Power supply safety state monitoring and early warning method and system for liquid-cooled cable

Abstract

The invention relates to the technical field of cable states, and particularly discloses a power supply safety state monitoring and early warning method and system for a liquid-cooled cable, and the method comprises the steps: constructing a temperature region self-adaptive fusion model system which comprises an ultralow-temperature monitoring unit, a normal-temperature monitoring unit and an ultrahigh-temperature monitoring unit; characteristic parameter differences of the liquid-cooled cable and the cooling liquid under different working conditions are adapted based on a clear temperature zone division standard, and through an integrated design of a temperature zone adaptive fusion model, multi-source data intelligent processing and hierarchical linkage protection, accuracy, real-time performance and reliability of power supply safety state monitoring of the liquid-cooled cable are significantly improved. A powerful technical guarantee is provided for safe and efficient operation of a liquid-cooled cable system, and the problem that due to the fact that a traditional power supply safety state monitoring and early warning method for the liquid-cooled cable is insufficient in coverage, the recognition precision of an established model is insufficient is solved.

Description

Technical Field

[0001] This application relates to the field of cable condition technology, and specifically discloses a method and system for monitoring and early warning of the power supply safety status of liquid-cooled cables. Background Technology

[0002] New energy vehicles are gradually replacing traditional gasoline vehicles and occupying an increasingly larger market share due to their significant advantages such as low emissions, low travel costs, and low noise. However, core pain points such as slow charging speed and low charging efficiency remain key bottlenecks restricting the further penetration of new energy vehicles, especially in commercial vehicles and long-distance travel scenarios where the need for rapid charging is more urgent.

[0003] To solve the problem of fast charging, the industry has mainly developed two technical approaches: one is to increase the voltage platform of the vehicle, thereby reducing the charging current and reducing line losses by using higher voltage; the other is to increase the current transmission capacity of the charging pile, thereby directly increasing the charging power.

[0004] Regardless of the method used, charging cables, as the core carrier of energy transmission, face stringent performance tests. As charging power climbs to hundreds of kilowatts or even megawatts, cables generate a large amount of Joule heat during the transmission of high current. If this heat cannot be dissipated in time, it will not only accelerate the aging of the cable insulation layer and shorten its service life, but may also cause safety hazards such as cable burnout, overheating failure of the charging gun, and charging pile malfunction. In severe cases, it may even lead to fire accidents, threatening the safety of people and property.

[0005] Against this backdrop, liquid-cooled high-power charging cable technology has emerged as a core solution to the heat dissipation problem of high-power charging. This technology integrates a liquid-cooling circulation system inside the cable, utilizing the flow characteristics of cooling media such as special coolants or insulating oil to absorb the heat generated during cable conduction in real time. This keeps the cable within a reasonable operating temperature range, effectively overcoming thermal damage to various components of the charging link caused by heat generation, and providing a safety guarantee for high-current, high-power fast charging.

[0006] Compared with traditional charging cables, liquid-cooled charging cables have significant advantages: while achieving the same current transmission capacity, they have a smaller diameter and lighter weight, making them easier to wire and carry; at the same time, the integrated design of cooling and conductivity functions enables them to adapt to long-term high-power charging scenarios, greatly improving charging efficiency and stability, and have gradually become the mainstream technology in the field of fast charging for new energy vehicles.

[0007] However, the structural complexity and special working environment of liquid-cooled charging cables pose new challenges to the monitoring of their power supply safety status, and traditional monitoring technologies are no longer suitable for their application requirements.

[0008] Traditional safety monitoring of charging cables focuses primarily on extracting low-frequency signals during abnormal operation, enabling only preliminary monitoring of serious faults such as overload and short circuits. Furthermore, it completely neglects the unique characteristics of liquid-cooled systems, exhibiting three major drawbacks: First, low accuracy in abnormal state identification; it cannot accurately capture subtle signal changes caused by cooling system faults such as cooling medium leakage and poor circulation, or conductive side faults such as partial insulation damage and poor contact of the conductive core, leading to false alarms and missed alarms. Second, long fault location warning time; the low-frequency signal response is delayed, often only detecting anomalies after the fault has caused a certain degree of damage, failing to provide advance warning and proactive protection, and making it unsuitable for the high-power, high-reliability application scenarios of liquid-cooled cables. Third, a single monitoring dimension; it only monitors the state of the conductive core, ignoring the decisive role of the cooling system's operational stability in overall safety, and failing to form a collaborative monitoring closed loop.

[0009] In fact, the safe operation of liquid-cooled cables is the result of the coordinated action of the conductive core and the cooling system. Any abnormality in either component can trigger a chain of risks. Abnormalities in the flow rate, temperature, and pressure of the cooling medium, as well as blockages and leaks in the circulation pipeline, will directly lead to a sharp drop in heat dissipation efficiency, which in turn will cause the conductive core to overheat, the insulation layer to age rapidly, and even break down and burn. On the other hand, faults such as poor contact or local damage to the conductive core will also aggravate local heat generation, exceeding the heat dissipation capacity of the cooling system and creating safety hazards.

[0010] Traditional monitoring methods lack targeted collaborative monitoring designs, making it impossible to accurately correlate abnormalities between conductivity and cooling states, or to fully reflect the overall safety level of liquid-cooled cables, thus failing to meet the safety protection requirements of high-power fast charging scenarios.

[0011] With the large-scale construction of fast charging networks for new energy vehicles, liquid-cooled charging cables have been widely used in public charging piles, battery swapping stations, port logistics vehicle charging, and other scenarios. Their safe operation is directly related to the reliability and safety of the entire charging network. If there is a lack of high-precision and high-efficiency safety status monitoring and early warning technology, once a fault occurs, it will not only lead to the interruption of charging services, but may also cause equipment damage, fires and other safety accidents, resulting in significant economic losses, and even hindering the large-scale promotion of liquid-cooled fast charging technology.

[0012] Therefore, given the structural characteristics and operating properties of liquid-cooled cables, and to overcome the limitations of traditional monitoring technologies, developing a method and system that can achieve coordinated monitoring of conductivity and cooling states, high-precision anomaly identification, rapid fault location, and timely early warning, accurately capturing weak fault signals, shortening early warning response time, and improving the active protection capability of liquid-cooled cable power supply safety has become an urgent technical challenge to be solved in the field of fast charging for new energy vehicles.

[0013] Chinese patent application CN119293564B discloses a method and system for monitoring and early warning of the power supply safety status of liquid-cooled cables. This monitoring and early warning method relies heavily on training with historical operating parameters, and the operating condition division is based on conventional scenarios. This leads to the problem of insufficient coverage of the model established by this method. On the one hand, under extreme climatic conditions such as high temperatures, such as above 40°C in the summer in Sichuan and Chongqing, and below -20°C in Northeast China, the physical properties of the coolant, such as viscosity or boiling point, and the performance of the cable insulation layer will change significantly. The Gaussian regression temperature prediction model of this method does not fully incorporate the parameter correction mechanism under extreme environments, which easily leads to an increase in the temperature field prediction deviation.

