple temperature threshold and timed heating control logic. The control strategy is relatively rigid, the execution process is open-loop, and it lacks self-learning and adaptive capabilities. This crude control method not only leads to unnecessary energy consumption and shortens the vehicle's driving range, but may also accelerate the aging of relay devices due to frequent or unnecessary heating, and even cause localized thermal stress damage due to uneven heating.

[0052] Based on this, this application provides a method, system, electronic device, and vehicle for anti-icing and de-icing of a high-voltage relay. The method constructs an icing risk prediction result using environmental parameters, relay temperature, high-voltage battery pack temperature, weather forecast data, and historical relay operation records. Anti-icing and de-icing actions are then performed in both the vehicle's dormant and active states. After de-icing is completed, the system can verify the de-icing effect through temperature and electrical characteristics, and update the risk prediction model based on the processing results, thus forming a closed-loop control process of "perception, prediction, decision-making, execution, verification, and learning." In other words, this application constructs a complete intelligent closed-loop system covering multiple stages, including environmental perception and risk prediction, predictive energy scheduling, power-on self-test and multi-dimensional diagnosis, adaptive de-icing decision-making and execution, fault handling and system degradation, secondary verification of effects, and system learning.

[0053] More specifically, the technical solution of this application achieves technological breakthroughs and system optimizations mainly through the following core dimensions: Dynamic prediction of icing risk through multi-source information fusion: The system integrates multi-source data such as ambient temperature, relative humidity, high-voltage battery pack temperature and historical operating parameters. Through parallel calculation using a physical heat conduction model and a statistical regression model, it obtains the probability of icing risk in real time and merges them to generate a comprehensive risk level, achieving high-precision and advanced risk prediction.

[0054] Adaptive De-icing Decision Based on Risk-Benefit Multi-Objective Game Theory: This approach models multiple objectives, including vehicle safety, energy consumption limits, startup time, and component lifespan, as a multi-constraint game problem. The controller adaptively selects the optimal de-icing strategy and heating parameters based on real-time icing risk, remaining battery charge, and vehicle operating mode.

[0055] Intelligent adaptive heating control with multi-actuator collaboration: The first heating unit attached to the surface of the outer shell, the second heating unit set on the base, and the flow path adjustment unit that controls the flow rate of the heat medium work together, and the heating power is dynamically adjusted by combining an adaptive closed-loop control algorithm. While ensuring accurate and efficient heating, the local thermal stress of the relay is reduced.

[0056] Multimodal cross-validation evaluation of de-icing effect: A multimodal validation system was established, including temperature recovery rate, system insulation resistance, and closed contact resistance. The de-icing effect was comprehensively judged by weighted scoring, avoiding the problem of incomplete de-icing that may be caused by open-loop control.

[0057] System self-evolution and knowledge solidification based on machine learning: After each de-icing event, the system uses the feature data from the de-icing process to update the configuration parameters of the risk prediction model and adaptive control strategy online, enabling the system to have continuous intelligence and autonomous evolution capabilities.

[0058] Root cause analysis and in-depth diagnosis of failures: When the system malfunctions or fails to defrost, a progressive cause-effect graph reasoning method is used to conduct multi-level analysis from the direct cause to the design root cause, accurately locate the cause of failure and generate a response strategy.

[0059] Tiered and phased fault safety handling and system degradation: Different safety threat levels are defined according to the severity of the fault, and standardized safety control modes from emergency shutdown and function limitation to warning observation are matched to maximize the electrical safety of the high-voltage system.

[0060] Personalized strategy matching based on the urgency of starting: Using the urgency level of vehicle starting and the allowable output power of the battery pack as decision inputs, the heating power curve is dynamically adjusted to achieve the optimal balance between rapid start-up requirements and device protection.

[0061] Predictive energy dispatch and proactive sleep management: When the vehicle is in sleep mode, it is periodically woken up and the risk of icing is calculated. Under the premise of meeting the safety threshold of remaining power, the auxiliary circuit is connected to carry out preventive low-power anti-icing and heat preservation to actively eliminate the risk of icing.

[0062] The intelligent anti-icing system architecture with a closed-loop process: It has built a closed-loop control architecture covering the entire life cycle of "perception-prediction-decision-execution-verification-learning". The data and control decisions of each link are closely coupled, realizing efficient self-management of high-voltage relay anti-icing and de-icing.

[0063] In other words, the embodiments of this application significantly improve the accuracy of icing judgment and greatly reduce false triggering and missed triggering by multi-source information fusion and dynamic risk prediction; through multi-objective adaptive decision-making and dynamic closed-loop control, energy consumption is optimized and start-up waiting time is shortened while ensuring the reliability of de-icing; through post-event multi-dimensional verification and event-driven online update mechanism, the system can adapt to different user habits and changeable extreme climates, and achieve a comprehensive improvement in the safety, economy, efficiency and long-term reliability of high-voltage relays in anti-icing under complex cold environments.

[0064] Thus, this application embodiment can achieve a complete intelligent closed-loop solution that addresses the pain points of traditional anti-icing solutions, such as reliance on a single threshold, rigid control strategies, open-loop execution, and lack of self-learning capabilities, while meeting high-voltage electrical safety standards.

[0065] It should be understood that the anti-icing and de-icing method for high-voltage relays provided in this application embodiment can be applied to pure electric vehicles, hybrid vehicles, plug-in hybrid vehicles, range-extended electric vehicles, or other vehicles equipped with high-voltage battery packs and high-voltage relays.

[0066] Optionally, the high-voltage relay may include a main positive relay, a main negative relay, a pre-charge relay, a charging relay, or other relays located in the vehicle's high-voltage circuit. The following explanation primarily uses the main relay between the power battery and the vehicle's high-voltage load as an example.

[0067] It is worth mentioning that the "anti-icing and heat preservation operation" in this application embodiment refers to the preventive heating at a power lower than the preset rated power of the heating component when the vehicle is in a dormant state and the risk of icing is predicted, so as to keep the temperature of the high-voltage relay above the dew point temperature or frost point temperature, or slow down the temperature of the high-voltage relay from falling into the icing risk range.

[0068] In this application, "de-icing operation" refers to dynamically heating the vehicle according to the target heating power curve when the vehicle receives a wake-up command and the high-voltage relay meets the de-icing trigger conditions, in order to eliminate the condensation, frost, or ice that has formed, or to restore the high-voltage relay to a state that meets the closing requirements.

[0069] Anti-icing and heat preservation actions differ from de-icing actions in terms of triggering timing, control targets, and heating power. Anti-icing and heat preservation actions are mainly used for active prevention during dormant states, while de-icing actions are mainly used for restoring the vehicle's condition after it has been awakened.

[0070] Based on the above description, the anti-icing and de-icing system for high-voltage relays provided in the embodiments of this application will be described by way of example.

[0071] Figure 1A schematic diagram of the architecture of the high-voltage relay anti-icing and de-icing system provided in an embodiment of this application is shown. Figure 1 As shown, in one embodiment, the anti-icing and de-icing system 100 of the high-voltage relay includes a sensing module 101, a controller 102, a heating component 103, a battery management system (BMS) 104, an on-board communication terminal 105, and an on-board terminal 106.

[0072] The sensing module 101 is used to acquire environmental parameters of the vehicle and temperature data of the high-voltage relay. The sensing module 101 may include an ambient temperature and humidity sensor, a relay temperature sensor, and a high-voltage battery pack temperature sensor.

[0073] For example, the ambient temperature and humidity sensor is used to collect the ambient air temperature and relative humidity of the environment in which the vehicle is located. Optionally, the ambient temperature and humidity sensor can be set in the air intake area of ​​the vehicle's front compartment, the area of ​​the exterior rearview mirror, the area of ​​the air conditioning's external air circulation intake, or other locations that can characterize the external environment of the vehicle.

[0074] The relay temperature sensor can be a negative temperature coefficient thermistor (NTC resistor), a platinum resistance temperature sensor, a thermocouple, or a digital temperature sensor. The relay temperature sensor is attached to the surface of the high-voltage relay housing and forms a thermal connection with the housing via thermally conductive adhesive, a thermally conductive pad, or a mechanical clamping structure. The relay temperature sensor is preferably located near the contact cavity, the base connection area, or in a location prone to condensation.

[0075] The high-voltage battery pack temperature can be obtained by the battery management system 104 based on one or more temperature sensors inside the battery pack (i.e., the high-voltage battery pack temperature sensor in the figure) and transmitted to the controller via the controller area network (CAN). In addition, the battery management system 104 can also obtain other battery status information, such as the state of charge (SOC).

[0076] The controller 102 can be an independent anti-icing and de-icing controller, or it can be implemented by a battery management system controller, vehicle controller, power domain controller, thermal management controller, or central computing platform. The controller 102 may include a processor, volatile memory, non-volatile memory, sensor interface, heating drive interface, and vehicle communication interface.

[0077] Non-volatile memory can store dew point calculation parameters, configuration parameters of heat conduction logic, configuration parameters of statistical regression logic, risk level thresholds, safety reserve power thresholds, heating power parameters, effect evaluation thresholds, retry limit values, and relay historical operation records.

[0078] The vehicle-mounted communication terminal 105 can be a T-Box, used to acquire weather forecast data for a preset time period in the future for the vehicle's location. The weather forecast data may include ambient air temperature prediction series and ambient relative humidity prediction series, and may also include precipitation probability, wind speed, air pressure, and weather type.

[0079] The vehicle terminal 106 may include an instrument panel, a central control display, a T-Box, a mobile terminal application, or a remote service platform, for outputting de-icing status, start-up waiting status, and fault warning information.

[0080] The heating element 103 is deployed in the area where the high-voltage relay is located and is connected to the controller. In one embodiment, the heating element 103 includes a first heating unit, a second heating unit, and a flow path regulating unit.

[0081] The first heating unit is attached to the surface of the high-voltage relay housing. The first heating unit can be a PTC heating film, a flexible resistance heating film, or other planar heating device, used to transfer heat to the high-voltage relay housing and contact cavity.

[0082] The second heating unit is located in the base of the high-voltage relay. The second heating unit can be located inside the base, on the lower surface of the base, or between the base and the mounting plate, and is used to heat the bottom of the relay and the sealed connection area.

[0083] The flow path regulating unit is used to control the flow rate of the heat medium guided to the vicinity of the high-voltage relay. The flow path regulating unit may include a fan, damper, flow control valve, or circulating pump. The heat medium may be air heated by the thermal management system or a circulating liquid.

[0084] The first heating unit, the second heating unit, and the flow path adjustment unit can be configured together, or at least one of them can be configured according to the vehicle structure. The controller can independently control each actuator to improve the temperature uniformity between different areas of the high-voltage relay.

[0085] In one specific implementation, the sensing module 101 is used to acquire environmental parameters of the vehicle and temperature data of the high-voltage relay. The controller 102 is used to determine the dew point temperature based on the environmental parameters, calculate the difference between the temperature data of the high-voltage relay and the dew point temperature, and determine the icing risk status of the high-voltage relay based on the difference. The heating element 103 is deployed in the area where the high-voltage relay is located and is connected to the controller.

[0086] The controller 102 is also used to control the battery management system 104 to output energy to drive the heating component 103 to perform anti-icing and heat preservation operation on the high-voltage relay when the vehicle is in a dormant state and the icing risk condition meets the first preset condition; and to control the heating component 103 to perform de-icing operation on the high-voltage relay when the vehicle receives a wake-up command and the high-voltage relay meets the second preset condition.

[0087] The above system modules can be implemented using software, hardware, or a combination of both. The module division is for illustrative purposes only and does not limit the specific organization of the controller program.

[0088] Based on the above-mentioned anti-icing and de-icing system for high-voltage relays, the anti-icing and de-icing method flow of high-voltage relays provided in the embodiments of this application will be described by way of example below.

[0089] Figure 2 A schematic flowchart illustrating an anti-icing and de-icing method for a high-voltage relay provided in an embodiment of this application is shown. Figure 2 As shown, this method can be derived from... Figure 1 The high-voltage relay shown is executed by the controller in the anti-icing and de-icing system, or by other controllers in the vehicle with data processing capabilities, including steps S201 to S204.

[0090] S201: Obtain the vehicle's environmental parameters and the temperature data of the high-voltage relay, determine the dew point temperature based on the environmental parameters, and calculate the difference between the high-voltage relay's temperature data and the dew point temperature.

[0091] S202, Based on this difference, determine the icing risk status of the high-voltage relay.

[0092] S203, when the vehicle is in a dormant state and the icing risk condition meets the first preset condition, the battery management system outputs energy to drive the heating component to perform anti-icing and heat preservation operation on the high-voltage relay.

[0093] S204, when the vehicle receives a wake-up command and the high-voltage relay meets the second preset condition, the heating component is controlled to perform a de-icing operation on the high-voltage relay.

[0094] Based on steps S201-S204 above, this technical solution acquires the vehicle's environmental parameters and the actual temperature data of the high-voltage relay, determines the dew point temperature based on the environmental parameters, and then calculates the physical difference between the relay temperature and the dew point temperature. This allows for accurate identification of icing risk based on real water vapor condensation physical conditions, overcoming the problems of false triggering and missed triggering caused by static single temperature thresholds. Furthermore, this technical solution determines the icing risk state based on the difference, achieving a quantitative and forward-looking assessment of the icing trend in the relay's microenvironment. For example, when the vehicle is in a dormant state and meets the first preset condition (high icing risk), the battery management system outputs energy to drive the heating component to perform anti-icing and heat preservation operations, achieving low-power active prevention during dormancy and preventing severe freezing of the relay contacts during long-term vehicle parking. When the vehicle receives a wake-up command and meets the second preset condition, the heating component performs de-icing operations, achieving efficient and targeted de-icing during the wake-up phase. This completes the closed-loop control of the entire process, from active prevention during dormancy to efficient de-icing during wake-up, balancing the absolute safety of the high-voltage relay operation with the optimal utilization rate of the vehicle's electrical energy.

[0095] Figure 3 A schematic diagram illustrating the overall flow of another high-voltage relay anti-icing and de-icing method provided in an embodiment of this application is shown. Figure 3 As shown, in a specific implementation, the anti-icing and de-icing method of the high-voltage relay is as follows: Figure 1 The complete process of the anti-icing and de-icing system of the high-voltage relay shown is as follows: Phase 1: Environmental perception and icing risk prediction.

[0096] When the vehicle's high voltage is de-energized, the system automatically triggers and enters "Phase 1: Environmental Perception and Icing Risk Prediction." In this phase, the system first collects environmental parameters (including ambient air temperature and relative humidity), the current temperature data of the high-voltage relay, the temperature of the high-voltage battery pack, and future weather data. Next, the system calculates the dew point temperature under the current microenvironment based on the environmental parameters, thereby determining the physical temperature difference between the actual temperature of the high-voltage relay and the dew point temperature. Then, the system inputs the physical temperature difference, the temperature evolution trend over time, and the heat source influence of the high-voltage battery pack into a preset risk prediction model to calculate the risk level used to indicate the possibility of relay icing and performs real-time risk level determination.

[0097] In one scenario, if the assessment result is high risk, the system transitions from Phase 1 to "Phase 2: Predictive Energy Dispatch and Active Prevention." During this phase, the system implements preventative strategies (e.g., controlling the BMS to activate the auxiliary power supply circuit to maintain heating at a low power level below the preset rated power, keeping the relay temperature above the freezing point). After the preventative anti-icing action is completed, the system enters a silent sleep state.

[0098] In another scenario, if the assessment result is medium or low risk, the system skips the active prevention steps of Phase Two and directly enters a silent dormant state.

[0099] 2. Listening for and determining wake-up events during the hibernation period.

[0100] After the system enters silent sleep mode, the high-voltage circuit of the entire vehicle is disconnected, the system maintains a low-power listening state, and performs wake-up event type determination on any wake-up events that occur: If the wake-up event is the arrival of a timed wake-up cycle, it means that the vehicle is still in the silent storage period. The system is periodically woken up and returns to stage 1 to perform environmental perception and icing risk prediction again, forming a closed loop of "monitoring-assessment-prevention-hibernation" during the hibernation period.

[0101] If the wake-up event is triggered by user operation (e.g., user remotely sending a vehicle use command via APP, unlocking with a physical key, or pressing the brake pedal, etc.), the system exits the hibernation state and enters "Phase 3: Power-on self-test and multi-dimensional feature diagnosis".

[0102] 3. Phase Three: Power-on self-test and multi-dimensional feature diagnosis.

[0103] In Phase 3, the system performs a self-check on various physical and electrical characteristics of the high-voltage relay (such as auxiliary contact status, high-voltage circuit insulation resistance, coil drive circuit resistance, etc.) and outputs the relay health status diagnosis result according to the diagnostic algorithm. Based on this diagnosis result, the system makes the following branch judgment: Branch 1 (Status: Healthy): If the diagnosis result indicates that the relay has no risk of icing and its electrical characteristics are normal, the system directly enters "Phase 4: Normal Closure, Vehicle Start-up Successful". Subsequently, the high-voltage relay closes, and the vehicle is powered on and running normally.

[0104] Branch 2 (Status: Confirmed Fault): If the diagnostic results indicate that the relay has irreversible physical damage or a serious electrical fault, the system will skip the de-icing stage and directly proceed to "Stage 8: Fault Safety Handling and System Degradation", locking the high-voltage closing command and sending an alarm to the vehicle terminal.

[0105] Branch 3 (Status: Suspected Icing): If the diagnostic results indicate that the relay parameters deviate abnormally and the environmental physical conditions meet the characteristics of icing, the system determines that it is in a suspected icing state and enters "Phase 5: Adaptive De-icing Decision and Execution".

[0106] 4. Phases 5 to 8: De-icing, verification, and fail-safe degradation loop.

[0107] In "Phase 5: Adaptive De-icing Decision and Execution", the controller calculates the optimal target heating power curve by combining the current start-up urgency level and the maximum allowable output power of the high-voltage battery pack, and coordinates the heating components to perform de-icing operation on the high-voltage relay with dynamically adjusted heating power.

[0108] After the de-icing operation is executed, the system enters the "Phase 6: Fault Handling and System Degradation" phase (this phase serves as the carrier for post-de-icing status verification and temporary error prevention mechanisms), and the verification results are judged: If the verification result is passed, it indicates that the de-icing is complete and the relay has resumed normal electrical function. The system then enters "Phase 7: Secondary Verification of Effect and System Learning", using historical data such as temperature rise characteristics, time, and energy consumption during this de-icing process to update the risk prediction model and control parameters online (self-learning). Subsequently, it controls the high-voltage relay to close, enabling the vehicle to power on and operate normally.

[0109] If the verification result is unsuccessful, it indicates that the relay is still stuck due to ice or has not completely melted. The system then needs to further determine whether the number of retry attempts has exceeded the limit (i.e., whether the current number of de-icing retry attempts has reached the preset limit).

[0110] If the number of retries does not exceed the limit (No): The system returns to stage 5, adjusts the heating strategy, and re-executes the de-icing operation to conduct a new round of de-icing and verification.

[0111] If the number of retries has exceeded the limit (Yes): The system determines that the fault cannot be eliminated after multiple heating and de-icing. In order to prevent the relay from being damaged by forced closure or causing a high-voltage short circuit safety hazard, the system directly enters "Phase 8: Fault Safety Handling and System Degradation" and executes the degradation strategy (such as permanently locking the relay closure, disconnecting the high-voltage circuit, and sending a serious fault warning to the instrument or remote vehicle terminal).

[0112] 5. Cyclic reset.

[0113] Once the vehicle successfully powers on and operates normally through stage 4 or stage 7, after its operation ends and the vehicle is powered off again, the vehicle control system resets and begins a new cycle of detection and protection.

[0114] In the embodiments of this application, in order to achieve accurate and forward-looking assessment of the risk of icing in high-voltage relays and avoid false triggering and missed triggering caused by a single threshold judgment, this application... Figure 3 The "Phase 1: Environmental Perception and Icing Risk Prediction" section has been designed with multiple dimensions and details.

[0115] Combination Figure 4 , Figure 4 It shows Figure 3A flowchart illustrating the environmental perception and icing risk prediction process. The system is activated when the vehicle meets specific triggering conditions, and the specific execution steps are as follows: Process start conditions When the system detects that the vehicle's high-voltage power supply is disconnected and enters a sleep state (i.e., the high-voltage circuit is cut off and the system enters a low-power standby mode), or when the preset wake-up cycle inside the system (e.g., but not limited to every 30 minutes) is reached, the controller is woken up and immediately triggers and starts the workflow of stage 1.

