A new energy vehicle high-voltage accident grading response and emergency rescue operation guidance method and system

Abstract

The application provides a new energy vehicle high-voltage accident grading response and emergency rescue operation guiding method and system, and is applied to the technical field of data processing. The application collects multi-source data of high-voltage system voltage and current, battery state, generates a standardized correlation sequence after denoising, and then converts into a three-dimensional risk atlas, analyzes the fault tree to subdivide the electric leakage, overvoltage and thermal runaway scenes, and constructs a type accident grading model. Then, the hazard weight factor is extracted to build a dynamic evaluation matrix, and the key response index is obtained by collaborative calculation of multi-source data. According to the accident stage, the response level is divided by parameter, and the characteristic matrix is generated; by comparing real-time and historical data, the correction factor is generated by using the accident curve slope. Combined with the vehicle type, architecture and environment grouping, the key factors are screened by using the support vector machine, the personalized rescue model is built by fusing multiple constraints, and finally the real-time grading response instruction and scene rescue guidance are output.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for graded response and emergency rescue operation guidance for high-voltage accidents in new energy vehicles. Background Technology

[0002] Current technologies for handling high-voltage accidents in new energy vehicles mainly fall into two categories, both of which have significant limitations:

[0003] Single-parameter threshold warning scheme: This scheme monitors single parameters such as high voltage and battery temperature using onboard sensors, and sets fixed thresholds (e.g., alarm when voltage exceeds 420V, power cut-off when temperature exceeds 60℃) to achieve simple risk warnings. For example, one technology monitors battery temperature through a BMS and triggers a thermal runaway warning when the temperature is ≥65℃. However, it relies solely on the temperature parameter and does not consider related factors such as ambient humidity and collision signals, which can easily lead to false alarms or missed alarms.

[0004] A standardized emergency response plan: Employing a unified rescue process to address different types of high-voltage accidents, such as leakage, overvoltage, or thermal runaway, all operations follow a fixed procedure of "cutting off the high-voltage power supply → waiting for professional rescue." For example, one emergency guideline requires dispatching fire trucks to the scene for all high-voltage accidents, failing to differentiate between accident levels (e.g., Level I low-risk leakage versus Level III thermal runaway) and scenarios (e.g., urban roads and tunnels), leading to wasted rescue resources or delayed responses.

[0005] Furthermore, the accuracy of risk assessment is low. Existing technologies mostly adopt a uniform classification standard (such as dividing risks into low / medium / high levels), without designing specific classification logic for the differences in the types of high-voltage accidents (leakage, overvoltage and overcurrent, thermal runaway). For example, the core risk of thermal runaway accidents is the "thermal propagation rate," while the core risk of overvoltage and overcurrent accidents is the "component burnout rate." Existing models classify both using the same indicators (such as risk level scores), leading to the "violent stage" of thermal runaway accidents and the "development stage" of overvoltage and overcurrent accidents being misclassified as the same level, making it impossible to match differentiated response strategies.

[0006] The existing emergency response strategies are mostly static and fixed, lacking dynamic adjustment mechanisms. They are not dynamically adjusted based on the real-time development trend of the accident (such as a sudden increase in risk) and historical data. For example, a certain technology stipulates that "rescue must arrive at the scene within 10 minutes for a Level II accident." However, when the accident level suddenly escalates from Level II to Level III within 5 minutes (such as thermal runaway from the smoke stage to the flame stage), resources are still dispatched according to the original 10-minute time limit, resulting in a delay in rescue. At the same time, the historical accident patterns of the same vehicle model and the same scenario are not considered (such as the average time for thermal runaway of a certain vehicle model to enter the severe stage is 8 minutes), resulting in poor strategy adaptability.

[0007] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore includes information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0008] According to one aspect of this application, a method for graded response and emergency rescue operation guidance for high-voltage accidents in new energy vehicles is provided, comprising: acquiring real-time voltage and current data, battery status data, collision signal data, and environmental temperature and humidity data of the high-voltage system of the new energy vehicle; removing instantaneous pulse interference and sensor acquisition noise from the high-voltage circuit; generating a standardized high-voltage-battery-environment correlation data sequence; converting the standardized high-voltage-battery-environment correlation data sequence into a three-dimensional accident risk distribution map; further subdividing the core risk scenarios of high-voltage leakage, overvoltage and overcurrent, and thermal runaway based on a fault tree analysis algorithm; constructing a graded model for high-voltage accidents in new energy vehicles based on the differences in the hazard diffusion characteristics of different accident types; extracting hazard weight factors for different accident scenarios based on the graded model for high-voltage accidents in new energy vehicles to construct a dynamic risk assessment matrix; performing collaborative calculations on multi-source monitoring data to obtain a combination of key response trigger indicators under the target accident level; and so on. The risk diffusion rate, energy release intensity, and personnel contact risk value are extracted during the development stage of a high-voltage accident to classify emergency response levels and generate a multi-dimensional accident feature matrix. Real-time accident data is compared with historical accident data of the same vehicle model and scenario to identify abnormal signals such as sudden escalation of accident levels and risk-response mismatch. Dynamic correction factors for response strategies are generated using the slope of the accident development curve. Scenarios are grouped based on new energy vehicle model type, high-voltage system architecture, and accident environment. A support vector machine algorithm is used to screen key influencing factors in the multi-dimensional accident feature matrix and dynamic correction factors for response strategies. A personalized high-voltage accident emergency rescue model is constructed by integrating emergency response time limits, rescue resource allocation constraints, and personnel protection standards. Based on the target response parameters output of the personalized high-voltage accident emergency rescue model, and combined with the temporal relationship between high-voltage accident development and rescue resource scheduling time, real-time graded response instructions and scenario-specific emergency rescue operation guidelines are generated.

[0009] Another aspect of this application discloses a high-voltage accident classification response and emergency rescue operation guidance device for new energy vehicles, comprising: an acquisition module for acquiring real-time voltage and current data, battery status data, collision signal data, and ambient temperature and humidity data of the high-voltage system of the new energy vehicle, removing instantaneous pulse interference and sensor acquisition noise from the high-voltage circuit, and generating a standardized high-voltage-battery-environment correlation data sequence; a processing module for converting the standardized high-voltage-battery-environment correlation data sequence into a three-dimensional accident risk distribution map, and then, based on a fault tree analysis algorithm, subdividing the core risk scenarios of high-voltage leakage, overvoltage and overcurrent, and thermal runaway, and constructing a classification model for high-voltage accidents of new energy vehicles by combining the differences in the hazard diffusion characteristics of different accident types; and extracting hazard weight factors for different accident scenarios based on the classification model to construct a dynamic risk assessment matrix, performing collaborative calculations on multi-source monitoring data, and obtaining key response triggering indicators under the target accident level. The system combines data from various sources to identify key factors. For example, it extracts risk diffusion rate, energy release intensity, and personnel contact risk values ​​based on the development stage of a high-voltage accident, classifies emergency response levels, and generates a multi-dimensional accident feature matrix. It compares real-time accident data with historical accident data of the same vehicle model and scenario to identify abrupt increases in accident level and abnormal signals indicating risk-response mismatch. It then uses the slope of the accident development curve to generate dynamic correction factors for response strategies. Furthermore, it groups scenarios based on new energy vehicle model type, high-voltage system architecture, and accident environment. A support vector machine algorithm is used to screen key influencing factors in the multi-dimensional accident feature matrix and dynamic correction factors for response strategies. Finally, it integrates emergency response time limits, rescue resource allocation constraints, and personnel protection standards to construct a personalized high-voltage accident emergency rescue model. Based on the target response parameters output from the personalized high-voltage accident emergency rescue model, and combined with the temporal relationship between high-voltage accident development and rescue resource scheduling time, it generates real-time graded response instructions and scenario-specific emergency rescue operation guidelines.

[0010] According to another aspect of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a second processor, implements the above-described method for graded response and emergency rescue operation guidance for high-voltage accidents in new energy vehicles.

[0011] This application provides a method and system for graded response and emergency rescue operation guidance for high-voltage accidents in new energy vehicles. The server constructs a full-process graded response and rescue guidance scheme for high-voltage accidents (leakage, overvoltage and overcurrent, thermal runaway) in new energy vehicles. First, multi-source data such as high-voltage voltage and current, and battery status are collected. After noise reduction, standardized correlation sequences are generated and converted into a three-dimensional risk map. Fault tree analysis is used to subdivide the scenarios, and a graded model is built by combining the characteristics of hazard diffusion.

[0012] A dynamic assessment matrix is ​​constructed by extracting hazard weighting factors, and key response indicators are obtained through collaborative calculation of multi-source data. Response levels are classified according to accident stages, generating a feature matrix. Real-time and historical data are compared, and response correction factors are generated using the slope of the accident curve. Key factors are screened using support vector machines by grouping by vehicle type, architecture, and environment, and multiple constraints are integrated to build a personalized rescue model. Finally, real-time graded response instructions (such as the three-level instruction for thermal runaway in tunnels) and scenario-specific operation guidelines (such as the leakage insulation handling steps for passenger vehicles) are output, solving the problems of coarse scenarios and rigid responses in existing technologies, and achieving accurate accident grading and efficient rescue.

