New energy automobile charging safety monitoring method, system, equipment and medium

Abstract

The invention relates to a new energy automobile charging safety monitoring method, system and device and a medium. The method comprises the following steps: acquiring battery data and charging pile data of an automobile; performing standardization processing on the battery data and the charging pile data of the automobile to obtain standardized automobile battery data and standardized charging pile data; fusing the standardized automobile battery data and the standardized charging pile data to obtain automobile pile fusion data; performing feature extraction on the vehicle pile fusion data to obtain vehicle battery feature parameters and charging pile feature parameters; inputting the automobile battery characteristic parameters and the charging pile characteristic parameters into a power battery voltage prediction model to obtain a power battery voltage prediction value; obtaining a fault diagnosis result according to the power battery voltage predicted value and the power battery voltage; and comprehensively evaluating the charging safety risk of the automobile according to the fault diagnosis result to obtain a safety risk evaluation result. By adopting the method, the charging safety monitoring accuracy and the charging safety can be improved.

Description

Technical Field

[0001] This invention belongs to the field of new energy vehicle technology, and in particular relates to a method, system, equipment and medium for monitoring the charging safety of new energy vehicles. Background Technology

[0002] With the rapid development of new energy vehicle technology and the continuous increase in the number of new energy vehicles on the road, charging safety has become an increasingly important concern. Currently, new energy vehicle charging safety monitoring mainly relies on experimental or simulation data, and the research objects are mostly limited numbers of individual battery cells or battery packs. Traditional technologies typically involve conducting various tests on the power battery under laboratory conditions to obtain its characteristic parameters under different operating conditions, and then constructing a charging safety monitoring model.

[0003] However, laboratory conditions differ significantly from actual charging environments. Real-world vehicle data suffers from noise interference, data quality issues, and variable operating conditions. Directly utilizing characteristic parameters under laboratory conditions cannot fully express the safety characteristics during actual charging, leading to limitations in the practical application of existing monitoring methods. Furthermore, in terms of charging fault diagnosis, there is a lack of research on charging voltage fault diagnosis models based on vehicle-charging pile collaborative operation data. Charging safety involves both new energy vehicles and charging piles. A comprehensive assessment of charging safety risks requires consideration of the power battery status of new energy vehicles, the operating status of charging piles, and the coupling effect between the two. Current technologies lack in-depth research on the integration of data resources and safety indicators between new energy vehicles and charging piles, failing to meet the needs of practical charging safety monitoring. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, system, equipment, and medium for monitoring the charging safety of new energy vehicles that can improve charging safety, in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a method for monitoring the charging safety of new energy vehicles, including: Acquire vehicle battery data and charging pile data; vehicle battery data includes power battery voltage, current and temperature parameters; charging pile data includes charging pile output voltage, charging pile current, charging pile temperature and insulation resistance value; The battery data of the vehicle and the charging pile data are standardized to obtain standardized vehicle battery data and standardized charging pile data. Standardized vehicle battery data and standardized charging pile data are merged to obtain vehicle-charging pile fusion data; Feature extraction is performed on the vehicle-charging pile fusion data to obtain vehicle battery feature parameters and charging pile feature parameters; The characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile are input into the preset power battery voltage prediction model to obtain the predicted value of the power battery voltage. The power battery voltage prediction model includes a convolutional layer, a bidirectional gate control loop unit, an attention mechanism layer and a fully connected layer. Based on the predicted value of the power battery voltage and the power battery voltage, fault diagnosis is performed on the power battery of the vehicle to obtain the fault diagnosis results. Based on the fault diagnosis results, a comprehensive evaluation of the charging safety risks of the vehicle is conducted to obtain the safety risk evaluation results; the safety risk evaluation results include a safety risk evaluation score.

[0006] In one embodiment, the characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile are input into a preset power battery voltage prediction model to obtain a predicted power battery voltage value, including: The characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile are concatenated to obtain a joint feature vector; the joint feature vector is used as the input of the power battery voltage prediction model. Local features are extracted from the joint feature vector through convolutional layers, and a convolutional feature map is output. Temporal features are extracted from the convolutional feature map by a bidirectional gate-controlled recurrent unit, resulting in a temporal feature vector. The temporal feature vector includes a forward hidden state and a backward hidden state. The temporal feature vector is weighted by an attention mechanism layer to obtain a weighted feature vector; The weighted feature vector is input into the fully connected layer to obtain the predicted value of the power battery voltage.

