Heat management assembly energy efficiency optimization system driven by car networking data

Abstract

The invention relates to the technical field of vehicle thermal management control, and discloses a thermal management assembly energy efficiency optimization system driven by vehicle networking data. The system comprises a data acquisition module, an energy efficiency analysis module, a communication interface module, a cloud strategy module and a vehicle-mounted control module. The data acquisition module initializes an energy efficiency monitoring environment and continuously acquires an operation parameter sequence; the energy efficiency analysis module compares the parameter sequence with an energy efficiency reference library to generate an energy efficiency deviation index, and extracts a current scene feature vector when the index exceeds a dynamic threshold value; the communication interface module packages the feature vector into a data packet and sends the data packet to the cloud; the cloud strategy module starts a strategy derivation engine after receiving the data packet, and generates an optimization instruction sequence based on scene matching; the vehicle-mounted control module receives and executes the instruction sequence to adjust thermal management assembly parameters. According to the invention, online monitoring, intelligent diagnosis and cloud collaborative optimization of the energy efficiency of the thermal management assembly are realized, and the energy utilization efficiency of the system under different working conditions is improved.

Description

Technical Field

[0001] This invention relates to the field of vehicle thermal management control technology, specifically to a vehicle network data-driven thermal management assembly energy efficiency optimization system. Background Technology

[0002] The control strategies of vehicle thermal management assemblies generally rely on preset fixed parameter tables or basic feedback control algorithms. The system monitors key parameters through sensors and compares them with static thresholds stored in the onboard controller, thereby triggering corresponding control actions. This rule-based control method can meet basic temperature regulation functions under standard operating conditions, but its design boundaries are relatively fixed.

[0003] Vehicles operate in complex and variable environments, including urban congestion, highway cruising, and extreme temperatures. Existing static control strategies struggle to accurately perceive and assess the system's overall energy efficiency under different scenarios. Fixed thresholds cannot effectively identify component performance degradation or energy efficiency deterioration under non-standard operating conditions, potentially leading to response lag or inaccurate judgments. Furthermore, limited computing resources of the onboard controller hinder complex multi-variable collaborative optimization calculations, typically allowing only single adjustments to local parameters and failing to achieve optimal global energy efficiency for the entire system, resulting in unnecessary energy losses.

[0004] Vehicle-to-everything (V2X) technology provides vehicles with real-time data interaction capabilities, but existing thermal management control technologies have failed to fully utilize this potential. The core issue lies in the lack of a solution capable of real-time, intelligent diagnosis of system energy efficiency status and precise triggering of optimization mechanisms, as well as a robust decision support system capable of generating refined and collaborative control commands based on specific scenario characteristics. Existing technological solutions are insufficient in terms of the accuracy of perception and the intelligence of decision-making, thus hindering further improvements in overall thermal management energy efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide a vehicle-to-everything (V2X) data-driven thermal management system for energy efficiency optimization, in order to solve the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides a vehicle-to-everything (V2X) data-driven thermal management assembly energy efficiency optimization system, the system comprising: The data acquisition module is used to initialize the energy efficiency monitoring environment after the vehicle thermal management assembly is started, and to continuously collect the operating parameter sequence of the thermal management assembly; The energy efficiency analysis module is used to compare the operating parameter sequence with the preset energy efficiency benchmark library in real time, generate an energy efficiency deviation index, and activate the abnormal data capture process when the energy efficiency deviation index exceeds the dynamic threshold to extract the feature vector of the current operating scenario. The communication interface module is used to encapsulate the feature vector into an optimization request data packet and send it to the cloud policy server through the vehicle network gateway; The cloud-based strategy module is used to receive the optimization request data packet, start the strategy derivation engine, and generate an adapted optimization instruction sequence based on scene feature matching. The vehicle control module is used to receive the optimized instruction sequence, parse and execute parameter adjustment operations on the thermal management assembly.

[0007] Preferably, the process of initializing the energy efficiency monitoring environment by the data acquisition module specifically includes: Create independent monitoring instances for each critical component of the thermal management assembly, including water pumps, radiators, and electric fans; Configure a dedicated data sampling frequency and accuracy level for each monitoring instance; Establish a data link between the monitoring instance and the vehicle-mounted sensor network, and verify the integrity of data transmission; Load the energy efficiency assessment rule set into the local cache. The energy efficiency assessment rule set contains the expected energy efficiency values ​​under different operating conditions. Start the background monitoring task, which periodically verifies the running status of each monitoring instance.

[0008] Preferably, the process by which the energy efficiency analysis module generates the energy efficiency deviation index specifically includes: The actual operating parameters for the current sampling period are obtained from the monitoring instance, including temperature readings, pressure readings, and power consumption readings. Calculate the multidimensional difference between the actual operating parameters and the expected energy efficiency value; The multidimensional differences are normalized to obtain the standardized deviation value; Based on the weighting coefficients defined in the energy efficiency assessment rule set, the standardized deviation values ​​are weighted and fused, and the fused result is output as the energy efficiency deviation index.

[0009] Preferably, the process of activating the abnormal data capture process by the energy efficiency analysis module specifically includes: When the energy efficiency deviation index continuously exceeds the dynamic threshold for a preset duration, the high-speed data recording mode is triggered. Record all control signals and status feedback of the thermal management assembly during abnormal periods with millisecond-level precision; Feature extraction is performed on the recorded data stream to identify recurring abnormal patterns; The feature vector is constructed based on the identified abnormal patterns, and the feature vector includes pattern encoding, duration stamp, and severity level; The feature vector is bound to the current vehicle identification code and then temporarily stored in the security buffer.

[0010] Preferably, the process of the cloud-based strategy module starting the strategy derivation engine includes: After receiving the optimization request data packet, decrypt and format verify it; The feature vector is extracted and input into a policy decision tree model, which contains a historical optimization case library; In the policy decision tree model, a multi-level matching query is performed to find the historical scene most similar to the current feature vector; Obtain the successful optimization solutions corresponding to the historical scenarios and make timeliness corrections for them; Generate the optimized instruction sequence, which includes specific parameter adjustment values ​​and execution order.

[0011] Preferably, the update mechanism of the strategy decision tree model includes: Regularly collect thermal management assembly optimization feedback data from multiple vehicle terminals; The collected optimization feedback data is cleaned and labeled to form a training sample set; The branch weights and node parameters of the policy decision tree model are updated using an incremental learning algorithm; Verify the matching accuracy of the updated policy decision tree model on the test dataset; Once the matching accuracy meets the requirements, the updated strategy decision tree model will be deployed to the production environment.

[0012] Preferably, the process of parsing and optimizing the instruction sequence by the vehicle control module includes: After receiving the optimized instruction sequence, check its digital signature and validity period; The instructions are parsed one by one and converted into control commands that can be executed by the underlying devices. Verify that the parameter range of each control command is within the safety limits; Arrange the execution queue of control commands according to the timing relationship defined in the instruction sequence; Set up execution status monitoring points for each control command to collect execution feedback.