[0116] 1. First step: Synchronous acquisition of data from multiple sensors.

[0117] To eliminate timing misalignment caused by sampling delays and bus transmission delays between different sensors, and to ensure that the physical quantities used to calculate temperature differences are based on the same time reference, the system first performs synchronous acquisition of data from multiple sensors: Step 1.1: The controller sends hardware-level synchronization trigger signals to sensors deployed throughout the vehicle to ensure that all associated sensors sample data at the same microsecond interval. Subsequently, the controller reads the following key physical quantities in parallel via a parallel bus or dedicated channel: Step 1.2, collect the ambient air temperature outside the vehicle (denoted as ). ).

[0118] Step 1.3, collect the relative humidity of the external environment (denoted as ). ).

[0119] Step 1.4: Collect the temperature sensor reading attached to the housing of the high-voltage relay (denoted as...). This reading is close to the surface of the relay's metal casing and is used to accurately characterize the microscopic local temperature of the relay's internal contacts.

[0120] Step 1.5: Obtain the current internal temperature of the high-voltage battery pack from the battery management system via the CAN local area network bus or other vehicle communication bus (denoted as ). Because the battery pack has a large thermal mass, its radiated and conducted heat is an important heat source affecting the microscopic thermal environment around the relay.

[0121] Step 1.6: Data Packaging and Timestamp Marking. The controller encapsulates and packages the raw data from different sources and in different formats into a unified sensor monitoring data packet according to a predetermined protocol format, and uses the high-precision clock inside the controller to mark the data packet with a unified sampling timestamp.

[0122] Step 1.7: The data cleaning and preprocessing controller cleans the data packets, including verifying the validity of each physical quantity (removing out-of-bounds values) and performing low-pass filtering to denoise, so as to output high signal-to-noise ratio reference physical data.

[0123] 2. Second step: Calculate key physical characteristics.

[0124] Based on the cleaned baseline physical data, the controller performs real-time analytical calculations of thermodynamic and physical characteristics: Step 1.8, Calculate the dew point temperature using the ambient air temperature collected in Step 1.2. And the ambient relative humidity collected in step 1.3 As input, the dew point temperature of the current outside air is accurately calculated using a preset dew point physical calculation formula (such as the Goff-Glach formula or the Antoine modified formula). ).

[0125] Step 1.9, Calculate the critical temperature difference. Calculate a critical temperature difference value to directly characterize the risk of water vapor condensation: using the relay housing temperature obtained in Step 1.4. Subtract the dew point temperature calculated in step 1.8 The real-time temperature difference value is obtained. The closer this difference is to zero or less than zero, the higher the probability of condensation and subsequent icing on the relay surface.

[0126] Step 1.10: Calculate the instantaneous rate of temperature change and analyze the high-voltage relay housing temperature over the most recent preset time period (e.g., within a few minutes). The system uses historical data to calculate the instantaneous rate of temperature change. If the instantaneous rate is negative and its absolute value exceeds a preset rate threshold, the relay is determined to be experiencing a rapid cooling trend, indicating that the risk of condensation and icing is accumulating rapidly.

[0127] 3. The third step: Obtain prediction auxiliary information.

[0128] To account for future environmental variables and differences in vehicle hardware, the controller obtains forward-looking auxiliary reference information through a communication network: Step 1.11: The cloud-based weather forecast data acquisition controller connects to the cloud server through the vehicle-mounted telematics unit (T-Box) to acquire refined weather forecast data for the next preset time period (e.g., the next 6 hours) based on the vehicle's current geographical location, focusing on extracting the future evolution prediction curves of temperature and humidity.

[0129] Step 1.12: Location and Spatial Parameter Acquisition. Obtain the vehicle's current Global Positioning System (GPS) coordinates and altitude. The climatic humidity characteristics of different regions and the air pressure changes caused by different altitudes are used to adaptively correct the parameters of the dew point calculation formula and icing prediction model.

[0130] Step 1.13: Read the relay's full life cycle historical operation and health record. The controller reads the personalized health record of the currently tested relay (including historical cumulative number of engagements, records of abnormal temperature differences, historical maximum contact resistance baseline, etc.) from the vehicle's non-volatile memory to correct the boundary conditions of the prediction model.

[0131] 4. Fourth step: Run the three prediction models in parallel and integrate them with the decision.

[0132] It should be understood that the risk prediction model in this embodiment includes a physical heat conduction model, a statistical analysis model, and an empirical prediction model.

[0133] After acquiring real-time physical characteristics and prediction aids, the controller runs three different dimensions of mathematical and physical models in the risk prediction model in parallel within the computing architecture to improve the robustness of risk assessment.

[0134] Step 1.14: Run the "Physical Heat Conduction Model" based on one-dimensional or multi-dimensional transient heat conduction equations, combined with the heat source of the high-voltage battery pack. The heat transfer effect and the shell temperature Thermodynamic simulation was performed on the real-time temperature evolution of the internal contact cavity of the relay to calculate the probability of icing risk based on the physical mechanism. .

[0135] Step 1.15: Run the "Statistical Analysis Model" based on a multiple regression algorithm or a random forest algorithm, using the current critical temperature difference. Using factors such as temperature change rate, altitude, and current ambient humidity as input features, the statistical probability of icing is calculated based on historical statistical patterns. .

[0136] Step 1.16: Run the "Empirical Prediction Model". Based on the acquired health record data and expert experience mapping rules, combined with the future weather forecast temperature and humidity curves, assess the deterioration trend of the vehicle during continuous parking and calculate the empirical probability of icing. .

[0137] After the model calculations in steps 1.14, 1.15, and 1.16 above are completed, the risk probabilities and feature data output by each model are incorporated into the multi-model fusion decision module (corresponding to...). Figure 3The real-time risk level determination logic in the module is used to output the final icing risk level through a weighted fusion algorithm or fuzzy control decision, so as to guide the subsequent active prevention action in stage 2 or the power-on diagnosis process in stage 3.

[0138] Based on the above description, the data acquisition and dew point calculation in this embodiment will be explained by way of example.

[0139] In one implementation, acquiring the vehicle's environmental parameters and the temperature data of the high-voltage relay may include: The system acquires the ambient air temperature and relative humidity of the vehicle's environment and uses these as environmental parameters; it also acquires the readings of the temperature sensor attached to the housing of the high-voltage relay and uses these readings as the temperature data of the high-voltage relay.

[0140] The controller can send synchronous sampling commands to each sensor and configure a unified timestamp for the collected data. Considering that vehicle sensors typically transmit data through different buses or interfaces, "synchronous sampling" can be hardware synchronous sampling or software time alignment by the controller based on the timestamps of each data point. It is not required that each sensor complete sampling at exactly the same physical moment.

[0141] The controller can perform range verification, rate of change verification, message age verification, filtering, and outlier removal on the collected data. For example, the effective range of ambient relative humidity can be set to 0%RH to 100%RH; if the humidity reading exceeds this range, or the data age of the corresponding message exceeds a preset value, the data will be marked as invalid.

[0142] The controller determines the dew point temperature based on the ambient air temperature and relative humidity. In one specific embodiment, the dew point temperature is calculated using the following formula:

[0143]

[0144] in, This indicates the relative humidity of the environment, expressed in %; This indicates the ambient air temperature, expressed in degrees Celsius. This indicates the dew point temperature, expressed in degrees Celsius. This represents the natural logarithm function.

[0145] The difference between the relay temperature data and the dew point temperature is expressed as:

[0146] in, This indicates the temperature of the high-voltage relay housing. This represents the difference between the relay housing temperature and the dew point temperature.

[0147] when When the voltage decreases, the high-voltage relay approaches a state where condensation occurs; when At this time, the temperature of the high-voltage relay casing is not higher than the dew point temperature, posing a risk of condensation. If the relay temperature drops further below the freezing point, there is a risk of frosting or icing.

[0148] To compensate for temperature sensor errors, the deviation between the outer casing temperature and the relay's internal temperature, and localized thermal bridging effects, a preset temperature difference threshold can be set to a value greater than 0. For example, the preset temperature difference threshold can be set to 2℃. When the difference is less than or equal to 2℃, the risk of icing is identified in advance.

[0149] Immediately, setting the threshold too low may cause the controller to activate anti-icing and insulation only after condensation has already formed; setting the threshold too high may increase the frequency of false triggers and increase sleep energy consumption. Therefore, the preset temperature difference threshold can be calibrated through environmental chamber tests for the corresponding vehicle model and relay assembly.

[0150] Next, we will provide an illustrative example of how the trend of time evolution can be used to determine high risk.

[0151] In one implementation, determining the icing risk status of a high-voltage relay based on the difference includes: The system acquires weather forecast data for a preset future time period and combines it with the temperature of the vehicle's high-voltage battery pack to construct a time evolution trend of the difference within the future time series; and determines the icing risk status of the high-voltage relay as high-risk when the time evolution trend indicates that the difference is less than or equal to a preset temperature difference threshold.

[0152] For example, the preset future time period can be 6 hours. The controller obtains the predicted ambient air temperature and relative humidity data for the next 6 hours through the T-Box. The weather forecast data can include a forecast point every hour, and the controller obtains the ambient temperature and humidity sequences at minute intervals through linear interpolation.

[0153] The controller also acquires the current temperature of the high-voltage battery pack. Since the high-voltage battery pack continues to act as a heat source for the area where the high-voltage relay is located even after the vehicle is powered off, the temperature of the high-voltage battery pack can affect the future temperature trend of the high-voltage relay.

[0154] The first in the future time series The difference in minutes can be expressed as:

[0155] in, For the first The relay predicts the temperature within minutes. According to the first The dew point temperature is calculated from the ambient temperature and relative humidity over several minutes.

[0156] The time evolution trend indicates that the difference is less than or equal to the preset temperature difference threshold. This can be that at least one difference in the future time series is less than or equal to the preset temperature difference threshold, or that the difference is continuously less than or equal to the preset temperature difference threshold within a preset duration.

[0157] For example, the preset temperature difference threshold is 2℃, and the preset duration is 20 minutes. If the forecast shows that the temperature difference between the 80th and 110th minutes in the future is consistently less than or equal to 2℃, the controller will determine the icing risk status as high-risk. Using the duration condition can reduce misjudgments caused by fluctuations in individual weather forecast points.

[0158] In one specific implementation, temperature conditions can be combined with the difference condition for judgment. For example, when the future difference is less than or equal to 2°C and the future relay temperature is less than or equal to 0°C, it is determined to be a high-risk level; when the difference is less than or equal to 2°C but the relay temperature is always higher than 0°C, it can be determined to be a condensation risk or a medium-risk level.

[0159] The heat conduction logic in the embodiments of this application will be described by way of example.

[0160] In some implementations, the controller can input the difference, environmental parameters, high-voltage battery pack temperature, and preset relay historical operation records into the risk prediction model, and use the heat conduction logic in the risk prediction model to calculate the first risk probability.

[0161] Figure 5 A schematic diagram of the heat conduction logic operation process is shown. (For example...) Figure 5 As shown, the operation process of the physical heat conduction logic specifically includes the following steps: Step 1: Data input and preprocessing.

[0162] The controller first reads and prepares the input data required by the model: 1.1 Read the current measured temperature data: including the current relay housing temperature (Unit: °C) Current ambient air temperature (Unit: °C), and the current high-voltage battery pack temperature. (Unit: °C)

[0163] 1.2 Obtaining Future Weather Forecast Data: The system acquires the ambient temperature forecast sequence for the next 6 hours (containing 6 hourly data points, denoted as...) via vehicle-to-everything (V2X) network (in-vehicle T-Box). to The controller performs linear interpolation on the hourly forecast data, converting it into a minute-by-minute forecast sequence of ambient temperature for the next 360 minutes. This includes the corresponding relative humidity forecast sequence. and predicted relative humidity series (in ).

[0164] 1.3 Read historical temperature data: Read the relay housing temperature from non-volatile memory 1 hour ago. (Unit: °C) Historical ambient air temperature series in minutes over the past hour (Unit: °C), and the historical temperature sequence of the high-voltage battery pack in minutes over the past hour. (Unit: °C)

[0165] 1.4 Read calibration model parameters: Read the preset temperature response coefficient (unit: Typical value ) and environmental heat dissipation weight (Units not specified, typical value) ).

[0166] Step 2: Core model calculation and master prediction.

[0167] The controller's prediction time domain for the next 6 hours (i.e., the total number of minutes predicted) Within a minute, iterative predictions are made based on the laws of heat conduction: System initialization: Set the initial value of the relay housing temperature prediction for minute 0. .

[0168] Iterative calculation step by step: For the th minute( The controller performs the following processing in a loop: 2. Estimate the battery pack temperature for the current minute: The controller establishes a high-voltage battery pack temperature decay estimation model. In a simplified implementation, the high-voltage battery pack temperature... The following formula can be used to estimate:

[0169] It is worth mentioning that the above-mentioned decrease of 0.01℃ per minute is only a simplified example. In actual applications, the controller can establish a more accurate nonlinear battery temperature decay model based on the current thermal capacity of the battery pack, ambient temperature, vehicle parking time, and thermal management status.

[0170] 2.1. Calculate the first... Comprehensive temperature and pressure difference per minute :

[0171] In the formula, the first term represents the heat conducted from the high-voltage battery pack to the relay as a heat source, and the second term represents the heat dissipated from the relay to the external environment. Used to balance the weighting relationship between the two.

[0172] 2.2. Predict the relay housing temperature for the next minute. :

[0173] 2.3. After the iteration is completed, the controller obtains the predicted relay housing temperature sequence for the next 6 hours. .

[0174] 2.4. Calculate the minute-by-minute predicted dew point temperature: using the predicted relative humidity sequence and predict ambient temperature The controller first calculates the intermediate variables. :

[0175] Then, the dew point temperature is predicted minute by minute. :

[0176] This yields the predicted dew point temperature sequence for the next 360 minutes. .

[0177] 2.5. Calculate the first risk probability (physical icing risk probability). The controller compares the predicted casing temperature with the predicted dew point temperature every minute. If the conditions are met... If the high-voltage relay casing is at risk of icing for that minute, the number of risk minutes is accumulated. Physical icing risk probability Represented as:

[0178] in, The range is .

[0179] Step 3: Model Confidence Assessment To evaluate the accuracy of the heat conduction prediction model under the current local thermal environment of the vehicle, the controller was validated using short-term backtesting data from historical measurements: 3.1 Short-term backtracking calculation: Using the measured relay casing temperature from 1 hour ago (i.e., 60 minutes ago) as the initial value, let... For the first time in history minute( ), using the measured ambient temperature sequence of the past hour and battery pack temperature sequence Perform iterations:

[0180]

[0181] After 60 iterations, the backtracking predicted temperature value for the current moment is obtained. .

[0182] 3.2 Calculate the prediction error Calculate the absolute error between the retrospectively predicted temperature and the currently measured casing temperature:

[0183] 3.3 Calculate the physical confidence level Based on prediction error (Unit: °C) Perform segmented confidence level assessment: like ℃, then confidence level ; like At ℃, the confidence level decreases linearly with the error: ; like ℃, then confidence level .

[0184] Output of the physical heat conduction model: The final output of the physical heat conduction prediction model is a binary tuple. For example, when the output tuple is The physical meaning of this is: the physical heat conduction model predicts that the high-voltage relay will be at risk of freezing for 33.3% of the time period (i.e., 120 minutes) within the next 6 hours, and the model has 100% confidence in this prediction.

[0185] Example calculation analysis: Assume the current baseline physical conditions are as follows: ℃, ℃, ℃, and the historical temperature 1 hour ago is known. ℃, calibration model parameters are , .

[0186] The master forecast calculation for the first minute of the future: The comprehensive temperature and pressure difference at minute 0 is:

[0187] The predicted casing temperature in the first minute is:

[0188] Confidence backtesting assessment: If backtesting is performed using data from the past hour, the predicted temperature value for the current moment can be calculated. If the temperature is ℃, then the prediction error is:

[0189] Due to error ℃, the controller determines the physical confidence level of this prediction. .

[0190] Parameter calibration and optimization methods: To ensure the model closely matches the specific vehicle relay mounting structure and material properties, the model parameters... and The following results were obtained through environmental chamber testing and calibration: Environmental chamber test: The relay assembly and high-voltage battery pack simulation structure are placed in an environmental chamber and different ambient temperatures are set (e.g., -10℃, 0℃, 10℃, 20℃, etc.).

[0191] Multi-channel data acquisition: During the experiment, the measured sequence of relay housing temperature was recorded synchronously at a high sampling rate. , ambient temperature measurement sequence and simulated or actual battery pack temperature measurement sequence.

[0192] Least squares parameter fitting: The experimental data are fitted and optimized using the least squares method. Let the model predict the temperature as... The objective function for parameter fitting is:

[0193] Find the objective function through optimization algorithms. The parameter value that is minimized is obtained, thereby outputting the optimal calibration parameter. and .

[0194] Validation and segmented parameter setting: After parameter calculation, the model accuracy is cross-validated using independent experimental data not involved in the fitting. To further improve accuracy, the controller can be segmented and different combinations of calibration parameters can be called according to different measured ambient temperature ranges (such as extremely cold ranges, near-zero temperature ranges, etc.). .

[0195] In addition to executing the physical heat conduction logic through the risk prediction model, this embodiment of the application also runs a statistical analysis model in parallel to execute statistical regression logic to calculate the second risk probability. The system then performs a multi-dimensional confidence assessment on the prediction results and outputs the confidence score. .

[0196] Next, the statistical regression logic in this embodiment will be explained by way of example.

[0197] Figure 6 This diagram illustrates the statistical regression logic performed by the statistical analysis model. Figure 6 As shown, the controller can also calculate the second risk probability using the statistical regression logic of the statistical analysis model in the risk prediction model. For example, the execution process of the statistical regression logic specifically includes the following steps: Step 1: Data input and feature construction.

[0198] 1.1 Reading Current Feature Data: The controller acquires seven core feature parameters in real time to characterize the vehicle's current state and historical trends: Current temperature difference The physical difference (unit: °C) between the current high-voltage relay casing temperature and the ambient air dew point temperature is the most direct physical characteristic reflecting the risk of immediate icing.

[0199] Current ambient air temperature (Unit: °C): Characterizes the heat dissipation boundary conditions between the current external environment of the vehicle and the local micro-environment of the relay.

[0200] Current ambient relative humidity (Unit: %): Used for analytical calculation of dew point temperature, determining the physical critical line for freezing.

[0201] Current high voltage battery pack temperature (Unit: °C): Represents the temperature of the maximum active heat source within the high-pressure system.

[0202] Current moment The range of values ​​is The integer represents the corresponding hour of the day, used to capture the periodic fluctuations in temperature caused by the alternation of day and night.

[0203] Historical temperature range (Unit: °C): Defined as the temperature within a preset number of days in the past (e.g., 7 days) relative to the current time. The statistical average of the actual temperature differences recorded at the same hour.

[0204] Temperature difference trend (Unit: °C / hour): Calculated as the rate of change of relay temperature difference over the past hour. When this value is negative and its absolute value increases, it indicates that the relay temperature is rapidly approaching the dew point temperature, and the risk of icing increases significantly.

[0205] 1.2 Reading Historical Datasets: The controller maintains a structured runtime feature database for model parameter learning. This database contains: Historical Feature Matrix :for A matrix of dimension, where This represents the number of valid historical records. Each row of the matrix corresponds to a combination of 7 feature parameters and a bias term of 1 at a specific historical moment:

[0206] Historical tag vector : for length of A column vector. Each element represents the actual measured relay temperature difference value after a preset prediction span (e.g., 3 hours) following the corresponding historical moment. :

[0207] Model parameter construction and training: The configuration parameters of a statistical regression model are represented by a weight vector. :

[0208] In the formula, For the intercept weights, the rest... These correspond to the weight coefficients of the seven features mentioned above. The controller uses the least squares method to solve for the optimized weight parameters:

[0209] To prevent matrix inversion from becoming infeasible or overfitting due to multicollinearity of features, in one implementation, the controller employs a... Ridge regression algorithm for regularization term solution of configuration parameters:

[0210] In the formula, The regularization coefficient is . For An identity matrix of dimension 1. Note: Based on the number of valid historical samples. During the insufficient cold start phase, the statistical regression logic first calls the initial configuration parameters generated during the development phase based on bench tests of the same model relay and equivalent installation structure; as the actual mileage and data of the vehicle accumulate, the controller periodically adjusts... and Perform online updates to resolve the problem. To achieve adaptive evolution of individualized vehicles.