[0007] In one embodiment, standardized vehicle battery data and standardized charging pile data are fused to obtain vehicle-charging pile fused data, including: Obtain vehicle charging order data; charging order data includes vehicle identification number, order creation time, charging start time, and charging end time; Based on the vehicle identification number and order creation time, standardized vehicle battery data is matched with charging order data to obtain preliminary matching data; Based on the charging start time and charging end time, the preliminary matching data is matched by time window to obtain time window matching data; Data alignment and stitching are performed on the time window matching data to obtain preliminary vehicle-pile fusion data; The initial vehicle-pile fusion data is denoised to obtain the vehicle-pile fusion data.

[0008] In one embodiment, feature extraction is performed on the vehicle-charging pile fusion data to obtain vehicle battery feature parameters and charging pile feature parameters, including: The vehicle-charging station fusion data is segmented using a sliding window to obtain multiple charging segment data. Feature parameters are extracted from the data of each charging segment to obtain the characteristic parameters of the vehicle battery and the charging pile. The characteristic parameters of the vehicle battery include the mean voltage, the variance of the current, and the temperature slope. The characteristic parameters of the charging pile include the output voltage fluctuation, the stability of the output current, and the rate of change of insulation resistance.

[0009] In one embodiment, based on the predicted value of the power battery voltage and the power battery voltage, a fault diagnosis is performed on the vehicle's power battery to obtain a fault diagnosis result, including: The predicted value of the power battery voltage and the residual value of the power battery voltage are calculated using the following formula to obtain the residual sequence: ; in, For the first The residual value at each time step. For the first The power battery voltage at each time step For the first Predicted power battery voltage at each time step; The residual mean and residual standard deviation of the residual series can be obtained using the following formulas: ; in, The mean of the residuals, The standard deviation of the residuals. For window size, For the first in the window One residual value; Based on the residual mean and residual standard deviation, the upper limit and lower limit of the dynamic threshold are obtained using the following formula: ; in, This is the upper limit of the dynamic threshold. This is the lower limit of the dynamic threshold. The mean of the residuals, The standard deviation of the residuals; The dynamic threshold range is obtained based on the upper limit and lower limit of the dynamic threshold. When the residual value exceeds the dynamic threshold range, a fault anomaly result is obtained; the fault anomaly result includes the abnormal time point and the anomaly type. Based on the abnormal fault results, fault diagnosis results are obtained; the fault diagnosis results include fault type, fault occurrence frequency and fault duration.

[0010] In one embodiment, based on the fault anomaly result, a fault diagnosis result is obtained, including: Cluster analysis is performed on abnormal time points in the fault and abnormal results to obtain abnormal event clusters; Feature extraction is performed on the cluster of abnormal events to obtain fault frequency features and fault duration; the fault frequency features are used to characterize the number of times a fault occurs within a preset time period; the fault duration is used to characterize the time period during which the abnormal event occurs. Based on the fault frequency characteristics and fault duration, the fault type classification results are obtained; the fault type classification results are used to indicate the type of fault. Based on the fault type classification results, the fault diagnosis results are obtained.

[0011] In one embodiment, based on the fault diagnosis results, a comprehensive evaluation of the charging safety risks of the vehicle is performed to obtain a safety risk evaluation result, including: Based on the fault diagnosis results, risk assessment indicators are obtained; the risk assessment indicators include vehicle battery characteristic parameters and charging pile characteristic parameters. Use the following formula to calculate the overall weight of each risk assessment indicator: ; in, For the first The comprehensive weight of each risk assessment indicator, To determine the matrix element, the first element is represented as the first element. The risk assessment indicator is relative to the first The importance of each risk assessment indicator This represents the total number of risk assessment indicators. To determine the order of a matrix; The overall weights of each risk assessment indicator are adjusted to obtain the adjusted overall weights of each risk assessment indicator. Based on the revised comprehensive weights of each risk assessment indicator and the individual risk assessment indicators, the safety risk assessment score is obtained using the following formula: ; in, As a safety risk assessment score, For the revised first The comprehensive weight of each risk assessment indicator, For the first The quantitative value of each risk assessment indicator. This represents the total number of risk assessment indicators. Based on the safety risk assessment score, a safety risk level is obtained; the safety risk level includes the first safety level, the second safety level, and the third safety level. Based on the safety risk level, a safety risk assessment result is obtained; the safety risk assessment result is used to indicate the safety risk of the power battery when the vehicle is charging.