[0013] Preferably, the process of the vehicle control module performing parameter adjustment includes: Control commands are sent sequentially to the actuators of the thermal management assembly according to the execution queue; Monitor the execution results of each control command and record the difference between the actual adjustment value and the target value; When an adjustment anomaly is detected, activate the backup adjustment plan or restore to the previous stable state; Collect response data of the thermal management assembly during the adjustment process and generate an optimization effect report; The optimization effect report is uploaded to the cloud strategy server.

[0014] Preferably, the process of calculating the multidimensional difference between the actual operating parameters and the expected energy efficiency value, and normalizing the multidimensional difference to obtain the standardized deviation value includes: For the actual operating parameters of each key component, extract the current values ​​of temperature reading, pressure reading, and power consumption reading respectively; Obtain the expected temperature, expected pressure, and expected power consumption under the corresponding operating conditions from the energy efficiency expected value; Calculate the absolute difference between the current value and the expected value for each reading to form the original difference vector; Each difference value in the original difference vector is scaled, and a preset normalization coefficient is used to map the difference value to the range of zero to one. Based on the physical meaning and importance of each reading, an adjustment factor is applied to calibrate the mapped difference value to obtain the standardized deviation value.

[0015] Preferably, the process of triggering a high-speed data recording mode when the energy efficiency deviation index continuously exceeds the dynamic threshold for a preset duration, and recording all control signals and status feedback of the thermal management assembly during the abnormal period with millisecond-level precision, includes: Monitor the time series of the energy efficiency deviation index, and start a timer when the index value is continuously higher than the dynamic threshold. After the timer reaches the preset duration, a trigger signal is generated and the high-speed data logger is activated; The high-speed data logger is configured to sample all control signal channels and status feedback channels of the thermal management assembly at millisecond intervals; The sampled signal data is sorted by timestamp and compressed and stored in a temporary buffer. Simultaneously, the start time, end time, and peak index of abnormal events are recorded for subsequent feature extraction.

[0016] Compared with the prior art, the beneficial effects of the present invention are: By comparing continuous sequences of operating parameters with a pre-defined energy efficiency benchmark library in real time, an energy efficiency deviation index is generated. A dynamic threshold is then used for judgment, enabling a more refined and adaptive perception of the system's energy efficiency status. This method overcomes the poor adaptability of fixed thresholds under complex operating conditions and effectively distinguishes between normal fluctuations and true energy efficiency degradation trends. The abnormal data capture process is activated only when the deviation exceeds the dynamic threshold, extracting the feature vector of the current operating scenario. This event-driven mechanism improves data processing efficiency, avoids invalid data transmission between the vehicle system and the cloud, and makes the allocation of optimized resources more accurate and timely.

[0017] Based on a cloud-based strategy inference engine, the received feature vectors undergo deep analysis, generating adapted optimized instruction sequences through scene feature matching. Leveraging the powerful computing capabilities and abundant global data of the cloud, the engine can execute complex optimization algorithms, thereby formulating refined control strategies with multi-parameter linkage that far exceed the computing capabilities of onboard systems. The issued instruction sequences, rather than single instructions, coordinate the timing-based actions of multiple actuators within the thermal management assembly, achieving a smooth and rapid transition from the current suboptimal operating point to a higher energy efficiency point. This scenario-based personalized strategy generation enables the thermal management assembly to dynamically adapt to various real-world road conditions and environments, achieving optimal global energy efficiency. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the working principle of the vehicle-to-everything (V2X) data-driven thermal management assembly energy efficiency optimization system described in this invention. Figure 2 A flowchart for initializing the energy efficiency monitoring environment for the data acquisition module; Figure 3 A flowchart for generating energy efficiency deviation indicators for the energy efficiency analysis module; Figure 4 This is a monitoring chart for anomaly detection and data capture in the thermal management system. Figure 5 This is a diagram showing the matching results of the cloud-based strategy decision tree. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1This invention provides a vehicle-to-everything (V2X) data-driven thermal management system for energy efficiency optimization. The system includes: a data acquisition module that initializes the energy efficiency monitoring environment by configuring a sensor network and establishing a data flow channel to ensure continuous acquisition of operating parameter sequences, which cover key indicators such as temperature, pressure, and power consumption. An energy efficiency analysis module receives the operating parameter sequences and compares them in real-time with a preset energy efficiency benchmark library. This benchmark library stores standard energy efficiency values ​​under different operating conditions. An energy efficiency deviation index is calculated to quantify the difference between the current operating state and the ideal state. When the energy efficiency deviation index exceeds a dynamic threshold, the energy efficiency analysis module activates an abnormal data capture process. This process involves high-speed data recording and feature extraction to generate a feature vector for the current operating scenario. The feature vector includes elements such as pattern encoding and timestamps.

[0021] The communication interface module encapsulates feature vectors into optimization request data packets. These packets, using a standard protocol format, are sent to the cloud-based policy server via the vehicle network gateway, ensuring reliable and secure data transmission. Upon receiving the optimization request data packets, the cloud-based policy module activates its policy inference engine. This engine matches historical data with scenario features to generate a suitable sequence of optimization instructions, containing specific parameter adjustment values ​​and execution order. The vehicle control module receives the optimization instruction sequence, parses the instruction content, and converts it into underlying control commands. Ultimately, it executes parameter adjustments to the thermal management assembly, including regulating water pump speed and fan power to improve energy efficiency.

[0022] Example 1: See Figure 2 In practical implementation, an independent monitoring instance is created for each key component of the thermal management assembly, including the water pump, radiator, and electric fan. Each monitoring instance is assigned a unique identifier and allocated memory resources for storing real-time data. A dedicated data sampling frequency and accuracy level are configured for each monitoring instance. The data sampling frequency is set according to the component characteristics; for example, high-frequency sampling is used for the water pump, while medium-frequency sampling is used for the radiator. The accuracy level ensures the accuracy of data reading and avoids noise interference. A data link is established between the monitoring instance and the vehicle sensor network. The data link is connected via CAN bus or Ethernet. A verification mechanism is used to verify the integrity of data transmission, ensuring that no data packets are lost or erroneous. An energy efficiency assessment rule set is loaded into the local cache. This rule set is downloaded from the cloud or pre-loaded into the vehicle storage and contains expected energy efficiency values ​​under different operating conditions, such as standard power consumption values ​​at different temperatures. A background monitoring task is started. This task runs at a fixed period, periodically verifying the operating status of each monitoring instance. The verification process includes checking whether the data stream is continuous and whether the instance is responsive. If an anomaly is detected, a restart or alarm is triggered. The entire initialization process is completed within seconds of the thermal management assembly starting up, ensuring that the energy efficiency monitoring environment is ready.

[0023] In implementation, monitoring instances are created using an object-oriented approach. Each instance encapsulates data acquisition methods and state variables, and instances communicate via message queues to avoid resource conflicts. The data sampling frequency is configured based on the dynamic characteristics of the components; for example, the sampling frequency increases when an electric fan is running at high speed to capture transient changes. The accuracy level is set according to sensor specifications; for example, a temperature sensor uses 0.1 degrees Celsius accuracy. Establishing the data link involves loading the driver and configuring the network protocol. To verify data transmission integrity, test data packets are sent and received to ensure link stability. During the loading of the energy efficiency assessment rule set, the rule file is parsed and mapped to a memory structure. The rule set is stored according to operating conditions for easy retrieval. Background monitoring tasks are started using multi-threading technology. The main thread is responsible for scheduling, while sub-threads execute specific verification tasks. Verification includes CPU utilization and memory usage, ensuring reasonable allocation of system resources. After initialization, the data acquisition module enters continuous operation, collecting and temporarily storing the sequence of operating parameters in a buffer for use by the energy efficiency analysis module.