[0211] Step 2: Calculation of primary forecast and linear regression.

[0212] 2.1 Feature vector construction, such as constructing the current feature vector The controller assembles the above 7 input features with a bias term (constant 1) to construct the feature vector of the current column:

[0213] 2.2 Future Temperature Difference Prediction Calculation: The controller utilizes the current feature vector and weight vector Perform a product operation to predict the relay temperature difference value at a preset time in the future (e.g., 3 hours later). Linear prediction formula:

[0214] 2.3 Risk Probability Calculation, i.e., Second Risk Probability Mapping: To map the continuously changing temperature difference prediction value to an intuitive risk probability, the controller uses the Sigmoid mapping function to calculate the second risk probability. :

[0215] In the formula, The preset safe temperature difference threshold (e.g., 2.0℃). A steepness coefficient (e.g., 2.0). When predicting temperature difference. Significantly greater than hour, Approaching 0 (safe); when the predicted temperature difference is below hour, Increase rapidly until the predicted temperature difference is much lower than expected. hour, Approaching 1 (high risk of freezing).

[0216] Step 3: Confidence Assessment and Multidimensional Validation The controller evaluates the reliability of the statistical prediction results from three different dimensions and calculates the statistical confidence level. : 3.1 Data Adequacy Assessment : Verify the amount of historical valid data accumulated in the training database Whether the confidence requirement is met. The calculation formula is:

[0217] In the formula, This is a preset minimum sample size requirement threshold (e.g., 168 samples, corresponding to the number of samples taken per hour over 7 days).

[0218] 3.2 Data stationarity assessment Assess the degree of thermodynamic environmental fluctuations during the vehicle's recent operating phase. The controller calculates the standard deviation of the actual temperature difference sequence of the relays over the past 24 hours. The score is calculated using the following formula:

[0219] The more stable the temperature environment ( The smaller the value, the higher the validity of historical statistical patterns, and the higher the score.

[0220] 3.3 Feature Correlation Assessment The controller checks whether the current feature state is contained within the known historical experience space. The controller then calculates the current feature vector. (Removing constant terms) and historical feature matrix The cosine similarity of all historical records in the data is used to filter out the highest values. Similarity values ​​(e.g.) The average similarity score is calculated. The closer this average similarity is to 1.0, the higher the match between the current operating condition and historical operating conditions, and the higher the feature relevance score. The higher.

[0221] 3.4 Overall Confidence Calculation: The controller performs a weighted sum of the scores from the above three dimensions to obtain the overall confidence score of the statistical regression logic. :

[0222] Output of the statistical regression model. The final output of this statistical regression model is a pair of tuples describing the icing trend: .in, This represents the probability of icing, ranging from 0 to 1. The confidence level of the model for the above probabilities is 0-1.

[0223] Example Calculation Analysis Assume the currently collected feature parameters are: ℃, ℃, , ℃, , ℃, ℃ / hour. Preset parameters are: , , , , , , , (Assuming here) (calibrated weights), and safety threshold ℃, steepness coefficient .

[0224] Predicted temperature difference calculation:

[0225] Calculate the probability of the second risk:

[0226] Calculate the overall confidence level: Assuming the following is calculated: number of historical samples ( ) Standard deviation of temperature difference over the past 24 hours The average of the top 10 maximum cosine similarities The overall confidence level is:

[0227] Ultimately, the statistical regression logic outputs the following binary tuple: .

[0228] Next, we will provide an example of risk fusion decision-making.

[0229] The first risk probability is obtained from the physical heat conduction model. With confidence level and the second risk probability output by the statistical regression model. With confidence level Then, the controller enters the multi-model fusion module to calculate the comprehensive risk level used to indicate the final icing risk state of the high-voltage relay. .

[0230] In one implementation, the controller determines the overall risk level using a dynamically weighted fusion formula based on confidence levels.

[0231] It should be understood that when sensor malfunctions or special operating conditions cause the denominator to... At this time, the controller will trigger control degradation logic, directly based on the current temperature difference. Safety protection judgment is made based on the preset temperature difference threshold.

[0232] In another implementation, the controller may also use a weighted fusion formula to calculate the overall risk level:

[0233] In the formula, the weighting coefficients satisfy... In the early stages of vehicle operation or when the number of historical valid samples is less than a threshold, the controller sets physical logic to dominate the weighting, such as... , As the vehicle history sample database becomes more comprehensive and the statistical regression model achieves the required accuracy after cross-validation, the controller automatically increases... The proportion.

[0234] Risk level assessment and hysteresis stabilization control The controller calculates the overall risk level. Then, it is compared with the preset risk level threshold: when At that time, the icing risk status of the high-voltage relay was determined to be low risk level; when At that time, the icing risk status of the high-voltage relay was determined to be at a medium risk level; when At that time, the icing risk status of the high-voltage relay is determined to be high-risk level (meeting the first preset condition).

[0235] To prevent high-frequency oscillations in the control output caused by slight fluctuations in the overall risk level near the risk assessment threshold, the controller implements hysteresis control for risk level switching: when the overall risk level... Rise to A high-risk status is entered when the risk level is 1 or above, and only when... Descending to Only when the high-risk status is exited and the anti-icing and insulation functions are deactivated will the operation be lifted.

[0236] Next, an example of offline degradation prediction will be provided.

[0237] When weather forecast data is unavailable, the controller enters an offline degradation prediction mode to avoid the system relying entirely on wireless communication or cloud-based weather data.

[0238] The controller can calculate the changing trend based on the ambient air temperature sequence, ambient relative humidity sequence, and high-voltage battery pack temperature sequence of the most recent hour, and construct future environmental parameters based on the changing trend.

[0239] For example, the rate of change of ambient air temperature can be expressed as:

[0240] The rate of change of relative humidity can be expressed as:

[0241] Future ambient temperatures can be projected based on a decreasing trend:

[0242] Future relative humidity can be predicted using the following formula:

[0243] in, For the future time, and These are the decay time constants for the temperature trend and the humidity trend, respectively.

[0244] The controller can also read historical environmental parameters stored locally and use these historical sequences to correct offline forecast results. In offline degradation forecast mode, the safety margin for risk assessment can be increased. For example, the preset temperature difference threshold can be increased from 2℃ to 3℃, or the comprehensive risk threshold for high-risk levels can be reduced from 0.6 to 0.55 to decrease the possibility of missed assessments due to missing weather forecasts.

[0245] After communication is restored, the controller reacquires weather forecast data and switches the offline prediction sequence to the weather forecast sequence through a weighted transition method to avoid sudden changes in the prediction results.

[0246] Figure 7 A schematic diagram illustrates the process of an empirical prediction model performing offline degradation prediction. For example... Figure 7 As shown, the operation process of the empirical prediction model in the risk prediction model specifically includes the following steps: Step 1: Data input and feature standardization.

[0247] 1.1 Reading Current Feature Data: The controller acquires the 7 core feature parameters at the current moment in real time to form a feature vector. To ensure consistency and low coupling of the model input, these seven feature parameters are completely consistent with the input features of the aforementioned statistical regression model, specifically:

[0248] 1.2 Building and Loading the Historical Case Library: The controller loads the historical case library from non-volatile memory. This historical case library serves as the knowledge base for the empirical prediction model and includes: Feature matrix :for A matrix of dimensions, where each row represents the feature vector of a historical case before standardization at the time of recording. Total number of historical cases:

[0249] Tag vector : for length of A column vector, where each element corresponds to the future actual temperature difference value of a historical case after a preset time span (e.g., 3 hours). :

[0250] 1.3 Loading Model Configuration Parameters: The controller reads the preset model configuration parameters, including: number of nearest neighbors. (usually selected) (odd numbers between); safe temperature difference threshold (e.g., 2.0℃); mean vector calculated from historical case database. and standard deviation vector (Each includes scaling parameters for 7 corresponding features); maximum distance threshold used for confidence assessment (e.g., 10.0); and the maximum threshold for distance from standard deviation. (e.g., 2.0).

[0251] Step 2: Distance calculation and K-nearest neighbor search.

[0252] 2.1 Feature Standardization Processing: Due to the significant differences in the physical dimensions and numerical ranges of various feature parameters (e.g., humidity range is...), The temperature difference is generally in (between), to prevent features with large numerical ranges from dominating distance calculations, the controller adjusts the current feature vector. Standardize and eliminate dimensions:

[0253] Similarly, the elements in the historical case matrix are also converted into a standardized matrix using the corresponding mean and standard deviation. .

[0254] 2.2 Euclidean Distance Calculation: The controller traverses the historical case library and calculates the standardized current feature vector. With the feature vector of each historical case (in Multidimensional Euclidean distance between ) The distance calculation formula is as follows:

[0255] 2.3 K-Nearest Neighbor Search and Ranking: The controller calculates the current state and all... Distance set of historical cases Sort the data in ascending order of distance and extract the data with the smallest distance. Each historical case serves as the nearest neighbor set of the current state.

[0256] 2.4 Calculation of the third risk probability (empirical risk probability): The controller statistically calculates this... Among the nearest neighbor historical cases, the number of actual results less than the safe temperature difference threshold. (i.e., satisfy) (Number of neighbors). Third risk probability. Calculate according to the following formula:

[0257] Step 3: Confidence assessment.

[0258] To evaluate the reference value of the nearest neighbor cases found for the current operating conditions, the controller calculates the empirical confidence score from three dimensions: distance, consistency, and temporal distribution diversity. : 3.1 Calculation of average distance to neighbors: Calculate the selected... The average of the nearest neighbor distances The smaller the average distance, the higher the physical fit between historical cases and current working conditions.

[0259] 3.2 Distance Consistency Calculation: Calculate the... The standard deviation of the nearest neighbor distance The smaller the standard deviation, the higher the degree of clustering of nearest neighbor cases in the feature space, meaning that the empirically derived boundary conditions are more consistent.

[0260] 3.3 Neighbor Diversity Assessment: The controller obtains this... The distribution of nearest neighbor cases along the time dimension. If the nearest neighbor cases are sparsely distributed across multiple distinct dates (e.g., more than or equal to 3 distinct historical dates), it indicates that this type of feature belongs to a stable and universally regular physical state, and the controller will assign a diversity score. Set to 1.0; if most neighbors are generated on the same day, it is determined that there is data randomness, and the diversity score is reduced.

[0261] 3.4 Calculation of Comprehensive Empirical Confidence: The controller calculates the weighted sum of the above three evaluation results according to preset weights to obtain the empirical confidence. :

[0262] In the formula, the average neighbor distance score accounts for 60% of the weight, the distance consistency score accounts for 20% of the weight, and the neighbor temporal diversity score accounts for 20% of the weight.

[0263] Output of the empirical prediction model. The final output of the empirical prediction model is a tuple describing the risk of similar historical operating conditions: .in, This represents the probability of icing risk based on statistical analysis of similar historical cases, ranging from 0 to 1. This represents the model's confidence level for this prediction, ranging from 0 to 1. For example, (0.6, 0.85) indicates that among the K most similar cases in history, the risk occurred in 60% of cases, and the model has an 85% confidence level in this prediction.

[0264] Example calculation and analysis are as follows: Assume the currently collected feature parameters are: ℃, ℃, , ℃, , ℃, ℃ / hour. Model parameters are set as follows: , ℃, , .

[0265] Feature standardization and neighbor search: After the input features are standardized, the current standardized feature vector is obtained. The controller calculates the Euclidean distance between the sample and 1000 samples in the historical database, sorts them, and selects the 10 nearest neighbor cases. The corresponding normalized distance sequence is as follows: .

[0266] Risk probability calculation: Read the actual temperature difference label values ​​from the historical records of these 10 nearest neighbor cases; the corresponding sequence is as follows: Because the temperature difference is less than the safe temperature threshold. The number of samples at ℃ is 6 (i.e. ),therefore:

[0267] Calculate the overall empirical confidence level: Calculate the average distance between 10 neighbors: ; Calculate the standard deviation of the distance: ; Read the distribution of neighboring dates and determine that they come from 4 different dates. .

[0268] The overall confidence level is:

[0269] The final empirical prediction model outputs the following binary tuple: .

[0270] In some implementations, to overcome the limitations of a single prediction model under extreme or variable conditions, the risk prediction model of this application deploys a multi-model fusion decision module. This module normalizes and dynamically weights the outputs of the parallel-running physical heat conduction model, statistical regression model, and empirical prediction model, and outputs the final icing risk level and control decision instructions based on the multi-dimensional feature matrix.

[0271] Figure 8 A diagram illustrating the operation of the multi-model fusion decision module is shown. Figure 8As shown, the operation process of the multi-model fusion decision module specifically includes the following steps: Step 1: Input and data validation.

[0272] The controller first obtains the raw output data and global parameters of the three prediction models, and then performs data validation: 1.1 Physical Model Output: Obtain the binary output of the physical heat conduction model. .in, The physical temperature difference (in °C) between the relay housing temperature and the dew point temperature at a future preset time (e.g., 3 hours later) predicted by the physical model. The instantaneous confidence level of the physical model for this prediction (range: ).

[0273] 1.2 Statistical Model Output: Obtain the binary tuples output by the statistical regression model. .in, The predicted temperature difference (in °C) for the same future time as predicted by the statistical regression model. The instantaneous confidence level of the statistical model for this prediction (range: ).

[0274] 1.3 Empirical Model Output: Obtain the binary tuples output by the empirical prediction model. .in, The probability of icing risk at the same future time predicted by the empirical model (range: ); The instantaneous confidence level of the empirical model for this prediction (range: ).

[0275] Loading global parameters and hyperparameters: safe temperature difference threshold (Default is 2.0℃); Sigmoid mapping steepness coefficient (Default is 2.0); Dynamic weight synthesis hyperparameters (Immediate confidence weighting coefficient, e.g., 0.5) (Historical accuracy weighting coefficient, for example, 0.3) (Work condition matching degree weighting coefficient, for example, 0.2), and the following constraints are met: ; A very small positive constant used to avoid a denominator of zero (Default value is 0.001).

[0276] Step 2: Result normalization and risk conversion.

[0277] Since the predicted values ​​output by physical and statistical models are continuous temperature difference values, while the empirical model outputs probability values, the controller first uses a nonlinear transformation formula to normalize the temperature difference prediction values ​​into a risk probability of a uniform dimension: 2.1 Physical Model Risk Probability Transformation:

[0278] It is worth mentioning that when Significantly greater than the safety threshold hour, If the value is positive and the absolute value is large, the mapping probability is... Approaching 0 (safe state); when Significantly smaller than When, mapping probability Approaching 1 (high-risk state).

[0279] 2.2 Statistical Model Risk Probability Transformation:

[0280] 2.3 Empirical Model Risk Probability: Due to the output of the empirical model... The value already represents a risk probability, and the controller will directly adopt it.

[0281] Step 3: Dynamic weight calculation and weighted fusion.

[0282] The controller combines instantaneous confidence, historical performance, and real-time operating conditions, employing a three-dimensional coupling to calculate the final normalized weight coefficients of each model: 3.1 Calculate the freshness weight (base weight) based on immediate confidence:

[0283]

[0284]

[0285] 3.2 Calculate the accuracy weighting adjustment coefficient based on historical performance: The controller adjusts the accuracy weighting coefficient based on the historical performance of each model. The mean absolute error in the prediction , and Calculate the corresponding prediction accuracy coefficient:

[0286] Then, the historical accuracy weighting coefficient , , Perform normalization calculations:

[0287] 3.3 Calculate the matching degree weight coefficient based on the current operating condition features: The controller extracts the current real-time operating condition feature vector. It then sequentially searches and matches the data with a preset working condition matching rule table to determine the working condition matching degree coefficient. : Rule 1: Extreme Condensation Risk Conditions. When the current ambient air temperature is detected... ℃ and relative humidity When activated, rule 1 is set to a matching coefficient of 1. At this point, the decision-making weight of the physical heat conduction model is increased.

[0288] Rule 2: Common Steady-State Operating Conditions. When the absolute value of the rate of change of temperature difference is detected... When the temperature is ℃ / hour and the number of valid historical case records is greater than 1000, activate rule 2 and set the matching degree coefficient to ℃ / hour. At this point, the proportion of decision-making based on case-driven experience models should be increased.

[0289] Rule 3: Data-driven advantageous operating conditions. When the current ambient air temperature is detected to meet the... When the relative humidity changes steadily, activate rule 3 and set the matching coefficient to [value missing]. At this point, the proportion of decision-making based on linear regression statistical models should be strengthened.

[0290] Rule 4: Default average operating condition. If none of the above three rules are successfully matched, the default operating condition matching coefficient is [value missing]. .

[0291] 3.4 Calculate the final comprehensive weight: The controller uses hyperparameters to calculate the weights of the above three dimensions. , , By performing a linear combination, the final unnormalized combined weights of each model are calculated:

[0292] Next, the final combined weights of each model are normalized to obtain the final weight coefficients used for fusion calculation. , , :

[0293] The final output of the fusion computation: The risk probabilities and confidence scores of the three models are weighted and summed using the final weighting coefficients to output the final fusion risk probability at the system level. and the overall confidence level of the fusion system :

[0294]

[0295] Step 4: Risk level determination and decision instruction generation.

[0296] 4.1 Input Feature Association: The controller obtains the final fusion risk probability. Current measured physical temperature difference Confidence level of fusion system And the contribution of each model to the final decision (including their respective final weights and original output data).

[0297] 4.2 Risk Level Determination Logic: The controller determines risk levels based on a priority order of "high risk, medium risk, low risk". High risk level: If the current status meets the condition " "or" If any of the values ​​in "℃" are displayed, the controller will determine that the current high-voltage relay has a high risk of freezing.

[0298] Medium risk level: If no high risk assessment has been triggered, and the current status simultaneously meets the following conditions: "and" "Based on these two conditions, the controller determines that the risk of icing is at a medium risk level."

[0299] Low risk level: If the current state simultaneously meets the following conditions: "and" "℃", the controller determines that the risk of icing is low.

[0300] Abnormal situation handling: If the above conditions cannot be matched due to missing data or unclear edge conditions, the controller will execute a conservative defense strategy and will determine the risk status as medium risk level by default.

[0301] 4.3 Details of Decision Command Generation and Execution: Based on the determined icing risk status, the controller dynamically generates and issues a sequence of execution commands: High-risk level instruction (immediately activate proactive defense): The controller monitors at the highest frequency. (Sampling and evaluating once per second) Perform system status tracking; request the BMS to activate the auxiliary low-pressure heating circuit, turn on the PTC heater and other preheating devices in the relay compartment, and control the heating power to within... Within the target dynamic power range, the target will raise the local environment temperature of the relay to above 8°C within 30 minutes; activate the relay compartment dehumidification system equipped in the vehicle; send a red icing risk warning signal to the vehicle terminal (instrument panel or central control screen); set the maximum safety protection limit: the PTC single heating time limit is set to 2 hours, and the maximum temperature rise limit of the relay shell is set to 15°C to prevent heat damage; set the upgrade warning strategy: if the risk level does not degrade after 30 minutes of continuous anti-icing heating, send a fault record to the vehicle communication terminal and alert the driver.

[0302] Medium-risk level command (preparing for defense and high-frequency monitoring): The system will increase the monitoring frequency to [a higher level]. Focus on monitoring temperature difference trends The system will take measures to prevent sudden changes in the temperature and humidity of the vehicle. The heating components will remain in a preheating standby state (power output will not be implemented for the time being), and the baseline of control parameters required to start preheating at any time will be calculated and locked. A yellow warning icon will be sent to the vehicle terminal, and the current abnormal state will only be recorded in the background system log to avoid interfering with the user's normal vehicle use. An upgrade warning strategy will be set: if the risk level rises to high risk within 1 hour after entering the medium risk state, the system will immediately and automatically switch to the active defense control mode.

[0303] Low-risk instruction (maintaining normal energy-saving monitoring): The system enters energy-saving standby mode, shuts down unnecessary edge computing modules to reduce power consumption, and reduces the monitoring and evaluation frequency to normal. (Collect and evaluate every 10 seconds); Maintain the heating components and dehumidification system in a standby state; Set a periodic re-inspection strategy: In normal standby state, the controller will call the multi-model fusion module again every 10 minutes to perform a complete icing risk calculation.