[0013] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0014] Figure 1 This document shows a flowchart illustrating a method for graded response and emergency rescue operation guidelines for high-voltage accidents in new energy vehicles, provided in an embodiment of this application.

[0015] Figure 2 This paper presents a schematic diagram of a high-voltage accident graded response and emergency rescue operation guidance device for new energy vehicles provided in an embodiment of this application. Detailed Implementation

[0016] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0017] The following is combined Figure 1 This application describes a method for graded response and emergency rescue operation guidance for high-voltage accidents in new energy vehicles, based on exemplary embodiments thereof. In one embodiment, this application also proposes a method and system for graded response and emergency rescue operation guidance for high-voltage accidents in new energy vehicles. Figure 1 As shown, this method is applied to a server and includes:

[0018] S101 acquires real-time voltage and current data, battery status data, collision signal data, and ambient temperature and humidity data of the high-voltage system of new energy vehicles, eliminates instantaneous pulse interference in the high-voltage circuit and sensor acquisition noise, and generates a standardized high-voltage-battery-environment related data sequence.

[0019] In one implementation, for real-time voltage and current data acquisition of the high-voltage system, voltage and current values ​​of the high-voltage circuit (such as the high-voltage distribution box, motor controller, and battery pack main circuit) are collected in real time using voltage and current sensors built into the high-voltage system. When the new energy vehicle is in motion, the output voltage of the high-voltage distribution box is collected every 100ms, obtaining data such as "DC380V, DC378.5V"; the input current of the motor controller is collected simultaneously, obtaining data such as "80A, 78.2A", reflecting the real-time power transmission status of the high-voltage system.

[0020] For battery status data acquisition, relying on the Battery Management System (BMS), core status parameters of the battery pack are collected, including individual cell voltage, total battery pack voltage, battery temperature, remaining charge (SOC), and state of health (SOH). Example: Collect the voltage of 288 individual cells in a ternary lithium battery pack, obtaining data such as "3.65V, 3.64V, 3.66V"; collect the temperature of the central area of ​​the battery pack, obtaining data such as "25℃, 25.3℃"; simultaneously record the battery pack SOC as "85%" and SOH as "92%", reflecting the battery's energy storage and health level.

[0021] For collision signal data acquisition, the vehicle's built-in collision sensors (such as the front bumper collision sensor and side door collision sensors) and airbag controller collect acceleration signals and collision trigger signals during a collision. Example: When a minor rear-end collision occurs, the front bumper collision sensor detects a longitudinal acceleration of "15g," triggering a collision signal output of "1" (representing a collision). If no collision occurs, the collision signal outputs "0," used to determine whether the accident was induced by a collision.

[0022] For collecting ambient temperature and humidity data, temperature and relative humidity data of the vehicle's environment are collected by temperature and humidity sensors installed on the exterior of the vehicle (such as below the windshield) and near the battery pack. Example: In an open-air parking lot during summer, the ambient temperature sensor collects data of "38℃" and the relative humidity sensor collects data of "65%"; in a cold winter environment, the collected data are "-5℃" and "80%", used to analyze the impact of environmental factors on the stability of the high-voltage system.

[0023] To address potential transient pulse interference from high-voltage circuits and sensor acquisition noise in the collected data, filtering algorithms (such as Kalman filtering and moving average filtering) are employed to ensure data accuracy. High-voltage systems are prone to transient pulse interference during switching of devices (such as IGBTs), causing abnormal spikes in voltage / current data. A sudden spike of "DC500V" (normal voltage range is DC350V-400V) in the collected high-voltage data was identified as transient pulse interference. A Kalman filter algorithm was used to smooth this data segment, correcting the abnormal spike to "DC379V," aligning it with the trend of surrounding normal data. During signal transmission, sensors may generate random noise due to electromagnetic interference, causing minor fluctuations in the data. The temperature data collected by the battery temperature sensor fluctuated frequently around "25℃", showing irregular changes of "25.1℃, 24.9℃, 25.2℃, 24.8℃" (the actual temperature did not fluctuate significantly), which was determined to be acquisition noise. The moving average filtering algorithm was used to take the average of the temperature data collected for 5 consecutive times to obtain "25℃", thus eliminating the influence of random noise.

[0024] After noise removal, multi-source data are correlated and integrated according to a unified format and timestamps to generate a structured "high-voltage-battery-environment correlation data sequence." This ensures that the data can be directly used for subsequent risk mapping construction. An example is shown below: Using "timestamp + high-voltage voltage + high-voltage current + battery SOC + battery temperature + collision signal + ambient temperature + ambient humidity" as the unified format, data for each dimension are aligned and correlated according to timestamps. For example, the correlation data corresponding to the timestamp "2024-05-2014:30:00.100" is "DC379V, 79A, 85%, 25℃, 0, 38℃, 65%"; the correlation data corresponding to the timestamp "2024-05-2014:30:00.200" is "DC378.5V, 78.2A, 85%, 25℃, 0, 38℃, 65%", forming a continuous standardized data sequence and achieving spatiotemporal correlation of multi-source data.

[0025] S102, after converting the standardized high-voltage-battery-environment related data sequence into a three-dimensional accident risk distribution map, subdivides the core risk scenarios of high-voltage leakage, overvoltage and overcurrent, and thermal runaway based on the fault tree analysis algorithm, and constructs a classification model of high-voltage accidents of new energy vehicles by type, combined with the differences in the hazard diffusion characteristics of different accident types.

[0026] In one implementation, a standardized high-voltage-battery-environment correlation data sequence is converted into a three-dimensional accident risk distribution map to generate basic high-voltage accident risk visualization data. Using "high-voltage system component location (X-axis: front to rear of vehicle, Y-axis: left to right side of vehicle, Z-axis: bottom to top of vehicle)" as the three-dimensional coordinates and "risk level (0-10 points, higher score indicates higher risk)" as the color mapping standard, the high-voltage voltage, battery temperature, and ambient humidity data for a continuous 10 minutes in the standardized data sequence are converted. For example, the battery pack area (X=2.5m, Y=0.5m, Z=0.3m) has a risk level of 7 points due to the battery temperature reaching 38℃ and the ambient humidity of 65%, and is displayed in orange in the corresponding map. The high-voltage distribution box area (X=1.2m, Y=0.5m, Z=0.6m) has a stable voltage of 378V, and is marked with a risk level of 2 points, and is displayed in blue in the corresponding map. Finally, a three-dimensional map that can clearly distinguish between high and low risk areas is formed, generating basic high-voltage accident risk visualization data containing the risk value of each coordinate point and color mapping rules.

[0027] A three-dimensional accident risk distribution map was imported into high-voltage system fault simulation software for simulation calculation, generating multi-dimensional high-voltage accident risk field simulation data. High-voltage system fault simulation software (such as PSCAD / EMTDC, MATLAB / Simulink) was selected. After importing the above three-dimensional map, simulation parameters were set (simulation duration 30 minutes, time step 0.1 seconds) to simulate the process of risk increase in the battery pack area. When the battery temperature rose from 38℃ to 45℃, the software calculated that the risk diffusion rate from the battery pack to the surrounding motor controller (X=1.8m, Y=0.5m, Z=0.4m) was 0.02m / s. Simultaneously, simulation data in dimensions such as "risk diffusion rate, risk value of each component at each time point, and risk impact range" were output. For example, at the 15th minute, the risk level in the motor controller area rose to 5 points, generating multi-dimensional high-voltage accident risk field simulation data containing time, space, and risk parameters.

[0028] Fault tree analysis (FBA) algorithms are used to subdivide core risk scenarios such as high-voltage leakage, overvoltage and overcurrent, and thermal runaway in multi-dimensional high-voltage accident risk field simulation data, generating scenario-based risk data. Specifically, the FBA algorithm decomposes risk sources from the dimensions of high-voltage system structure and fault propagation paths, establishing hierarchical analysis models for the triggering conditions and impact range of different core risk scenarios to achieve precise risk scenario classification. For the "high-voltage leakage" risk scenario, risk sources are decomposed from the dimensions of high-voltage system structure (high-voltage wiring harness, connectors, insulation layer) and fault propagation paths (insulation layer damage → exposed wiring harness → current leakage → contact with vehicle body), establishing a hierarchical model. The top-level event is "high-voltage leakage accident," the intermediate events are "insulation layer damage" and "exposed wiring harness," and the bottom-level events are "insulation layer aging (use time exceeding 5 years)" and "vibration-induced damage (vehicle passing over bumpy road surface)." Based on simulation data, the triggering conditions for this scenario are determined to be "insulation layer damage area > 0.5cm² and ambient humidity > 60%," and the affected area is "an area with a radius of 1m centered on the damage point." Similarly, the scenarios of "overvoltage and overcurrent" (triggering conditions: high voltage > 420V or current > 150A, affected area: high-voltage distribution box and connecting wiring harness) and "thermal runaway" (triggering conditions: battery temperature > 60℃ and SOC > 80%, affected area: battery pack and surrounding 2m area) are decomposed, and finally, scenario-based risk data containing the risk sources, triggering conditions, and affected areas of each scenario are generated.