[0024] In some embodiments, when creating an independent monitoring instance for each critical component of the thermal management assembly, the initialization of the monitoring instance includes memory allocation and parameter settings. Memory allocation dynamically requests a fixed-size buffer from the heap, and parameter settings include the sampling period and alarm threshold. The monitoring instance structure includes a sensor handle, a data buffer pointer, and status flags. The sensor handle is used to access specific hardware, the data buffer pointer points to a circular buffer, and the status flags indicate whether the instance is ready. It can be understood that the creation order of monitoring instances follows component priority: water pump monitoring instances are created first, followed by radiator monitoring instances, and electric fan monitoring instances are created last. This order ensures that monitoring of critical components is initiated first. In specific implementations, when configuring a dedicated data sampling frequency and accuracy level for each monitoring instance, the data sampling frequency is set based on the component's thermal inertia. Water pumps have low thermal inertia, so the sampling frequency is high; radiators have high thermal inertia, so the sampling frequency is low. The accuracy level is configured to match the sensor type: analog sensors are configured with a high accuracy level, and digital sensors are configured with a standard accuracy level. The configuration process is completed by reading and writing configuration registers, which are stored in non-volatile memory and automatically loaded when the system starts.

[0025] Optionally, when establishing a data link between the monitoring instance and the vehicle-mounted sensor network, a master-slave architecture is adopted, with the monitoring instance acting as the master node and the sensors as slave nodes. The link establishment process includes hardware initialization and protocol handshake. Hardware initialization configures the baud rate and data bits of the communication interface, and the protocol handshake exchanges device identifiers and synchronization signals. A CRC checksum algorithm is used to verify the integrity of data transmission, with a checksum appended to each data packet, and the receiving end verifies the correctness of the checksum. In specific implementations, when loading the energy efficiency assessment rule set into the local cache, the energy efficiency assessment rule set is described in XML format, containing the mapping relationship between operating conditions and expected energy efficiency values. The loading process includes file parsing and data structure construction. The file parser reads the XML file and extracts key fields, and the data structure construction creates a hash table in memory to quickly query the expected energy efficiency value using the operating condition parameter as the key. It can be understood that when the background monitoring task is started, it runs as a system daemon process with a priority lower than real-time tasks but higher than ordinary application tasks. The monitoring task's workflow includes status collection and health checks. Status collection reads the heartbeat signals of each monitoring instance, and health checks verify whether the instance's memory usage exceeds limits. The task exception handling mechanism includes an automatic restart process, where a watchdog timer triggers system recovery when a monitored task crashes.

[0026] In practical implementation, after creating an independent monitoring instance for each key component of the thermal management assembly, the lifecycle management of the monitoring instance includes three phases: creation, operation, and destruction. The creation phase allocates resources and registers the instance in the management list; the operation phase continuously collects data and handles anomalies; and the destruction phase releases resources and removes the instance from the management list. The data acquisition method for the monitoring instance adopts an interrupt-driven mode. When sensor data is ready, an interrupt service routine is triggered, and the interrupt service routine stores the data in a buffer. In some embodiments, when a dedicated data sampling frequency and accuracy level are configured for each monitoring instance, the dynamic adjustment of the data sampling frequency is based on changes in operating conditions. When the thermal management assembly is detected to enter a high-load mode, the sampling frequency of all instances is automatically increased. Accuracy level switching is implemented through software configuration. High-precision mode enables the sensor's built-in filtering function, while standard-precision mode uses raw data. Configuration parameters are stored in EEPROM, supporting runtime modification and immediate effect. After establishing a data link between the monitoring instance and the vehicle sensor network, data link maintenance includes periodic heartbeat detection and fault switching. Heartbeat detection sends ping packets to confirm the connection status, and fault switching enables a backup communication path when the link is interrupted.

[0027] Optionally, when loading the energy efficiency assessment rule set to the local cache, the update mechanism of the energy efficiency assessment rule set supports incremental download. When new rules are released in the cloud, only the changed parts are downloaded through differential updates. The storage area of ​​the local cache is divided into multiple sectors. New rules are written to the spare sector, and the active sector is switched after verification. It can be understood that after starting the background monitoring task, the scheduling strategy of the background monitoring task is based on time-slice rotation. Each monitoring instance's verification operation is allocated a fixed time slice to prevent a single instance from consuming too many resources. The monitoring task's logging function writes the running status to flash memory, and the log files are named by timestamp for easy follow-up problem tracking. In specific implementation, the complete process of the data acquisition module initializing the energy efficiency monitoring environment includes hardware self-test and software initialization. The hardware self-test verifies whether the sensors and communication interfaces are normal, and the software initialization loads configuration parameters and starts the task. The sign of successful initialization is that all monitoring instances enter the ready state and the background monitoring task starts outputting heartbeat signals. The handling process for initialization failure includes retry and degraded operation. Retry is performed up to three times, and degraded operation disables the monitoring function of non-critical components.

[0028] In some embodiments, when creating independent monitoring instances for each critical component of the thermal management assembly, the anomaly handling mechanism of the monitoring instances includes timeout retransmission and data verification. Timeout retransmission re-requests data when it is lost, and data verification uses parity bits. The communication interface of the monitoring instances supports hot-swapping, automatically re-identifying and re-establishing a connection when sensors are replaced. After configuring a dedicated data sampling frequency and accuracy level for each monitoring instance, the smooth adjustment of the sampling frequency uses a first-order low-pass filter to avoid data jumps caused by frequency abrupt changes. The adaptive adjustment of the accuracy level is based on signal quality, automatically increasing the accuracy level when the signal-to-noise ratio decreases. Optionally, when establishing a data link between the monitoring instance and the vehicle sensor network, the encrypted transmission of the data link uses the AES algorithm, and the encryption key is updated periodically to prevent cracking. The bandwidth management of the data link uses a token bucket algorithm to limit the maximum data rate of a single instance and avoid network congestion. When loading the energy efficiency assessment rule set into the local cache, the version management of the energy efficiency assessment rule set uses semantic version numbers. Changes in the major version number indicate incompatible updates, while changes in the minor version number indicate functional enhancements.

[0029] In practical implementation, after the background monitoring task is started, its performance analysis function calculates the CPU usage time of each monitoring instance and generates a resource usage report. The remote configuration function of the background monitoring task allows modification of monitoring parameters through a diagnostic interface; modification requires authentication and authorization. The integrity and reliability of the energy efficiency monitoring environment initialized by the data acquisition module are ensured through multiple checks, including memory checksums and code signing. Memory checksums verify data integrity, and code signing verifies program authenticity. The timing constraints of the initialization process are strictly defined, with a maximum execution time limit for each step; exceeding this limit results in initialization failure. The data buffer management of the monitoring instances employs a dual-buffering technique: a front-end buffer for acquisition and a back-end buffer for processing, with intermittent buffer swapping to avoid data contention. The data sampling frequency synchronization mechanism uses a global clock source, ensuring that all instances' sampling actions are based on the same time base, guaranteeing data timestamp consistency. The hardware implementation of the accuracy level depends on the characteristics of the sensor itself; high-precision sensors offer better linearity and temperature drift compensation.