[0012] Secondly, this application also provides a new energy vehicle charging safety monitoring system, including: The data acquisition module is used to acquire battery data and charging pile data of the vehicle; the battery data of the vehicle includes the power battery voltage, current and temperature parameters; the charging pile data includes the charging pile output voltage, charging pile current, charging pile temperature and insulation resistance value. The data standardization processing module is used to standardize the battery data of the vehicle and the charging pile data to obtain standardized vehicle battery data and standardized charging pile data. The vehicle-charging pile data fusion module is used to fuse standardized vehicle battery data and standardized charging pile data to obtain vehicle-charging pile fusion data. The feature parameter extraction module is used to extract features from the vehicle-charging pile fusion data to obtain vehicle battery feature parameters and charging pile feature parameters; The battery voltage prediction module is used to input the characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile into the preset power battery voltage prediction model to obtain the predicted value of the power battery voltage. The power battery voltage prediction model includes a convolutional layer, a bidirectional gate control loop unit, an attention mechanism layer and a fully connected layer. The fault diagnosis module is used to diagnose faults in the vehicle's power battery based on the predicted value of the power battery voltage and the power battery voltage, and obtain the fault diagnosis results. The safety evaluation module is used to comprehensively evaluate the charging safety risks of automobiles based on the fault diagnosis results, and obtain the safety risk evaluation results, which include a safety risk evaluation score.

[0013] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect.

[0014] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect.

[0015] The aforementioned method, system, equipment, and medium for monitoring the charging safety of new energy vehicles acquire multi-dimensional data from vehicle batteries and charging piles. This data is then standardized to eliminate the impact of data discrepancies, and finally fused to form vehicle-pile integrated data to break down data silos. Subsequently, feature parameters are extracted and input into a voltage prediction model containing structures such as convolutional layers and bidirectional gate control loop units. Combined with actual voltage data, fault diagnosis is achieved, and finally, a safety risk assessment is completed based on the diagnostic results. This method effectively solves problems such as significant differences between laboratory and actual charging environments and noise interference with real-vehicle data. It fully integrates vehicle-pile data resources, fuses safety indicators, considers the coupling effect between the two, avoids the limitations of traditional monitoring methods, improves the accuracy and applicability of charging safety monitoring, meets the safety monitoring needs in actual charging scenarios, and provides reliable protection for the charging safety of new energy vehicles. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies 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.

[0017] Figure 1 This is a flowchart illustrating a method for monitoring the charging safety of new energy vehicles in one embodiment. Figure 2 This is a schematic diagram of the structure of a new energy vehicle charging safety monitoring system in one embodiment; Figure 3 This is a schematic diagram of a computer device in one embodiment. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] In one embodiment, such as Figure 1 As shown, a method for monitoring the charging safety of new energy vehicles is provided. This embodiment illustrates the application of this method to a safety monitoring terminal (hereinafter referred to as the terminal). It can be understood that this method can also be applied to a server, or to a system including both a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps: S1 acquires vehicle battery data and charging station data.