[0029] Characteristic parameters of hazard diffusion rate, radius of influence, and energy release intensity for different accident types, including high-voltage leakage, overvoltage and overcurrent, and thermal runaway, were collected to construct a typological hazard diffusion characteristic library and generate accident characteristic constraint data. Specifically, the typological hazard diffusion characteristic library customized parameters for the current conduction risk of high-voltage leakage, the component burnout risk of overvoltage and overcurrent, and the thermal propagation and gas leakage risk of thermal runaway. Characteristic parameters for the three types of accidents were collected through laboratory simulations and statistical analysis of historical accident data. For "high-voltage leakage," the hazard diffusion rate was collected as "0.1 m / s (dry environment), 0.2 m / s (humid environment)," the influence radius was "1.5 m (voltage 380V), 2 m (voltage 400V)," and the energy release intensity was "50 W (leakage current 1A), 100 W (leakage current 2A)." Focusing on the risk of current conduction, customized "correlation parameters between leakage current and conduction distance under different humidity levels" were developed. For "overvoltage and overcurrent," the hazard diffusion rate was collected as "0.05 m / s (overvoltage 10%), 0.1 m / s (overvoltage 20%)," the influence radius was "0.8 m (overcurrent 10%), 1.2 m (overcurrent 20%)," and the energy release intensity was "2..." For the risk of component burnout, "00W (10% overpressure) and 300W (20% overpressure)" were used to customize "correlation parameters between different overpressure and overcurrent levels and component burnout time". For "thermal runaway", the hazard diffusion rate was collected as "0.3m / s (initial stage) and 0.8m / s (intense stage)", the influence radius was "2m (initial stage) and 5m (intense stage)", and the energy release intensity was "500W (initial stage) and 2000W (intense stage)". For the risk of thermal spread and gas leakage, "correlation parameters between thermal spread rate and toxic gas concentration at different temperatures" were customized. Finally, an accident-type hazard diffusion characteristic library was constructed to generate accident characteristic constraint data containing the above customized parameters.

[0030] By integrating scenario-based risk data and accident characteristic constraint data, a classification model for high-voltage accidents in new energy vehicles is constructed. The scenario-based risk data for the "high-voltage leakage" scenario (triggering conditions: insulation layer damage area > 0.5 cm² and ambient humidity > 60%, impact range: radius 1 m) is integrated with the corresponding accident characteristic constraint data (diffusion rate 0.2 m / s, energy release intensity 100 W) to set risk level classification standards—when the triggering conditions are met, the diffusion range < 0.5 m, and the energy release intensity < 80 W, it is classified as Level I (low risk); when the diffusion range is 0.5-1 m and the energy release intensity is 80-120 W, it is classified as Level II (low risk). Level II (medium risk); when the diffusion range is >1m and the energy release intensity is >120W, it is Level III (high risk); similarly, data are integrated and classification standards are set for the "overvoltage and overcurrent" and "thermal runaway" scenarios respectively. For example, in the "thermal runaway" scenario, the initial stage (diffusion rate 0.3m / s, influence range <2m) is Level II, and the severe stage (diffusion rate >0.5m / s, influence range >3m) is Level III. Finally, a classification model for high-voltage accidents of new energy vehicles can be formed to assess the risk level for the three types of accidents respectively.

[0031] S103, based on the classification model of high-voltage accidents of new energy vehicles, extracts the hazard weight factors of different accident scenarios to construct a dynamic risk assessment matrix, performs collaborative calculation on multi-source monitoring data, and obtains the combination of key response trigger indicators under the target accident level.

[0032] In one implementation, a hazard weighting factor extraction mechanism is constructed. Based on a classification model of high-voltage accidents of new energy vehicles, core parameters such as hazard diffusion rate, energy release intensity, and personnel contact risk value are extracted from different accident scenarios such as high-voltage leakage, overvoltage and overcurrent, and thermal runaway as hazard weighting factors. The weights of each factor are assigned by the analytic hierarchy process to generate a set of hazard factors with weight coefficients. From the "high voltage leakage" scenario, extract the hazard diffusion rate (e.g., 0.2 m / s in a humid environment), energy release intensity (e.g., 100W when the leakage current is 2A), and personnel contact risk value (e.g., 8 points out of 10 when a person comes into contact with the leakage point). From the "overvoltage and overcurrent" scenario, extract the hazard diffusion rate (e.g., 0.1 m / s when the overvoltage is 20%), energy release intensity (e.g., 300W when the overvoltage is 20%), and personnel contact risk value (e.g., 6 points when approaching a burnt-out component). From the "thermal runaway" scenario, extract the hazard diffusion rate (e.g., 0.8 m / s during the severe stage), energy release intensity (e.g., 2000W during the severe stage), and personnel contact risk value (e.g., 10 points when approaching a thermal runaway battery pack). Using the analytic hierarchy process (AHP), five high-voltage system experts were invited to score the weights of each factor. The weights for each factor were determined as follows: in the "high-voltage leakage" scenario, the hazard diffusion rate was 0.3, the energy release intensity was 0.4, and the personnel contact risk value was 0.3; in the "overvoltage and overcurrent" scenario, the weights were 0.2, 0.5, and 0.3, respectively; and in the "thermal runaway" scenario, the weights were 0.4, 0.4, and 0.2, respectively. This resulted in a set of hazard factors with weighted coefficients, such as "High-voltage leakage: hazard diffusion rate (0.3), energy release intensity (0.4), personnel contact risk value (0.3)".

[0033] A dynamic risk assessment matrix construction framework is designed, using the set of hazard factors as the row dimension and the accident development stage as the column dimension. Hazard weight factors in each scenario are mapped to corresponding positions in the matrix according to their stages. The matrix elements are initialized and calibrated using historical accident risk level correlation data to generate a dynamic risk assessment matrix with stage attributes. For the "high-voltage leakage" scenario, the set of hazard factors (hazard diffusion rate, energy release intensity, and personnel contact risk value) is used as the matrix row dimension, and the accident development stages (initial stage: insulation layer just damaged; development stage: leakage range expands; severe stage: leakage affects the human body) are used as the column dimension. Each factor is mapped to its corresponding position according to the stage. For example, the initial stage hazard diffusion rate is 0.1 m / s (weight 0.3), energy release intensity is 50W (weight 0.4), and personnel contact risk value is 3 points (weight 0.3). At the same time, combined with historical high-voltage leakage accident data of the same vehicle model (such as the historical average risk quantification value of 2.8 points in the initial stage), the matrix elements are initialized and calibrated, and finally a dynamic risk assessment matrix with stage attributes is generated. Each cell in the matrix contains "factor value + weight + calibrated basic risk value", such as the initial stage hazard diffusion rate cell is labeled "0.1 m / s (0.3), calibrated basic risk value 0.8 points".

[0034] A multi-source monitoring data collaborative calculation model was established. Real-time voltage and current data of the high-voltage system, battery status data, collision signal data, and environmental temperature and humidity data were imported into a dynamic risk assessment matrix. A weighted summation algorithm was used to perform collaborative calculations on the multi-source data at the same stage within the matrix, outputting real-time risk quantification values ​​for different stages under each accident scenario. For the development stage of the "high-voltage leakage" scenario, real-time monitoring data was imported, specifically: high-voltage system voltage 385V, battery temperature 36℃, collision signal 0 (no collision), and environmental humidity 70%. Combining the weights of each factor in the matrix for this stage (hazard diffusion rate 0.3, energy release intensity 0.4, personnel contact risk value 0.3), a weighted summation algorithm was used to calculate: the real-time hazard diffusion rate is 0.18 m / s, corresponding to a risk score of 0.18 / 0.2×3=2.7 points (0.2 m / s is the maximum diffusion rate for this scenario, corresponding to full risk). The real-time energy release intensity is 90W, corresponding to a risk score of 90 / 100×4=3.6 points (100W is the maximum value, corresponding to a full score of 4 points); the real-time personnel contact risk value is 7 points, corresponding to a risk score of 7 / 10×3=2.1 points (10 points is the maximum value, corresponding to a full score of 3 points). The total real-time risk quantification value is 2.7×0.3+3.6×0.4+2.1×0.3=0.81+1.44+0.63=2.88 points, and the output is "High-voltage leakage - development stage: real-time risk quantification value 2.88 points".