[0030] In practical implementation, the data link redundancy design includes primary and backup dual links. In the event of a primary link failure, automatic switchover to the backup link is performed transparently to the upper layers. The energy efficiency assessment rule set caching strategy adopts the LRU algorithm, where recently used rules are kept in memory and less frequently used rules are evicted to storage. Adaptive scheduling of background monitoring tasks is based on system load; the monitoring cycle is extended when the system load is high and shortened when the load is low. Error handling during the initialization of the energy efficiency monitoring environment by the data acquisition module includes tiered alarms, logging of minor errors, and triggering system protection for serious errors. The initialization process progress is indicated by LED codes, with different codes corresponding to different initialization stages, facilitating on-site diagnosis.

[0031] Example 2: See Figure 3The process of generating the energy efficiency deviation index by the energy efficiency analysis module begins with obtaining the actual operating parameters of the current sampling period from the monitoring instance. The actual operating parameters include temperature readings, pressure readings, and power consumption readings. These readings are read through the buffer of the data acquisition module to ensure data timeliness. The multidimensional difference between the actual operating parameters and the expected energy efficiency value is calculated. The multidimensional difference is calculated separately for each key component. For example, for the water pump component, the absolute difference between the current value and the expected value of the temperature reading, pressure reading, and power consumption reading is calculated. The multidimensional difference is normalized to obtain the standardized deviation value. The normalization process uses a preset normalization coefficient to map the difference value to the range of zero to one to avoid the influence of dimensions. According to the weight coefficients defined in the energy efficiency assessment rule set, the standardized deviation value is weighted and fused. The weight coefficients are set based on the importance of the component. For example, the power consumption reading has a higher weight. The fused result is output as the energy efficiency deviation index. In practical implementation, for the actual operating parameters of each key component, the current values ​​of temperature, pressure, and power consumption readings are extracted. These current values ​​are obtained from the real-time data stream of the monitoring instance, using direct memory access to read the sensor buffer, reducing CPU overhead. The expected temperature, expected pressure, and expected power consumption under the corresponding operating conditions are obtained from the energy efficiency expectation values. These expected values ​​are obtained by querying the energy efficiency assessment rule set based on the current vehicle speed and ambient temperature; the query process is accelerated using a hash index. The absolute difference between the current value and the expected value of each reading is calculated, forming an original difference vector. This original difference vector contains differences in multiple dimensions, such as temperature, pressure, and power consumption. Each difference value in the original difference vector undergoes a scaling transformation, using a preset normalization coefficient to map the difference value to a range of zero to one. The normalization coefficient is derived from historical data statistics, such as the maximum permissible deviation value. Based on the physical meaning and importance of each reading, an adjustment factor is applied to calibrate the mapped difference value. The adjustment factor is set by expert experience; for example, the adjustment factor for temperature readings is smaller because its change is slower. Finally, a standardized deviation value is obtained.

[0032] In some embodiments, when calculating the multidimensional difference between actual operating parameters and expected energy efficiency values, a parallel processing architecture is used. The difference calculation for each parameter type is assigned to an independent thread, and threads exchange data via shared memory. The calculation of absolute differences uses floating-point subtraction, and the result is taken as the absolute value to prevent negative values ​​from affecting the calculation. The original difference vector is stored in an array structure, with array indices corresponding to parameter indices. It can be understood that the normalization process involves mathematical transformations. The normalization coefficients are pre-stored in a configuration file, which is in key-value pair format, where the key is the parameter name and the value is the normalization coefficient value. The scaling transformation formula is expressed as:

[0033] in: This is the preliminary standardized value of parameter j. It is the actual reading of parameter j. It is the expected value of parameter j. These are the normalized coefficients of parameter j. Based on the historical maximum deviation of parameter j, ensure It falls within the zero to one range. The calibration procedure applies an adjustment factor. Standard deviation after calibration Adjustment factor Based on the sensitivity determination of parameter j, parameters with higher sensitivity have larger adjustment factors. In the weighted fusion stage, weight coefficients are loaded from the energy efficiency assessment rule set, the weight coefficient matrix is ​​aligned with the parameter list, and the fusion calculation uses dot product operations. The energy efficiency deviation index... The formula is expressed as:

[0034] in: It is the weight coefficient of parameter j. This is the total number of parameters.

[0035] In practical implementation, when obtaining the actual operating parameters for the current sampling period, the sampling period is triggered by a system timer. The timer interrupt service routine reads sensor data, and the actual operating parameters are temporarily stored in a circular buffer. The buffer size accommodates data from multiple sampling periods to prevent data loss. The process of obtaining the expected energy efficiency value includes operating condition matching. Operating condition parameters include vehicle speed, ambient temperature, and engine load. The matching algorithm uses nearest neighbor search to find the most similar operating condition record in the energy efficiency assessment rule set. When calculating the absolute difference, the difference value calculation uses hardware acceleration units, such as FPGA coprocessors, to improve the calculation speed. The construction of the original difference vector dynamically allocates memory, and the vector dimension changes with the number of parameters. In the normalization process, the scaling operation is optimized using a lookup table method. The mapping results of common difference values ​​are pre-calculated and stored in the lookup table to reduce the amount of real-time computation. The application of the calibration adjustment factor uses multiplication instructions. The adjustment factor is stored in a constant array, and the array index corresponds to the parameter number.

[0036] In some embodiments, when calculating multidimensional differences, batch processing is supported, processing multiple sampling point data at once to improve throughput. Batch processing uses the SIMD instruction set to calculate the differences of multiple parameters in parallel. The compressed storage of the original difference vector uses differential encoding to reduce memory usage; differential encoding records the changes in continuously sampled difference values. In normalization processing, the normalization coefficients for scaling are dynamically adjusted according to changes in the operating environment, such as increasing the normalization coefficients under high-temperature conditions. The adaptive update of the calibration adjustment factor is based on feedback learning, which uses gradient descent to optimize the adjustment factor value. The weight coefficients of the weighted fusion can be adjusted at runtime; the weight coefficients are modified through a configuration interface, and the changes take effect immediately. In specific implementations, the process of obtaining actual operating parameters includes data alignment, which ensures that the reading timestamps of different sensors are synchronized. The synchronization mechanism uses a global clock source. The query optimization of the energy efficiency expectation value uses a Bloom filter to quickly filter out mismatched operating conditions and reduce query time. When calculating absolute differences, overflow checks for difference values ​​use saturation operations to prevent numerical anomalies. The calculation precision of normalization processing uses high-precision floating-point numbers to avoid rounding errors. The application condition branch of the calibration adjustment factor selects different adjustment factor sets based on parameter quality. The post-processing of the weighted fusion output includes limiting processing, which restricts the energy efficiency deviation index to a range of zero to one hundred for easier intuitive understanding.