[0304] Step 5: Structured Design of Module Output After completing the above calculations and decisions, the multi-model fusion decision module outputs a standard structured data packet. This data packet specifically includes the following: Part 1 (Basic Information): This includes the system timestamp of the current prediction execution, the software algorithm version number, the future prediction span value, and the calculation time for this iteration. Part 2 (Fusion Result): This includes the final fused probability of icing risk. Overall confidence level The first part (model details) includes the current risk level assessment result, the boundary conditions of the rules triggering the current risk level, and the current measured physical temperature difference and relative safe temperature difference margin of the relay. The second part (model details) includes the specific weight allocation for the three models, their respective original confidence scores and risk level outputs, and the weight coefficients currently used in the statistical regression model. The system includes: 1) the nearest neighbor matching index found by the empirical model; 2) the weighted average of the current feature matching; 3) the weighted average of the current feature matching; 4) the weighted average of the current feature matching; 5) the weighted average of the current feature matching; 6) the weighted average of the current feature matching; 7) the weighted average of the current feature matching; 8) the weighted average of the current feature matching; 9) the weighted average of the current feature matching; 10) the weighted average of the current feature matching; 11) the weighted average of the current feature matching; 12) the weighted average of the current feature matching; 13) the weighted average of the current feature matching; 14) the weighted average of the current feature matching; 15) the weighted average of the current feature matching; 16) the weighted average of the current feature matching; 17) the weighted average of the current feature matching; 18) the weighted average of the current feature matching; 19) the weighted

[0305] Next, we will provide an exemplary description of ice prevention and insulation during dormancy.

[0306] In some implementations, when the vehicle is in a dormant state and the icing risk condition meets a first preset condition, the battery management system outputs energy to drive the heating component to perform anti-icing and heat preservation operations on the high-voltage relay, including: After the vehicle disconnects from the high-voltage power supply, the current remaining charge of the high-voltage battery pack is obtained; and if the current remaining charge is greater than the preset safe reserve charge threshold and the icing risk status indicates that there is an icing risk, the battery management system is triggered to connect the auxiliary power supply circuit and control the heating components to heat at a maintenance power lower than the preset rated power.

[0307] Optionally, the first preset conditions may include: the vehicle is in a dormant state, the icing risk level is high, the current remaining charge of the high-voltage battery pack is greater than the safe reserve charge threshold, the heating components are not faulty, and the high-voltage battery pack meets the output energy conditions.

[0308] The safety reserve capacity threshold is used to reserve energy required for vehicle starting and essential low-voltage functions. For example, the safety reserve capacity threshold can be set to 20% SOC. Anti-icing and heat preservation are permitted when the current remaining charge of the high-voltage battery pack is greater than 20% SOC; anti-icing and heat preservation are stopped when the remaining charge drops to 20% SOC. To avoid frequent start-stop cycles near the threshold, anti-icing and heat preservation can be reinstated only when the remaining charge recovers to above 22% SOC.

[0309] The auxiliary power supply circuit may include a high-voltage battery pack, a DC / DC converter, a low-voltage bus, a heating drive switch, fuses, and heating components. Alternatively, it may employ a structure where the high-voltage battery pack supplies power to the heating components via a high-voltage heating drive module. The battery management system activating the auxiliary power supply circuit does not imply activation of the complete high-voltage main circuit used for vehicle propulsion.

[0310] In one example, the heating element has a preset rated power of 30W, and the anti-icing and heat preservation uses a maintenance power of 5W to 15W. The controller can adjust the average power of the heating element via pulse width modulation.

[0311] Under the corresponding relay heat capacity and -10°C environmental conditions, if the maintenance power is below 5W, the input heat may be less than the heat dissipated by the relay to the environment, making it difficult to prevent the relay temperature from dropping below the dew point temperature. If the maintenance power is above 15W and operates continuously, it will increase energy consumption during vehicle hibernation and may also cause a large temperature difference between the area near the heating element and other areas of the relay. Therefore, 5W to 15W is suitable as the hibernation maintenance power range in this embodiment.

[0312] It should be understood that 5W and 15W are example parameters derived from the corresponding relay structure. For heating components with larger heat capacity, different heat dissipation conditions, or different rated power, other sustaining power ranges can be determined through thermal balance tests.

[0313] The controller can perform closed-loop power adjustment based on the temperature difference. For example, the target temperature difference can be set to 3°C. When the actual temperature difference is below 3°C, the holding power is increased; when the actual temperature difference is above 3°C, the holding power is decreased; when the actual temperature difference is above 4°C for 10 consecutive minutes, heating is stopped. Heating is restarted when the actual temperature difference drops below 2°C again.

[0314] The anti-icing and heat preservation operation can be stopped when any of the following conditions are met: the icing risk level is restored to a low risk level; the current remaining power drops to the safe reserve power threshold; the heating component malfunctions; the temperature sensor data is invalid; or the vehicle receives a wake-up command.

[0315] Figure 9 A schematic diagram illustrating the specific process of predictive energy scheduling and proactive prevention is shown. Combined with... Figure 3 and Figure 9 In this embodiment, when the system determines that the icing risk status meets the first preset condition (i.e., it is determined to be at a high risk level, or meets the medium risk monitoring requirements) in "Stage 1", the controller moves from Stage 1 to "Stage 2: Predictive Energy Scheduling and Active Prevention". This stage aims to actively eliminate or mitigate the icing risk of the high-voltage relay by coordinating the scheduling of multiple heat sources and adaptive power control, while ensuring the overall vehicle energy efficiency. The detailed execution flow of Stage 2 is as follows: Step 1: Optimization and decision-making regarding control objectives.

[0316] The controller first calculates refined control target parameters based on the fusion output results from Phase 1: 1.1 Risk Information Analysis: The controller reads the overall risk level and final risk probability output from Phase 1. And the contribution weights of the three sub-models (physical model weights) Statistical model weights Empirical model weights ), which serves as the underlying input for developing preventative strategies.

[0317] 1.2 Weighted Control Target Calculation: To balance physical boundary safety and the effectiveness of historical experience, the controller uses a weighted algorithm to calculate the target value for this action: Target temperature calculate:

[0318] In the formula, 7.0℃ and 8.0℃ are preset physical and statistical safety reference temperatures. The relay temperature (in °C) of similar success cases retrieved by the empirical model.

[0319] Target heating rate calculate:

[0320] In the formula, The heating rate (unit: °C / min) is for similar successful cases.

[0321] 1.3 Heat Source Priority Determination and Energy Dispatch: The controller monitors the high-voltage battery pack temperature, state of charge (SOC), and system high and low voltage in real time, and executes heat source dispatching logic according to the principle of optimal energy efficiency. Priority rule: Battery pack residual heat > Relay-specific heater > PTC auxiliary heater > Air conditioning heating system.

[0322] Decision logic: If the temperature of the high-voltage battery pack is detected... If the vehicle is charging and the temperature is below ℃, the controller prioritizes using residual heat from the battery coolant or radiant heat for insulation; otherwise, it activates the relay-specific heater as the main heat source.

[0323] Step 2: Adaptive control execution.

[0324] The controller enters the corresponding execution branch based on the real-time risk level and implements closed-loop feedback control: Branch 2.1: High-Risk Active Defense Control Gradient heating power control: The controller implements a segmented heating strategy, starting with low pulse power in the initial stage to avoid electrical and thermal shock; during the main heating stage, the power is dynamically adjusted to match the temperature rise. When approaching When the temperature reaches a stable state, reduce the power to prevent overshoot.

[0325] Predictive-corrective closed-loop control: Utilizing the physical heat conduction model from stage one, it predicts future temperature trends in real time. If the predicted temperature will exceed 13°C in 5 seconds, the heating power is reduced by 30% in advance; if the predicted temperature difference is 1 minute later... If the temperature is still below 2°C, the power will be increased by 20% as compensation.

[0326] Multi-heat source collaborative management: While ensuring the output of the main heat source, the system automatically introduces battery waste heat assistance according to the energy flow status to minimize the additional power consumption of anti-icing operation.

[0327] Human-machine interaction feedback: Sends a red icing risk warning to the vehicle terminal and displays the anti-icing heating progress in real time on the central control interface.

[0328] Branch 2.2: Preparation for Enhanced Monitoring of Medium-Risk Areas Monitoring frequency shift: The system will increase the frequency of sensor sampling and risk assessment from the conventional... Upgraded to Focus on monitoring the trend of temperature difference changes .

[0329] System preheating standby: The controller drives the heating components to perform a self-test program and pre-calculates the optimal heating parameter package based on the current operating conditions, and the heating system enters a zero-power standby state.

[0330] Silent warning: The instrument panel only displays a yellow warning, which does not interfere with the driver's operation, and the system background maintains high sensitivity for monitoring.

[0331] Branch 2.3: Low-risk energy efficiency optimization monitoring: Sensor polling optimization: Main temperature sensor maintenance When operating at low frequency, non-core auxiliary sensors enter sleep mode.

[0332] Intelligent wake-up mechanism: The controller sets three wake-up thresholds: one is wake-up due to abnormal trends (such as...). ℃ / s); second, predictive risk wake-up (such as a sudden increase in risk level calculated by a simplified model to above 0.3); third, fixed-cycle timed wake-up (such as every 5 minutes).

[0333] Power consumption minimization control: turn off the power supply of the heating drive circuit, reduce the operating frequency of the main control chip, and run only a low-level simplified prediction algorithm.

[0334] Step 3: Intelligent exit judgment and post-exit monitoring.

[0335] The controller continuously monitors exit conditions during operation to achieve a smooth transition from a preventative state to routine monitoring. Exit condition determination: The following four indicators must be met simultaneously: Measured relay temperature Current physical temperature difference ℃; Overall risk probability reassessed in Phase One The above conditions remained stable within the preset 5-minute judgment window.

[0336] Gradual exit from control sequence: T1 (30s): Reduce the heating power to 50% of the current value and verify the temperature stability; T2 (60s): Reduce the power to 20% and observe whether the temperature difference shows a downward trend; T3 (120s): Maintain a small power of 5% for pulsed maintenance, and finally completely shut down the heater after confirming that there is no risk of rebound.

[0337] Post-monitoring period management: After a complete exit, the system enters a 5-minute high-frequency post-monitoring period. If the temperature stabilizes, return to the hibernation mode of Phase 1; if the temperature drops rapidly and abnormally, immediately restart the risk assessment.

[0338] Fault mode handling (system-level protection): Sensor redundancy protection: If a single temperature sensor fails, the controller calls nearby hotspot sensors to perform numerical estimation; if multiple failures occur, it forces a switch to a conservative anti-icing mode based on time and fixed power.

[0339] Heater failure protection: If the main heater fails, the controller attempts to switch to the backup auxiliary heat source; if the entire system fails to heat, the high-voltage closing logic is locked, and a fault record is pushed to the vehicle communication terminal (T-Box) and the driver is notified.

[0340] Calculation example (scene reconstruction): Assuming it's 3 AM, the ambient temperature is -2℃, and the humidity is 92%, the system outputs a high-risk instruction in Phase 1. ), model weights are The empirical model matched similar cases with a target temperature of 8.5℃ and a rate of ℃ / min.

[0341] Target parameter calculation: Target temperature ℃; Target rate ℃ / min.

[0342] Energy dispatch execution: Since the battery temperature is 18℃ and it is in slow charging state, the system prioritizes the use of waste heat generated by the battery pack charger for preheating. At the same time, the relay heater starts at 30% power and finally completes the anti-icing and heat preservation closed-loop process within 30 minutes, so that the relay temperature stabilizes at 8.2℃ (temperature difference reaches 4.0℃) and successfully exits the prevention state.

[0343] Next, we will provide an exemplary description of the wake-up self-test and the connection between anti-icing and de-icing.

[0344] In some implementations, after the vehicle receives a wake-up command, the controller performs a power-on self-test. The wake-up command can be triggered by vehicle unlocking, starting button press, remote start, scheduled departure, charging request, or high-voltage power-on request from the vehicle controller.

[0345] The second preset condition may include at least one of the following conditions: the current difference is less than or equal to the preset temperature difference threshold; the overall risk level reaches a high risk level; the temperature of the high-voltage relay is lower than the preset safe temperature; the insulation resistance does not meet the closing condition; the operating characteristic of the high-voltage relay indicates suspected icing; or the anti-icing and heat preservation during the dormant period has not achieved the target.

[0346] During dormancy, ice prevention and insulation take priority over de-icing during awakening. Specifically: If the anti-icing and heat preservation during the dormant period has enabled the high-voltage relay to meet the safe closing conditions, then after the vehicle is awakened, it does not need to perform a high-power de-icing operation and can directly enter the normal closing process. If the risk is reduced by anti-icing and insulation during dormancy, but the safe closure conditions are still not met upon waking, then the de-icing operation will be initiated. If the anti-icing and heat preservation during the dormancy period is not executed due to insufficient remaining power, system failure, or changes in risk prediction results, and the second preset condition is met upon wake-up, then the de-icing operation will be executed first, and the closing of the high-voltage relay will be temporarily suspended. The controller is not allowed to output a normal closing command until the de-icing effect has been verified.

[0347] For example, when the vehicle is woken up, if the high-voltage relay temperature is 3℃, the temperature difference is 4℃, and the insulation resistance meets the preset conditions, it can directly enter the normal closing process. If the high-voltage relay temperature is -4℃, the temperature difference is -1℃, and the overall risk level is 0.85, then the second preset condition is met, and the de-icing process is initiated.

[0348] In the embodiments of this application, in order to achieve closed-loop evolution of the system control strategy and overcome the negative impact of changes in the vehicle environment and component aging (such as thermal decay of the heating film and oxidation of relay contacts) on anti-icing performance, the system constructs an online self-learning and continuous parameter optimization engine. This engine logically inherits... Figure 3 "Phase 7: Secondary Validation of Results and System Learning" corresponds to the specific execution algorithm flow. Figure 10 The optimized control flow shown (which can also be called the self-learning optimization stage in the self-learning subsystem) is illustrated.

[0349] Figure 10 A schematic diagram of the online self-learning and continuous optimization process is shown. For example... Figure 10 As shown, the online self-learning and continuous optimization process specifically includes the following execution steps: 1. Triggering mechanism.

[0350] The self-learning and optimization process can be triggered by any of the following conditions: timed trigger: the system automatically wakes up and executes according to a preset time period (e.g., once a month); event trigger: when the system detects a continuous performance degradation of a preset degree in key electrical parameters (such as contact resistance, longer closing time) or thermal response indicators of the relay during daily power-on self-test or secondary verification, it is automatically triggered.

[0351] 2. Step 1: Data collection and cleaning.

[0352] 1.1 Collection of historical data from multiple sources.

[0353] Once the self-learning engine is triggered, it first collects multi-source data from various dimensions from onboard storage media (such as non-volatile memory NVM) or cloud databases: Data collection phase one (sensing and prediction): including the timestamp of each prediction; input feature vectors (including current temperature difference, ambient air temperature, ambient relative humidity, battery pack temperature, current time, historical temperature differences for the same period, and temperature difference trend); and the raw outputs of each sub-model (the temperature difference predicted by the physical model and its physical confidence level). Statistical model prediction of temperature difference and its statistical confidence level Empirical model risk probability and its empirical confidence level The final output of the fusion decision module (fusion risk probability) Overall confidence level (This includes determining the risk level and executing decision-making instructions), as well as the contribution weights of each model.

[0354] Phase Two (Energy Dispatch and Prevention) Data Collection: This includes the start and end times of each anti-icing / de-icing control cycle; multi-dimensional time-series characteristic curves during the prevention process (including time-series signal sequences of relay temperature, heating power, target temperature, ambient temperature, and ambient humidity); actual result data (i.e., the relay temperature difference actually measured at the predicted time point in the future); actual icing risk label values ​​(used to characterize whether the vehicle actually iced during this cycle, 1 indicates icing, 0 indicates no icing); abnormal safety event records (including logs of over-temperature, over-current, and communication failures); and energy consumption assessment data (including total power consumption and average power of a single operation).

[0355] Collect auxiliary background data, including basic vehicle information (VIN code, software version, geographical location parameters); environmental and meteorological background information (historical climate data, weather forecast deviation records); and vehicle operating status (driving, charging, and stationary status indicators, battery pack SOC, and total vehicle mileage).

[0356] 1.2 Data cleaning and standardization.

[0357] Because the sampling rates of each sensor are different and there is occasional interference, the system needs to clean the collected data: Multi-source time series alignment: The system aligns asynchronously sampled time series data to the same standard minute-level time base. If data is missing at a corresponding time point, the system uses a linear interpolation algorithm for resampling.

[0358] In the formula, ,and and These are adjacent valid measured data points.

[0359] Outlier identification and removal: The system sets physical safety boundaries, stipulating that if the temperature value exceeds... Relative humidity exceeds or heating power exceeds If this occurs, it is directly determined as a hard fault in the sensor. For occasional spike data within the boundary, the system adopts... The three-standard-deviation principle was used to statistically identify and eliminate such cases.

[0360] Missing value imputation: For a small number of missing values ​​in the data stream, the system uses the mean of adjacent valid values ​​for interpolation imputation; for large-area data loss (missing ratio exceeding a preset threshold), the entire historical record is directly removed; in some implementations, the system can also run an auxiliary prediction model to fit and estimate the missing features using other related features.

[0361] 1.3 Multi-level data storage and compression The system classifies and manages the cleaned historical data: high-precision time-series data within the most recent preset number of days (e.g., 7 days) is stored in local storage; historical data exceeding this time limit is asynchronously uploaded to the cloud server for long-term storage via the vehicle-mounted T-Box after lossy feature compression (e.g., using a fitting algorithm based on inflection point extraction to compress time-series curves).

[0362] 3. Data sufficiency verification.

[0363] Before entering the performance evaluation, the controller performs a data sufficiency check: determining whether the number of valid historical records within the evaluation period reaches the preset minimum evaluation data volume threshold. (For example strip): If not satisfied (No): The system terminates the current self-learning process and re-enters the waiting state until the next trigger cycle; If the conditions are met: the system officially enters "Step Two: Performance Evaluation and Problem Diagnosis".

[0364] 4. Second step: Performance evaluation and problem diagnosis.

[0365] 2.1 Evaluation of the accuracy of the prediction model The system quantitatively calculates the prediction errors of the three parallel prediction models during the evaluation period: Error assessment of physical heat conduction model: calculation of the first The predicted temperature difference error value for this test:

[0366] Then calculate the mean absolute error and root mean square error :

[0367] Error assessment of statistical regression models: Similarly, calculate the error sequence of the statistical regression model. and output the mean absolute error. and root mean square error .

[0368] Empirical prediction model probability error assessment: Since the output of the empirical model is the risk probability Actually, testing labels (1 indicates icing has occurred, 0 indicates no icing). The system uses a binary classification cross-entropy loss function to evaluate its probability bias:

[0369] And calculate the average cross-entropy loss value:

[0370] 2.2 Comprehensive Evaluation of the Effects of Integrated Decision-Making The system constructs a confusion matrix to evaluate the degree of agreement between high-risk warning judgments (corresponding to positive warnings) and actual icing results: TP (True Positive): The system determines it to be high-risk, and freezing actually occurs; FP (False Positive): The system judges it as high risk, but no freezing actually occurs; FN (False Negative, Missed Report): The system judges it as medium / low risk, but freezing actually occurs; TN (True Negative): The system determines it to be of medium / low risk, and no freezing actually occurred.

[0371] Based on this matrix, the system calculates the following core evaluation indicators:

[0372]

[0373]

[0374]

[0375]

[0376]

[0377] 2.3 Multi-dimensional working condition subdivision assessment.

[0378] To diagnose the failure characteristics of the model in specific sub-scenarios, the system performs the following interval slicing analysis: Temperature range analysis: This includes analyzing the ambient temperature range. Divided into , , , Four temperature ranges were used to statistically analyze the performance of each model within each range. And decision accuracy.

[0379] Humidity range analysis: This involves analyzing the relative humidity of the environment. Divided into , , The three intervals are analyzed in detail to assess the accuracy of condensation and icing prediction under high humidity conditions.

[0380] Time interval analysis: sampling time The model was divided into four time periods: early morning (0-6 am), morning (6-12 am), afternoon (12-18 pm), and evening (18-24 pm) to identify the systematic impact of diurnal thermal radiation rhythm on model performance.

[0381] 2.4 Generate a performance report The system integrates the above evaluation indicators to generate a performance evaluation report that includes an overall performance overview, a comparison of the performance of each model, performance differences under different operating conditions, identification of poor operating conditions, and targeted improvement suggestions.