[0020] For example, the safety monitoring terminal establishes a data interaction link with the battery management system of the new energy vehicle and the communication module of the charging pile to obtain battery data of the vehicle and charging pile data. The vehicle's battery data includes power battery voltage, current, and temperature parameters; the power battery voltage reflects the potential difference between the positive and negative electrodes and is a fundamental parameter for determining whether the battery is in the normal charging and discharging range; the current characterizes the rate of charge flow inside the battery and is directly related to charging power and battery heat generation; the temperature parameter covers the temperature of individual battery cells and the overall temperature of the battery pack, as excessively high or low temperatures will affect battery performance and safety. The charging pile data includes charging pile output voltage, charging pile current, charging pile temperature, and insulation resistance. The charging pile output voltage must match the voltage required by the vehicle battery to ensure a stable charging process; the charging pile current determines the charging speed and must be controlled within the battery's tolerance range; charging pile temperature monitoring can promptly detect overheating problems in the charging pile's internal circuitry; the insulation resistance value is used to assess the electrical insulation performance between the charging pile and the vehicle, and an excessively low resistance value may cause leakage risks.

[0021] S2 standardizes the vehicle's battery data and charging pile data to obtain standardized vehicle battery data and standardized charging pile data.

[0022] For example, the safety monitoring terminal standardizes vehicle battery data and charging pile data. This standardization eliminates the impact of differences in dimensions and numerical ranges between different data types, laying a unified foundation for subsequent data fusion and analysis. Specifically, parameters such as voltage, current, and temperature in vehicle battery data differ in units and magnitudes from those in charging pile data, such as output voltage, current, temperature, and insulation resistance. For instance, voltage might be measured in volts (V), current in amperes (A), temperature in degrees Celsius (°C), and insulation resistance in megaohms (MΩ). Directly using the raw data for analysis could lead to some parameters being over-amplified or under-amplified in calculations, affecting the accuracy of the analysis results. The safety monitoring terminal maps various data types to the same data range. By calculating the mean and standard deviation of the raw data for each parameter and then converting it into dimensionless standardized data according to a preset standardization formula, standardized vehicle battery data and standardized charging pile data are obtained, making different types of data comparable and fusionable.

[0023] S3 merges standardized vehicle battery data and standardized charging pile data to obtain vehicle-charging pile fusion data.

[0024] For example, the safety monitoring terminal integrates standardized vehicle battery data and standardized charging pile data to break down data silos between vehicles and charging piles, constructing a comprehensive vehicle-charging pile fusion data system that reflects the charging process. During charging, the vehicle battery status and the charging pile's operational status are strongly correlated. Specifically, changes in the charging pile's output current directly affect the battery's charging current and temperature changes; relying solely on single data points makes it difficult to comprehensively assess charging safety. The safety monitoring terminal uses data fusion technology to first analyze the correlation between the two types of standardized data, identifying matching battery and charging pile data at the same time. Then, it employs data splicing and integration strategies to correlate battery parameters reflecting the same charging moment or stage with charging pile parameters, forming vehicle-charging pile fusion data that includes multi-dimensional status information for both the vehicle and the charging pile. This fusion data not only includes core parameters such as battery voltage, current, and temperature but also integrates the charging pile's output characteristics and insulation performance parameters, providing a more comprehensive and accurate reflection of the overall operational status during charging.

[0025] S4 extracts features from the vehicle-charging pile fusion data to obtain vehicle battery feature parameters and charging pile feature parameters.

[0026] For example, the safety monitoring terminal performs feature extraction on the vehicle-charging pile fusion data, that is, it mines key information that can characterize the operating status of the vehicle battery and the charging pile from the massive amount of fusion data to obtain the characteristic parameters of the vehicle battery and the charging pile. The vehicle-charging pile fusion data contains a large amount of time-series raw data, and directly using it for analysis will result in problems such as data redundancy and lack of emphasis on key information. Among them, the characteristic parameters of the vehicle battery include power battery voltage parameters, current characteristic parameters, temperature characteristic parameters, state of charge characteristic parameters, capacity degradation rate characteristic parameters, internal resistance growth rate characteristic parameters, and voltage consistency characteristic parameters. Power battery voltage parameters reflect battery health by analyzing voltage change trends; current characteristic parameters monitor current fluctuations to determine charging stability; temperature characteristic parameters track temperature change rates and warn of overheating risks; state of charge (SOC) characteristic parameters, calculated based on charging time and current data, reflect the battery's current charge level; capacity degradation rate characteristic parameters assess battery aging by comparing current capacity with initial capacity; internal resistance growth rate characteristic parameters are calculated based on voltage and current data, and increased internal resistance is a key indicator of battery performance decline; voltage consistency characteristic parameters analyze voltage differences between individual battery cells, as excessive differences can affect overall battery performance and safety. Charging pile characteristic parameters include output voltage, current, temperature, insulation resistance, charging efficiency, and fault frequency characteristic parameters. Output voltage characteristic parameters monitor voltage stability to avoid damage to the battery caused by voltage fluctuations; output current characteristic parameters assess the smoothness of current output; temperature characteristic parameters monitor the charging pile's own heating; insulation resistance characteristic parameters track changes in insulation performance; charging efficiency characteristic parameters are obtained by calculating the ratio of actual charging amount to theoretical charging amount; and fault frequency characteristic parameters are used to characterize the number of historical faults that have occurred in the charging pile.