[0037] Optionally, the complete process for generating energy efficiency deviation indicators by the energy efficiency analysis module includes error handling, including handling missing parameters and data anomalies. When parameters are missing, the last valid value is used as a substitute; when data is anomaly, re-collection is triggered. Performance monitoring records the execution time of the indicator generation process, and the algorithm is optimized when the execution time exceeds the limit. In specific implementation, the acquisition frequency of actual operating parameters is consistent with the sampling frequency of the monitoring instance, and the sampling frequency is configurable, ranging from milliseconds to seconds. The storage format of the energy efficiency expectation value is optimized to columnar storage to improve query efficiency. Intermediate results of multidimensional difference calculations are cached for debugging and verification. Normalization coefficients are grouped and managed according to parameter type: temperature parameters in one group, pressure parameters in another. Calibration adjustment factors are stored with encrypted protection to prevent tampering. The weighted fusion calculation is pipelined, with pipeline stages including data retrieval, calculation, and write-back.

[0038] In practical implementation, the energy efficiency analysis module initialization includes loading configuration parameters, which are read from the file system in FAT32 format. The actual operating parameter acquisition and synchronization mechanism uses semaphores to ensure data integrity. Read-write locks are used for concurrent control of energy efficiency expectation value queries, ensuring safe multi-threaded access. Numerical stability checks are performed on the calculated values ​​of multi-dimensional differences to ensure input values ​​are not invalid. Normalization coefficients are hot-updated without interrupting service. Calibration adjustment factors are backed up and restored, with backups stored in non-volatile memory. Redundant calculations are used to verify the output of weighted fusion, comparing the consistency of two results. A publish-subscribe model is used for the energy efficiency deviation index, with subscribers registering callback functions.

[0039] Example 3: The process of activating the abnormal data capture process of the energy efficiency analysis module starts when the energy efficiency deviation index continuously exceeds the dynamic threshold for a preset duration. The preset duration is set by system configuration parameters, such as five seconds, to ensure the continuity of the abnormal state rather than instantaneous fluctuations. The time series of the energy efficiency deviation index is monitored. When the index value is continuously higher than the dynamic threshold, a timer is started. The timer is implemented using a hardware timer with millisecond-level accuracy. After the timer reaches the preset duration, a trigger signal is generated and the high-speed data logger is activated. The high-speed data logger is a dedicated hardware module or software component, configured to sample all control signal channels and status feedback channels of the thermal management assembly at millisecond intervals. The control signal channels include PWM output and digital commands, and the status feedback channels include analog readings from temperature and pressure sensors. The sampled signal data is sorted by timestamp and compressed and stored in a temporary buffer. The compression algorithm uses LZ77 to reduce storage space usage. At the same time, the start time, end time, and peak index of the abnormal event are recorded for subsequent feature extraction. In practice, feature extraction is performed on the recorded data stream. The feature extraction process uses digital signal processing techniques, such as Fast Fourier Transform or wavelet analysis, to identify recurring abnormal patterns. These abnormal patterns include periodic oscillations, step changes, or trend drifts. Based on the identified abnormal patterns, feature vectors are constructed. The feature vectors contain pattern codes, duration stamps, and severity levels. The pattern codes are enumerated values ​​representing different anomaly types, the duration stamps record the duration of the anomaly, and the severity level quantifies the severity of the anomaly. After the feature vectors are bound to the current vehicle identification number, they are temporarily stored in a secure buffer. The secure buffer uses encrypted storage and access control to prevent unauthorized access.

[0040] In some embodiments, when monitoring the time series of energy efficiency deviation indicators, the time series data is stored in a circular buffer. The buffer size accommodates data from the most recent few minutes. A sliding window algorithm is used to determine if data continuously exceeds a dynamic threshold. The sliding window size matches a preset duration; for example, a five-second window checks if five consecutive sampling points exceed the threshold. The start and stop of the timer are managed by a state machine, whose states include idle, monitoring, and triggered states. State transitions are based on changes in indicator values. After generating a trigger signal, the activation process of the high-speed data logger includes allocating memory resources and initializing acquisition parameters. Memory resources are allocated from the system heap. Acquisition parameters include the sampling rate and channel configuration. It can be understood that when sampling at millisecond intervals, the sampling interval is fixed at one millisecond to ensure high-precision data capture. Sampling actions are executed by interrupt service routines with the highest interrupt priority to avoid data loss. Sampling of the control signal channel uses a digital input capture unit, while sampling of the state feedback channel uses an ADC converter with a resolution of 12 bits. The signal data is sorted based on hardware timestamps provided by the system clock counter. Compressed storage uses streaming compression, and the compressed data block size is optimized for cache line alignment to improve access efficiency.

[0041] In practice, the feature extraction process segments the data stream, with each segment having a configurable length, such as a one-second segment. Segmentation uses overlapping windows to reduce edge artifacts. Anomaly pattern recognition employs a pattern matching algorithm, which predefines template patterns learned from historical anomaly data. Feature vectors are constructed using a structure representation. Structure members include an integer pattern code, a floating-point duration stamp, and an integer severity level. The severity level is calculated based on the deviation magnitude and duration, expressed by the following formula:

[0042] in: It is a severe level. This refers to the number of sampling points during the abnormal period. It is the mode weight of the i-th sampling point. It is the energy efficiency deviation index value of the i-th sampling point, and the mode weight. Different weight values ​​correspond to different pattern codes obtained from the pattern encoding mapping. The vehicle identification number (VIN) binding process is performed after the feature vector is generated. The VIN is read from the vehicle's ECU, and a digital signature is generated using the HMAC algorithm to ensure binding integrity. The temporary storage operation in the security buffer uses atomic writes to avoid multi-threaded conflicts. The buffer size is fixed, and the oldest data is overwritten when it is full.

[0043] Optionally, the activation conditions for the anomaly data capture process are configurable. Preset durations and dynamic thresholds can be adjusted via a configuration interface, with adjustments taking effect immediately without requiring a system restart. The sampling channels of the high-speed data logger can be dynamically enabled or disabled, selecting the relevant channels based on the anomaly type to reduce unnecessary data collection. The feature extraction algorithm supports plug-in expansion, with new algorithms dynamically loaded to enhance the ability to identify new patterns. It can be understood that the pattern weight is crucial in the severity level calculation formula. The settings are based on pattern importance; patterns with higher importance have larger weights. For example, periodic oscillations have a weight of 1.2, and step changes have a weight of 1.0. These weight values ​​are stored in the configuration file. Number of sampling points. The determination is based on the duration of the anomaly, the duration stamp is converted into the number of sampling points, and the energy efficiency deviation index value. Read from the time series buffer. Feature vectors are serialized using JSON format for easy network transmission, and the serialized data is stored encrypted.

[0044] In practical implementation, a high-precision timer is used for the timer implementation. The timer interrupt handling function checks the indicator value, and a counter mechanism is used for continuous threshold judgment. The counter increments when the indicator exceeds the threshold and resets when the indicator falls below the threshold. Memory allocation for the high-speed data logger uses DMA technology, direct memory access reduces CPU intervention, and acquisition parameter configuration is completed by writing to registers. The signal data compression algorithm is executed in real time, with an adjustable compression ratio, balancing compression speed and ratio. Abnormal event time recording uses UTC timestamps, synchronized with the network time protocol. Digital signal processing for feature extraction uses dedicated libraries, such as the FFTW library, to improve computational performance. The enumerated values ​​for pattern encoding are predefined, including normal, oscillating, and step codes. The encryption of the security buffer uses the AES-256 algorithm, and the key is rotated periodically.