[0382] 5. Step 3: Parameter Optimization and Adjustment 3.1 Retraining the Statistical Regression Model Weights The system is based on the collected data. The real dataset of valid samples (feature matrix) Actual future temperature difference label vector Retrain the regression weight vector: Feature standardization: for the matrix Each feature column in the dataset is Z-score standardized.

[0383] Ridge regression weights: Construct an extended feature matrix that includes the bias intercept term. (The first column is all 1s), solve for the new weight parameters. :

[0384] In the formula, This is the regularization coefficient, used to control overfitting; It is the identity matrix; This is the transpose of the extended matrix.

[0385] Training process: Data is sorted according to... The set is divided into a training set and a validation set, and the solution is obtained on the training set. On the validation set Minimize as the goal, through Cross-validation selects the optimal regularization coefficient. .

[0386] 3.2 Online updates of the experience model case library To maintain the timeliness and conciseness of the case study library, the self-learning engine dynamically updates the case study library: New case addition conditions: must simultaneously meet the following: complete control process data; clear actual results (whether icing occurs); sufficient Euclidean distance difference between the feature vector of the new case and the cases in the existing case library; and good control action effect (i.e., successful anti-icing and reasonable energy consumption).

[0387] Case deletion criteria: The case must meet any of the following conditions: the case has been stored for more than 180 days; the case has a sensor malfunction that causes feature blurring; or the case's output results deviate significantly from those of most similar cases in the library.

[0388] Case Library Mean and Standard Deviation Update: After updating the case library, the system updates the mean and standard deviation of the seven features. and standard deviation Perform a recalculation and update:

[0389] In the formula, This represents the total number of cases in the current case library. For the first The first case Each feature value.

[0390] 3.3 Optimization of Fusion Weight Parameters During the parameter optimization phase, the system aims to find the optimal combination of fused hyperparameters. The optimization problem can be expressed as:

[0391]

[0392] The system uses a grid search algorithm or a Bayesian optimization algorithm to sample multiple sets of parameter combinations within the aforementioned constrained parameter space, uses historical backtracking data to simulate the fusion decision effect, sorts them according to the comprehensive score, and selects the set of parameters with the highest score as the final optimization result.

[0393] 3.4 Optimization of Control Parameters and Safety Thresholds PID control parameter self-tuning: Based on the overshoot and response time during the temperature rise control process, the system uses the Ziegler-Nichols closed-loop tuning method to fine-tune the heating PID parameters. Temporarily set the integral and derivative actions to 0, and gradually increase the proportional gain. Until the system casing temperature response curve shows a sustained constant amplitude oscillation; Record the critical gain at this point. and oscillation period ; Calculate the new PID parameters using the following formula:

[0394] In some implementations, the system can also use reinforcement learning algorithms to directly tune the optimal PID parameters from historical control processes.

[0395] Protection threshold optimization: The system analyzes historical over-temperature logs and, in conjunction with the relay's limit specifications, adaptively lowers or raises the safe temperature rise threshold.

[0396] 6. Step Four: Verification and Deployment 4.1 Comparison and significance test of new and old parameters Before the parameter and model update packages are distributed and deployed, rigorous offline comparison and verification must be performed: A / B test design: In cloud simulation test platform or real vehicle dual controller comparison test, the test samples are divided into experimental group (running a new model with new parameters) and control group (running the old model), and parallel computing is performed under the same working condition data source.

[0397] Comparison metrics: including various models The percentage reduction, the improvement in fusion decision accuracy, the reduction in false alarm and false negative rates, and the changes in average heating time and energy consumption.

[0398] Statistical significance test: The system uses two independent samples. The test was used to examine the difference in the overall scores between the experimental group and the control group, and the calculated two-tailed test probability was obtained. The value must satisfy:

[0399] In other words, the new model can only pass validation if it is confirmed at a 95% confidence level that its performance is significantly better than that of the old model.

[0400] 4.2 Safety and Stability Regression Testing The new parameters need to be simulated under extreme cold and high humidity conditions, and a comprehensive regression test needs to be completed to ensure that the updated model will not experience performance degradation or thermal control loop oscillation under normal operating conditions where it previously performed well.

[0401] 4.3 Update Package Generation After successful verification, the self-learning system generates a software update package in standard JSON format, which includes: new weights for the statistical model. Update the difference list in the case library (including adding and deleting cases); add new fusion hyperparameters. ; Optimized PID control parameters and protection thresholds; Updated metadata (version number, generation timestamp, applicable VIN number / vehicle model, data verification and checksum).

[0402] 4.4 OTA Deployment and Rollback Mechanism Phased Deployment Strategy: The system will be deployed in stages via over-the-air (OTA) updates: first deployed to... The pilot vehicle was observed and operated for 3 days; after confirming no abnormalities, it was deployed to... The vehicles were observed for 3 days before being extended to other areas. The vehicles; ultimately achieving a full upgrade.

[0403] Rollback mechanism: If occasional fault codes or abnormal degradation of performance indicators appear in the vehicle monitoring log after deployment, the controller will immediately trigger a rollback command to restore the model parameter package of the previous stable version.

[0404] 7. Complete example of computation and self-learning.

[0405] 7.1 Scene Setting and Current State Data collection period: The past 30 days, the effective cumulative historical data volume This is the first complete control record.

[0406] Current model parameters: Old values ​​of statistical regression weights ; Old fusion weight (Instant confidence level) (Historical performance) (Working condition matching degree); The experience case library contains 500 historical cases. Old PID control parameters: .

[0407] 7.2 Performance Evaluation Results Based on the performance evaluation logic calculations in Phase 3, the performance of each model is as follows: Physical Model Statistical model The empirical model has an average cross-entropy loss of 0.15; the fusion decision performance is as follows: accuracy is... The false alarm rate is The false negative rate was .

[0408] Calculate the overall score of the current decision:

[0409] 7.3 Parameter Optimization Calculation Process Statistical regression weight retraining: self-learning engine extracts this For a set of effective feature samples, after feature standardization, a regularization coefficient is set. The new weights are obtained by solving the problem using ridge regression:

[0410] The mean absolute error of the statistical model was evaluated on the validation set. from Reduce to .

[0411] Fusion weight optimization: The system uses grid search to find the optimal three-dimensional weights and obtain the optimal combination. Under this weighting, the backtest composite score from Upgraded to .

[0412] Case library update: The system automatically added 50 new highly representative cases and removed 30 old cases, updating the case library to 520 cases.

[0413] 7.4 Validation and Deployment Decisions The system conducted A / B gray-scale testing (the test period was 7 days, with 50 vehicles in each of the control and experimental groups): Experimental group (applying new parameters): Decision accuracy improved to The false alarm rate was reduced to The false negative rate decreased to The actual test score reached .

[0414] Control group (maintaining old parameters): Decision accuracy was The false alarm rate is The false negative rate was The overall score is .

[0415] Significance verification: The experimental group showed a significant improvement in overall score compared to the control group. Two independent samples Test results The difference was confirmed to be statistically significant.

[0416] Deployment: The system generates an OTA update parameter package with version number V1.1, completes the initial distribution to 10 test vehicles, and after confirming that no security restrictions are triggered within 3 days, pushes it to all vehicles for deployment.

[0417] 8. Definition of System-Level Operating Thresholds and Optimization Trigger Parameters To ensure the robustness of system operation, the key threshold constraints involved in Phase 3 are shown in Table 1: Table 1 Key Threshold Constraints

[0418] To ensure the safe, reliable, and stable operation of the high-voltage relay icing risk prediction and active defense system throughout the vehicle's entire lifecycle, and to establish a continuously self-evolving user service closed loop, this application embodiment also designs "Phase 4: Real-time system health monitoring, assessment, and operation and maintenance mechanism." This phase serves as the foundation for system operation and is deeply coupled with the aforementioned three phases, forming a closed loop integrating perception, prediction, control, self-learning, and operation and maintenance.

[0419] Figure 11 A flowchart illustrating the real-time system health monitoring, assessment, and operation and maintenance mechanism is provided. Figure 11 As shown, the detailed execution flow of stage 4 is as follows: Step 1: Real-time system health monitoring The system health monitoring module performs high-frequency, real-time, multimodal omnidirectional scanning of four dimensions: vehicle hardware, software, data quality, and predictive performance.

[0420] 1.1 Hardware Health Monitoring The controller performs second-level continuous monitoring of all associated physical entities in the entire system: Monitoring objects include: Sensor subsystems, including housing temperature sensors, ambient temperature and humidity sensors, battery pack temperature sensors, and loop voltage / current sensors; Actuator subsystems, including relay-specific heating films / heaters, ventilation valve controllers, and differential pressure actuators for the cabin seals; Communication subsystems, including onboard CAN bus, onboard Ethernet, and 4G / 5G onboard wireless communication modules (T-Box); and Power supply subsystems, including low-voltage battery voltage, system supply current, and auxiliary power status of the high-voltage power distribution unit (PDU).

[0421] Monitoring metrics and scoring operators: 1. Sensor health :

[0422] In the formula, Scoring the data validity (determined by verifying the reasonable range of physical limits); A performance score for latency from request to response; The sensor calibration deviation score is calculated based on the last calibration time and the current zero drift value.

[0423] 2. Actuator health :

[0424] In the formula, Scoring is given for the degree of agreement between the actual heat output and the desired control command. Scoring the physical delay of the command being issued to the power device to start up and heat up; The remaining service life score is calculated based on the cumulative operating time and power consumption integral.

[0425] 3. Communication health :

[0426] In the formula, Score the success rate of bus frame or cloud data packet transmission; Scoring of end-to-end communication delay; The transmission error rate of data frames is scored.

[0427] Hardware Comprehensive Health Score :

[0428] 1.2 Software Operation Status Monitoring The controller monitors the activity level of each algorithm process and the system resource utilization status within the monitoring software container and the main control chip. Monitoring targets include the main perception and prediction process, the control and actuation process, the data compression process, and the communication process. Additionally, it includes business metrics such as memory usage, CPU utilization, remaining storage space, kernel error logs and prediction frequency, control success rate, and decision accuracy.

[0429] Monitoring metrics and scoring operators: 1. Process health :

[0430] In the formula, Score the process's liveness status; A penalty score is assigned for the number of times an abnormal crash triggers an automatic restart. The health status is categorized into memory usage and CPU utilization.

[0431] 2. System stability score :

[0432] In the formula, , , These represent the occurrence rates of critical errors (such as kernel crashes), general non-fatal errors, and warning events in the background operation log within the unit's statistical time window, respectively.

[0433] Software Comprehensive Health Score :

[0434] 1.3 Real-time monitoring of data quality To ensure the accuracy and reliability of the feature parameters input to the model and to prevent "dirty data" caused by sensor drift and external interference from distorting the model's decisions, a data quality monitoring module was designed for the controller. Monitoring dimensions: 1. Data integrity: Real-time monitoring of signal loss rate (Defined as the ratio of the number of missing data points to the theoretical total number of sampled points) and the maximum consecutive missing length.

[0435] 2. Data accuracy: Monitor the proportion of outliers that exceed the preset reasonable range, and perform physical and logical consistency verification on multi-channel related signals (such as the logical relationship between temperature rise rate and heating power).

[0436] 3. Data Timeliness: Monitor the time delay from data collection to memory processing and the freshness score of timestamps. .

[0437] Overall Data Quality Score :

[0438] In the formula, The percentage of data points identified as abnormal physical glitch.

[0439] 1.4 Real-time monitoring of predictive performance The controller establishes a self-diagnostic closed loop to dynamically evaluate the prediction accuracy and fusion decision-making effect of the three models online. 1. Real-time prediction error monitoring: Calculates the average absolute error of the physical model's most recent 100 predictions. The mean absolute error of the statistical model's most recent 100 predictions And the real-time cross-entropy loss of the empirical model's most recent 100 predictions.

[0440] 2. Confidence Trend Monitoring: Monitor the mean fluctuation trend of the confidence scores of each sub-model output, and the overall confidence score. The trend is recorded, and the frequency of low-confidence events is also recorded.

[0441] 3. Real-time monitoring of decision-making effectiveness: The real-time accuracy, false alarm rate, and false negative rate of the most recent 50 decisions are statistically analyzed within a sliding window.

[0442] Performance degradation grading warning rules: Yellow Alert (Mild Decline): Any predictive performance indicator that occurs for 3 consecutive days The above decline; Orange alert (moderate decline): Any predictive performance indicator that occurs for two consecutive days The above decline; Red Alert (Severe Decline): Any predictive performance indicator that occurs on the same day The above describes a precipitous drop.

[0443] 2. Second step: System health status assessment The system health assessment module calculates the overall system health score using a weighted composite algorithm, and classifies the overall system status accordingly. Comprehensive Health Score calculate:

[0444] In the formula, the system predicts the performance health score. It accounts for the largest weight (30%), hardware and software each account for 25%, and data quality accounts for 20%.

[0445] Health Status Grading Standards: Health status ( All hardware devices are functioning normally, the software is running stably, the data quality is good, and the predictive performance meets the standards. The system is operating under normal maintenance mode.

[0446] Sub-health state ( Some non-core indicators showed slight anomalies, which do not affect the core anti-icing control, but require attention and prevention. The system has adopted an early warning maintenance mode.

[0447] Fault status ( (This indicates that critical system functions have been compromised or there is an electrical safety risk. The system has entered emergency response mode.)

[0448] 3. Third step: Tiered operation and maintenance strategy Based on the assessed health status, the controller executes the corresponding maintenance response procedures: 3.1 Normal Operation and Maintenance Mode When the system is in a healthy state, the controller performs routine maintenance and optimization tasks: 1. Daily data report generation: Regularly send operation and maintenance data to the cloud, and automatically generate daily system operation summaries (daily reports), weekly performance trend analysis (weekly reports), and monthly comprehensive evaluation reports (monthly reports).

[0449] 2. Preventive maintenance reminders: Based on the cumulative working time of the heater, statistics of vehicle use in harsh environments, and calendar time, the system predicts hardware lifespan and generates regular maintenance reminders.

[0450] User interaction optimization: Silently collect user operation logs, analyze user intervention preferences under different freezing risks, and automatically optimize interface layout and prompt information.

[0451] Local knowledge base updates: Automatically extracts processing experience from historical successful self-learning processes and updates the fault diagnosis library and processing guidelines.

[0452] 3.2 Early Warning and Maintenance Mode When the system is in a sub-healthy state, the controller executes an active correction and diagnostic preparation procedure: 1. Early warning level determination: Level 1 Warning (Attention Level): Only a single non-core indicator shows a slight abnormality; Level 2 Warning (Alert Level): Multiple non-core indicators are abnormal, or a single key indicator shows a significant abnormality; Level 3 Warning (Alert Level): Key indicators show a severe decline, approaching the fault threshold. Handling Procedure: Warning information is pushed out in a tiered manner: Level 1 warnings are only recorded in the system's local log; Level 2 warnings are uploaded to the vehicle's backend system; Level 3 warnings are sent as an immediate notification to the vehicle's operations backend system via the network.

[0453] 2. Automatic diagnosis and cause analysis: The self-diagnosis module is activated, compares historical fault characteristics, automatically locates nodes that may cause degradation (such as "sensor drift" or "high communication latency"), and generates a preliminary diagnostic report.

[0454] 3. Remote assistance preparation: Establish a remote diagnostic connection in the cloud backend, collect RawData (raw time series data) of relevant modules at high frequency, and prepare to issue a repair plan.

[0455] 4. Spare parts and service scheduling preparation: If hardware aging is diagnosed, the system will automatically match the spare parts part number corresponding to the vehicle model in the background, pre-schedule service network resources, and generate a maintenance plan.

[0456] 3.3 Emergency Response Mode When the system determines that it is in a fault state, the controller initiates the safety fault tolerance and emergency repair response procedure: 1. Fault Classification: Level 1 Fault (Minor): Some non-core functions are downgraded, but the high-voltage safety of the whole vehicle and the core predictive anti-icing function are not affected; Level 2 Fault (Severe): The anti-icing heater fails or the key sensor fails, resulting in the failure of some anti-icing functions; Level 3 Fault (Critical): The system experiences a serious shutdown, short circuit, or there are major high-voltage safety risks such as relay sticking.

[0457] Handling Procedure: Emergency Alarm and Notification: A red fault alarm pops up on the vehicle's human-machine interface, and alarms are sent to the user and customer service via APP push and automatic technical calls.

[0458] 2. Safety mode switching: The system immediately disconnects from the conventional multi-model fusion control and switches to the redundant backup system, enabling a safety-limited simplified control strategy (such as fixed duty cycle heating based on ambient temperature) to ensure the basic safety of the high-voltage relay.

[0459] 3. Remote Diagnosis and Repair: The cloud-based technical support team remotely collects frozen frame data from the fault site, performs in-depth correlation analysis, and attempts to perform remote software reset or patch repair.

[0460] 4. On-site service dispatch: If the hardware damage is determined to be irreversible, the system will automatically dispatch the order to the nearest authorized service station, dispatch the corresponding spare parts, and push the repair appointment progress to the vehicle owner's mobile terminal in real time.

[0461] Fault recovery confirmation and closed loop: After the on-site maintenance is completed, the controller performs a comprehensive functional and performance-level loop test. After the driver and the system automatically verify that there are no errors, the fault work order is closed and the system returns to the first step of health monitoring.

[0462] 4. Fourth step: User interaction and service.

[0463] 4.1 Multi-dimensional User Interface The system is designed with a multi-level human-machine interface adapted to different user roles: Driver Interface (Simple Level): Integrated into the LCD instrument panel, it only displays green (safety), yellow (warning), and red (fault) status indicator lights, and provides brief text prompts and emergency one-button operation when necessary. Owner Interface (Detailed Level): Integrated into the vehicle's central control screen and mobile APP, it displays an overview of the system's health status, historical anti-icing records, energy-saving statistics charts, and interactive settings menus. Maintenance Interface (Professional Level): Accessible only to authorized technicians, it displays underlying sensor voltage and resistance values, communication error rate, actuator life curves, PID real-time parameters, and a manual test and diagnostic interface.

[0464] 4.2 Personalized configuration and adaptive learning.

[0465] Configuration parameters: Users can customize warning preferences (warning level, notification channel, do-not-disturb period), control strategy preferences (heating intensity preference, energy-saving / powerful mode selection, degree of automation control) and display themes.

[0466] Adaptive learning mechanism: The controller continuously records the user's temporary adjustment behavior (such as manually turning on the maximum heating multiple times under high humidity), and automatically fine-tunes the user's default anti-icing start temperature difference threshold through a reinforcement learning mechanism.

[0467] 4.3 Proactive services and strategy push.

[0468] Push service types include: lifespan-based regular maintenance reminders, software OTA update push notifications; energy-saving and safe driving optimization suggestions based on user habits; and value-added service recommendations.

[0469] Push control strategy: The system limits the push frequency to ensure accurate push notifications at appropriate times, such as when the vehicle is stationary and powered on, to avoid excessively disturbing users.

[0470] 4.4 Closed-loop collection of user feedback.

[0471] The system provides feedback channels such as in-vehicle voice control, user interface access, and QR code scanning. Collected feedback data is automatically categorized into three types: problems, suggestions, and praise, and then automatically distributed to the corresponding technical support, customer service, or product management departments based on this category. The processing results are then fed back to the user in real time via the app, forming a closed-loop management system.

[0472] 5. Step Five: Data Analysis and Continuous Improvement.

[0473] 5.1 Multidimensional Analysis of Operation and Maintenance Data The cloud system regularly aggregates and analyzes operational data from the vehicle fleet to monitor product line quality. Reliability metrics: Calculate the system's mean time between failures (MTBF), mean time to repair (MTTR), and overall system availability.

[0474] Maintenance efficiency indicators: accuracy of statistical early warning, accuracy of fault location, and first-time repair rate.

[0475] Satisfaction index: Collect and analyze user satisfaction ratings for service response and problem resolution.

[0476] 5.2 Continuous Improvement Mechanism (PDCA Cycle) The system executes a PDCA continuous improvement cycle in the cloud: collecting operation and maintenance and feedback data, analyzing and locating common technical problems and algorithm defects, formulating optimization and upgrade plans, implementing and verifying them in a small fleet, and finally achieving standardized release of relevant control parameters for all vehicle models through OTA updates.