[0014] On the other hand, the model training samples mostly focus on stable fast charging conditions, and do not cover sudden conditions such as sudden changes in charging power and instantaneous blockage of the coolant circulation loop. Under sudden conditions of thermal runaway, the model has difficulty accurately capturing the weak feature signals of such scenarios. In addition, as mentioned above, this method relies too much on selecting feature variables based on changes in the temperature field and constructing a dataset based on these feature variables. This feature variable selection mechanism is not specific enough to liquid-cooled cables of different materials, such as copper alloys or aluminum composite conductors. As the thermally conductive materials age, the feature correlation will drift, resulting in a decrease in the recognition accuracy of the subsequently established model.

[0015] In view of this, the present invention provides a method and system for monitoring and early warning of the power supply safety status of liquid-cooled cables, so as to solve the above problems. Summary of the Invention

[0016] The purpose of this invention is to solve the problem that the traditional power supply safety status monitoring and early warning method for liquid-cooled cables has insufficient coverage, resulting in insufficient recognition accuracy of the established model.

[0017] To achieve the above objectives, the present invention provides the following basic solution:

[0018] A method for monitoring and early warning of the power supply safety status of liquid-cooled cables includes the following steps: Step S1: Constructing a temperature-zone adaptive fusion model system, which includes three sets of temperature-zone-specific monitoring units, respectively adapted to ultra-low temperature conditions, normal temperature conditions, and ultra-high temperature conditions. Each specific monitoring unit integrates a fusion prediction module, an intelligent decision-making module, and a feature dynamic calibration module. Based on the multi-source data fusion logic, the model architecture is optimized to match the characteristic parameters of the liquid-cooled cable and coolant under different temperature zones; the ultra-low temperature condition corresponds to an ambient temperature ≤-25℃, the normal temperature condition corresponds to an ambient temperature of -25℃-40℃, and the ultra-high temperature condition corresponds to an ambient temperature ≥40℃; Step S2: Collecting the full life cycle operating parameters of the liquid-cooled cable, measured temperature field data, and scene-specific data. After signal enhancement and feature filtering preprocessing, a temperature-zone-specific training dataset is constructed. A transfer learning mechanism is introduced to optimize the model training process, and the three sets of temperature zones are respectively... The dedicated monitoring unit undergoes training and iterative calibration to obtain a mature monitoring unit that meets the preset accuracy requirements; Step S3: Real-time acquisition of the operating environment parameters of the liquid-cooled cable through multi-dimensional environmental sensing components, and automatic matching of the corresponding temperature zone dedicated monitoring unit by the temperature zone intelligent switching module in combination with the temperature gradient change trend, enabling uninterrupted adaptive monitoring of the entire temperature zone; Step S4: Real-time operating data of the liquid-cooled cable is acquired through multi-sensor fusion technology, input into the currently adapted temperature zone dedicated monitoring unit, and the fusion prediction module outputs accurate temperature field distribution data, the intelligent decision-making module realizes the capture of weak fault signals and diagnosis of sudden operating conditions, and the feature dynamic calibration module corrects the feature correlation weights in real time to offset the effects of material aging and parameter drift; Step S5: Graded early warning information is generated based on the diagnostic results of the temperature zone dedicated monitoring unit, and the intelligent adjustment module of the liquid cooling system is linked to trigger corresponding protective actions, while cross-temperature zone monitoring data interaction is realized through the inter-model data collaboration mechanism.

[0019] Furthermore, in step S1, the ultra-low temperature operating condition is adapted to the scenario of a sudden increase in coolant viscosity and a change in cable terminal interface pressure; the normal temperature operating condition is adapted to the scenario of conventional fast charging operation; and the ultra-high temperature operating condition is adapted to the scenario of a decrease in coolant boiling point and accelerated thermal aging of insulation layer.

[0020] Furthermore, the ultra-low temperature dedicated monitoring unit integrates a coolant viscosity-temperature-pressure correlation model, optimizes the kernel function based on the Kalman filter-Gaussian process regression fusion algorithm, and introduces an interface pressure correction factor to compensate for prediction deviations caused by insulation performance degradation under low temperature conditions; the ultra-high temperature dedicated monitoring unit embeds an insulation layer thermal aging rate prediction module, combines BM3D image-level denoising technology to identify overheating signals, and achieves multi-source sensor data fusion optimization through the Kalman filter-Gaussian process regression fusion algorithm to enhance the ability to predict local overheating faults; the room temperature dedicated monitoring unit adopts a spectrum coding and lightweight network collaborative architecture to balance monitoring accuracy and real-time computing efficiency, adapting to large-scale fast charging scenario deployment.

[0021] Furthermore, the full life cycle operating parameters of the liquid-cooled cable mentioned in step S2 include coolant flow rate, cable core current, terminal interface pressure, and ambient temperature and humidity; the scenario-specific data include extreme temperature zone characteristic data, emergency operating condition simulation data, and material and aging correlation data.

[0022] Furthermore, the extreme temperature characteristic data includes the viscosity-pressure coupling curve of the coolant at ultra-low temperatures and the dielectric loss-temperature correlation data of the insulation layer at ultra-high temperatures; the sudden operating condition simulation data generates power surge, instantaneous blockage of the coolant circuit, and terminal partial discharge scenario data through multi-physics field coupling simulation, and expands the scarce fault samples by combining adversarial generative networks; the material and aging correlation data includes the thermal conductivity decay curve of copper alloy and aluminum composite conductors, insulation layer aging cycle parameters, and coolant full life cycle performance decay data.

[0023] Furthermore, in step S2, the preprocessing is as follows: the SSA-VMD-MCS algorithm is used to eliminate signal noise and frequency drift, the improved mRMR algorithm is used to screen highly correlated and low-redundancy features, and the transfer learning mechanism is used to transfer the model parameters of similar scenarios to reduce sample dependence.

[0024] Furthermore, the multi-sensor fusion technology mentioned in step S4 includes integrated phase-sensitive optical time-domain reflectometry distributed fiber optic sensing technology, infrared thermal imaging technology, and ultra-high frequency sensing technology, which simultaneously acquire temperature field, vibration wave, partial discharge, and thermal imaging signals, and realize fault location through data-level fusion.