[0027] S5: Input the characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile into the preset power battery voltage prediction model to obtain the predicted value of the power battery voltage.

[0028] Specifically, the safety monitoring terminal inputs the characteristic parameters of the vehicle battery and the charging pile into a preset power battery voltage prediction model to accurately predict future voltage changes in the power battery and obtain the predicted power battery voltage value. The power battery voltage prediction model combines various deep learning technologies, with each layer performing different functions: the convolutional layer, by setting multiple convolutional kernels, extracts local features from the input characteristic parameters, capturing local correlation information between characteristic parameters, such as the correlation between voltage and current changes over a short period; the output convolutional feature map integrates key local features; the bidirectional gated recurrent loop (BRL)... The Bi-GRU (Bi-GRU) unit, a recurrent neural network unit capable of simultaneously utilizing forward and backward temporal information, is used to extract temporal features from the convolutional feature map. Since battery voltage changes during charging exhibit significant temporal variation, the forward hidden state captures temporal information from the start of charging to the current moment, while the backward hidden state captures temporal information from the current moment to the end of charging. The combined temporal feature vector comprehensively reflects the temporal pattern of voltage changes. The attention mechanism layer calculates the importance weights of each element in the temporal feature vector, giving higher attention to key temporal features and assigning greater weight to temporal features at voltage abrupt changes, thus strengthening important information and obtaining a weighted feature vector. The fully connected layer maps the weighted feature vector to a preset output dimension. Through the calculation of multiple neurons, the final predicted value of the power battery voltage is output, providing a comparative basis for subsequent fault diagnosis.

[0029] S6, based on the predicted value of the power battery voltage and the power battery voltage, performs fault diagnosis on the vehicle's power battery and obtains the fault diagnosis results; For example, the safety monitoring terminal performs fault diagnosis on the vehicle's power battery based on the difference between the predicted and actual power battery voltage. First, the residual value between the two is calculated, resulting in a residual sequence. The terminal determines a dynamic threshold range by calculating the mean and standard deviation of the residual sequence. When the residual value exceeds the dynamic threshold range, it is judged as a fault anomaly, and the anomaly time point and anomaly type are recorded. Based on these fault anomaly results, further analysis yields fault diagnosis results, including fault type, fault occurrence frequency, and fault duration. Through fault diagnosis using predictive models and dynamic thresholds, potential faults in the power battery during charging can be detected in a timely and accurate manner, improving charging safety.

[0030] S7, based on the fault diagnosis results, conducts a comprehensive evaluation of the charging safety risks of the vehicle and obtains the safety risk evaluation results.

[0031] Specifically, based on fault diagnosis results, the safety monitoring terminal comprehensively evaluates the safety risks of vehicle charging through multi-dimensional index quantification and weighted calculation, resulting in a safety risk assessment result including a safety risk assessment score. Information such as fault type, fault frequency, and fault duration in the fault diagnosis results have varying degrees of impact on charging safety. For example, the risk of internal short circuit faults is far higher than that of minor voltage fluctuations; the higher the frequency and duration of faults, the higher the safety risk. The safety monitoring terminal first converts the information in the fault diagnosis results into quantitative indicators, then assigns weights to each indicator based on its impact on safety; a weighted summation is then used to calculate the safety risk assessment score, with a higher score indicating a higher safety risk; finally, the safety risk level is determined by combining the score's range, forming a complete safety risk assessment result that provides clear safety decision-making basis for users or charging management systems.