[0045] See Figure 4This chart presents the real-time monitoring and anomaly detection of the energy efficiency deviation index during the operation of the vehicle-to-everything (V2X) thermal management system. Multiple curves clearly illustrate the changes in the system's energy efficiency status at different time points. The main curve reflects the dynamic changes in the system's actual energy efficiency deviation, exhibiting regular fluctuations and sudden anomalies over time. Dynamic threshold reference lines and static baselines are specifically included in the chart, visually displaying the comparison between the system's energy efficiency status and preset standards. Within specific time periods, the chart highlights anomaly detection areas identified by the system with clearly filled color areas; these areas represent states where the energy efficiency deviation consistently exceeds the reasonable range. Several special trigger points are also marked in the chart, representing key moments when the system determines that an abnormal state persists and initiates high-speed data recording. The entire chart, through its carefully designed use of different colors and line styles, vividly demonstrates the complete workflow of the system from normal monitoring to anomaly identification and triggering data capture, reflecting the system's continuous monitoring capability of thermal management energy efficiency status and its rapid response mechanism to anomalies.

[0046] Example 4: The process of the cloud-based strategy module starting the strategy derivation engine begins with receiving an optimization request data packet. This data packet, uploaded through the communication interface module, contains feature vectors and vehicle identification information. Upon receipt, it is decrypted and its format verified. Decryption uses an asymmetric encryption algorithm such as RSA, while format verification checks the data packet length and field integrity. Feature vectors are extracted and input into the strategy decision tree model. The strategy decision tree model contains a historical optimization case library, which stores a large number of historical scenarios and their corresponding optimization schemes. Multi-level matching queries are performed in the strategy decision tree model. Multi-level matching is based on the similarity calculation of feature vectors to find the historical scenario most similar to the current feature vector. Successful optimization schemes corresponding to historical scenarios are obtained and their timeliness is corrected. Timeliness correction takes into account factors such as component aging or environmental changes, generating an optimization instruction sequence containing specific parameter adjustment values ​​and execution order. The optimization instruction sequence is encapsulated in a structured format such as XML or JSON. In practice, the strategy decision tree model's update mechanism operates periodically, with an update cycle configurable, such as weekly. It collects thermal management assembly optimization feedback data from multiple vehicle terminals. This feedback data includes energy efficiency indicators before and after adjustments, as well as execution logs. The collected feedback data is cleaned and labeled. The cleaning process removes duplicate or invalid records, while the labeling process adds scene labels and effect scores to form a training sample set. This training sample set is used for model retraining. Incremental learning algorithms are used to update the branch weights and node parameters of the strategy decision tree model. Incremental learning avoids full retraining, saving computational resources. The updated strategy decision tree model's matching accuracy on the test dataset is verified. The test dataset is randomly divided from historical data, and the matching accuracy threshold is set to a preset value, such as 95%. When the matching accuracy meets the requirements, the updated strategy decision tree model is deployed to the production environment. The deployment process employs a rolling update strategy to minimize service interruptions.

[0047] In some embodiments, when receiving optimization request packets, the transmission of the packets uses a security protocol such as HTTPS. The decryption key is obtained from a key management service. Format verification includes checksum verification and pattern checking. Checksum verification uses the CRC32 algorithm, and pattern checking is based on a predefined pattern. After extracting feature vectors, the feature vectors are normalized to ensure consistent input scale. The normalization method uses Min-Max scaling. Before being input into the policy decision tree model, the feature vectors are converted to the model input format, such as floating-point arrays. The multi-level matching query of the policy decision tree model adopts a tree traversal algorithm, starting from the root node, selecting branches based on feature values, until a leaf node returns the matching result. Similarity calculation uses Euclidean distance or cosine similarity, expressed by the formula:

[0048] in: It is a similarity score. It is the k-th component of the current eigenvector. It is the k-th component of the historical scene feature vector. This refers to the feature vector dimension; a higher similarity score indicates a higher matching degree. Successful optimization schemes for historical scenarios are retrieved using database queries, with the query condition being a similarity score exceeding a threshold. Timeliness correction applies a linear adjustment factor, calculated based on a time decay function. The generation of the optimization instruction sequence includes parameter validation and sorting logic. Parameter validation ensures values ​​are within a safe range, and the sorting logic is based on dependencies.

[0049] It is understandable that the data collection for updating the policy decision tree model is achieved through a message queue. The message queue buffers the data stream to prevent peak loads. The cleaning and labeling processes are executed automatically. The cleaning rules include range checks and consistency verification. The labeling tool uses semi-supervised learning to reduce human intervention. Referring to Table 1, the structured storage of the training sample set uses a relational database. Table 1 records the sample data.

[0050] Table 1: Training Sample Set Structure Table

[0051] The incremental learning algorithm uses online gradient descent, with an adjustable learning rate when updating branch weights. Node parameter adjustments are based on the gradient of the loss function. The validation process uses cross-validation, and the matching accuracy is calculated as the number of correct matches divided by the total number of test samples. When deploying to a production environment, new model version control uses semantic version numbers, and a rollback mechanism is activated when validation fails.

[0052] In practical implementation, decryption and format verification failure handling includes logging and retry mechanisms. When decryption fails, the error reason is recorded, and retries are limited to a maximum of three times. When format verification fails, the data packet is discarded and a negative response is sent. Before feature vectors are input into the policy decision tree model, missing values ​​are handled using interpolation, which is calculated based on adjacent feature values. Multi-level matching queries are optimized using an index structure; the index accelerates similarity calculation, and the historical scene database is sharded for storage, improving query performance. Timeliness correction factors are calculated based on the time difference, which is the difference between the current time and the historical scene time; the correction formula applies exponential decay. The serialization of instruction sequences is optimized using Protocol Buffers format to reduce transmission size.

[0053] Optionally, the update cycle of the policy decision tree model is dynamically adjusted based on the data inflow rate; the cycle is shortened when the rate is high. Data collection updates support filtering conditions, specifying vehicle type or operating condition range. Parallel processing of cleaning and labeling uses the MapReduce framework to improve processing efficiency. Hierarchical sampling is used for the training sample set to ensure class balance. Incremental learning algorithms include random forest incremental learning or decision tree incremental learning. Time series partitioning is used for the validation dataset to avoid data leakage. Health checks during deployment include model loading tests and inference latency tests; traffic is switched after the health checks are passed.

[0054] In some embodiments, the decryption operation of receiving optimization request packets is accelerated using a hardware security module, which provides encryption and decryption services. Field checks for format validation include enumeration value validation and range validation. The input cache for feature vectors caches recent query results; the cache size is configurable, reducing the number of model calls. Multi-level matching queries in the policy decision tree model support approximate matching, which uses locality-sensitive hashing to improve query speed. Optimization schemes for historical scenarios are stored in a distributed file system, providing high availability. The decay function parameters for timeliness correction are configurable and adjusted based on component lifespan. The template for generating optimization instruction sequences is based on the Velocity template engine, which dynamically generates instruction content.