[0025] Furthermore, in step S4, the working mechanism of the feature dynamic calibration module includes the following steps: First, a multi-material feature database is established. During model initialization, feature weights are automatically matched according to the cable conductor material, and the monitoring requirements of different cable materials are quickly adapted through transfer learning. Second, based on the model aging theory, the correlation between feature variables and temperature field prediction results is monitored in real time. When the correlation decays to a preset threshold, automatic calibration is triggered. The feature weights are iteratively updated through Kalman filtering, and the calibration cycle is adaptively adjusted according to the operating conditions and environmental parameters. Finally, the performance parameters of the cable and coolant are collected periodically, a model accuracy decay prediction curve is constructed, and the monitoring data and calibration process are recorded in combination with encrypted logs. In step S3, the temperature zone intelligent switching module adopts a predictive switching logic. It starts the preheating of the target temperature zone dedicated monitoring unit in advance based on the trend of ambient temperature change, so as to achieve seamless switching. In step S5, the data collaboration mechanism realizes cross-unit data synchronization through a shared encrypted database. Abnormal features automatically trigger the adaptation of warning thresholds of other temperature zone units, so as to achieve consistency of warnings when switching temperature zones.

[0026] Furthermore, the graded early warning strategy described in step S5 is as follows: the early warning is divided into three levels according to the severity of the fault, including general early warning, severe early warning and emergency early warning. General early warning corresponds to fluctuations in charging power and slight abnormalities in coolant flow; severe early warning corresponds to a sudden drop in coolant flow rate and a slow rise in local temperature; emergency early warning corresponds to partial discharge of the terminal and a sudden temperature change exceeding the threshold. The intelligent adjustment module of the linkage liquid cooling system under each level triggers protective actions such as power reduction charging, dynamic flow adjustment and emergency shutdown.

[0027] Based on the power supply safety status monitoring and early warning method for liquid-cooled cables disclosed in this application, this application also discloses a power supply safety status monitoring and early warning system for liquid-cooled cables to implement the above method, including the following modules: A multi-dimensional sensing module: integrating a phase-sensitive optical time-domain reflectometry distributed fiber optic sensor, an infrared thermal imaging sensor, an ultra-high frequency sensor, and an environmental temperature and humidity sensor, used to simultaneously collect the liquid-cooled cable temperature field, vibration wave, partial discharge, terminal interface pressure, coolant flow rate, and environmental parameters, providing multi-source raw data for subsequent processing; A temperature zone intelligent switching module: connected to the multi-dimensional sensing module, receiving environmental temperature and temperature gradient change data, and using predictive switching logic to preheat the target temperature zone's dedicated monitoring unit in advance, achieving seamless adaptive switching under ultra-low temperature, normal temperature, and ultra-high temperature conditions; A temperature zone adaptive fusion model module: containing three sets of temperature zone-specific monitoring units, each unit integrating a Kalman filter-Gaussian process regression fusion prediction module, an intelligent decision-making module, and a feature dynamic calibration module, respectively adapted to different temperature zone conditions, and implementing... The system performs tasks such as temperature field prediction, emergency condition diagnosis, and feature weight calibration. The data processing and storage module uses the SSA-VMD-MCS algorithm and an improved mRMR algorithm to perform signal denoising and feature screening preprocessing, and builds a shared encrypted database to store full lifecycle operation data, scenario-specific data, model training parameters, and calibration logs. The liquid cooling system intelligent adjustment module, as a linkage execution component, receives instructions from the graded early warning and linkage control module, precisely adjusts the coolant flow rate and circulation speed, and coordinates with the charging system to adjust the charging power. The graded early warning and linkage control module receives diagnostic results from the temperature zone adaptive fusion model module, generates early warning information according to the early warning strategy, and outputs it visually. It also links with the liquid cooling system intelligent adjustment module to trigger protective actions such as reduced power charging, dynamic adjustment of coolant flow rate, and emergency shutdown based on the early warning level. The edge computing and communication module deploys lightweight edge computing units to undertake part of the model's inference tasks and builds a stable communication link to achieve data interaction between modules and information synchronization with the remote operation and maintenance platform.

[0028] The principle and effect of this solution are as follows:

[0029] 1. Compared with the prior art, the liquid-cooled cable power supply safety status monitoring and early warning method and system disclosed in this invention addresses the technical pain points of existing liquid-cooled cable monitoring technologies, such as poor adaptability across the entire temperature range, insufficient fault prediction accuracy, and susceptibility to environmental and aging effects. Through the construction of a temperature range adaptive fusion model system, multi-source data fusion, and intelligent algorithm collaborative design, it achieves high-precision safety monitoring and efficient early warning of liquid-cooled cables throughout their entire life cycle and under all operating conditions, with significant overall technical effects.

[0030] 2. Compared with existing technologies, this invention innovatively constructs a temperature zone adaptive fusion model system comprising three dedicated monitoring units for ultra-low temperature, normal temperature, and ultra-high temperature. Based on a clear temperature zone division standard, it adapts to the differences in characteristic parameters of liquid-cooled cables and coolants under different operating conditions, solving the technical problem that the single architecture of traditional monitoring models cannot simultaneously cover extreme temperature zones and normal operating conditions. Specifically, the ultra-low temperature dedicated monitoring unit effectively compensates for prediction deviations caused by insulation performance degradation under low-temperature environments through a coolant viscosity-temperature-pressure correlation model and an interface pressure correction factor. The ultra-high temperature dedicated monitoring unit embeds an insulation layer thermal aging rate prediction module. Simultaneously, the intelligent temperature zone switching module adopts predictive switching logic, preheating the target temperature zone unit in advance, achieving seamless and uninterrupted adaptive monitoring across the entire temperature zone. This completely overcomes the bottlenecks of existing technologies in low monitoring accuracy and delayed switching in extreme temperature zones, and can be widely adapted to diverse application scenarios such as extreme high and low temperature environments and large-scale fast charging.

[0031] 3. Compared with existing technologies, this invention combines multi-dimensional environmental sensing components with multi-sensor fusion technology to simultaneously collect multi-source signals such as operating parameters throughout the entire lifecycle of liquid-cooled cables, temperature field data, scene-specific data, vibration waves, partial discharge, and thermal imaging. After denoising using the SSA-VMD-MCS algorithm and feature filtering preprocessing using an improved mRMR algorithm, the quality of the original data is significantly improved. Simultaneously, a transfer learning mechanism is introduced to optimize the model training process, and adversarial generative networks are used to expand scarce fault samples, reducing the model's dependence on massive labeled samples and ensuring the training accuracy and generalization ability of the monitoring unit. In the data processing stage, a Kalman filter-Gaussian process regression fusion algorithm is used to achieve multi-source data fusion optimization. Combined with a feature dynamic calibration module, feature association weights are corrected in real time, effectively offsetting the impact of material aging, parameter drift, and environmental interference on the monitoring results. Compared with traditional monitoring technologies, it can accurately capture weak fault signals and sudden operating conditions, achieving accurate prediction of temperature field distribution and early prediction of faults such as local overheating and partial discharge, significantly improving diagnostic accuracy and reliability.