[0032] The aforementioned method for monitoring the charging safety of new energy vehicles effectively addresses the limitations of monitoring caused by noise interference, uneven data quality, and differences between laboratory and actual environments by acquiring and standardizing data from both the vehicle and charging pile ends. The vehicle-charging pile data fusion and feature extraction process deeply integrates vehicle-charging pile data resources and incorporates safety indicators, fully considering the coupling effect between the two and compensating for the shortcomings of existing technologies in utilizing vehicle-charging pile collaborative data. Combined with a voltage prediction model and fault diagnosis mechanism containing structures such as convolutional layers and bidirectional gate control loop units, it can accurately identify battery faults. Subsequent safety risk assessment provides quantitative evidence for charging safety, comprehensively improving the accuracy and reliability of monitoring in actual charging scenarios and meeting the actual charging safety monitoring needs.

[0033] In an optional embodiment, the characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile are input into a preset power battery voltage prediction model to obtain the predicted value of the power battery voltage, including the following steps: S11, the characteristic parameters of the car battery and the characteristic parameters of the charging pile are concatenated to obtain a joint feature vector.

[0034] For example, the safety monitoring terminal first verifies the data format of the vehicle battery characteristic parameters and the charging pile characteristic parameters to ensure that the data dimensions and data types of the two types of parameters are consistent, avoiding splicing failures due to format differences. The vehicle battery characteristic parameters include multiple dimensions such as voltage, current, and temperature, while the charging pile characteristic parameters include output voltage, output current, and insulation resistance. The safety monitoring terminal matches the characteristic parameters of the two sets of parameters one by one according to a preset parameter sorting rule. Specifically, it arranges the vehicle battery voltage characteristic parameters and the charging pile output voltage characteristic parameters adjacently, and then uses a data splicing algorithm to merge the two types of parameters into a unified joint feature vector. This joint feature vector integrates key feature information from both the vehicle and the charging pile, comprehensively reflecting the coordinated operation status of the battery and the charging pile during the charging process, and providing complete input data for subsequent feature processing at each layer of the model.

[0035] S12 extracts local features from the joint feature vector through a convolutional layer and outputs a convolutional feature map. For example, when a safety monitoring terminal extracts local features from a joint feature vector using convolutional layers, it presets multiple convolutional kernels of different sizes, such as 3×1 and 5×1 kernels, based on the data dimension and feature type of the joint feature vector. Each convolutional kernel extracts a local feature pattern, such as the correlation between voltage and current changes over a short period of time, or the coordinated change between temperature and insulation resistance. The convolutional kernels perform sliding calculations on the joint feature vector, mapping local features to feature values ​​through matrix multiplication and activation function processing. The feature values ​​output by multiple convolutional kernels are integrated to form a convolutional feature map. The convolutional feature map condenses key local information in the joint feature vector and removes some redundant data.

[0036] S13 extracts temporal features from the convolutional feature map by controlling the loop unit through a bidirectional gate, thus obtaining a temporal feature vector.

[0037] For example, the safety monitoring terminal extracts temporal features from the convolutional feature map through a bidirectional gated loop unit. Since the feature data in the convolutional feature map is arranged in chronological order, the forward computation process of the bidirectional gated loop unit progresses backward from the feature data at the start of charging, capturing the positive change pattern in the time series; the backward computation process traces back from the feature data at the current charging moment, capturing the reverse correlation information in the time series. The forward hidden state and the backward hidden state store the forward and reverse temporal feature information, respectively. The temporal feature vector formed by concatenating the two can fully reflect the change trend of the convolutional feature map in the time dimension, providing key temporal features for the subsequent attention mechanism layer.