[0035] A key response trigger indicator screening mechanism is proposed. Using the target accident level as a constraint, real-time risk quantification values ​​are compared against thresholds to identify monitoring parameters corresponding to quantification values ​​exceeding preset risk thresholds. The correlation between parameters is analyzed using Pearson correlation coefficients to eliminate redundant parameters and generate a candidate set of key response trigger indicators with correlation labels. The target accident level is set as Level II (medium risk), with a corresponding risk threshold of 2.5-4 points for the "high-voltage leakage" scenario. The calculated real-time risk quantification value of 2.88 points falls within the threshold range, and the corresponding monitoring parameters (ambient humidity 70%, high-voltage system voltage 385V, battery temperature 36℃) are selected. The correlation between parameters was analyzed using Pearson correlation coefficient. The correlation coefficients were calculated as follows: ambient humidity and hazard diffusion rate: 0.82 (strong correlation); high-voltage system voltage and energy release intensity: 0.75 (strong correlation); battery temperature and personnel contact risk value: 0.31 (weak correlation). The weakly correlated parameter "battery temperature" was removed, and a candidate set of key response trigger indicators with correlation labels was generated, such as "ambient humidity (correlation with diffusion rate: 0.82) and high-voltage system voltage (correlation with energy release intensity: 0.75)".

[0036] A model for optimizing indicator combinations is constructed. Based on the monitoring frequency, data accuracy, and response timeliness of indicators within the candidate set, a particle swarm optimization algorithm is used to optimize the combination of indicators. The risk warning accuracy of the indicator combination is used as the objective function to iteratively optimize and generate the target key response trigger indicator combination under the target accident level. In the optimization process of key response trigger indicator combinations for the Level II target accident level of high-voltage accidents involving new energy vehicles, the core performance parameters of the indicators within the candidate set must be used as the basis, combined with the iterative optimization logic of the particle swarm optimization algorithm, to ensure that the final indicator combination meets the target requirement of a risk warning accuracy rate of ≥95%. Specifically, the monitoring frequency, data accuracy, and response timeliness parameters of the environmental humidity and high-voltage system voltage indicators within the candidate set are all determined based on the actual needs of the high-voltage accident monitoring scenario for new energy vehicles and the characteristics of the sensor technology.

[0037] Regarding the environmental humidity index, considering that the impact of environmental humidity changes on the risk of high-voltage leakage accidents is relatively slow (e.g., a humid environment needs to persist for a period of time before significantly increasing the risk of leakage), a temperature and humidity sensor with an accuracy of ±2% (such as the SHT31 sensor) is selected, and a monitoring frequency of 1 time / second is set. This not only captures the trend of humidity changes in a timely manner but also avoids data redundancy. The response time is set to 0.5 seconds because the sensor data needs to go through signal amplification and analog-to-digital conversion processes from acquisition and transmission to system processing. This duration can meet the real-time requirements of humidity data for Level II accidents, ensuring that the data can support risk warning decisions.

[0038] For high-voltage system voltage indicators, the high-voltage system voltage is a core real-time parameter reflecting overvoltage, overcurrent, and high-voltage leakage accidents. Voltage fluctuations can cause changes in risk level within milliseconds (e.g., instantaneous overvoltage can quickly lead to component burnout). Therefore, a high-voltage voltage sensor with an accuracy of ±0.5V (such as the LV25-P type sensor) is selected, and a high-frequency monitoring frequency of 1 time / 100ms (i.e., 10 times / second) is set to accurately capture instantaneous voltage fluctuations. The response time is controlled within 0.1 seconds. By optimizing the sensor signal transmission link (such as using high-speed transmission via CAN bus), it is ensured that voltage data can be fed back to the risk assessment system in real time, providing support for rapid early warning of Level II accidents.

[0039] With a risk warning accuracy rate of ≥95% as the objective function, the particle swarm optimization algorithm is used to optimize the combination of candidate indicators. The algorithm simulates the movement of particles in the solution space (i.e., the selection and adjustment of indicator combinations) to gradually approach the optimal solution. The specific iterative steps and logic are as follows: In the initial stage of the algorithm, a single indicator, "monitoring environmental humidity alone," is selected as the initial solution (initial particle position). The warning effect of this combination is verified by retrieving historical Level II high-voltage accident data (such as 500 sets of Level II high-voltage leakage accident data). In the historical data, when identifying Level II accidents solely through changes in environmental humidity, some omissions (e.g., Level II accidents caused by voltage anomalies but humidity below the threshold were not warned) and misjudgments (e.g., false alarms when humidity meets the standard but there is no leakage risk) occur because key information such as voltage anomalies cannot be captured. The final calculated warning accuracy rate is 88%, which does not meet the target requirement of ≥95%, requiring further iterative optimization.

[0040] Based on the accuracy deviation of the initial combination (88%-95%=-7%), the algorithm adjusts the particle movement direction (i.e., increases the dimension of the indicator combination) and uses "joint monitoring of ambient humidity and high-voltage system voltage" as the new particle position. This combination integrates the advantages of two types of indicators: ambient humidity data can help determine the trend of increasing leakage risk, and high-voltage system voltage data can identify voltage anomalies in real time, forming a complementary monitoring logic.

[0041] For the combination of "environmental humidity + high-voltage system voltage joint monitoring," 500 sets of historical Level II accident data were used again for verification: For Level II high-voltage leakage accidents, when humidity ≥ 60% (triggering leakage risk condition) and voltage shows abnormal fluctuations of ±10V, the system can provide accurate early warnings; for Level II overvoltage and overcurrent accidents, when voltage ≥ 400V (overvoltage threshold) and humidity data helps to eliminate environmental interference factors, false alarms can be reduced. The final calculated early warning accuracy rate of this combination is 96.5%, meeting the objective function requirement of "≥95%". The algorithm then stopped iterating, and this combination was determined to be the optimal key response trigger indicator combination for Level II accidents.

[0042] The final generated combination of "ambient humidity (monitoring frequency 1 time / second, accuracy ±2%) + high-voltage system voltage (monitoring frequency 1 time / 100ms, accuracy ±0.5V)" is highly compatible with the risk characteristics of Level II accidents. From the perspective of the risk level characteristics of Level II accidents, they belong to medium-risk accidents and require early warning in the initial stage of risk diffusion (e.g., before the leakage range expands or the voltage anomaly worsens). This combination of high-frequency voltage monitoring can capture voltage anomalies in the early stages of risk, while humidity monitoring can help determine the risk diffusion trend. The combination of the two can cover the core risk factors of Level II accidents. From the perspective of emergency response requirements, the emergency response corresponding to Level II accidents needs to be initiated quickly after the warning (e.g., cutting off the local high-voltage circuit and notifying maintenance personnel to arrive on site). The high data accuracy of this combination (voltage ±0.5V, humidity ±2%) ensures the reliability of the warning results and avoids over- or delayed responses due to data errors, providing key indicator support for accurate emergency response to Level II accidents.

[0043] S104 extracts the risk diffusion rate, energy release intensity, and personnel contact risk value according to the development stage of high-voltage accidents, classifies the emergency response level of accidents, and generates a multi-dimensional accident feature matrix.

[0044] In one implementation, hazard weighting factors are extracted from the high-voltage leakage, overvoltage and overcurrent, and thermal runaway scenarios in the classification model of high-voltage accidents in categorized new energy vehicles, generating a set of hazard factors with weighted coefficients. From the "high-voltage leakage" scenario, the following are extracted: hazard diffusion rate of 0.2 m / s in a humid environment, energy release intensity of 100 W at a leakage current of 2 A, and personnel contact risk value of 8 points (out of 10) when a person comes into contact with the leakage point. From the "overvoltage and overcurrent" scenario, the following are extracted: hazard diffusion rate of 0.1 m / s at 20% overvoltage, energy release intensity of 300 W at 20% overvoltage, and personnel contact risk value of 6 points when near a burnt-out component. From the "thermal runaway" scenario, the following are extracted: hazard diffusion rate of 0.8 m / s during the severe stage, energy release intensity of 2000 W during the severe stage, and personnel contact risk value of 10 points when near a thermally runaway battery pack. Five high-voltage system experts were invited to score the data using the analytic hierarchy process (AHP). The weights of the three factors in the "high-voltage leakage" scenario were determined to be 0.3, 0.4, and 0.3; in the "overvoltage and overcurrent" scenario, they were 0.2, 0.5, and 0.3; and in the "thermal runaway" scenario, they were 0.4, 0.4, and 0.2. Finally, a set of hazard factors with weighted coefficients was generated, such as "overvoltage and overcurrent: hazard diffusion rate (0.2), energy release intensity (0.5), personnel exposure risk value (0.3)".