[0032] 4. Compared with existing technologies, the temperature-zone-specific monitoring units constructed in this invention integrate three major modules: fusion prediction, intelligent decision-making, and dynamic feature calibration, forming a closed-loop monitoring logic of "data acquisition - predictive analysis - decision diagnosis - dynamic calibration". The dynamic feature calibration module, based on a multi-material feature database and model aging theory, can automatically adapt feature weights according to the cable conductor material, monitor the correlation between feature variables and temperature field prediction results in real time, trigger adaptive calibration and record calibration logs, construct a model accuracy decay prediction curve, and achieve full lifecycle traceability and maintenance of monitoring accuracy. Simultaneously, the system constructs a cross-temperature zone data collaboration mechanism through a shared encrypted database, achieving data synchronization and early warning threshold adaptation for monitoring units in different temperature zones, ensuring consistent early warnings during temperature zone switching, and avoiding false alarms and missed alarms caused by operating condition changes. Compared to existing technologies that can only achieve short-term operational status monitoring and cannot address the accuracy decay problem caused by long-term aging, this invention can effectively extend the stable operation cycle of the monitoring system and adapt to the full lifecycle safety management needs of liquid-cooled cables. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 A flowchart illustrating a power supply safety status monitoring and early warning method for liquid-cooled cables according to an embodiment of this application is shown. Detailed Implementation

[0035] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0036] Implementation, for example Figure 1 As shown: This embodiment takes the liquid-cooled cable power supply system of a new energy fast charging station as the application scenario. In this scenario, the liquid-cooled cable adopts a copper alloy conductor, a cross-linked polyethylene insulation layer, and is equipped with an ethylene glycol-water mixed coolant. It needs to cover the extreme low temperature in winter (-30℃), the high temperature in summer (45℃), and the normal temperature in daily life (25℃) to realize the real-time monitoring and early warning of the cable safety status during fast charging. The following describes the implementation process of the present invention in detail with specific data, algorithm details and calculation steps.

[0037] First, in accordance with the system framework requirements, this monitoring and early warning system is built, specifically as follows: A liquid-cooled cable power supply safety status monitoring system is built, with the following specific configurations for each module: Multi-dimensional sensing module: integrating a phase-sensitive optical time-domain reflectometry distributed fiber optic sensor (measurement accuracy ±0.1℃, spatial resolution controlled at 1m), an infrared thermal imaging sensor (frame rate 25fps, temperature range -40℃-150℃), an ultra-high frequency sensor (detection frequency band 300MHz-3GHz), and temperature and humidity sensors (temperature accuracy ±0.2℃, humidity accuracy ±2%RH). It simultaneously collects cable temperature field, vibration waves, partial discharge, terminal interface pressure, coolant flow rate, and environmental parameters.

[0038] Intelligent temperature zone switching module: Built-in temperature gradient prediction algorithm, sampling period of 100ms. When the temperature change rate of three consecutive samples is ≥0.5℃ / s, the target temperature zone unit preheating is started, and the switching delay is ≤500ms.

[0039] Temperature Zone Adaptive Fusion Model Module: Three dedicated monitoring units are constructed, all developed based on the Python TensorFlow framework and deployed on edge computing nodes. The CPU is an Intel Core i7-12700H and the GPU is an NVIDIA RTX 3060. Each unit integrates a Kalman filter-Gaussian process regression fusion prediction module, an intelligent decision-making module, and a feature dynamic calibration module.

[0040] Data processing and storage module: The SSA-VMD-MCS algorithm and the improved mRMR algorithm are used for data preprocessing. A MySQL shared encrypted database is built with AES-256 encryption. The database stores full lifecycle running data, scenario-specific data and model parameters.

[0041] Tiered early warning and linkage control module: Configured with a visual early warning terminal, it links the variable frequency pump of the liquid cooling system and the fast charging power controller to realize the linkage between early warning output and protective actions.

[0042] Edge computing and communication module: Deploys lightweight edge computing units to handle 80% of model inference tasks; adopts 5G+Ethernet dual-link communication, with data transmission rate ≥100Mbps and latency ≤20ms.

[0043] Intelligent adjustment module for liquid cooling system: controls the coolant circulation pump, with a flow rate adjustment range of 5-50L / min, and works in conjunction with the fast charging power supply to achieve continuously adjustable power from 0-800kW.

[0044] The above-mentioned system is used to implement a method for monitoring and early warning of the power supply safety status of liquid-cooled cables. The method includes the following steps: Step S1: Construct a temperature zone adaptive fusion model system. The temperature zone adaptive fusion model system includes three sets of temperature zone-specific monitoring units, which are adapted to ultra-low temperature conditions, normal temperature conditions and ultra-high temperature conditions respectively. Each specific monitoring unit integrates a fusion prediction module, an intelligent decision-making module and a feature dynamic calibration module. Based on the multi-source data fusion logic, the model architecture is optimized to match the characteristic parameters of liquid-cooled cables and coolants under different temperature zones.

[0045] Step S2: Collect the full life cycle operation parameters, measured temperature field data and scene-specific data of the liquid-cooled cable. After signal enhancement and feature screening preprocessing, construct a temperature zone-specific training dataset. Introduce a transfer learning mechanism to optimize the model training process. Train and iteratively calibrate the three groups of temperature zone-specific monitoring units respectively to obtain mature monitoring units that meet the preset accuracy requirements.

[0046] The full life cycle operating parameters of the liquid-cooled cable mentioned in step S2 include coolant flow rate, cable core current, terminal interface pressure, and ambient temperature and humidity; the scenario-specific data include extreme temperature zone characteristic data, emergency operating condition simulation data, and material and aging correlation data.

[0047] The extreme temperature characteristic data includes the viscosity-pressure coupling curve of coolant at ultra-low temperatures and the dielectric loss-temperature correlation data of insulation layer at ultra-high temperatures; the sudden operating condition simulation data generates power mutation, instantaneous blockage of coolant circuit, and terminal partial discharge scenario data through multi-physics field coupling simulation, and expands the scarce fault samples by combining adversarial generative networks; the material and aging correlation data includes the thermal conductivity decay curve of copper alloy and aluminum composite conductor, insulation layer aging cycle parameters, and coolant full life cycle performance decay data.

[0048] The preprocessing in step S2 is as follows: the SSA-VMD-MCS algorithm is used to eliminate signal noise and frequency drift, the improved mRMR algorithm is used to screen highly correlated and low-redundancy features, and the transfer learning mechanism is used to transfer the model parameters of similar scenarios to reduce sample dependence.