[0477] To allocate R&D resources rationally, the priority of improvement tasks is determined by the following formula:

[0478] 5.3 Knowledge Management and Automatic Updates The system establishes a dynamic knowledge management system that includes a fault knowledge base, a best practice base, and a typical case base. The system supports manual data entry and has a self-learning function that automatically extracts correlation features from successfully repaired fault data. Through regular expert review and effectiveness verification, the system achieves continuous automatic updates to the operations and maintenance knowledge base.

[0479] Next, an example of dynamic de-icing control will be provided.

[0480] In some implementations, when the vehicle receives a wake-up command and the high-voltage relay meets a second preset condition, the heating component is controlled to perform a de-icing operation on the high-voltage relay, including: The system acquires the vehicle's emergency start level and the maximum permissible output power of the high-voltage battery pack; determines the target heating power curve based on the emergency start level and the maximum permissible output power; and controls the heating components to dynamically adjust the heating power of the high-voltage relay based on the target heating power curve.

[0481] Optionally, the urgency level can include low urgency level, medium urgency level, and high urgency level.

[0482] For example, if the scheduled departure time is more than 10 minutes away from the current time, it can be classified as a low urgency level; if the user unlocks the car normally or presses the start button, it can be classified as a medium urgency level; if the user selects the quick start mode or the vehicle has an emergency power supply requirement, it can be classified as a high urgency level.

[0483] The maximum permissible output power of the high-voltage battery pack is determined by the battery management system based on the high-voltage battery pack temperature, current remaining charge, individual cell voltage, health status, and fault status.

[0484] The target heating power satisfies:

[0485] in, Target heating power; The power curve corresponding to the startup urgency level; The maximum power that the high-voltage battery pack is allowed to be allocated to the heating components; This is the preset rated power of the heating element.

[0486] At low urgency levels, you can first heat at 10W for 3 minutes, and then adjust the power between 10W and 20W according to the rate of temperature recovery.

[0487] At medium-urgent levels, heating can be done at 25W for the first 2 minutes, then reduced to 15W, and adjusted according to temperature feedback.

[0488] At high emergency levels, heating can be used at 30W if both the high-voltage battery pack and the heating components allow; when the relay temperature approaches the target temperature, the power is reduced to 10W to reduce temperature overshoot.

[0489] The controller can also employ a combination of model predictive control and adaptive proportional-integral-derivative (AID) control. The model predictive controller predicts the relay temperature under different heating powers based on heat conduction logic, and selects the power curve that can shorten the defrosting time within power and temperature constraints. The AID controller adjusts the heating power based on the deviation between the actual and predicted temperatures.

[0490] During de-icing, the controller can control the first heating unit, the second heating unit, and the flow path regulating unit separately. For example, the first heating unit can be controlled to increase the relay housing temperature at a higher power, while the second heating unit is controlled to heat the base. Once the ice begins to melt, the flow rate of the heat medium is increased to promote dehumidification. When the insulation resistance has not yet recovered, heating continues at a lower power while maintaining ventilation.

[0491] In some implementations, when the system outputs a diagnostic status of "Status: Suspected Icing" in Stage 3, the controller immediately activates and runs the "Stage 5: Real-time Situation Assessment, Intelligent De-icing, and Effect Verification" process. This stage, as a direct safety intervention, aims to implement refined and adaptive de-icing control on the high-voltage relays suspected of icing, and to perform dual-index verification of the actuation effect. Finally, a clear system status result is output to Stage 6 to guide the vehicle's start-up.

[0492] Next, an exemplary description of the de-icing effect evaluation will be provided.

[0493] In some embodiments, after the heating element performs a de-icing operation on the high-voltage relay, the method further includes: Collect at least one of the following: temperature rise rate, insulation resistance, and closed contact resistance of the high-voltage relay; determine the de-icing effect assessment result based on the collected temperature rise rate, insulation resistance, or closed contact resistance; and update the configuration parameters of the risk prediction model based on the data from this de-icing operation if the de-icing effect assessment result indicates successful de-icing.

[0494] The rate of temperature recovery can be expressed as:

[0495] The controller can compare the measured temperature recovery rate with the expected temperature recovery rate at the corresponding heating power. For example, at a heating power of 25W, the expected temperature recovery rate is no less than 0.5℃ / minute. If the rate is below this threshold for three consecutive minutes, it can be determined that the heating effect is insufficient or that there is a malfunction in the heating element.

[0496] A temperature recovery rate threshold that is too low may fail to detect heating failures in a timely manner; a threshold that is too high may lead to misjudgments under extremely low ambient temperatures or in structures with large heat capacity. Therefore, the temperature recovery rate threshold can be dynamically determined based on the current ambient temperature, heating power, and heat conduction model.

[0497] Insulation resistance can be obtained from the vehicle's insulation monitoring device. After de-icing, the controller determines whether the insulation resistance meets the safety requirements of the vehicle's high-voltage system. For example, an insulation resistance threshold can be set to 500 Ω / V. For an 800V high-voltage system, the corresponding example threshold is 400kΩ. The actual threshold should be determined based on applicable vehicle high-voltage safety standards and vehicle design requirements.

[0498] The contact resistance can be tested by performing a test closure test through a current-limiting diagnostic circuit or a pre-charge circuit after pre-verification of temperature and insulation conditions. The contact resistance can be expressed as:

[0499] in, This is the terminal voltage after the high-voltage relay is closed. For testing current.

[0500] For example, if the reference contact resistance of a high-voltage relay is 1mΩ, the effectiveness verification threshold can be set to 1.5mΩ. If the measured contact resistance is less than or equal to 1.5mΩ, the contact condition meets the example verification conditions; if it is higher than this threshold, there may be incomplete contact closure, obstructed mechanical movement, or residual water film in the contact area.

[0501] In one implementation, the de-icing effect can be determined by combining other safety conditions, provided that at least one of the temperature recovery rate, insulation resistance, or closed contact resistance meets the corresponding condition. Preferably, dual-mode verification is performed using both temperature recovery rate and insulation resistance, and the closed contact resistance is detected using a current-limiting method after both have passed.

[0502] The de-icing performance rating can be expressed as:

[0503] in, , and These represent the normalized scores for temperature rise rate, insulation resistance, and closed contact resistance, respectively. .

[0504] For example, you can set , , When the overall score is not lower than 0.8 and the insulation resistance meets the mandatory safety threshold, the de-icing is considered successful.

[0505] In implementations with the corresponding hardware, the system may also include an optical sensing unit for acquiring image features of the area where the high-voltage relay is located.

[0506] In one configuration, the high-voltage relay housing has an electrically insulated transparent observation area, and a miniature imaging unit is located outside the transparent observation area, thereby avoiding the direct placement of ordinary imaging devices inside the high-voltage sealed contact cavity.

[0507] In another configuration, insulated optical fibers are used to transmit optical information from inside the relay or the base area to the imaging unit outside the high-voltage isolation area.

[0508] In another configuration, an infrared thermal imaging unit is placed outside the high-voltage relay to identify local low-temperature areas, ice-covered areas, and melting processes by analyzing the temperature distribution on the relay surface.

[0509] The controller can extract frost coverage ratio, reflectivity changes, edge texture, or infrared temperature distribution from images, and cross-validate the optical verification results with temperature recovery rate, insulation resistance, or closed contact resistance.

[0510] Next, we will provide an example of model updates.

[0511] If the de-icing effect evaluation result indicates that the de-icing was successful, the controller will write the data of this de-icing operation into the relay's historical operation file.

[0512] The data collected during this de-icing operation may include ambient air temperature before and after de-icing, ambient relative humidity, relay temperature, high-voltage battery pack temperature, current remaining power, first risk probability, second risk probability, overall risk level, target heating power curve, actual heating power, heating duration, cumulative heating energy, temperature recovery rate, insulation resistance, closed contact resistance, and de-icing effect evaluation results.

[0513] The controller can update the configuration parameters of the statistical regression model using new samples. For linear regression models, least squares fitting can be re-executed after accumulating a preset number of new samples, or online updates can be performed using recursive least squares.

[0514] The controller can also correct the temperature response coefficient in the heat conduction logic based on the error between the actual temperature curve and the predicted temperature curve. and environmental heat dissipation weight .

[0515] To prevent a single instance of outlier data from causing significant changes in model parameters, the magnitude of each parameter update can be limited. For example, after a single update... or The change in parameters relative to the previous parameters should not exceed 5%. Update results exceeding this limit can be temporarily stored as candidate parameters and verified in subsequent events before being implemented into control.

[0516] The following provides an example of de-icing failure and safety downgrade.

[0517] In some implementations, when the de-icing effect evaluation result indicates that de-icing has failed, the controller determines whether the number of retry heating attempts has reached a preset limit.

[0518] If the preset limit for the number of attempts is not reached, the heating strategy will be adjusted according to the reason for the failure, and the de-icing process will be repeated. For example: If the temperature recovery rate is insufficient, increase the power of the first heating unit or the second heating unit; When the relay temperature has recovered but the insulation resistance is insufficient, reduce the heating power and increase the flow rate of the heat medium to prolong the dehumidification process; When the maximum allowable output power of the high-voltage battery pack is insufficient, reduce the heating power and extend the heating time; If the contact resistance is too high, the current limiting test closure verification should be performed again if the temperature and insulation conditions meet the requirements.

[0519] The preset limit for the number of attempts can be set to 3, i.e., the initial de-icing and two retries. If the limit is set too low, the vehicle may be locked after a single occasional measurement fluctuation; if the limit is set too high, it may increase waiting time, energy consumption, and relay thermal load. Therefore, the limit can be calibrated by combining the rated power of the heating element, the single heating time, the remaining charge of the high-voltage battery pack, and the relay's allowable temperature.

[0520] In some implementations, if the de-icing effect evaluation result indicates that de-icing has failed and the number of retry heating attempts has reached a preset limit, the closing command of the high-voltage relay is locked, and a fault warning message is sent to the vehicle's on-board terminal.

[0521] Locking the high-voltage relay's closing command means that the controller prohibits the output of a normal closing drive signal to the high-voltage relay, or sends a closing prohibition status to the battery management system and the vehicle controller, in order to prevent the high-voltage circuit from being forcibly connected when there is an insulation abnormality, suspected icing, or the relay's operation is obstructed.

[0522] Fault warning information may include high-voltage relay de-icing failure, high-voltage system temporarily unable to be powered on, suggestion to wait for ambient temperature to rise or contact maintenance, number of retry attempts, abnormal detection indicators, and fault timestamp.

[0523] The locked state cannot be released simply by a regular user repeatedly sending wake-up commands. The controller can release the lock according to a preset recovery strategy after detecting improved environmental conditions, restored insulation resistance, and successful re-verification; in the case of hardware failure or multiple verification failures, it can require release of the lock through an authorized diagnostic device.

[0524] Based on the above description Figure 12 A flowchart illustrating real-time situation assessment, intelligent de-icing, and effect verification is shown. Figure 12 As shown, the detailed execution process of stage 5 includes the following steps: 1) First step: Real-time situation assessment and decision-making Upon receiving a "suspected icing" signal, the system first performs a rapid real-time situation assessment before executing physical de-icing actions to determine the optimal immediate course of action: 1. Collect real-time environmental and vehicle status data, i.e., real-time status data collection.

[0525] controller with High sampling rate data acquisition of the latest environmental parameters, vehicle status, and electrical parameters ensures real-time decision-making: Temperature and humidity data: Real-time temperature of relay housing surface. Real-time ambient temperature Real-time temperature of high-voltage battery pack and real-time ambient relative humidity Electrical condition data: Real-time insulation resistance of high-voltage circuit Relay contact resistance and system leakage current Vehicle physical status: battery state of charge (SOC), vehicle operating condition (stationary / driving / charging), and the user's currently set operating mode (energy-saving mode / standard mode / performance mode).

[0526] 2. Recalculate the current icing risk, i.e., real-time icing and electrical risk reassessment.

[0527] Because of the transmission delay between the diagnostic judgment in Phase 3 and the actual execution in Phase 5, the controller uses the latest data to perform a second real-time calculation on icing and electrical risks: Calculate temperature margin :

[0528] In the formula, the real-time dew point temperature The following simplified formula is used for calculation:

[0529] like This indicates that the surface temperature of the relay housing is below the dew point temperature, making it extremely easy for water vapor to condense and posing a very high risk of freezing.

[0530] Assess electrical risk components: based on the rate of change of insulation resistance Analyze contact resistance deviation and leakage current transient fluctuations to assess the risk of electrical short circuits or poor contact caused by icing.

[0531] Calculate the real-time risk composite value Normalized output of real-time comprehensive risk value (0 indicates no risk, 1 indicates high risk).

[0532] 3. Assess resource availability and security.

[0533] The controller retrieves the following safety boundaries to determine whether the current system meets the safety conditions for performing active heating: Battery power constraints: high-voltage or low-voltage battery... This ensures the vehicle's basic starting energy consumption. Heating circuit status: The main heater and backup heater have no open-circuit or short-circuit fault codes and are in a ready state. Temperature safety range: High-voltage battery pack temperature. Within the safe operating temperature range ( Driving status constraints: The vehicle must be stationary; de-icing and heating must not be activated while the vehicle is in motion. Fault code barrier: The system has no unprocessed critical fault codes in the power stage or high-voltage safety stage.

[0534] 4. Check user commands and system status.

[0535] It retrieves whether the user has performed a manual cancellation operation within the most recent preset number of seconds, reads the current user's operating mode limit priority (such as the limit on heating power in energy-saving mode), and checks whether the vehicle has higher-level safety alarms (such as collision warnings or high-voltage disconnection requests).

[0536] 5. Two-branch decision rule The controller, taking into account the above evaluation dimensions, executes the following decision logic: Immediate execution conditions (must be met simultaneously): Real-time comprehensive risk value ;and Furthermore, the heater is functioning normally; the battery temperature is within a safe range; the vehicle is stationary; and there are no user cancellation commands or high-level system alarms. Decision: Immediately initiate the "Adaptive Intelligent De-icing Process".

[0537] Conditions for postponement of execution (meeting any one of them is sufficient): Real-time comprehensive risk value ;or The following may occur: a heater malfunction; an abnormal battery temperature; the vehicle is not stationary; the user has performed a cancellation operation; or a high-level alarm exists in the system. Decision results The system has entered "wait / monitoring operation mode".

[0538] 2). Second step: Adaptive intelligent de-icing process.

[0539] When the decision is "Execute immediately", the controller initiates the closed-loop temperature control program: 1. Develop an adaptive heating strategy The controller is based on the real-time comprehensive risk value. Based on the State of Charge (SOC) and the user's operating mode, the most suitable heating strategy is matched: Lower-level heating strategy parameter definition: Aggressive strategy: Heating power set to Target temperature The maximum heating time per cycle is limited to 5 minutes. Standard strategy: Heating power set to... Target temperature The maximum heating time per cycle is limited to 8 minutes. Conservative strategy: Set the heating power to [value missing]. Target temperature The maximum heating time for a single cycle is limited to 12 minutes.

[0540] Strategy matching rules: If and Activate aggressive strategies to quickly eliminate freezing; like and or satisfy and Activate standard strategies; like and Alternatively, the user may have actively selected the energy-saving mode, activating the conservative strategy.

[0541] 2. Perform adaptive power control During the heating process, the controller monitors the real-time temperature change of the relay housing and calculates the temperature deviation:

[0542] It also performs three-stage dynamic power adjustment: when At that time, the controller drives the heating module to rapidly heat the device at the maximum target power of the selected heating strategy.

[0543] when When this happens, the controller will reduce the heating power to a moderate rated power (e.g., half of the target power). To ensure a smooth transition and prevent temperature overshoot caused by thermal inertia.

[0544] when At this time, the controller enters a high-precision adjustment stage and outputs low pulse power to maintain the temperature.

[0545] If the actual measured temperature The controller immediately adjusted the heating power to And stop heating.

[0546] 3. Real-time monitoring and security protection During de-icing, the three-in-one safety monitor operates in parallel to ensure system safety: Temperature safety protection: The surface temperature of the relay must never exceed the limit temperature of 15℃. If the temperature reaches 13℃, the system will force degradation. The heating power is set; if the temperature reaches 15℃, the heating drive circuit will be disconnected unconditionally.

[0547] Time safety protection: When the cumulative heating time reaches the maximum heating time limit corresponding to the currently selected strategy, the system will automatically stop heating and output an anti-icing timeout warning.

[0548] Power safety protection: If the battery SOC unexpectedly drops below 15% during the heating process, the controller will immediately interrupt the de-icing process to prioritize the protection of the start-up battery.

[0549] Sensor redundancy failure protection: If an open-circuit fault occurs in the temperature sensor, the controller immediately switches to a method based on contact resistance. Indirect electrical monitoring of the rate of change and temperature rise estimation model; seamless switching to the backup heater if the main heating circuit hardware fails; if communication is interrupted, the controller maintains the current power for safe heating for 60 seconds before forcibly shutting off.

[0550] 3). Third step: Verification of de-icing effect using two indicators After heating stops and the system remains stationary for 30 seconds, the controller performs dual verification tests on temperature recovery and electrical characteristics to assess the de-icing quality. 3.1 Temperature recovery index verification Calculate the final temperature difference after the heating operation ends:

[0551] like The temperature recovery level was rated as "excellent"; like It was judged as "good"; like It was determined to be "qualified"; like It was determined to be "unqualified".

[0552] 3.2 Verification of Electrical Characteristics Calculate the insulation resistance recovery rate of high-voltage circuits :

[0553] In the formula, This is the actual insulation resistance value measured after heating. This refers to the system's nominal rated insulation resistance value.

[0554] like The electrical restoration level was rated as "excellent"; like It was judged as "good"; like It was determined to be "qualified"; like It was determined to be "unqualified".

[0555] 3.3 Comprehensive Result Judgment Matrix The controller inputs the temperature recovery level and electrical recovery level into the following decision matrix to obtain the final comprehensive de-icing result: Table 2 Decision Matrix

[0556] 3.4 Post-judgment self-learning processing If the judgment result is "success" (complete success / basic success): the system will save the feature vector, strategy parameters and temperature rise slope of this control as a success case to the experience case library of stage three, and use it to fine-tune the adaptive weights.

[0557] If the judgment result is "failure" (partial success / failure): the system starts the fault analysis program, generates a failure log, stores the failure characteristics in the fault database, and simultaneously increases the initial confidence penalty term of the three-phase physical model online.

[0558] 4). Fourth step: Waiting / monitoring the operating mode.

[0559] If the decision made in the first evaluation step is "postpone execution", the controller enters a low-power wait / monitor mode: High-frequency state polling: the controller uses The frequency is used to continuously monitor ambient temperature and humidity, relative humidity, relay surface temperature, insulation resistance, and battery SOC.

[0560] Reassess Triggering Conditions: During monitoring, if any of the following conditions are met, the system will immediately exit the waiting mode and re-initiate the de-icing assessment: Real-time risk value Re-increased to ; Previously limited resource conditions have been restored (e.g., the connection of the charging cable allows the SOC of the high-voltage battery to recover). above); The vehicle owner manually issued a de-icing command.

[0561] Risk resolution determination: If in monitoring mode, the risk value is... For 5 consecutive minutes, it remained consistently below The controller determines that the current icing risk has been actively eliminated by changes in ambient temperature and humidity. The system outputs a "Risk Cleared" signal to stage six and exits stage five.

[0562] 5). Fifth step: Output the data packet structure to stage six.

[0563] Regardless of which branch path is executed, Phase 5 outputs a standard data structure package to Phase 6 (Initiation Decision and Safety Hint Phase) at the end: Successful data packet: Core status code: De-icing successful, allowing normal closure of the high-voltage relay and vehicle start-up. Related data fields: Overall performance verification score, total de-icing heating time, and cumulative battery power consumption. Suggested action guidance: It is recommended that the driver complete the vehicle start-up within the preset safe start-up time window (e.g., 15 minutes).

[0564] Failure data packet: Core status code: De-icing failed, restricting relay closure to prevent short circuit or sticking. Related data fields: The source of the failure (e.g., insulation not restored, overheating alarm), executed power curve, system fault code. Suggested actions: Prompts the owner to try manual forced de-icing, connecting to an external charging facility for supplemental power, or calling on-site repair. Emergency measures include limiting the vehicle's high-power electrical output.

[0565] Risk Relief Data Packet: Core Status Code: The risk has been automatically eliminated, and normal vehicle starting is permitted. Related Data Fields: Current ambient temperature margin value, duration of continuous monitoring. Suggested Action Guidance: The vehicle can be started normally, and the system silently records this event.