[0045] A dynamic matching process is performed between the hazard factor set and the accident development stage. A matrix is ​​constructed by mapping hazard factors by row dimension and accident stages by column dimension. Elements are calibrated using historical data to generate a dynamic risk assessment matrix with stage attributes. The hazard factor set (hazard diffusion rate, energy release intensity, and personnel exposure risk value) for the "thermal runaway" scenario is used as the row dimension, and the accident development stages (initial stage: abnormal battery temperature rise; development stage: smoke appearance; severe stage: flame eruption) are used as the column dimension. The initial stage is mapped to a hazard diffusion rate of 0.3 m / s (weight 0.4), an energy release intensity of 500 W (weight 0.4), and a personnel exposure risk value of 4 points (weight 0.2); the development stage is mapped to a hazard diffusion rate of 0.5 m / s (weight 0.4), an energy release intensity of 1200 W (weight 0.4), and a personnel exposure risk value of 7 points (weight 0.2); and the severe stage is mapped to a hazard diffusion rate of 0.8 m / s (weight 0.4), an energy release intensity of 2000 W (weight 0.4), and a personnel exposure risk value of 10 points (weight 0.2). By combining 100 sets of historical thermal runaway accident data (such as an average risk quantification value of 3.2 points in the initial stage), the matrix elements are initialized and calibrated, and finally a dynamic risk assessment matrix with stage attributes is generated. The matrix cells are labeled with "factor value + weight + calibration base risk value". For example, the cell for the hazard diffusion rate in the initial stage of thermal runaway is labeled with "0.3m / s (0.4), calibration base risk value 1.2 points".

[0046] Multi-source monitoring data and dynamic risk assessment matrix are processed collaboratively. A weighted summation algorithm is used to calculate the data at the same stage to generate real-time risk quantification values ​​for different stages of each scenario. The real-time risk quantification values ​​are compared and filtered with the target accident level threshold to remove redundant parameters and label the correlation of indicators, generating a candidate set of key response trigger indicators with correlation labels. For the development stage of the "high-voltage leakage" scenario, real-time monitoring data (high-voltage system voltage 382V, battery temperature 35℃, collision signal 0, ambient humidity 68%) was imported. Combined with the weights of this stage in the matrix (hazard diffusion rate 0.3, energy release intensity 0.4, personnel contact risk value 0.3), a weighted summation algorithm was used to calculate: the real-time hazard diffusion rate is 0.17m / s, corresponding to a risk score of 0.17 / 0.2×3=2.55 points; the real-time energy release intensity is 85W, corresponding to a risk score of 85 / 100×4=3.4 points; the real-time personnel contact risk value is 6.5 points, corresponding to a risk score of 6.5 / 10×3=1.95 points. The total real-time risk quantification value = 2.55×0.3+3.4×0.4+1.95×0.3=0.765+1.36+0.585=2.71 points. The target accident level II (medium risk) threshold was set at 2.5-4 points, and corresponding monitoring parameters were selected (ambient humidity 68%, high-voltage system voltage 382V, battery temperature 35℃). Through Pearson correlation coefficient analysis, the correlation between ambient humidity and hazard diffusion rate was 0.81, the correlation between high-voltage system voltage and energy release intensity was 0.76, and the correlation between battery temperature and personnel contact risk value was 0.29. Redundant "battery temperature" was removed, and a candidate set with correlation labels was generated, such as "ambient humidity (correlation 0.81), high-voltage system voltage (correlation 0.76)".

[0047] Based on the monitoring frequency, data accuracy, and response timeliness of the candidate indicator set, a combination optimization process is performed. A particle swarm optimization algorithm is used to optimize the early warning accuracy, generating the optimal combination of key response trigger indicators for the target accident level. For Level II "high-voltage leakage" accidents, the candidate indicator set includes "ambient humidity" with a monitoring frequency of 1 time / second, accuracy ±2%, and response timeliness of 0.5 seconds; and "high-voltage system voltage" with a monitoring frequency of 1 time / 100ms, accuracy ±0.5V, and response timeliness of 0.1 seconds. With "early warning accuracy ≥ 95%" as the objective function, the algorithm initially selected "ambient humidity monitoring alone" as the initial solution and verified it by calling 500 sets of historical Class II high-voltage leakage data. Due to the inability to capture voltage anomalies, the early warning accuracy was only 87%. After adjusting to "ambient humidity + high-voltage system voltage joint monitoring" and verifying it again, accurate early warning was given when humidity ≥ 60% and voltage fluctuation ± 10V, the false alarm rate was reduced to 3.2%, and the early warning accuracy reached 96.8%, meeting the target requirements. Finally, the optimal combination was generated: "ambient humidity (monitoring frequency 1 time / second, accuracy ± 2%) + high-voltage system voltage (monitoring frequency 1 time / 100ms, accuracy ± 0.5V)".

[0048] S105 compares real-time accident data with historical accident data of the same vehicle type and scenario to identify abnormal signals such as a sudden increase in accident level and an imbalance in risk-response matching, and uses the slope of the accident development curve to generate a dynamic correction factor for the response strategy.

[0049] In one implementation, real-time accident data of high-voltage accidents involving new energy vehicles is aligned with historical accident data of the same model and scenario. Corresponding historical data samples are matched according to accident type and development stage to generate a basic data comparison set. If a "high-voltage leakage accident (accident type) of a small pure electric passenger vehicle (model: a certain brand Model A)" occurs and is in the "development stage (leakage range expands to 0.8m)," then 200 sets of high-voltage leakage accident data for this model in the past 3 years are selected from the historical database, and 80 sets of data also in the "development stage" are extracted as historical samples. The collection dimensions of the real-time accident data (such as leakage current per 100ms, environmental humidity, and risk level) are unified with the dimensions of the historical sample data to ensure complete matching of time intervals and parameter types. Finally, a basic data comparison set containing "real-time high-voltage leakage development stage data + 80 sets of historical data of the same model, scenario, and stage" is generated.

[0050] The data comparison dataset is used to analyze the differences between the real-time accident level change trend and historical data from the same period. This process identifies sudden increases in accident levels exceeding the normal range within a short period, generating a sudden increase in accident level identification results. In the aforementioned high-voltage leakage accident data comparison dataset, the historical trend for accident level changes (development stage) is "0.2 points increase every 5 seconds (maximum risk level 10 points)," with a normal increase of no more than 2.4 points per minute. Real-time accident data shows that the accident level rose from 3.5 points (initial development stage) to 6.8 points within 30 seconds, an increase of 3.3 points, exceeding the normal increase by 0.9 points. Furthermore, the rate of increase in level for three consecutive monitoring cycles (100ms each) was more than twice the historical average, thus identifying a "sudden increase in accident level signal." A sudden increase in accident level identification result is generated, labeled "Accident type: high-voltage leakage, development stage, sudden increase period: 2 minutes 10 seconds - 2 minutes 40 seconds, level increase: 3.3 points."

[0051] The matching degree between real-time risk parameters and corresponding response measures is verified. By comparing with the historical best risk-response matching model, imbalances in response measures that are lagging or excessive are identified, generating risk-response matching imbalance identification results. For the above-mentioned high-voltage leakage accident, the real-time risk parameters are "leakage current 1.8A, risk level 6.8 (Level II medium risk)", and the currently initiated response measure is "only disconnecting the local power supply of the high-voltage circuit". The historical best risk-response matching model is retrieved, and the model shows that the optimal response measure corresponding to "leakage current 1.5-2.0A, risk level 6-7 (Level II medium risk)" is "disconnecting the local high-voltage power supply + starting the insulation monitoring instrument for real-time monitoring + notifying maintenance personnel to arrive on site within 5 minutes". The comparison revealed that the current response measures lacked the "insulation monitoring" and "personnel dispatch" steps, which constitutes an imbalance of "delayed response measures". If the real-time risk parameters are "leakage current 0.5A, risk level 2.2 points (Level I low risk)", but the measures of "cutting off the high-voltage power supply of the whole vehicle + firefighters arriving on site" are initiated, then it is considered "excessive response measures". The final result of risk-response matching imbalance identification is generated and marked "accident type: high-voltage leakage, imbalance type: delayed response, missing measures: insulation monitoring, personnel dispatch".

[0052] The results of identifying sudden escalation of accident levels and risk-response mismatch are integrated and processed to extract the accident development time series data corresponding to the two types of abnormal signals. The slope of the accident development curve is calculated, and a slope quantification value is generated. The "sudden escalation signal" and "response lag signal" of the above-mentioned high-voltage leakage accidents are integrated to extract the accident development time series data within 1 minute corresponding to the two types of signals—time nodes 0 seconds (level 3.5 minutes), 10 seconds (4.2 minutes), 20 seconds (5.1 minutes), 30 seconds (6.8 minutes), 40 seconds (7.2 minutes), 50 seconds (7.5 minutes), and 60 seconds (7.8 minutes). Using "time (seconds)" as the X-axis and "accident level" as the Y-axis, linear fitting is performed on the data to obtain the accident development curve equation "Y=0.07X+3.48". The slope of the curve is calculated to be 0.07 (i.e., the level increases by 0.7 points every 10 seconds), generating a slope quantification value of "0.07 points / second".