[0049] Step S3: Real-time collection of operating environment parameters of liquid-cooled cable by multi-dimensional environmental sensing components, and automatic matching of corresponding temperature zone dedicated monitoring units by temperature zone intelligent switching module in combination with temperature gradient change trend, enabling uninterrupted adaptive monitoring of the entire temperature zone.

[0050] In step S3, the temperature zone intelligent switching module adopts predictive switching logic, and starts the target temperature zone dedicated monitoring unit to preheat in advance based on the trend of ambient temperature change, so as to achieve seamless switching.

[0051] The specific steps of the temperature zone intelligent switching module using predictive switching logic are as follows: First, assuming a sampling period of 100ms, ambient temperature data is continuously collected through a multi-dimensional sensing module; then, the temperature change rate 'a' is calculated in real time. When the temperature change rate 'a' for three consecutive samples is ≥0.5℃ / s or ≤-0.5℃ / s, it is determined to be a temperature zone crossing trend; second, a preheating start command is immediately sent to the dedicated monitoring unit of the target temperature zone to preload model parameters, sensor channels, and computing resources; finally, when the ambient temperature reaches the temperature zone threshold, the switching is completed instantaneously, with no data interruption and no monitoring blind spots throughout the process, achieving seamless adaptive monitoring across the entire temperature zone.

[0052] Step S4: Collect real-time operating data of liquid-cooled cable through multi-sensor fusion technology, input it into the current temperature zone-specific monitoring unit, and output accurate temperature field distribution data by the fusion prediction module. The intelligent decision-making module realizes the capture of weak fault signals and diagnosis of sudden working conditions. The feature dynamic calibration module corrects the feature correlation weight in real time to offset the effects of material aging and parameter drift.

[0053] The multi-sensor fusion technology mentioned in step S4 includes integrated phase-sensitive optical time-domain reflectometry distributed fiber optic sensing technology, infrared thermal imaging technology, and ultra-high frequency sensing technology. It simultaneously collects temperature field, vibration wave, partial discharge, and thermal imaging signals, and achieves fault location through data-level fusion.

[0054] Regarding the output of accurate temperature field distribution data by the fusion prediction module: Specifically, the fusion prediction module adopts a Kalman filter Gaussian process regression fusion algorithm. The data input includes multi-source sensor data such as distributed fiber optic temperature measurement, infrared thermal imaging, UHF, and temperature and humidity. Based on the steps of the Kalman filter algorithm, the state prediction step can be obtained. Then, through gain calculation, the state of the state prediction step is updated, and then the error is updated. Finally, the squared exponential kernel function is selected to perform regression fitting on the filtered data, and the output generates the axial-radial-surface three-dimensional temperature field distribution data of the cable, realizing accurate temperature prediction of the entire cable link.

[0055] In step S4, the working mechanism of the feature dynamic calibration module includes the following steps: First, a multi-material feature database is established. During model initialization, feature weights are automatically matched according to the cable conductor material, and the monitoring requirements of different cable materials are quickly adapted through transfer learning. Second, based on the model aging theory, the correlation between feature variables and temperature field prediction results is monitored in real time. When the correlation decays to a preset threshold, automatic calibration is triggered. The feature weights are iteratively updated through Kalman filtering, and the calibration cycle is adaptively adjusted according to the operating conditions and environmental parameters. Finally, cable and coolant performance parameters are collected periodically, a model accuracy decay prediction curve is constructed, and monitoring data and calibration process are recorded in encrypted logs.

[0056] Regarding the establishment of a multi-material feature database, the model automatically matches feature weights based on the cable conductor material during initialization, and quickly adapts to the monitoring needs of different cable materials through transfer learning. The first step is to establish a multi-material feature database, such as a material database for common cable conductors like copper alloys, aluminum alloys, aluminum composites, and tin-plated copper; a feature item database for thermal conductivity, resistivity, temperature coefficient, aging rate, and viscosity-pressure coupling coefficient; and a storage format database using a MySQL encrypted database with AES-256 encryption, indexed and stored by material type.

[0057] Then, during model initialization, feature weights are automatically matched based on the cable conductor material. The cable conductor material identifier is read during model initialization, and the corresponding material feature weight vector is retrieved from the database. For example, for copper alloy: flow rate 0.3, current 0.4, pressure 0.2, temperature 0.1.

[0058] Next, the monitoring requirements for cables of different materials are quickly adapted through transfer learning. Regarding transfer learning: As is well known, transfer learning has a source domain, which in this application is the well-trained liquid-cooled cable monitoring model for fast charging stations. Through this model, mature convolutional layer weight parameters, feature extractor parameters, and kernel function parameters are obtained. Then, for new cable models, only minor adjustments are needed to achieve rapid convergence and adapt to material differences.

[0059] Step S5: Generate graded early warning information based on the diagnostic results of the temperature zone-specific monitoring unit, trigger corresponding protective actions by linking the intelligent adjustment module of the liquid cooling system, and realize cross-temperature zone monitoring data interaction through the data collaboration mechanism between models.

[0060] The graded early warning strategy described in step S5 is as follows: the early warning is divided into three levels according to the severity of the fault, including general early warning, severe early warning and emergency early warning. General early warning corresponds to fluctuations in charging power and slight abnormalities in coolant flow; severe early warning corresponds to a sudden drop in coolant flow rate and a slow rise in local temperature; emergency early warning corresponds to partial discharge of the terminal and a sudden temperature change exceeding the threshold. The intelligent adjustment module of the linkage liquid cooling system under each level triggers protective actions such as power reduction charging, dynamic flow adjustment and emergency shutdown.

[0061] The temperature zone division criteria mentioned in step S1 are as follows: the ultra-low temperature condition corresponds to an ambient temperature ≤ -25℃, which is suitable for scenarios of sudden increase in coolant viscosity and change in cable terminal interface pressure; the normal temperature condition corresponds to an ambient temperature of -25℃ to 40℃, which is suitable for conventional fast charging operation scenarios; the ultra-high temperature condition corresponds to an ambient temperature ≥ 40℃, which is suitable for scenarios of decrease in coolant boiling point and accelerated thermal aging of insulation layer.

[0062] Specifically: Ultra-low temperature dedicated monitoring unit: integrates coolant viscosity-temperature-pressure correlation model, optimizes kernel function based on Kalman filter-Gaussian process regression fusion algorithm, wherein the Gaussian process regression kernel function uses square exponential kernel, and after Kalman filter optimization, it compensates for prediction deviation caused by insulation performance degradation at low temperature.

[0063] Ultra-high temperature dedicated monitoring unit: Embedded insulation layer thermal aging rate prediction module, combined with BM3D image-level denoising technology to process infrared thermal imaging signals, enhances the ability to predict local overheating faults, and can identify local temperature anomalies ≥0.5℃.