[0038] S14, the temporal feature vector is weighted through the attention mechanism layer to obtain the weighted feature vector; For example, the safety monitoring terminal weights the time-series feature vector through an attention mechanism layer. By identifying feature elements in the time-series feature vector that have a significant impact on the prediction of the power battery voltage and assigning them higher weights, the model's prediction accuracy is improved. The safety monitoring terminal first calculates the attention score of each element in the time-series feature vector. The attention score is obtained through a preset attention calculation function, which comprehensively considers the element's own feature value, its correlation with neighboring elements, and its position information in the entire time sequence. Specifically, the time-series feature element corresponding to the voltage mutation moment will receive a higher attention score. The terminal assigns weights to each element in the time-series feature vector based on the attention scores, with elements having higher scores receiving higher weights. Then, each element is multiplied by its corresponding weight to obtain the weighted feature elements. All weighted feature elements are integrated to form a weighted feature vector. Through this weighting process, the time-series features that play a crucial role in voltage prediction are strengthened, while the influence of irrelevant or secondary features is weakened. This allows subsequent fully connected layers to focus on core features for calculation, improving the accuracy of the power battery voltage prediction.

[0039] S15, the weighted feature vector is input into the fully connected layer to obtain the predicted value of the power battery voltage.

[0040] For example, the safety monitoring terminal inputs the weighted feature vector into a fully connected layer. This fully connected layer consists of multiple layers of neurons, each connected to all neurons in the layer above it, enabling deep feature transformation and integration of the weighted feature vector. Specifically, the first layer of neurons performs linear transformation and nonlinear activation on the weighted feature vector, mapping it to a new feature space and extracting more abstract feature information. The intermediate layer neurons further process the features output from the previous layer, gradually optimizing the feature representation and enhancing the ability to capture voltage change patterns. The last layer of neurons maps the multi-layered processed features to a single output dimension, and the output value is the predicted voltage of the power battery. The fully connected layer continuously optimizes parameters through preset weight parameters and bias terms, combined with the backpropagation algorithm, ensuring that the output predicted value accurately matches the actual voltage change trend of the power battery.

[0041] In an optional embodiment, standardized vehicle battery data and standardized charging pile data are fused to obtain vehicle-charging pile fused data, including the following steps: S21, obtain the vehicle's charging order data.

[0042] Specifically, the safety monitoring terminal establishes a data communication connection with the charging operation management platform to acquire vehicle charging order data. This charging order data is structured data recording key information for a single charging service, including the Vehicle Identification Number (VIN, a unique identifier for each vehicle), order creation time, charging start time, and charging end time. The VIN is the core identifier linking the vehicle to the charging order, allowing for precise location of the specific new energy vehicle. The order creation time records the time the user initiates the charging request, reflecting the initial application moment for the charging service. The charging start time is the time when the charging pile actually begins supplying power to the vehicle, marking the official start of the charging process. The charging end time is the time when the charging pile stops supplying power, marking the termination of the charging process. By acquiring vehicle charging order data, not only can the time range of a single charging service be clearly defined, but the charging order can also be associated with the specific vehicle through the VIN, providing crucial evidence for subsequent matching of vehicle-charging pile data with order data, ensuring the accuracy and relevance of data fusion.

[0043] S22, Based on the vehicle identification number and order creation time, the standardized vehicle battery data is matched with the charging order data to obtain preliminary matching data; For example, the safety monitoring terminal matches standardized vehicle battery data with charging order data based on the vehicle identification number (VIN) and order creation time, establishing a correspondence between vehicle battery data and specific charging orders to obtain preliminary matching data. Specifically, since multiple vehicles may be charging simultaneously within the same time period, relying solely on time-based matching can easily lead to data confusion. The VIN, being unique, serves as a key identifier to distinguish different vehicles. The safety monitoring terminal first uses the VIN as an index to filter out the standardized battery data of the vehicles corresponding to each charging order; then, combining this with the order creation time, it further filters out standardized vehicle battery data within a preset time range before and after the order creation time, eliminating interference from battery data unrelated to the charging order. Through this dual matching of the VIN and order creation time, it ensures that each piece of standardized vehicle battery data can be accurately associated with its corresponding charging order.

[0044] S23, based on the charging start time and charging end time, perform time window matching on the preliminary matching data to obtain time window matching data.