[0055] In practical implementation, the updated data sources for the strategy decision tree model include real-time streaming and batch files. Real-time streaming uses Kafka for transmission, while batch files are uploaded periodically. The deduplication algorithm in the cleaning process is based on feature vector fingerprints, which are calculated using MD5 hashing. The scoring calculation in the annotation process uses objective metrics, such as the percentage improvement in energy efficiency. The training sample set is compressed using a columnar storage format to reduce storage space. The loss function of the incremental learning algorithm is defined as cross-entropy loss. Node parameter updates use backpropagation. Real-time alarms are used to monitor the verification accuracy, triggering manual review when the accuracy falls below a threshold. Version management for model deployment uses a Git-style workflow, ensuring traceability of version history.

[0056] See Figure 5This chart demonstrates the matching and analysis process of historical optimization cases by the strategy decision tree model in the cloud-based strategy module. The chart uses grouped bar charts to show the matching similarity between multiple historical scenarios and the current system state, along with their corresponding historical optimization effect scores. The two groups of bars represent the two key indicators of matching similarity and optimization effect score, respectively, revealing the reference value of different historical scenarios through intuitive comparison. The chart particularly highlights the location of the best-matching scenario and uses reference lines to mark the high-similarity baseline, clearly distinguishing highly relevant scenarios from general scenarios. The information box on the left side of the chart provides a summary of the key matching results, including specific information about the best-matching scenario and the timeliness of the data. The entire chart, through scientific data visualization, demonstrates how the cloud-based strategy module intelligently selects the most relevant scenarios from a large number of historical cases and comprehensively considers matching degree and historical effects to generate optimization decision schemes, reflecting a data-driven intelligent decision-making process and an effective mechanism for utilizing historical experience.

[0057] Example 5: The process of parsing and optimizing the instruction sequence by the vehicle control module starts after receiving the optimized instruction sequence. The receiving process is completed through the vehicle network gateway. It checks the digital signature and validity period. The digital signature verification uses an asymmetric encryption algorithm to verify the source credibility. The validity period check compares the current time with the instruction timestamp to prevent the instruction from expiring. The instruction content is parsed one by one and converted into control commands that can be executed by the underlying device. The control commands include setting register values ​​or sending CAN messages. It verifies whether the parameter range of each control command is within the safety limit. The safety limit is read from the vehicle configuration file to avoid overshooting and damage to the device. According to the timing relationship defined in the instruction sequence, the execution queue of the control commands is arranged. The timing relationship is set based on the device response time. The execution status monitoring point is set for each control command to prepare for collecting execution feedback. The monitoring point is implemented through a callback function. In practice, when performing parameter adjustment operations, control commands are sent sequentially to the actuators of the thermal management assembly according to the execution queue. The actuators include water pump drivers and fan controllers. The execution results of each control command are monitored, and the difference between the actual adjustment value and the target value is recorded. The difference value is used to evaluate the optimization effect. When an adjustment anomaly is detected, a backup adjustment plan is activated or the system is restored to the previous stable state. The backup adjustment plan is preset locally. The response data of the thermal management assembly during the adjustment process is collected to form an optimization effect report. The optimization effect report includes changes in energy efficiency indicators. The optimization effect report is uploaded to the cloud strategy server for model updates.

[0058] In some embodiments, when checking digital signatures and expiration dates, digital signature verification calls cryptographic library functions. The cryptographic library implements the RSA-PSS signature verification algorithm. Expiration date checking uses the system clock, synchronized with a network time server. Parsing instruction content uses a parser library, which maps JSON instructions to device commands. The conversion of device commands is based on a command dictionary, which stores the mapping relationship between command codes and operations. When verifying parameter ranges, security limits are cached in memory for fast access. Parameter range checks use comparison instructions, and errors are logged when parameters exceed the range. The execution queue is arranged based on a dependency graph, which describes the sequential constraints between commands. The critical path method is used to calculate the timing relationships. Registration of execution status monitoring points uses the observer pattern, which allows for the dynamic addition of monitoring points.

[0059] It is understandable that when sending control commands to the actuator, the sending interface adopts an asynchronous non-blocking method. Asynchronous non-blocking avoids thread blocking. The command sending result is notified through a callback. When monitoring the execution result, the actual adjustment value is read from the sensor feedback, and the target value is extracted from the command. The difference is calculated using the following formula:

[0060] in: It is the normalized difference value. This is the actual adjustment value. It is the target adjustment value. and These are the maximum and minimum allowed values ​​for the parameter.

[0061] In practical implementation, the complete process of parsing and optimizing instruction sequences includes an error recovery mechanism. This mechanism attempts to re-parse or use default values ​​when parsing fails. It also verifies the semantic correctness of command conversion, checking command existence and parameter integrity. Dynamic updates of security limits support hot updates via message notifications. The execution queue scheduling priority is adjustable, with higher-priority commands executed first. Data collection for execution status monitoring points uses a polling method, with the polling interval matching the command execution time. The communication protocol for sending control commands is standardized, using CANopen or Modbus. A command retry mechanism is enabled when transmission fails, with a limited number of retries. Timeout handling for monitoring execution results involves setting a timeout timer; command execution is considered failed after a timeout. Optionally, during the checking of digital signatures and validity periods, the signature key is rotated periodically every 90 days. The validity period tolerates time drift, allowing ±3 seconds. The parser for instruction content supports syntax extensions, implemented by loading plugins. Parameter range verification includes boundary value testing, which includes upper and lower bound verification. The execution queue is arranged to account for resource conflicts, and a locking mechanism is used for resource conflict detection. The callback functions for execution status monitoring points are lightweight and have an upper limit on execution time. The communication channels for sending control commands are redundantly backed up, switching to the backup channel in case of a primary channel failure.

[0062] In practical implementation, the pump driver control for parameter adjustment operations employs a PID algorithm, with PID parameters obtained from the instructions. The fan controller uses PWM modulation, with the PWM duty cycle specified by the instructions. Sensor data monitoring the execution results is filtered using a moving average method. The threshold for anomaly detection is adaptively adjusted based on historical anomaly frequencies. Backup adjustment schemes are stored with encrypted protection to prevent tampering. The state restoration to the previous stable state is saved every five minutes, with a saving frequency of once per second. Differential encoding is used for data compression in the optimization effect report, reducing data volume. Network bandwidth management for uploading the optimization effect report uses traffic shaping to ensure no impact on real-time control.

[0063] In some embodiments, the buffering mechanism for receiving the optimization instruction sequence uses double buffering to avoid data overwriting during parsing. The certificate chain verification for digital signature verification includes root certificate checking. Error handling for parsing instruction content includes exception capture, with stack traces recorded after exception capture. Security limit settings for verification parameter ranges are tiered, including warning and danger limits. The timing relationship of the execution queue is visualized for debugging purposes. The sampling frequency of monitoring data for execution status monitoring points is configurable, and the sampling frequency is related to the criticality of the command. A response mechanism is used to confirm the transmission of control commands, ensuring command delivery. In specific implementations, the actual adjustment value of the execution result is calibrated using sensor calibration curves stored in EEPROM. Adjustment anomalies are classified into overshoot anomalies and slow response anomalies, based on difference patterns. The selection strategy for alternative adjustment schemes is based on anomaly type, using a rule engine. The generation of optimization effect reports is executed asynchronously, without blocking the control loop. Data integrity of the uploaded optimization effect reports is verified using hash checking, with the hash checksum appended to the end of the report.