[0053] Based on the mapping calculation of the slope quantization value and the preset correction coefficient library, the weight of the correction factor is determined according to the slope magnitude, and the dynamic correction factor of the response strategy is generated. The preset correction coefficient library is set as follows: "Slope < 0.03 min / second (low growth rate), correction coefficient 0.8; 0.03 ≤ slope ≤ 0.07 min / second (medium growth rate), correction coefficient 1.2; slope > 0.07 min / second (high growth rate), correction coefficient 1.5", and the correction factor weight = slope quantification value × correction coefficient. The slope quantification value of the above high-voltage leakage accident is 0.07 min / second, corresponding to a correction coefficient of 1.2, and the calculated correction factor weight = 0.07 × 1.2 = 0.084. If the slope is 0.08 min / second (high growth rate), then the correction coefficient is 1.5, and the weight = 0.08 × 1.5 = 0.12. Finally, a dynamic correction factor for the response strategy is generated, labeled "Accident type: high-voltage leakage, correction factor weight: 0.084, purpose: to improve the timeliness of subsequent emergency response measures, such as shortening the personnel dispatch time from 5 minutes to 3 minutes".

[0054] S106 groups scenarios based on new energy vehicle type, high-voltage system architecture, and accident environment. It uses support vector machine algorithm to screen key influencing factors for multi-dimensional accident feature matrix and dynamic correction factors of response strategy. It integrates emergency response time limit, rescue resource allocation constraints, and personnel protection standard information to construct a personalized high-voltage accident emergency rescue model.

[0055] In one implementation, high-voltage accident scenarios are grouped based on the classification dimensions of new energy vehicle model type, high-voltage system architecture, and accident environment, generating scenario grouping results. Grouping is performed according to a three-dimensional combination of "vehicle type (small pure electric passenger vehicle / medium-sized pure electric commercial vehicle), high-voltage system architecture (centralized high-voltage architecture / distributed high-voltage architecture), and accident environment (urban roads / highways / tunnels)". For example, "small pure electric passenger vehicle (model) + centralized high-voltage architecture (architecture) + urban road (environment)" constitutes one scenario, where the high-voltage components are concentrated at the front of the vehicle, and urban road rescue resources are easily accessible; "medium-sized pure electric commercial vehicle (model) + distributed high-voltage architecture (architecture) + tunnel (environment)" constitutes another scenario, where the high-voltage components are dispersed on both sides of the vehicle, and ventilation is poor and rescue space is limited in the tunnel. Finally, all high-voltage accident scenarios are divided into 8 groups, each containing 20-30 historical accident samples, generating scenario grouping results labeled with "scenario number, vehicle type, architecture, and environment".

[0056] A support vector machine (SVM) algorithm was employed to screen key influencing factors in the multidimensional accident feature matrix and dynamic correction factors for response strategies. Through classification and regression analysis of the feature variables, low-correlation factors were eliminated, retaining those that played a core role in rescue decisions, thus generating a set of key influencing factors. Taking a scenario of "small pure electric passenger vehicles + centralized architecture + urban roads" as an example, the multidimensional accident feature matrix includes five types of features: "risk diffusion rate, energy release intensity, personnel contact risk value, battery temperature, and ambient humidity." These features, combined with the dynamic correction factors for the response strategy in this scenario (e.g., a correction factor weight of 0.084), were input into the SVM algorithm for training. Through classification and regression analysis, the correlation between each feature and rescue decisions (e.g., the number of rescue personnel and equipment type) was calculated. The factors "risk diffusion rate (correlation 0.92), energy release intensity (correlation 0.88), and personnel contact risk value (correlation 0.85)" are strongly correlated with rescue decisions, while "battery temperature (correlation 0.35) and ambient humidity (correlation 0.28)" are low-correlation factors and are removed. The final set of key influencing factors for this scenario is: "risk diffusion rate, energy release intensity, personnel contact risk value, and dynamic correction factor for response strategy".

[0057] Data fusion processing is performed on the set of key influencing factors, emergency response time limits, rescue resource allocation constraints, and personnel protection standards to establish the correlation mapping relationship between the various elements and generate a fused feature dataset. For the key influencing factor set (risk diffusion rate, energy release intensity, and dynamic correction factor for response strategy) in the "medium-sized pure electric commercial vehicle + distributed architecture + tunnel" scenario, three types of constraint information are incorporated: emergency response time limit (the tunnel scenario requires arrival at the scene within 10 minutes; exceeding this time limit will result in a risk diffusion range exceeding 5 meters), rescue resource allocation constraints (only two vehicles equipped with insulated rescue equipment are available for dispatch in this area), and personnel protection standards (for thermal runaway accidents in tunnels, level-two protective equipment against high temperatures and toxic gases is required). Establish a correlation mapping relationship: such as "risk diffusion rate > 0.5m / s → emergency response time limit compressed to 8 minutes → dispatch 2 insulated rescue vehicles → personnel put on level 2 protective equipment", integrate the mapping relationship with key factor data to generate a fusion feature dataset containing "factor values, constraints, and mapping rules", such as "risk diffusion rate 0.6m / s, emergency response time limit 8 minutes, resources 2 insulated vehicles, level 2 protection, correction factor 0.12".

[0058] A personalized emergency rescue model for high-pressure accidents is constructed based on a fusion feature dataset. The model uses scenario grouping results as input dimensions, key influencing factors as core parameters, and multiple constraint information as boundary conditions to achieve personalized adaptation of rescue strategies for different scenarios. Taking the scenario grouping results of "small pure electric passenger vehicles + centralized architecture + urban roads" as the model input dimension, the key influencing factors of this scenario (risk diffusion rate, energy release intensity, etc.) are used as core parameters, and emergency response time (15 minutes for urban roads), resource constraints (3 rescue vehicles), and protection standards (Level 1 protection) are used as boundary conditions to construct the model. When a high-voltage leakage accident occurs in this scenario, the model takes real-time data as input (risk diffusion rate 0.2m / s, energy release intensity 90W, correction factor 0.084) and calculates personalized rescue strategies by integrating the mapping rules in the feature dataset: "Dispatch one insulated rescue vehicle, rescue personnel wearing level one protective gear, arrive at the scene within 12 minutes, and prioritize cutting off the power supply to the high-voltage distribution box at the front of the vehicle." If the input is a "tunnel scenario," the model outputs an adaptive strategy of "dispatch two rescue vehicles, arrive within 8 minutes, and provide level two protection" based on the boundary conditions of that scenario, thus realizing personalized adjustments to rescue strategies for different scenarios.

[0059] S107, based on the target response parameter output of the personalized high-voltage accident emergency rescue model, combined with the high-voltage accident development time sequence pattern and the time correlation information of rescue resource scheduling, generates real-time hierarchical response instructions and scenario-specific emergency rescue operation guidelines.

[0060] In one implementation, for a Level II high-voltage leakage accident in the scenario of "small pure electric passenger vehicle + centralized high-voltage architecture + urban road", real-time data (risk diffusion rate 0.2m / s, energy release intensity 90W, response strategy dynamic correction factor 0.084) is input into a personalized model. The model combines the emergency response time limit (15 minutes), rescue resource constraints (3 insulated rescue vehicles) and personnel protection standards (Level 1 protection) of the scenario, and outputs target response parameters: "Rescue resource dispatch: 1 insulated rescue vehicle; personnel configuration: 2 maintenance personnel wearing Level 1 protective equipment; handling priority: prioritize cutting off the power supply of the high-voltage distribution box at the front of the vehicle; response time limit: arrive at the scene within 12 minutes; risk control target: control the leakage range within 0.5m within 1 hour".

[0061] By combining the development timeline patterns of different types of high-voltage accidents (such as risk spread rate and energy release rhythm) with the time information of rescue resource dispatch (such as resource location and dispatch time), a correlation analysis is conducted to ensure that response instructions and rescue operations match the dynamic development rhythm of the accident. For a Level III thermal runaway accident in the scenario of "medium-sized pure electric commercial vehicle + distributed high-voltage architecture + tunnel", the known development timeline pattern of thermal runaway accidents is "initial stage (0-10 minutes): smoke generation, heat spread rate 0.3 m / s; severe stage (10-20 minutes): flame eruption, heat spread rate 0.8 m / s"; the rescue resource dispatch time information is "the nearest rescue vehicle with fire extinguishing equipment is located 3 kilometers from the tunnel entrance, dispatching to the accident site takes 8 minutes, and rescue personnel need 2 minutes to put on level 2 protective equipment". After correlation analysis, it is determined that: rescue personnel must arrive at the scene and initial fire extinguishing operations must be completed within 10 minutes after the accident occurs (i.e., before the severe stage of thermal runaway). Therefore, the priority of rescue vehicle dispatch must be raised to the highest level, and the dispatch time must be reduced to within 7 minutes to ensure matching with the accident development timeline.