[0064] Dedicated monitoring unit for ambient temperature: It adopts a collaborative architecture of spectrum coding and lightweight MobileNetV3 network. Spectrum coding converts current and flow signals into 512-dimensional feature vectors, and the number of MobileNetV3 network parameters is controlled at 2.8M. The inference speed is ≥30 frames / s, balancing monitoring accuracy and real-time computing efficiency, and is suitable for large-scale deployment of fast charging stations.

[0065] Next is the data construction and processing: First, collect the full life cycle operating parameters of the liquid-cooled cable, measured temperature field data, and scenario-specific data to construct a temperature-zone-specific training dataset. In this embodiment, it is assumed that the sample size is 100,000 sets for each region. The preprocessing steps are as follows:

[0066] Data Acquisition: Collect full lifecycle operating parameters: coolant flow rate 5-50 L / min, cable core current 100-500 A, terminal interface pressure 0.1-0.5 MPa, ambient temperature and humidity -30℃-45℃, 0%-90% RH; Collect extreme temperature characteristic data: coolant viscosity-pressure coupling curve at ultra-low temperature (-30℃), insulation layer dielectric loss-temperature correlation data at ultra-high temperature (45℃); Sudden operating condition simulation data: Through ANSYS multiphysics coupling simulation, generate scenario data for power surges, instantaneous blockage of coolant circuits, and partial discharge at terminals, and expand scarce fault samples using Generative Adversarial Network (GAN), increasing the number of fault samples from 2000 to 20,000; Material and aging correlation data: thermal conductivity decay curve of copper alloy conductor, aging cycle parameters of cross-linked polyethylene insulation layer, and coolant full lifecycle performance decay data.

[0067] Preprocessing: The SSA-VMD-MCS algorithm is used to eliminate signal noise and frequency drift. First, the variational mode decomposition (VMD) parameters are optimized by the Sparrow Search (SSA) algorithm, with the decomposition level K=5, penalty factor α=2000, and noise tolerance σ=0.001, resulting in 5 intrinsic mode components. Then, Monte Carlo simulation is used to filter the effective components, ultimately improving the signal-to-noise ratio from 25dB to 48dB.

[0068] Then, the mRMR algorithm was improved to screen features: the mutual information between each feature and the target variable was calculated, and features with high correlation and low redundancy were screened out. The four core features of coolant flow rate, cable core current, terminal pressure and ambient temperature were retained, and the feature dimension was reduced from 12 dimensions to 4 dimensions.

[0069] The specific steps are as follows: First, construct the original feature set: typically 12 dimensions, including flow rate, current, pressure, ambient temperature and humidity, vibration, partial discharge, insulation loss, etc.; then, calculate mutual information: MI(x,y), which measures the correlation between feature x and target variable y (temperature field / fault label); calculate redundancy: mutual information between features is used to eliminate highly redundant features; establish an objective function through mutual information, and use the objective function to select and retain four core features: coolant flow rate, cable core current, terminal pressure, and ambient temperature. At this point, the dimension is reduced from 12 to 4, thus reducing the computational load without sacrificing accuracy.

[0070] Next, transfer learning optimization: transfer the parameters of the liquid-cooled cable monitoring model of similar fast charging stations, initialize the weights of this model, and reduce the number of model training iterations from 2000 to at least 800 to improve the convergence speed.

[0071] Specifically: As mentioned above, transfer learning has a source domain, which in this application is the well-trained monitoring model of liquid-cooled cables for fast charging stations. Through this model, mature convolutional layer weight parameters, feature extractor parameters, and kernel function parameters are obtained. The source model parameters are directly loaded to avoid random initialization. By freezing the bottom feature layer, fine-tuning is achieved by training only the top decision layer and output layer, thereby improving the training convergence speed and ultimately reducing the number of model training iterations from 2000 to at least 800.

[0072] Next, the preprocessed dataset was divided into training, validation, and test sets in a 7:2:1 ratio. Three temperature-specific monitoring units were then trained using these sets with the following parameters: learning rate 0.001, batch size 32, and 800 iterations. During training, a feature dynamic calibration module was used for iterative calibration. Training stopped when the prediction accuracy on the validation set reached ≥98.5%. The final mature monitoring units were obtained: ≥98.7% accuracy for the ultra-low temperature unit, ≥99.2% for the room temperature unit, and ≥98.6% for the ultra-high temperature unit, all meeting the preset accuracy requirements.

[0073] The multi-dimensional sensing module collects ambient temperature in real time, and the intelligent temperature zone switching module achieves predictive switching based on the temperature gradient change trend. A specific example is as follows:

[0074] First, the winter fast charging scenario: the ambient temperature drops from -24℃ to -26℃, the temperature change rate is -0.6℃ / s, and the switching conditions are met after 3 consecutive samplings. The temperature zone intelligent switching module starts the ultra-low temperature dedicated monitoring unit to preheat 200ms in advance, and completes the seamless switching. The monitoring is uninterrupted during the switching process, and the temperature field prediction deviation is ≤0.2℃.

[0075] Secondly, in the summer fast charging scenario: when the ambient temperature rises from 39℃ to 41℃, with a temperature change rate of 0.7℃ / s, the ultra-high temperature dedicated monitoring unit is activated for preheating. After switching, the thermal aging rate prediction module is immediately embedded to calculate the aging rate of the insulation layer in real time.

[0076] Next, data is collected synchronously using multi-sensor fusion technology. The specific calculation and processing steps are as follows: First, data acquisition: A phase-sensitive optical time-domain reflectometry distributed fiber optic sensor collects the axial temperature field data of the cable, an infrared thermal imaging sensor collects the surface temperature distribution image of the cable, and an ultra-high frequency sensor collects the partial discharge signal. The sampling period is set to 100ms. Second, data-level fusion: A Kalman filter-Gaussian process regression fusion algorithm is used to process the multi-source data. After fusion, accurate temperature field distribution data is output. For example, the deviation between the predicted and measured values ​​of the cable core temperature is only 0.15℃. Third, feature dynamic calibration: Calibration is performed according to the following steps. Calculation: During model initialization, feature weights are matched from a multi-material feature database based on the copper alloy conductor material. For example, coolant flow rate weight is 0.3, cable core current weight is 0.4, terminal pressure weight is 0.2, and ambient temperature weight is 0.1. Transfer learning is used to quickly adapt the model. Based on the model aging theory, the correlation between feature variables and temperature field prediction results is monitored in real time. When the correlation decays from the initial value to a preset threshold, automatic calibration is triggered. The feature weights are iteratively updated using Kalman filtering. After the update, the cable core current weight is 0.42, the ambient temperature weight is 0.12, and other weights remain unchanged. The calibration cycle is adaptively adjusted, i.e., the cycle is shortened.