[0064] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0065] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A vehicle-to-everything (V2X) data-driven thermal management assembly energy efficiency optimization system, characterized in that, The system includes: The data acquisition module is used to initialize the energy efficiency monitoring environment after the vehicle thermal management assembly is started, and to continuously collect the operating parameter sequence of the thermal management assembly; The energy efficiency analysis module is used to compare the operating parameter sequence with the preset energy efficiency benchmark library in real time, generate an energy efficiency deviation index, and activate the abnormal data capture process when the energy efficiency deviation index exceeds the dynamic threshold to extract the feature vector of the current operating scenario. The communication interface module is used to encapsulate the feature vector into an optimization request data packet and send it to the cloud policy server through the vehicle network gateway; The cloud-based strategy module is used to receive the optimization request data packet, start the strategy derivation engine, and generate an adapted optimization instruction sequence based on scene feature matching. The vehicle control module is used to receive the optimized instruction sequence, parse and execute parameter adjustment operations on the thermal management assembly.

2. The vehicle-to-everything (V2X) data-driven thermal management assembly energy efficiency optimization system as described in claim 1, characterized in that, The process of initializing the energy efficiency monitoring environment by the data acquisition module specifically includes: Create independent monitoring instances for each critical component of the thermal management assembly, including water pumps, radiators, and electric fans; Configure a dedicated data sampling frequency and accuracy level for each monitoring instance; Establish a data link between the monitoring instance and the vehicle-mounted sensor network, and verify the integrity of data transmission; Load the energy efficiency assessment rule set into the local cache. The energy efficiency assessment rule set contains the expected energy efficiency values ​​under different operating conditions. Start the background monitoring task, which periodically verifies the running status of each monitoring instance.

3. The vehicle-to-everything (V2X) data-driven thermal management assembly energy efficiency optimization system as described in claim 2, characterized in that, The process by which the energy efficiency analysis module generates the energy efficiency deviation index specifically includes: The actual operating parameters for the current sampling period are obtained from the monitoring instance, including temperature readings, pressure readings, and power consumption readings. Calculate the multidimensional difference between the actual operating parameters and the expected energy efficiency value; The multidimensional differences are normalized to obtain the standardized deviation value; Based on the weighting coefficients defined in the energy efficiency assessment rule set, the standardized deviation values ​​are weighted and fused, and the fused result is output as the energy efficiency deviation index.

4. The vehicle-to-everything (V2X) data-driven thermal management assembly energy efficiency optimization system as described in claim 3, characterized in that, The process of activating the abnormal data capture process in the energy efficiency analysis module specifically includes: When the energy efficiency deviation index continuously exceeds the dynamic threshold for a preset duration, the high-speed data recording mode is triggered. Record all control signals and status feedback of the thermal management assembly during abnormal periods with millisecond-level precision; Feature extraction is performed on the recorded data stream to identify recurring abnormal patterns; The feature vector is constructed based on the identified abnormal patterns, and the feature vector includes pattern encoding, duration stamp, and severity level; The feature vector is bound to the current vehicle identification code and then temporarily stored in the security buffer.

5. The vehicle-to-everything (V2X) data-driven thermal management assembly energy efficiency optimization system as described in claim 4, characterized in that, The process of the cloud-based strategy module starting the strategy derivation engine includes: After receiving the optimization request data packet, decrypt and format verify it; The feature vector is extracted and input into a policy decision tree model, which contains a historical optimization case library; In the policy decision tree model, a multi-level matching query is performed to find the historical scene most similar to the current feature vector; Obtain the successful optimization solutions corresponding to the historical scenarios and make timeliness corrections for them; Generate the optimized instruction sequence, which includes specific parameter adjustment values ​​and execution order.

6. The vehicle-to-everything (V2X) data-driven thermal management assembly energy efficiency optimization system as described in claim 5, characterized in that, The update mechanism of the policy decision tree model includes: Regularly collect thermal management assembly optimization feedback data from multiple vehicle terminals; The collected optimization feedback data is cleaned and labeled to form a training sample set; The branch weights and node parameters of the policy decision tree model are updated using an incremental learning algorithm; Verify the matching accuracy of the updated policy decision tree model on the test dataset; Once the matching accuracy meets the requirements, the updated strategy decision tree model will be deployed to the production environment.

7. The vehicle-to-everything (V2X) data-driven thermal management assembly energy efficiency optimization system as described in claim 5, characterized in that, The process of parsing and optimizing the instruction sequence by the vehicle control module includes: After receiving the optimized instruction sequence, check its digital signature and validity period; The instructions are parsed one by one and converted into control commands that can be executed by the underlying devices. Verify that the parameter range of each control command is within the safety limits; Arrange the execution queue of control commands according to the timing relationship defined in the instruction sequence; Set up execution status monitoring points for each control command to collect execution feedback.

8. The vehicle-to-everything (V2X) data-driven thermal management assembly energy efficiency optimization system as described in claim 7, characterized in that, The process by which the vehicle control module performs parameter adjustment operations includes: Control commands are sent sequentially to the actuators of the thermal management assembly according to the execution queue; Monitor the execution results of each control command and record the difference between the actual adjustment value and the target value; When an adjustment anomaly is detected, activate the backup adjustment plan or restore to the previous stable state; Collect response data of the thermal management assembly during the adjustment process and generate an optimization effect report; The optimization effect report is uploaded to the cloud strategy server.

9. The vehicle-to-everything (V2X) data-driven thermal management assembly energy efficiency optimization system as described in claim 3, characterized in that, The process of calculating the multidimensional difference between the actual operating parameters and the expected energy efficiency value, and normalizing the multidimensional difference to obtain the standardized deviation value includes: For the actual operating parameters of each key component, extract the current values ​​of temperature reading, pressure reading, and power consumption reading respectively; Obtain the expected temperature, expected pressure, and expected power consumption under the corresponding operating conditions from the energy efficiency expected value; Calculate the absolute difference between the current value and the expected value for each reading to form the original difference vector; Each difference value in the original difference vector is scaled, and a preset normalization coefficient is used to map the difference value to the range of zero to one. Based on the physical meaning and importance of each reading, an adjustment factor is applied to calibrate the mapped difference value to obtain the standardized deviation value.

10. The vehicle-to-everything (V2X) data-driven thermal management assembly energy efficiency optimization system as described in claim 4, characterized in that, The process of triggering a high-speed data recording mode to record all control signals and status feedback of the thermal management assembly during an abnormal period with millisecond-level precision when the energy efficiency deviation index continuously exceeds the dynamic threshold for a preset duration includes: Monitor the time series of the energy efficiency deviation index, and start a timer when the index value is continuously higher than the dynamic threshold. After the timer reaches the preset duration, a trigger signal is generated and the high-speed data logger is activated; The high-speed data logger is configured to sample all control signal channels and status feedback channels of the thermal management assembly at millisecond intervals; The sampled signal data is sorted by timestamp and compressed and stored in a temporary buffer. Simultaneously, the start time, end time, and peak index of abnormal events are recorded for subsequent feature extraction.

Images