[0566] 6). Scenario calculation example.

[0567] Assuming it's 3 AM and the ambient temperature is... The humidity is The system was diagnosed as "suspected freezing" in Phase 3, and then proceeded to Phase 5. Situation reassessment and decision-making: collecting the latest data Calculate the dew point temperature Temperature margin As the insulation resistance begins to decline, the real-time risk value is calculated. At this time, the battery charge was measured. The vehicle was stationary. The system met the "immediate execution condition" and decided to initiate de-icing.

[0568] Heating strategy matching: due to real-time risk value and The system employs an aggressive heating strategy: output heating power. The target is set as .

[0569] Adaptive control execution: Phase 1 (Rapid Heating): with Power heating, temperature rise rate is When the relay temperature drops from Rise to Temperature difference deviation ; Phase Two (Smooth Transition): When the temperature difference is below 3℃, the controller will reduce the heating power to [unclear]. The temperature continued to rise to At this time, the temperature difference ; Phase Three (Fine Temperature Monitoring and Exit): Entering low pulse duty cycle temperature monitoring, when the temperature reaches... When the temperature control target is met, the heater is turned off, and the total heating time is approximately [time missing]. minute.

[0570] Effect verification and judgment: Measurement was taken after heating was stopped and the mixture was left to stand for 30 seconds. Calculate the temperature difference The temperature recovery rating is "Excellent"; the measured insulation resistance recovery rate is... The electrical characteristic recovery level is "excellent". Judgment result If the overall de-icing is determined to be "completely successful", a "success" signal is sent to stage six and the operation log is output, ending the stage five process.

[0571] In this embodiment, after the system completes the de-icing operation and effect verification in Phase 5, the controller immediately initiates "Phase 6: Startup Decision, Graded Safety Degradation, and Fault Report Generation". The main responsibility of this phase is to analyze the execution results of Phase 5, execute graded driving restrictions (degradation control) within the safety boundary that the system cannot fully recover, and generate a standardized diagnostic report to provide detailed data input for the subsequent "Phase 7: Secondary Effect Verification and System Learning".

[0572] Figure 13 This diagram illustrates the process of initiation decision-making, tiered security degradation, and fault report generation. (For example...) Figure 13 As shown, the detailed execution flow of stage 6 is as follows: 1. Step 1: Analysis of the output results in Stage 5.

[0573] The controller first reads and parses the output data packet passed in Phase 5 to determine the subsequent control flow: Data Extraction: The controller extracts the result type fields and validation scores from the data packet. The system records the failure cause code and the real-time risk value during the monitoring period, and records the receiving timestamp to mark the starting point of the current processing flow.

[0574] Analysis rules: If the verification score The system determines the de-icing result as "successful"; if the verification score is... (And if the risk is not eliminated), the system determines the de-icing result as "failed"; if the automatic risk elimination condition is triggered and the system exits during stage five, the system determines the result as "risk eliminated".

[0575] 2. Second step: Triage mechanism.

[0576] 2.1 Processing of "Success" Results When the result is "successful", the controller performs a status reset and normalization recording: Record success log: Write a success log containing timestamp, successful de-icing flag, actual heating duration, total power consumption, and final relay status to the system's non-volatile memory.

[0577] Update system status: Rewrite the health status flag of the high-voltage relay to "normal", clear historical icing fault codes, reset the heater operating cycle count, and update the "last successful de-icing time" parameter. The system then prepares to enter phase seven.

[0578] 2.2 Handling of "Risk Relief" Results When the risk is determined to be "cleared", the controller executes a silent exit procedure: Record the risk resolution log: Record the specific reasons for risk resolution (such as increased temperature margin, decreased humidity, etc.), real-time risk score at the time of resolution, monitoring duration and other parameters, and mark the processing result as "natural elimination without heating".

[0579] Update system status: Clear suspected freezing indicators and add an automatic risk elimination counter.

[0580] An environmental self-healing analysis report will be generated: This report analyzes the coupled impact of temperature and humidity changes and vehicle self-heating (such as battery discharge self-heating and controller heating) on ​​relay temperature, providing background data for subsequent optimization of sensing parameters. Following this, we will proceed to Phase Seven.

[0581] 2.3 Handling of "Failure" Results When a "failure" is detected, the controller immediately initiates fault diagnosis and multi-level redundancy control: Root cause analysis: Based on the parameters fed back from stage five, the faults are categorized as follows: Thermodynamic failure: limited heating power, heating time reaching the limit but still not meeting the standard, damage to the main heater hardware, or excessive environmental convective heat loss.

[0582] Environmental limit failure: Ambient temperature below Or relative humidity consistently greater than .

[0583] System-level failure: Battery SOC below Limiting power supply, severe temperature sensor drift, or PID control overshoot oscillation.

[0584] Implement redundancy backup strategies: Based on the identified failure causes, the controller attempts to implement the following backup strategies (each strategy is tried at most once in a single loop): Backup Strategy 1 (Power Compensation): If the main heater is working properly but its power is insufficient, increase its drive power to the maximum safety limit. If the main heater fails, seamlessly switch to the backup auxiliary heating circuit, extend the heating time to 10 minutes, and lower the target temperature to the safety critical point. At the same time, the frequency of temperature monitoring is increased to prevent local hot spots from overheating.

[0585] Backup Strategy 2 (Insulation Interference): If excessive heat loss is detected, the controller closes all air intake grilles and vents of the vehicle via the local interconnection network (LIN bus), starts the auxiliary insulation fan, and simultaneously increases the current heating power compensation. .

[0586] Backup Strategy 3 (Segmented Pulse Heating): If the battery SOC is below the limit and power supply cannot be continuously provided, switch to the cyclic control mode of "pulse power supply, waiting for the battery to stabilize and then pulse heating again".

[0587] If any of the above alternative strategies improves the verification score after execution, If the backup plan is successful, it will be marked as "backup plan successful" and ready to enter phase seven; if the backup plan still fails to meet the target after execution, the controller will determine to enter the system degradation procedure.

[0588] 3. Third step: Execute system security downgrade.

[0589] 3.1 Determination of Downgrade Level If de-icing fails completely and the backup plan is ineffective, the controller will proceed based on the final verification score from Phase 5. Implement tiered downgrades: Downgrade Level 1 (Minor Restrictions): Applicable to (The de-icing effect is not ideal, but the physical conditions for opening and closing are basically met.)

[0590] Downgrade Level 2 (Moderate Restriction): Applicable (De-icing failed, and there is a high risk of mechanical sticking of the relay, but the electrical insulation has not broken down.)

[0591] Degradation Level 3 (Severe Restriction / Security Lockout): Applicable to (The relay is severely frozen and cannot operate, or the electrical insulation resistance is below the safety limit).

[0592] 3.2 Downgrade restrictions and user prompts Based on the determined downgrade level, the controller, in conjunction with the vehicle control unit (VCU), battery management system (BMS), and human-machine interface (HMI), performs the following joint control: Execute first-level downgrade control (minor restrictions): Vehicle restrictions: Limit the maximum discharge current of the high-voltage circuit, prohibit high-current conditions such as rapid acceleration, and prevent damage to relay contacts that have not been fully de-iced.

[0593] HMI prompt: The central control screen displays "Warning: The high-voltage relay performance has deteriorated, limiting rapid acceleration performance. Please arrange for inspection as soon as possible," allowing the vehicle to continue driving normally.

[0594] Implement the second level of downgrade control (moderate restriction): Vehicle Limitation: Limits the maximum drive power of the vehicle to the rated value. The maximum speed was safely locked at... The following completely disables DC fast charging.

[0595] HMI prompts: The instrument panel displays a yellow malfunction indicator light, and the central control screen displays "Warning: Relay malfunction. The vehicle speed and charging power have been limited. Avoid long-distance driving and go to the service station for repair immediately." Short-distance emergency driving is permitted.

[0596] Implement third-level downgrade control (severe restriction): Vehicle restrictions: The controller forcibly cuts off the high-voltage closing enable signal, prohibiting the vehicle from being powered on at high voltage, and the vehicle only retains the low-voltage auxiliary power supply system; the high-voltage electrical safety interlock to prevent misoperation is activated.

[0597] HMI prompts: A red critical fault warning pops up on the instrument panel and central control screen, indicating "Danger: Relay failure, vehicle safety protection has been activated and cannot be driven, please contact emergency assistance immediately".

[0598] Data logging: The controller writes the start time of the degradation, the degradation level, the specific limiting parameters, and the expected conditions for resolving the degradation into the system fault log.

[0599] 4. Fourth step: Fault report generation and distribution.

[0600] After determining the final control path, the controller automatically generates a fault diagnosis and analysis report for this anti-icing / de-icing event using a standardized template. The report mainly includes the following structured content: Event Summary: Assign a unique event identifier and record the trigger time, end time, and final decision-making result.

[0601] Fault details: Record all raw temperature, humidity, and insulation resistance data during the fifth stage of the encapsulation process, the redundant backup attempt process during the sixth stage, and the final degradation status.

[0602] Impact assessment analysis: Quantitatively assess the extent to which the current vehicle restricts users' daily travel, as well as the potential thermal impact on the battery pack and relay lifespan.

[0603] Repair guidance and suggestions: If the heater is determined to be faulty, the output will be: "1. Measure the resistance at both ends of the main heater (normal range is..."). ); 2. Check the 12V power supply of the control module; 3. Verify the PWM duty cycle signal; 4. If damaged, it is recommended to replace the heating component.

[0604] If sensor drift is detected, the output will be: "1. Check the temperature sensor connector; 2. Measure the resistance response at different temperatures; 3. Perform zero-point calibration; 4. Replace the sensor if necessary."

[0605] Key points of self-learning suggestions: Summarize the deviation characteristics of this icing diagnosis model, extract the abrupt changes in environmental temperature and humidity, and generate feature differentiation suggestions for adaptive updates of the Stage 7 model.

[0606] 5. Output to stage seven The fault reports, final system health status flags, degradation status parameters, and time-series data packets of the de-icing process generated in Phase 6 are encapsulated and synchronously output to "Phase 7: Secondary Verification of Effect and System Learning" to complete the closed-loop iteration of algorithm parameters.

[0607] In the embodiments of this application, in order to perform closed-loop quality verification of the anti-icing and de-icing control effects and realize the online autonomous iterative evolution of the system control algorithm and strategy library, the system constructs a "Stage 7: Secondary Verification of Effect and Continuous Evolutionary Learning of System" control module. Figure 13 A flowchart illustrating the process of secondary validation of effects and continuous evolutionary learning of the system is shown. For example... Figure 13 As shown, its core execution logic diagram is as follows: 1. Triggering and Classification Analysis Mechanism Phase 7's initiation depends entirely on the output of Phase 6. The controller first reads the standard data packet input to Phase 6 and extracts the processing result type field: If the result type is "successful processing" (i.e., anti-icing / de-icing operation was successful and the vehicle's functions were not restricted), the system will enter the "Secondary verification of successful cases and generation of strategy optimization suggestions" branch; If the result type is "failure handling" (i.e., de-icing failed and system security degradation has been performed), the system enters the "in-depth analysis of failure cases and generation of improvement measures" branch; If the result type is "risk relief" (i.e., the system did not perform heating action, and the risk of freezing was self-healed through environmental changes), the system enters the "risk relief mode learning and threshold dynamic optimization" branch.

[0608] Branch 1: Secondary Validation of Successful Cases and Generation of Strategy Optimization Suggestions Upon receiving a "successful processing" signal, the self-learning engine performs multi-dimensional verification and replicability evaluation of the control process to extract positive control experience: 2.1 Multi-dimensional cross-validation The system performs in-depth secondary verification of de-icing quality across three dimensions: time, space, and function, and calculates a comprehensive cross-validation score. Time dimension verification score :

[0609] Short-term stability score After monitoring heating stops The temperature rebound characteristic of the relay within minutes requires a temperature drop slope. And the standard deviation of temperature fluctuation .

[0610] Mid-term tracking score Monitoring after startup minute Within hours, under the first high-current discharge condition of the vehicle, is the rate of change of the relay's contact resistance... And the electrical insulation resistance remains at the nominal value. above.

[0611] Long-term impact rating : Check whether the relay will trigger the same icing fault again within the next 24 hours.

[0612] Spatial Dimension Verification Scoring :

[0613] Temperature uniformity score Calculate the maximum temperature difference at multiple points on the relay surface. The standard for excellence is .

[0614] Heat distribution score Assess whether the heat diffusion gradient to the high-voltage conductor busbar and the sealed chamber meets the design expectations, and check for the existence of local hot spots.

[0615] Temperature difference consistency score : Evaluate whether the temperature difference fluctuation between each temperature measuring point of the relay and the environment is acceptable. .

[0616] Functional dimension verification score :

[0617] Electrical performance rating Verify the contact resistance after closure. And insulation resistance value .

[0618] Mechanical performance rating The spring pressure and closing response time of the contact point are indirectly inverted by the change in contact resistance.

[0619] Reliability rating : Monitor the electrical stability of the device after 10 consecutive physical opening and closing operations under a high current load.

[0620] Overall score determination:

[0621] like The de-icing verification was deemed successful.

[0622] 2.2 Relay Physical Fatigue and Life Assessment To assess the negative impact of frequent heating on material aging, a dual-stress cumulative damage assessment model was established: Thermal stress damage increment :

[0623] In the formula, This refers to the thermal stress fatigue coefficient of the material. It is the thermal expansion matching coefficient; The initial ambient temperature; This represents the transient rate of temperature change.

[0624] Electrical stress damage increment :

[0625] In the formula, To control the operating current; For loop resistance; This refers to the heating time per cycle. This is an aging correction factor based on the cumulative number of historical relay actuations.

[0626] Life expectancy prediction update:

[0627] In the formula, This is the total damage capacity limit corresponding to the design life of the relay.

[0628] 2.3 Strategy Replicability and Optimization Suggestion Generation The system uses Euclidean distance to calculate the degree of matching between current operating conditions (including multi-dimensional vectors such as temperature, humidity, wind speed, and battery SOC) and a preset set of standard boundary conditions. If the conditions match and a verification score is achieved... Furthermore, since the energy consumption is lower than the historical average, the system sets this control parameter as the preferred recommended parameter for this specific scenario and increases the recommended proportion of this strategy in the fusion weight.

[0629] 3. Branch Two: In-depth analysis of failure cases and generation of improvement measures When a "failure handling" signal is received, the system initiates a diagnostic procedure to locate control defects and formulate compensatory control logic. 3.1 Five-layer root cause localization analysis The system constructs a causal reasoning model to perform hierarchical analysis of the collected failure data: First layer (phenomenon layer): Directly read the failure manifestations (such as verification scores below 0). (The surface temperature rise of the relay did not meet the standard).

[0630] The second layer (direct cause layer): calculates control deviations (such as insufficient actual heating power, premature triggering of heating time limit, excessive heat loss, etc.).

[0631] Third layer (system layer): Retrieve related component limitations (such as battery SOC being lower than...). This can lead to current limiting protection, excessive wind speed causing heat convection loss, or sensor calibration drift.

[0632] Fourth layer (design layer): Positioning algorithm strategy matching deviation (such as using the standard temperature control strategy in a low power and high wind speed scenario, underestimating the air convection heat dissipation coefficient).

[0633] Fifth layer (root cause layer): Identify knowledge gaps and model biases (such as the lack of a high wind speed correction factor in the heat loss model).

[0634] The system comprehensively evaluates the amount of relevant evidence, the strength of the causal chain, and the historical recurrence rate, and selects the factor with the highest confidence level as the root cause of the event.

[0635] 3.2 Counterfactual Decision-Making Inference and Digital Twin Verification The self-learning engine calls local or cloud-based digital twin models to reconstruct the fault scene: Counterfactual scenario construction: The original actual decision-making strategy is used as the benchmark, and then replaced in the twin simulation environment with "aggressive strategy (high power, short time)", "conservative strategy extension (low power, long time)" and "segmented pulse heating strategy" respectively.

[0636] Expected value assessment:

[0637] In the formula, and These represent the success probability of the simulation result and the corresponding quantitative value score (determined by the verification score). and This refers to the probability of safety risks such as battery over-discharge or heater overheating caused by this strategy, and the corresponding risk costs.

[0638] Deduction output: Select the alternative with the highest expected value as the recommended improvement strategy for this working condition.

[0639] 3.3 Fault Evolution Prediction and Improvement Measures Formulation Evolution prediction: Using hidden Markov models to predict the state evolution path of the ice-free relay (such as the deterioration path of short circuit caused by increased icing, the stable path of maintaining the confined state, or the self-healing path of environmental warming).

[0640] Improvement measures generated: Model layer improvement: Introduce a wind speed correction term into the heat loss model. ,in To estimate wind speed in real time, This is the wind speed correction factor; and the SOC safety protection threshold for strategy switching is set from... Dynamically improved to .

[0641] System control layer improvements: For scenarios of "low battery + high wind speed", a composite compensation control scheme of "pulse preheating + air intake grille closure + extended heating" is automatically generated, and the over-temperature protection action threshold under low battery conditions is lowered from... Dynamically move up to .

[0642] 4. Branch 3: Risk Relief Pattern Learning and Dynamic Threshold Optimization When a "risk cleared" signal is received, the system extracts the physical laws governing the automatic de-icing process to avoid unnecessary heating actions under similar self-healing conditions in the future. 4.1 Automatically disable driver mode recognition The controller collects time-series curves of the self-healing process and performs similarity comparison and classification: Temperature-driven type: characterized by the rate of increase in ambient temperature. (This often occurs at sunrise or when a vehicle enters a heated room).

[0643] Humidity-driven type: characterized by a decrease in ambient humidity rate (This often occurs in underground parking garages or highly ventilated environments).

[0644] Vehicle waste heat driven type: characterized by heat conduction from the battery pack or high-voltage drive motor, causing the relay compartment temperature to rise.

[0645] Complex type: The combined effect of multiple factors mentioned above.

[0646] The system calculates the overall matching degree between the current trend and known self-healing patterns. If the matching degree of all known patterns is lower than [a certain value], [the system will proceed as follows]. If so, the current operating condition sequence will be stored as a new pattern in the local environment database.

[0647] 4.2 Dynamic Optimization of Icing Detection Threshold To minimize heating energy consumption while ensuring safety, the system employs a Bayesian optimization algorithm to correct the judgment threshold online. The optimization problem can be described as follows:

[0648]

[0649] In the formula, False alarm rate; This refers to the false negative rate. Penalty coefficient to prevent underreporting (typical configuration) (to reflect the principle of prioritizing safety).

[0650] Optimization Example: When identified as temperature-driven or with residual vehicle heat, due to the extremely high probability of self-healing, the system lowers the real-time icing risk assessment threshold for subsequent similar scenarios from... Increased to respectively and This is to reduce the number of unnecessary preventative heating cycles.

[0651] 4.3 Case Knowledge Value Assessment The system scores the knowledge value of automatically resolved cases:

[0652] In the formula, Data integrity is scored; As a representative score of the model; Based on the scalability of the scenario; As a feature of innovation. If If the case is correct, it will be written into the system's core knowledge base; otherwise, it will be stored in the auxiliary log database.

[0653] 5. Unified system learning and optimization module Regardless of which branch the system executes, all outputs are ultimately fed into the system learning and evolution module to achieve global updates of model parameters and policies. 5.1 Adaptive Update of Model Parameters Risk prediction model update: Employ mini-batch gradient descent to measure the mean square error between the predicted risk value and the actual icing result. The objective is to minimize the weights, using the latest end-to-end data to update the weights online. The update step size is adjusted by a confidence factor, and the learning rate is set to... .

[0654] Heat loss model parameter update: Utilizing real-time time-series feedback data of "temperature-power" during the heating process, a recursive least squares method with a forgetting factor is employed (the forgetting factor is configured as follows). Online correction of heat conduction matrix parameters.

[0655] Control policy reinforcement learning update: The system constructs a Q-learning reinforcement learning mechanism, defining the environment and vehicle state as... The heating strategy selection action is defined as follows: Reward function Defined as:

[0656] The Q-value update equation is:

[0657] In the formula, the learning rate Discount factor .

[0658] 5.2 Strategy Base Iteration and Knowledge Graph Expansion Strategy scoring and selection: The system calculates the comprehensive score of each strategy in the strategy library.