[0077] Cable and coolant performance parameters are collected every 24 hours to construct a model accuracy attenuation prediction curve. This embodiment discloses the current model prediction curve, y=0.0002x+0.992, where x is the number of operating days, 0.0002 and 0.992 are considered modifiable coefficients, and y is the attenuation amount. The monitoring data and calibration process can be recorded in encrypted logs to ensure traceability of accuracy. Specifically, cable conductor temperature, insulation dielectric loss, coolant viscosity, and flow deviation are collected every 24 hours. The performance parameters are mapped to the [0,1] interval through normalization, and the accuracy attenuation trend is fitted using univariate linear regression of y=0.0002x+0.992.

[0078] The intelligent decision-making module captures weak fault signals and sudden operating conditions. Specific examples are as follows: Weak fault diagnosis: For example, during fast charging, the coolant flow rate is detected to slowly decrease from 30L / min to 28L / min, with a change rate of 0.05L / min·s, and the local temperature rises from 35℃ to 36.2℃. The thermal imaging signal is processed by the ultra-high temperature dedicated monitoring unit BM3D noise reduction technology to identify the local overheating signal and diagnose it as a slight blockage in the coolant circuit, which falls under the category of serious warning.

[0079] Emergency Condition Diagnosis: When the amplitude of the partial discharge signal in the terminal suddenly rises to 500pC, the ultra-high frequency sensor quickly captures the signal. The intelligent decision-making module combines the temperature change data, for example, a temperature rise of 2°C within 1 second, and diagnoses the terminal insulation as damaged. This falls under the category of emergency warning.

[0080] Based on the diagnostic results, a graded early warning message is generated, which triggers protective actions by linking the intelligent adjustment module of the liquid cooling system. At the same time, cross-temperature zone data collaboration is achieved through a shared encrypted database. Specific examples are as follows: General warning: Fast charging power fluctuation, such as a drop from 100kW to 105kW to 100kW, or slight abnormality in coolant flow, such as a drop from 30L / min to 29.5L / min to 30L / min. The graded early warning and linkage control module outputs a yellow warning.

[0081] When a yellow warning is issued, the intelligent adjustment module of the linked liquid cooling system will stabilize the coolant flow rate at 30L / min without adjusting the charging power, and the warning information will be synchronized to the remote operation and maintenance platform.

[0082] Severe Warning: When the coolant flow rate drops sharply, for example from 30L / min to 20L / min, or the local temperature rises slowly, for example from 35℃ to 38℃, an orange warning is issued, and the coolant flow rate is adjusted to 40L / min. At the same time, the charging power is reduced from 120kW to 100kW. The cross-temperature zone data coordination mechanism triggers the adaptation of the warning thresholds for the normal temperature and ultra-high temperature units, that is, the temperature warning threshold for the normal temperature unit is adjusted from 40℃ to 39℃.

[0083] Emergency Warning: If the amplitude of partial discharge in the terminal exceeds 500pC or the temperature suddenly exceeds the threshold (e.g., from 35℃ to 37.5℃ or a temperature rise of 2.5℃ within 1 second), a red warning will be issued, immediately triggering an emergency shutdown, cutting off the fast charging power supply, maintaining the maximum flow rate of the coolant system for cooling, and simultaneously encrypting and uploading the fault data to the remote operation and maintenance platform to coordinate with operation and maintenance personnel for on-site troubleshooting.

[0084] This method solves the problem that traditional liquid-cooled cable power supply safety status monitoring and early warning methods have insufficient coverage, resulting in insufficient recognition accuracy of the established model.

[0085] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0086] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units; some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0087] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0088] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. Alternatively, if the integrated unit of the present invention is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0089] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for monitoring and early warning of the power supply safety status of a liquid-cooled cable, characterized in that, Includes the following steps: Step S1: Construct a temperature zone adaptive fusion model system. The temperature zone adaptive fusion model system includes three sets of temperature zone-specific monitoring units, which are adapted to ultra-low temperature conditions, normal temperature conditions, and ultra-high temperature conditions, respectively. Each specific monitoring unit integrates a fusion prediction module, an intelligent decision-making module, and a feature dynamic calibration module. Based on the multi-source data fusion logic, the model architecture is optimized to match the characteristic parameters of liquid-cooled cables and coolants under different temperature zones. The ultra-low temperature condition corresponds to an ambient temperature ≤-25℃, the normal temperature condition corresponds to an ambient temperature of -25℃ to 40℃, and the ultra-high temperature condition corresponds to an ambient temperature ≥40℃. Step S2: Collect the full life cycle operation parameters, temperature field measured data and scene-specific data of the liquid-cooled cable. After signal enhancement and feature screening preprocessing, construct a temperature zone-specific training dataset. Introduce the transfer learning mechanism to optimize the model training process. Train and iteratively calibrate the three groups of temperature zone-specific monitoring units respectively to obtain mature monitoring units that meet the preset accuracy requirements. Step S3: Real-time collection of operating environment parameters of liquid-cooled cable through multi-dimensional environmental sensing components; automatic matching of corresponding temperature zone dedicated monitoring unit by temperature zone intelligent switching module in combination with temperature gradient change trend; uninterrupted adaptive monitoring of the entire temperature zone. Step S4: Collect real-time operating data of liquid-cooled cable through multi-sensor fusion technology, input it into the current temperature zone-specific monitoring unit, and output accurate temperature field distribution data by the fusion prediction module. The intelligent decision module realizes the capture of weak fault signals and diagnosis of sudden working conditions. The feature dynamic calibration module corrects the feature correlation weight in real time to offset the effects of material aging and parameter drift. Step S5: Generate graded early warning information based on the diagnostic results of the temperature zone-specific monitoring unit, trigger corresponding protective actions by linking the intelligent adjustment module of the liquid cooling system, and realize cross-temperature zone monitoring data interaction through the data collaboration mechanism between models.

2. The method for monitoring and early warning of the power supply safety status of a liquid-cooled cable according to claim 1, characterized in that, In step S1, the ultra-low temperature condition is adapted to the scenario of a sudden increase in coolant viscosity and a change in cable terminal interface pressure; the normal temperature condition is adapted to the scenario of conventional fast charging operation; and the ultra-high temperature condition is adapted to the scenario of a decrease in coolant boiling point and accelerated thermal aging of insulation layer.