[0062] Based on the target response parameters output by the personalized model, and combined with the correlation analysis results of accident development time and resource scheduling time, response levels are divided according to accident severity, generating clear and executable real-time hierarchical response instructions. For the aforementioned Level III thermal runaway accident, three levels of response instructions are generated: 1. Level 1 Instruction (0-2 minutes after accident): Send an instruction to the tunnel management department to "close the tunnel entrance of the accident section and guide surrounding vehicles to detour," and simultaneously send an instruction to the rescue center to "dispatch one rescue vehicle equipped with fire extinguishing equipment and one medical support vehicle, prioritizing their deployment to the accident site via the tunnel's emergency access route"; 2. Level 2 Instruction (2-8 minutes after accident): Send an instruction to rescue personnel to "wear Level 2 protective equipment, carry insulated pliers and high-temperature fire blankets, and arrive at the scene in approximately 8 minutes. Upon arrival, first check the concentration of toxic gases in the tunnel"; 3. Level 3 Instruction (8-10 minutes after accident): Send an instruction to on-site rescue personnel to "immediately upon arrival, cover the battery pack with a high-temperature fire blanket, and simultaneously cut off the vehicle's distributed high-voltage circuit power supply to control the spread of heat," ensuring that each level of instruction accurately matches the accident development stage and resource scheduling progress.

[0063] Based on the characteristics of different accident scenarios (such as vehicle structure, environmental limitations, and accident type), and combined with the target response parameters of personalized models, detailed emergency rescue operation guidelines for each scenario are generated, down to the operational steps, to ensure that rescue personnel can operate according to the guidelines. Specifically, 1. "Small pure electric passenger vehicle + centralized architecture + urban road" high-voltage leakage scenario guidelines: ① After arriving at the scene, rescue personnel should first use an insulation detector to detect the leakage area on the vehicle body and confirm that the leakage point is located near the high-voltage distribution box at the front of the vehicle; ② Wearing insulated gloves, use insulated pliers to disconnect the main power switch of the high-voltage distribution box (location: inside the fuse box on the left side of the front of the vehicle); ③ Start the insulation monitor to monitor the leakage current change in real time and ensure that the current drops below 0.1A; ④ Wrap the damaged part of the leakage wire harness with insulating tape, mark it with a "handled" label, and prohibit unauthorized personnel from touching it; 2. "Medium-sized pure electric commercial vehicle + distributed architecture + tunnel" thermal runaway scenario guidelines: ① Upon arrival at the scene After arriving at the scene, first use a toxic gas detector to check the CO concentration in the tunnel. If the concentration is >500ppm, the tunnel ventilation system must be turned on for 3 minutes before entering; ② Two rescuers work together, one using a high-temperature fire blanket to cover the battery pack (battery pack location: middle of both sides of the vehicle), and the other operating the distributed high-voltage circuit cut-off device (a total of 3 cut-off points, located at the front, middle, and rear of the vehicle); ③ After extinguishing the fire, use an infrared thermometer to monitor the battery pack temperature. If the temperature is >80℃, continue spraying cooling agent until the temperature drops below 40℃; ④ Set up a "Accident Rescue, No Entry" warning sign at the tunnel entrance to guide rescue vehicles to park in an orderly manner and avoid blocking the emergency passage.

[0064] In one implementation, such as Figure 2 As shown, this application also provides a graded response and emergency rescue operation guidance device for high-voltage accidents in new energy vehicles, including:

[0065] The acquisition module 201 is used to acquire real-time voltage and current data, battery status data, collision signal data and ambient temperature and humidity data of the high-voltage system of new energy vehicles, remove instantaneous pulse interference in the high-voltage circuit and sensor acquisition noise, and generate a standardized high-voltage-battery-environment related data sequence.

[0066] Processing module 202 is used to convert standardized high-voltage-battery-environment related data sequences into a three-dimensional accident risk distribution map. Based on fault tree analysis algorithms, it further subdivides core risk scenarios such as high-voltage leakage, overvoltage and overcurrent, and thermal runaway. Combining the differences in hazard diffusion characteristics among different accident types, it constructs a classification model for high-voltage accidents in new energy vehicles. Based on this classification model, it extracts hazard weight factors for different accident scenarios to construct a dynamic risk assessment matrix. It then performs collaborative calculations on multi-source monitoring data to obtain key response trigger indicator combinations for the target accident level. Finally, it extracts risk diffusion rate, energy release intensity, and personnel contact risk values ​​according to the high-voltage accident development stage, classifies accident emergency response levels, and generates a multi-dimensional accident feature matrix. This is compared with real-time data. Accident data is compared with historical accident data of the same vehicle model and scenario to identify abnormal signals such as sudden escalation of accident level and risk-response mismatch. The slope of the accident development curve is used to generate dynamic correction factors for response strategies. The scenarios are grouped according to the type of new energy vehicle, high-voltage system architecture and accident environment. The support vector machine algorithm is used to screen key influencing factors of the multi-dimensional accident feature matrix and dynamic correction factors of response strategies. Personalized high-voltage accident emergency rescue models are constructed by integrating emergency response time limits, rescue resource allocation constraints and personnel protection standards. Based on the target response parameters output of the personalized high-voltage accident emergency rescue model, combined with the correlation information between the high-voltage accident development time sequence and rescue resource scheduling time, real-time graded response instructions and scenario-specific emergency rescue operation guidelines are generated.

[0067] All embodiments in this application are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for the method, electronic device, electronic device, and readable storage medium for evaluating the graded response and emergency rescue operation guidelines for high-voltage accidents in new energy vehicles are basically similar to the above-described embodiments for the method of graded response and emergency rescue operation guidelines for high-voltage accidents in new energy vehicles, and are therefore described simply. Relevant parts can be referred to in the descriptions of the above-described embodiments for the method of graded response and emergency rescue operation guidelines for high-voltage accidents in new energy vehicles.

Claims

1. A new energy vehicle high-voltage accident grading response and emergency rescue operation guidance method, characterized in that, The application comprises the following steps: Collecting real-time voltage and current data of high-voltage systems of new energy vehicles, battery state data, collision signal data and environmental temperature and humidity data, eliminating transient pulse interference of high-voltage circuits and sensor collection noise, and generating standardized high-voltage-battery-environmental correlation data sequence; After converting the standardized high-voltage-battery-environmental correlation data sequence into a three-dimensional accident risk distribution map, the core risk scenarios of high-voltage leakage, overvoltage and overcurrent, and thermal runaway are subdivided based on the fault tree analysis algorithm, and a type-specific new energy vehicle high-voltage accident grading model is constructed by combining the hazard diffusion characteristics of different accident types, including converting the standardized high-voltage-battery-environmental correlation data sequence into a three-dimensional accident risk distribution map, and generating basic high-voltage accident risk visualization data; The three-dimensional accident risk distribution map is imported into a high-voltage system fault simulation software for simulation calculation, and multi-dimensional high-voltage accident risk field simulation data are generated; The high-voltage leakage, overvoltage and overcurrent, and thermal runaway core risk scenarios in the multi-dimensional high-voltage accident risk field simulation data are subdivided by using the fault tree analysis algorithm, and scenario-based risk data are generated, wherein the fault tree analysis algorithm decomposes the risk source from the high-voltage system structure and fault propagation path dimensions, establishes a hierarchical analysis model for the trigger conditions and influence range of different core risk scenarios, and realizes accurate division of the risk scenarios; the characteristic parameters of the hazard diffusion rate, influence radius and energy release intensity of different accident types of high-voltage leakage, overvoltage and overcurrent, and thermal runaway are collected to construct an accident typed hazard diffusion characteristic library, and accident characteristic constraint data are generated, wherein the accident typed hazard diffusion characteristic library customizes parameters for the current conduction risk of high-voltage leakage, the component burning risk of overvoltage and overcurrent, and the thermal spread and gas leakage risk of thermal runaway; the scenario-based risk data and the accident characteristic constraint data are fused to construct a type-specific new energy vehicle high-voltage accident grading model; Based on the type-specific new energy vehicle high-voltage accident grading model, the hazard weight factor of different accident scenarios is extracted to construct a dynamic risk assessment matrix, and the multi-source monitoring data are calculated cooperatively to obtain the key response trigger index combination under the target accident grade; According to the development stage of the high-voltage accident, the risk diffusion rate, energy release intensity and personnel exposure risk value are extracted to divide the accident emergency response level, and a multi-dimensional accident feature matrix is generated. The real-time accident data is compared with the historical accident data of the same vehicle type and the same scene to identify abnormal signals of accident grade sudden increase and risk-response mismatch, and a dynamic correction factor of response strategy is generated by using the slope of the accident development curve, including real-time accident data and historical accident data of the same vehicle type and the same scene of new energy vehicle high-voltage accident, data alignment processing is performed, corresponding historical data samples are matched according to the accident type and development stage, and a data comparison basic set is generated; the real-time accident grade change trend in the data comparison basic set is compared with the historical data of the same period, and difference analysis processing is performed, the sudden increase signal of the accident grade exceeding the normal increase amplitude in a short time is identified, and an accident grade sudden increase identification result is generated; the matching degree of real-time risk parameters and corresponding response measures is verified, the historical optimal risk-response matching model is compared, and the imbalance situation of lagging or excessive response measures is screened out, and a risk-response mismatch identification result is generated; the accident grade sudden increase identification result and the risk-response mismatch identification result are integrated, the accident development time sequence data corresponding to the two types of abnormal signals is extracted, the slope of the accident development curve is calculated, and a slope quantitative value is generated; based on the slope quantitative value and the preset correction coefficient library, the correction factor weight is determined according to the slope size, and a dynamic correction factor of response strategy is generated; The scene is grouped in combination with the new energy vehicle type category, high-voltage system architecture and accident occurrence environment, the support vector machine algorithm is used to screen the key influence factors of the multi-dimensional accident feature matrix and the dynamic correction factor of the response strategy, and the individualized high-voltage accident emergency rescue model is constructed by fusing the emergency response time limit, rescue resource configuration constraint and personnel protection standard information; Based on the target response parameter output of the individualized high-voltage accident emergency rescue model, in combination with the high-voltage accident development time sequence rule and rescue resource scheduling time correlation information, real-time hierarchical response instructions and scene-specific emergency rescue operation guidelines are generated.