[0659] If the score is below the set threshold for three consecutive times, the system will automatically mark it as a "high-risk strategy" and implement a lock-in restriction; if the new strategy generated by the strategy innovation mechanism passes 10 simulation verifications in shadow mode, it will be upgraded to a mass production recommended strategy.

[0660] Dynamic expansion of the knowledge graph: The system updates the locally stored relational knowledge graph and dynamically adds newly generated entities such as "environmental elements - vehicle status - control strategy - fault result" and their corresponding association weights to improve the performance of relational reasoning.

[0661] 6. Vehicle start status confirmation and data output After the system's self-learning and evolution process is completed, the controller performs a final confirmation of the vehicle's status based on the final constraint execution status of Phase Six and the secondary verification results of Phase Seven, and generates a unique, digitally signed, and tamper-proof vehicle status confirmation code. The judgment and prompt rules are shown in Table 3. Table 3 Prompt Rules Table

[0662] In the embodiments of this application, in order to ensure the safety of occupants and electrical safety in extreme vehicle-level faults or high-voltage failure scenarios, the system is designed with "Phase 8: Vehicle-level Safety Defense Line and Emergency Fault-Tolerant Intervention Mechanism". As the final safety barrier of the vehicle control chain, when Phase 6 or Phase 7 determines that there is an irreversible major electrical safety hazard in the system, or when the vehicle triggers a new fatal threat such as high-voltage adhesion or thermal runaway during operation, the controller will immediately and forcibly activate Phase 8.

[0663] Figure 14 A flowchart illustrating the vehicle-level safety defense system and emergency fault-tolerant intervention mechanism is shown. Figure 14As shown, the detailed execution process of stage 8 includes the following steps: 1. First step: Safety status assessment and urgency classification The system safety manager accesses the vehicle bus in real time to perform multi-source data fusion assessment of vehicle dynamics, electrical physical field, road environment and user status.

[0664] 1.1 Multi-source security status fusion assessment Vehicle dynamics and powertrain health assessment: Driving hazard rating :

[0665] In the formula, The overspeed risk coefficient has a value of ; The acceleration hazard score is determined by the ratio of the current acceleration to the maximum permissible acceleration. and The risk ratings correspond to steering angular velocity and braking deceleration, respectively.

[0666] Powertrain health :

[0667] In the formula, This is the battery temperature correction factor. It is the internal resistance degradation factor; , and The health scores are for the motor, electronic control system, and charging interface, respectively.

[0668] Environmental and climate dimension assessment: Road environment hazard level The road surface adhesion coefficient is estimated based on the current temperature and humidity, and then weighted and summed in combination with the surrounding traffic density and visibility detected by radar / camera.

[0669] Climate risk level :

[0670] In the formula, Given the current risk of icing; and These are the exceedance coefficients for rainfall intensity and ambient wind speed, respectively.

[0671] User-related dimensions and system health assessment: System overall health :

[0672] The system combines multi-dimensional indicators to evaluate the stability of software and hardware, bus communication and power supply, and quantitatively calculates the impact of the current fault on the safety of the whole vehicle based on safety impact, functional loss ratio and performance degradation ratio.

[0673] 1.2 Urgency Level Classification Based on comprehensive calculations, the controller categorizes the current overall vehicle threat level into five levels: Level 1 Threat (Red Alert, Stop Safely Immediately): Applicable to situations involving complete failure of the power / steering / braking system, high-voltage runaway, or fire risk.

[0674] Level 2 Threat (Orange Alert, Operation Restricted): Partial failure of critical high-voltage systems, degraded security protection or severely degraded performance.

[0675] Level 3 Threat (Yellow Alert, Degraded Operation): Auxiliary electrical system failure, moderate performance degradation, or poor user takeover status.

[0676] Level 4 Threat (Blue Alert, Warning Observation): Minor system anomalies or external environmental fluctuations require enhanced monitoring.

[0677] Level 5 Threat (Green Status, System Self-Recovery): Occasional and temporary anomalies that can be automatically self-healed online by the control system.

[0678] Threat Level Calculate using the following formula and round to the nearest integer:

[0679] 1.3 Recoverability Assessment and Strategic Decision-Making The system calculates the overall recovery probability based on the status of the hardware backup channel, the software hot-repair capability, the security of the current docking point, and the accessibility of communication rescue. And the estimated recovery time. Then, Pareto optimal strategy selection is performed between security benefits and implementation costs to determine the optimal intervention plan:

[0680] In the formula, The security improvement score is given after the strategy is implemented; To assess feasibility; and These are the costs of impact on users and the social traffic environment, respectively.

[0681] 2. Second step: Selection and execution of security processing mode The controller seamlessly switches to the corresponding physical intervention mode based on the assessed threat level: 2.1 Emergency Safety Stop Mode (Level 1 Threat Response) Execute adaptive progressive safe parking: If there are no vehicles behind, use the maximum safe deceleration to perform emergency braking to stop; if in a busy road section, perform progressive deceleration to stop, control the vehicle to slow down smoothly and guide it to the side of the road; if there is an emergency lane, activate automatic tracking to control the vehicle to stop at the side of the road.

[0682] Multiple safety protection interlocks: Electrical protection: The high-voltage busbar is immediately de-energized and active discharge is performed to maintain the low-voltage auxiliary power supply to ensure the power supply of critical sensors and T-Box.

[0683] Mechanical protection: Forcefully engages the electronic parking brake (EPB), locks the steering mechanism, and unlocks the central locking system to prevent intrusion from outside the vehicle.

[0684] Environmental protection: Automatically activates hazard warning lights and emergency lighting inside and outside the vehicle to ensure basic passenger comfort.

[0685] Fault isolation and system reconfiguration: Disconnect the power and communication links of the faulty submodule to prevent the fault from spreading. System reconfiguration is completed by switching to backup hardware, loading a security software image, or rebuilding the degradation control loop.

[0686] Rescue communication initiation: Automatically pushes an encrypted emergency data packet containing VIN code, precise GPS coordinates, fault level, high-voltage system status and vehicle owner status to the emergency rescue center (120 / 119) and vehicle service center via T-Box.

[0687] 2.2 Restrict operating mode (Level 2 threat response) Functional limitations: Only core safety functions such as power steering, basic hydraulic braking, necessary external lighting, and emergency communication are retained; maximum speed is limited to [limited to] ... Turn off unnecessary comfort features that consume a lot of electricity, such as seat heating and multimedia entertainment systems.

[0688] Dynamic performance adjustment:

[0689] In the formula, the safety factor It is a decreasing function of the current threat level and the degree of environmental hazard. The system also limits the intensity of kinetic energy recovery and lowers the intervention trigger threshold of ABS / ESP.

[0690] Safety boundary monitoring and user guidance: Real-time monitoring of longitudinal and lateral obstacle avoidance distances. Once the safety margin approaches the threshold, the system will issue a joint alarm through a combination of visual red flashing on the HUD, steering wheel vibration, and voice broadcast, and will actively intervene with the controller to correct the braking if necessary.

[0691] 2.3 Degraded Operation Mode (Level 3 Threat Response) Core functions are maintained: the highest priority resources are allocated to basic driving, steering and braking, and redundant backup monitoring is established for relevant main control modules.

[0692] Accessibility downgrade: Level 1 Degradation: Power is slightly limited, restricting rapid response.

[0693] Level 2 downgrade: Disable automatic control functions such as automatic parking and advanced driver assistance systems (ADAS), and simplify the LCD instrument panel display interface.

[0694] Level 3 Degradation: Completely shut down non-safety-level high-energy-consuming loads.

[0695] Redundant system hot-switching: If the main control bus or computing unit fails, the system will transfer control to the backup redundant controller in a seamless hot-switching manner to ensure the continuity of steering and braking control.

[0696] 3. Third step: System state stabilization and recovery mechanism After the security intervention is executed, the system monitoring module continuously analyzes the operating parameters to assess whether the system has the conditions for functional recovery: 3.1 System stability closed-loop determination The system periodically calculates safety, performance, and reliability scores. The system is considered to have entered a stable state if and only if the following conditions are met:

[0697] Otherwise, the system will maintain its current downgraded or restricted mode, prohibiting any function recovery operations.

[0698] 3.2 Progressive Functional Recovery Control When the hardware (stable voltage and no unrecoverable faults), software (no memory leaks or conflicts), and external environment all meet the recovery conditions, the self-learning engine initiates a gradual release of functions: The recovery sequence logic is as follows: functions related to life safety, core driving control, ADAS driver assistance, and comfort and convenience systems will be released in stages.

[0699] Two-way verification rollback mechanism: After each functional module is restored, automatic self-test and functional closed-loop test must be performed. If the self-test fails or abnormal fluctuations occur, the system immediately rolls back to the previous safety restriction state and records the current fault recovery conflict log.

[0700] 4. Fourth step: Generation and output of security incident reports Upon completion of Phase Eight, the system automatically generates a structured, digitally signed security event report and outputs final status data: Report content structure: Event Overview: Event UUID encoding, trigger source parsing, processing mode, and scope of impact.

[0701] Processing log includes the evolution trajectory of multi-source safety states, dynamic control curves for emergency deceleration and stopping, topological relationships of hardware isolation, and verification data for function recovery.

[0702] Effect evaluation: Quantitatively evaluate the gains and losses of this emergency intervention on vehicle voltage stability, chassis dynamic response, and user safety.

[0703] Follow-up prevention recommendations: Generate specific suggestions for replacing physical components, rolling back software versions, or upgrading cloud models.

[0704] Final data packet output: The controller generates a unique vehicle status confirmation code, along with a system evolution report and knowledge update package. This confirmation code serves as the core safety access key for repowering the vehicle or closing the high-voltage relay, controlling the vehicle to enter the corresponding driving / locking state, and is simultaneously uploaded to the cloud database for group security analysis.

[0705] To more clearly illustrate the technical solutions in the embodiments of this application, a complete example is provided below.

[0706] The vehicle disconnects from the high-voltage power supply and enters sleep mode. The controller collects the ambient air temperature as 2℃, the ambient relative humidity as 92%RH, the high-voltage relay housing temperature as 4℃, the high-voltage battery pack temperature as 8℃, and the current remaining power as 55% SOC.

[0707] The controller calculates the current dew point temperature using the dew point calculation formula and then calculates the current temperature difference. Since the current temperature difference is still higher than the preset temperature difference threshold of 2°C, the current state is not directly classified as high-risk.

[0708] The controller obtains the weather forecast for the next 6 hours via the T-Box. The weather forecast shows that the ambient air temperature will drop to -2℃ in the next 2 hours, while the relative humidity will remain above 90%RH.

[0709] The controller iteratively calculates the relay temperature for the next 360 minutes based on heat conduction logic, and also calculates the dew point temperature and its difference for each minute. The prediction results show that the difference will decrease to 2°C in the next 70 minutes, and will decrease to below 0°C in the next 100 minutes, and the relay temperature will be below 0°C.

[0710] The heat conduction logic calculation yielded a first risk probability of 0.82, while the statistical regression logic, combined with the relay's historical operating records, yielded a second risk probability of 0.74. Using weights of 0.6 and 0.4 for fusion, the overall risk level is:

[0711] The overall risk level is higher than 0.6, therefore the icing risk status is classified as high risk.

[0712] The vehicle is in sleep mode, and the current remaining battery charge is 55% SOC, which is greater than the 20% SOC safety reserve threshold. The controller triggers the battery management system to connect the auxiliary power supply circuit and controls the first heating unit to maintain heating at 8W.

[0713] When the temperature difference is below 2°C, the controller will maintain the power at 12W; when the temperature difference is above 3°C, the power will be reduced to 5W; when the temperature difference is above 4°C for 10 consecutive minutes, heating will stop.

[0714] Two hours later, the vehicle received the user's start command. At this time, the high-voltage relay casing temperature was 2.5℃, with a difference of 3.1℃, and the insulation resistance met safety requirements. Since the anti-icing insulation during the dormant period had ensured that the high-voltage relay met the closing conditions, the controller did not perform a high-power de-icing operation and directly entered the normal closing process.

[0715] In another embodiment, during vehicle hibernation, the remaining charge of the high-voltage battery pack is below the safe reserve charge threshold, therefore anti-icing and heat preservation are not performed. When the vehicle wakes up, the relay housing temperature is -5°C, the ambient relative humidity is 95%RH, the overall risk level is 0.9, and the controller determines that the high-voltage relay meets the second preset condition.

[0716] The vehicle's emergency start level is medium emergency, and the maximum power allowed to be allocated to the heating components from the high-voltage battery pack is 30W. The controller heats at 25W for the first 2 minutes, then adjusts the power to 15W based on the temperature recovery. Once the relay temperature reaches 3°C, the controller stops rapid heating and maintains 5W for 1 minute.

[0717] The controller detected a temperature rise rate of 1.2℃ / minute, and the insulation resistance met the safety threshold. Subsequently, a trial closure was performed via the current-limiting diagnostic circuit, and the measured closure contact resistance was 0.8mΩ, below the effectiveness verification threshold of 1.5mΩ. The controller confirmed successful de-icing, released the high-voltage relay closure restriction, and updated the configuration parameters of the risk prediction model based on the de-icing operation data.

[0718] This application also provides a vehicle that includes the above-mentioned high-voltage relay for anti-icing and de-icing.

[0719] The vehicle can be a pure electric vehicle (PEV / BEV), a battery electric vehicle (BEV), a hybrid electric vehicle (HEV), a plug-in hybrid electric vehicle (PHEV), a range-extended electric vehicle (REEV), a new energy vehicle, or other vehicles equipped with a high-voltage battery pack and a high-voltage relay.

[0720] In other embodiments, the vehicle may further include a high-voltage battery pack, a battery management system, a vehicle controller, an on-board communication terminal, an on-board terminal, and a high-voltage load. The controller can communicate with the battery management system, the vehicle controller, the on-board communication terminal, and the on-board terminal via an on-board communication network.

[0721] When the vehicle is in a dormant state, the system performs anti-icing and heat preservation based on environmental parameters and risk prediction results; when the vehicle receives a wake-up command, the system selects normal closure or de-icing based on the status of the high-voltage relay; after de-icing is completed, the effect is verified by temperature characteristics and / or electrical characteristics; if de-icing fails and the number of retries reaches the limit, the closure command is locked and a fault warning message is sent.

[0722] This application also provides an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program. The processor is communicatively connected to the memory and runs the computer program to implement some or all of the steps in the above-described method.

[0723] The electronic device may be a cockpit domain controller, body controller, vehicle controller, fragrance controller, central computing platform, in-vehicle information processing equipment, or other vehicle computing equipment.

[0724] In some embodiments, the above methods can be implemented by a computer program or instructions. The controller includes a processor and a computer-readable storage medium storing the computer program or instructions. When the processor executes the computer program or instructions, it implements one or more steps of the above-described anti-icing and de-icing methods for the high-voltage relay.

[0725] Computer-readable storage media may include read-only memory (ROM), random access memory (RAM), EEPROM, flash memory, or other storage media capable of storing programs and data.

[0726] This application embodiment can also provide a computer program product. This computer program product includes a computer program or instructions, which, when executed by a processor, implement one or more steps in the aforementioned anti-icing and de-icing methods for the high-voltage relay.

[0727] It should be understood that the application of this application is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Those skilled in the art can understand that implementing all or part of the processes of the above embodiments and making equivalent changes according to the claims of this application still fall within the scope of this application.

Claims

1. A method for preventing and de-icing a high-voltage relay, characterized in that, include: The vehicle's environmental parameters and the temperature data of the high-voltage relay are acquired. The dew point temperature is determined based on the environmental parameters, and the difference between the temperature data of the high-voltage relay and the dew point temperature is calculated. Based on the difference, the icing risk status of the high-voltage relay is determined; When the vehicle is in a dormant state and the icing risk state meets the first preset condition, the battery management system is controlled to output energy to drive the heating component to perform anti-icing and heat preservation operation on the high-voltage relay. as well as When the vehicle receives a wake-up command and the high-voltage relay meets the second preset condition, the heating component is controlled to perform a de-icing operation on the high-voltage relay.

2. The method according to claim 1, characterized in that, The step of determining the icing risk status of the high-voltage relay based on the difference includes: Obtain weather forecast data for a preset future time period, and combine it with the temperature of the vehicle's high-voltage battery pack to construct the temporal evolution trend of the difference within the future time series; and If the time evolution trend indicates that the difference is less than or equal to a preset temperature difference threshold, the icing risk status of the high-voltage relay is determined to be high-risk.

3. The method according to claim 2, characterized in that, The determination of the icing risk status of the high-voltage relay also includes: The difference, the environmental parameters, the high-voltage battery pack temperature, and the preset relay historical operation records are input into the preset risk prediction model; The first risk probability is calculated using the heat conduction logic in the risk prediction model, and the second risk probability is calculated using the statistical regression logic in the risk prediction model; and By combining the first risk probability and the second risk probability, a comprehensive risk level is generated to indicate the icing risk state.

4. The method according to any one of claims 1-3, characterized in that, When the vehicle is in a dormant state and the icing risk condition meets a first preset condition, controlling the battery management system to output energy to drive the heating component to perform anti-icing and heat preservation operations on the high-voltage relay includes: After the high-voltage power supply to the vehicle is disconnected, the current remaining charge of the high-voltage battery pack is obtained; and When the current remaining power is greater than the preset safe reserve power threshold and the icing risk status indicates that there is an icing risk, the battery management system is triggered to connect the auxiliary power supply circuit and control the heating component to heat at a maintenance power lower than the preset rated power.

5. The method according to any one of claims 1-3, characterized in that, When the vehicle receives a wake-up command and the high-voltage relay meets the second preset condition, controlling the heating component to perform a de-icing operation on the high-voltage relay includes: Obtain the vehicle's emergency start level and the maximum permissible output power of the high-voltage battery pack; The target heating power curve is determined based on the aforementioned start-up urgency level and the aforementioned maximum permissible output power; and The heating component is controlled to dynamically adjust the power of the high-voltage relay according to the target heating power curve.

6. The method according to claim 5, characterized in that, After the heating component is controlled to perform a de-icing operation on the high-voltage relay, the method further includes: Collect at least one of the following: temperature rise rate, insulation resistance, and closed contact resistance of the high-voltage relay; The de-icing effect evaluation result is determined based on the collected temperature recovery rate, insulation resistance, or closed contact resistance; and If the de-icing effect evaluation result indicates that the de-icing was successful, the configuration parameters of the risk prediction model are updated based on the data from this de-icing operation.

7. The method according to claim 6, characterized in that, The method further includes: If the de-icing effect evaluation result indicates that de-icing has failed and the number of retry heating attempts has reached a preset limit, the closing command of the high-voltage relay is locked, and a fault warning message is sent to the vehicle's on-board terminal.

8. The method according to any one of claims 1-3, characterized in that, The acquisition of the vehicle's environmental parameters and the temperature data of the high-voltage relay includes: The ambient air temperature and relative humidity of the vehicle's environment are obtained, and these are used as the environmental parameters; and The readings of the temperature sensor attached to the housing of the high-voltage relay are obtained, and the readings are used as the temperature data of the high-voltage relay.

9. An anti-icing and de-icing system for a high-voltage relay, characterized in that, include: The sensing module is used to acquire the vehicle's environmental parameters and the temperature data of the high-voltage relay; The controller is used to determine the dew point temperature based on the environmental parameters, calculate the difference between the temperature data of the high-voltage relay and the dew point temperature, and determine the icing risk status of the high-voltage relay based on the difference. as well as A heating element is deployed in the area where the high-voltage relay is located and is connected to the controller; The controller is also used to control the battery management system to output energy to drive the heating component to perform anti-icing and heat preservation operation on the high-voltage relay when the vehicle is in a dormant state and the icing risk state meets the first preset condition. And when the vehicle receives a wake-up command and the high-voltage relay meets the second preset condition, the heating component is controlled to perform a de-icing operation on the high-voltage relay.

10. The system according to claim 9, characterized in that, The heating component includes: The first heating unit is attached to the outer surface of the high-voltage relay housing; The second heating unit is disposed on the base of the high-voltage relay; and The flow path regulating unit is configured to control the flow rate of the thermal medium directed to the vicinity of the high-voltage relay.

11. A vehicle, characterized in that, The anti-icing and de-icing system includes the high-voltage relay as described in claim 9 or 10.

12. An electronic device, characterized in that, include: Memory, on which computer programs are stored; and A processor, coupled to the memory, is configured to run the computer program to implement the method as described in any one of claims 1 to 8.