3. The method for monitoring and early warning of the power supply safety status of a liquid-cooled cable according to claim 2, characterized in that, Ultra-low temperature dedicated monitoring unit: integrates a coolant viscosity-temperature-pressure correlation model, optimizes the kernel function based on Kalman filter-Gaussian process regression fusion algorithm, and introduces an interface pressure correction factor to compensate for prediction deviations caused by insulation performance degradation under low temperature conditions; Ultra-high temperature dedicated monitoring unit: embeds an insulation layer thermal aging rate prediction module, combines BM3D image-level denoising technology to identify overheating signals, and achieves multi-source sensor data fusion optimization through Kalman filter-Gaussian process regression fusion algorithm to enhance the ability to predict local overheating faults; Room temperature dedicated monitoring unit: adopts a spectrum coding and lightweight network collaborative architecture to balance monitoring accuracy and real-time computing efficiency, and is suitable for large-scale fast charging scenario deployment.

4. A method for monitoring and early warning of the power supply safety status of a liquid-cooled cable according to claim 1 or 3, characterized in that, The full life cycle operating parameters of the liquid-cooled cable mentioned in step S2 include coolant flow rate, cable core current, terminal interface pressure, and ambient temperature and humidity; the scenario-specific data include extreme temperature zone characteristic data, emergency operating condition simulation data, and material and aging correlation data.

5. The method for monitoring and early warning of the power supply safety status of a liquid-cooled cable according to claim 4, characterized in that, The extreme temperature characteristic data includes the viscosity-pressure coupling curve of coolant at ultra-low temperatures and the dielectric loss-temperature correlation data of insulation layer at ultra-high temperatures; the sudden operating condition simulation data generates power mutation, instantaneous blockage of coolant circuit, and terminal partial discharge scenario data through multi-physics field coupling simulation, and expands the scarce fault samples by combining adversarial generative networks; the material and aging correlation data includes the thermal conductivity decay curve of copper alloy and aluminum composite conductor, insulation layer aging cycle parameters, and coolant full life cycle performance decay data.

6. The power supply safety status monitoring and early warning method for a liquid-cooled cable according to claim 5, characterized in that, In step S2, the preprocessing is as follows: the SSA-VMD-MCS algorithm is used to eliminate signal noise and frequency drift, the improved mRMR algorithm is used to screen highly correlated and low-redundancy features, and the transfer learning mechanism is used to transfer the model parameters of similar scenarios to reduce sample dependence.

7. The power supply safety status monitoring and early warning method for a liquid-cooled cable according to claim 6, characterized in that, The multi-sensor fusion technology mentioned in step S4 includes integrated phase-sensitive optical time-domain reflectometry distributed fiber optic sensing technology, infrared thermal imaging technology, and ultra-high frequency sensing technology. It simultaneously collects temperature field, vibration wave, partial discharge, and thermal imaging signals, and achieves fault location through data-level fusion.

8. The method for monitoring and early warning of the power supply safety status of a liquid-cooled cable according to claim 7, characterized in that, In step S4, the working mechanism of the feature dynamic calibration module includes the following steps: First, a multi-material feature database is established. During model initialization, feature weights are automatically matched according to the cable conductor material. Through transfer learning, the monitoring needs of cables of different materials are quickly adapted. Secondly, based on the model aging theory, the correlation between feature variables and temperature field prediction results is monitored in real time. When the correlation decays to a preset threshold, automatic calibration is triggered. The feature weights are updated iteratively through Kalman filtering, and the calibration cycle is adaptively adjusted according to the operating conditions and environmental parameters. Finally, the performance parameters of the cable and coolant are collected regularly to construct a model accuracy attenuation prediction curve, and the monitoring data and calibration process are recorded in encrypted logs. In step S3, the temperature zone intelligent switching module adopts a predictive switching logic, which combines the trend of ambient temperature change to start the preheating of the target temperature zone dedicated monitoring unit in advance, so as to achieve seamless switching. In step S5, the data collaboration mechanism achieves cross-unit data synchronization through a shared encrypted database, and abnormal features automatically trigger the adaptation of warning thresholds in other temperature zone units, thereby achieving consistent warnings during temperature zone switching.

9. The method for monitoring and early warning of the power supply safety status of a liquid-cooled cable according to claim 8, characterized in that, The graded early warning strategy described in step S5 is as follows: the early warning is divided into three levels according to the severity of the fault, including general early warning, severe early warning and emergency early warning. General early warning corresponds to fluctuations in charging power and slight abnormalities in coolant flow. A severe warning corresponds to a sudden drop in coolant flow rate and a slow rise in local temperature. Emergency warnings correspond to partial discharge of the terminal and sudden temperature changes exceeding the threshold; the intelligent adjustment module of the linkage liquid cooling system at each level triggers protective actions such as power reduction charging, dynamic flow adjustment, and emergency shutdown.

10. A power supply safety status monitoring system for liquid-cooled cables, characterized in that, For implementing the power supply safety status monitoring and early warning method for liquid-cooled cables as described in any one of claims 1-9, the system includes the following modules: Multi-dimensional sensing module: integrates phase-sensitive optical time-domain reflectometry distributed fiber optic sensor, infrared thermal imaging sensor, UHF sensor and environmental temperature and humidity sensor, used to simultaneously collect temperature field, vibration wave, partial discharge, terminal interface pressure, coolant flow and environmental parameters of liquid-cooled cable, providing multi-source raw data for subsequent processing; Intelligent temperature zone switching module: It connects to the multi-dimensional sensing module to receive data on changes in ambient temperature and temperature gradient. It adopts a predictive switching logic to preheat the dedicated monitoring unit of the target temperature zone in advance, so as to achieve seamless adaptive switching under ultra-low temperature, normal temperature and ultra-high temperature conditions. Temperature Zone Adaptive Fusion Model Module: Includes three sets of temperature zone-specific monitoring units. Each unit integrates a Kalman filter-Gaussian process regression fusion prediction module, an intelligent decision-making module, and a feature dynamic calibration module, which are adapted to different temperature zone conditions to perform temperature field prediction, emergency condition diagnosis, and feature weight calibration tasks. Data processing and storage module: The SSA-VMD-MCS algorithm and the improved mRMR algorithm are used to perform signal denoising and feature selection preprocessing, and a shared encrypted database is built to store full lifecycle operation data, scene-specific data, model training parameters and calibration logs; Intelligent adjustment module for liquid cooling system: As a linkage execution component, it receives instructions from the graded early warning and linkage control module, precisely adjusts the coolant flow rate and circulation speed, and coordinates with the charging system to adjust the charging power; The graded early warning and linkage control module receives the diagnostic results from the temperature zone adaptive fusion model module, generates early warning information according to the early warning strategy and outputs it visually. At the same time, it links with the intelligent adjustment module of the liquid cooling system to trigger protective actions such as power reduction charging, dynamic adjustment of coolant flow, and emergency shutdown according to the early warning level. Edge computing and communication module: Deploys lightweight edge computing units to undertake part of the model inference tasks; and establishes stable communication links to realize data interaction between modules and information synchronization with the remote operation and maintenance platform.

Images