2. The method of claim 1, wherein, Based on the hierarchical model of new energy vehicle high-voltage accidents, the hazard weight factor of different accident scenes is extracted to construct a dynamic risk assessment matrix, and multi-source monitoring data is calculated to obtain a key response trigger index combination under the target accident grade, including: A hazard weight factor extraction mechanism is constructed, based on the hierarchical model of new energy vehicle high-voltage accidents, the hazard diffusion rate, energy release intensity and personnel exposure risk value are extracted as core parameters from different accident scenes of high-voltage leakage, overvoltage and overcurrent, and thermal runaway, and the weight of each factor is assigned by using the analytic hierarchy process, and a hazard factor set with weight coefficients is generated; A dynamic risk assessment matrix construction framework is designed, the hazard factor set is used as the row dimension of the matrix, the accident development stage is used as the column dimension, the hazard weight factors in each scene are mapped to the corresponding positions of the matrix according to the stage, and the matrix elements are initialized and calibrated in combination with the historical accident risk grade correlation data, and a dynamic risk assessment matrix with stage attributes is generated; A multi-source monitoring data collaborative computing model is established, real-time voltage and current data of high-voltage systems, battery state data, collision signal data and environmental temperature and humidity data are imported into a dynamic risk assessment matrix, and a weighted summation algorithm is used to collaboratively operate the multi-source data in the matrix at the same stage, and output the real-time risk quantification value at different stages under each accident scenario; A key response trigger index screening mechanism is proposed, and the real-time risk quantification value is compared with the threshold value under the constraint of the target accident level, and the monitoring parameters corresponding to the quantification value exceeding the preset risk threshold are screened out, the relevance between the parameters is analyzed through the Pearson correlation coefficient, redundant parameters are eliminated, and a candidate set of key response trigger indexes with correlation degree labels is generated; An index combination optimization model is constructed, the monitoring frequency, data accuracy and response timeliness of the indexes in the candidate set are used to optimize the combination of the indexes through a particle swarm optimization algorithm, the risk warning accuracy of the index combination is taken as the objective function, and the target key response trigger index combination under the target accident level is iteratively optimized and generated.

3. The method of claim 1, wherein, The risk diffusion rate, energy release intensity and personnel exposure risk value are extracted according to the development stage of high-voltage accidents, the accident emergency response level is divided, and a multi-dimensional accident feature matrix is generated, including: The hazard weight factor extraction and processing of high-voltage leakage, overvoltage and overcurrent, and thermal runaway scenarios in the high-voltage accident grading model of the new energy vehicle is performed, and a hazard factor set with weight coefficients is generated; The hazard factor set and the accident development stage are dynamically matched, a matrix is constructed by mapping the hazard factor in the row dimension and the accident stage in the column dimension, and the elements are calibrated combined with historical data to generate a dynamic risk assessment matrix with stage attributes; The multi-source monitoring data and the dynamic risk assessment matrix are collaboratively calculated, the weighted summation algorithm is used to operate the data at the same stage, and the real-time risk quantification value at different stages of each scenario is generated; the real-time risk quantification value is compared and screened with the target accident level threshold, redundant parameters are eliminated and the index correlation degree is labeled, and a candidate set of key response trigger indexes with correlation degree labels is generated; Based on the monitoring frequency, data accuracy and response timeliness of the index candidate set, combination optimization processing is performed, the particle swarm optimization algorithm is used to optimize the target function with the warning accuracy as the target, and the optimal key response trigger index combination under the target accident level is generated.

4. The method of claim 1, wherein, The scenarios are grouped according to the new energy vehicle type, high-voltage system architecture and accident environment, the support vector machine algorithm is used to screen the key influence factors of the multi-dimensional accident feature matrix and the response strategy dynamic correction factor, and the individualized high-voltage accident emergency rescue model is constructed by integrating the emergency response time limit, rescue resource allocation constraints and personnel protection standard information, including: Based on the classification dimensions of new energy vehicle type, high-voltage system architecture and accident environment, the high-voltage accident scenarios are grouped, and the scenario grouping results are generated; The support vector machine algorithm is used to screen the key influence factors of the multi-dimensional accident feature matrix and the response strategy dynamic correction factor, the classification and regression analysis of the characteristic variables are performed through the algorithm, the low correlation factors are eliminated, and the factors that play a core role in rescue decision-making are retained, and a key influence factor set is generated; The key influence factor set is fused with emergency response time limit, rescue resource allocation constraint and personnel protection standard information, and a correlation mapping relationship between elements is established to generate a fused feature data set; Based on the fused feature data set, a personalized high-voltage accident emergency rescue model is constructed, taking scene grouping results as model input dimensions, key influence factors as core parameters and multi-constraint information as boundary conditions, to realize personalized adaptation of the model to rescue strategies in different scenes.

5. A new energy vehicle high-voltage accident grading response and emergency rescue operation instruction device, characterized in that, The device for implementing the method of claim 1 comprises: An acquisition module is configured to acquire real-time voltage and current data of a high-voltage system of a new energy vehicle, battery state data, collision signal data and environmental temperature and humidity data, eliminate high-voltage loop transient pulse interference and sensor collection noise, and generate a standardized high-voltage-battery-environment correlation data sequence; A processing module is configured to convert the standardized high-voltage-battery-environment correlation data sequence into a three-dimensional accident risk distribution map, subdivide core risk scenarios of high-voltage leakage, overvoltage and overcurrent and thermal runaway based on a fault tree analysis algorithm, construct a type-specific new energy vehicle high-voltage accident grading model in combination with differences in hazard diffusion characteristics of different accident types, extract hazard weight factors of different accident scenarios based on the type-specific new energy vehicle high-voltage accident grading model to construct a dynamic risk assessment matrix, perform collaborative calculation on multi-source monitoring data, acquire a key response trigger index combination under a target accident grade, extract a risk diffusion rate, energy release intensity and personnel exposure risk value according to a high-voltage accident development stage, divide accident emergency response grades, and generate a multi-dimensional accident feature matrix; compare real-time accident data with historical accident data of the same vehicle model and scene to identify abnormal signals such as accident grade surge and risk-response mismatch, generate a response strategy dynamic correction factor using an accident development curve slope; group scenes in combination with new energy vehicle model categories, high-voltage system architectures and accident occurrence environments, use a support vector machine algorithm to screen key influence factors from the multi-dimensional accident feature matrix and the response strategy dynamic correction factor, fuse emergency response time limit, rescue resource allocation constraint and personnel protection standard information to construct a personalized high-voltage accident emergency rescue model, and output target response parameters of the personalized high-voltage accident emergency rescue model in combination with high-voltage accident development time sequence rules and rescue resource scheduling time correlation information to generate real-time grading response instructions and scene-specific emergency rescue operation guidelines.

6. An electronic device, comprising: It comprises: a first processor; and a memory for storing executable instructions of the first processor; The first processor is configured to execute the new energy vehicle high-voltage accident grading response and emergency rescue operation guideline method of any one of claims 1-5 by executing the executable instructions.

7. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the second processor to implement the new energy vehicle high-voltage accident grading response and emergency rescue operation guideline method of any one of claims 1-5.

Images