[0045] For example, the safety monitoring terminal performs time window matching on the preliminary matching data based on the charging start time and charging end time to accurately locate battery data that completely corresponds to a single charging process, thus obtaining time window matching data. The preliminary matching data may include preparatory data before charging begins and redundant data after charging ends. This data is not valid data during the charging process and needs to be filtered through the time window. Specifically, the safety monitoring terminal constructs a time window that perfectly matches the duration of a single charging process, using the charging start time in the charging order as the starting point and the charging end time as the ending point. Then, it extracts standardized vehicle battery data from the preliminary matching data whose timestamps fall within this time window. This data precisely corresponds to the changes in the vehicle's battery state during this charging process, eliminating invalid data before and after charging. This ensures that subsequent data fusion is based only on valid data during the charging process, improving the quality of data fusion.

[0046] S24, perform data alignment and stitching on the time window matching data to obtain preliminary vehicle-pile fusion data; For example, the safety monitoring terminal performs data alignment and splicing on time-window matching data. It integrates standardized vehicle battery data and corresponding standardized charging pile data within the time window according to their timestamps to obtain preliminary vehicle-charging pile fusion data. Specifically, since the collection frequencies of vehicle battery data and charging pile data may differ (battery data is collected every 1 second, and charging pile data every 2 seconds), direct splicing would result in data time misalignment. The safety monitoring terminal first processes the timestamps of both types of data uniformly, using an interpolation algorithm to complete the data with lower collection frequencies, ensuring complete synchronization of the timestamps of the two types of data. Then, according to the timestamp order, it splices the standardized vehicle battery parameters and standardized charging pile parameters at the same time point to form structured data entries containing timestamps, battery parameters, and charging pile parameters. All data entries are arranged in chronological order, constituting preliminary vehicle-charging pile fusion data, achieving precise alignment of vehicle and charging pile data in the time dimension.

[0047] S25. The initial vehicle-pile fusion data is denoised to obtain the vehicle-pile fusion data.

[0048] Specifically, the safety monitoring terminal performs noise reduction processing on the initial vehicle-to-pile fusion data to eliminate noise data introduced during data acquisition due to factors such as sensor errors and communication interference, resulting in clean vehicle-to-pile fusion data. The initial vehicle-to-pile fusion data may contain abnormal data points, such as voltage spikes or zero current caused by temporary sensor malfunctions, which do not conform to actual charging patterns. This noise data can affect the accuracy of subsequent feature extraction and analysis. The safety monitoring terminal uses preset noise reduction algorithms, such as moving average algorithms or outlier detection algorithms, to analyze the initial vehicle-to-pile fusion data point by point: the moving average algorithm smooths data fluctuations and eliminates small noises by calculating the average of multiple adjacent data points; the outlier detection algorithm identifies and removes abnormal data points that exceed a set normal data range. Through noise reduction processing, it is ensured that the final vehicle-to-pile fusion data can truly and accurately reflect the actual operating status of the vehicle and the charging pile during the charging process.

[0049] In an optional embodiment, feature extraction is performed on the vehicle-charging pile fusion data to obtain vehicle battery feature parameters and charging pile feature parameters, including the following steps: S31, the vehicle-charging pile fusion data is segmented by a sliding window to obtain multiple charging segment data; For example, the safety monitoring terminal performs sliding window segmentation on the vehicle-charging pile fusion data to divide the continuous time-series vehicle-charging pile fusion data into multiple time-related charging segment data. The vehicle-charging pile fusion data is a continuous data sequence arranged in chronological order. Directly extracting features from the entire data makes it difficult to capture the characteristic differences between different charging stages; for example, the battery and charging pile state changes differently in the initial, middle, and final stages of charging. The safety monitoring terminal presets a fixed-length sliding window and a sliding step size. The sliding window length is set according to the duration of typical stages in the charging process, ensuring that the data within each window reflects a complete local charging stage. The sliding step size controls the interval of window movement, avoiding excessive data overlap or omission. By successively moving the sliding window across the vehicle-charging pile fusion data sequence, the continuous data is divided into multiple non-overlapping or slightly overlapping charging segment data. Each segment data corresponds to a specific charging time period, facilitating subsequent feature extraction for different charging stages.

[0050] S32 extracts feature parameters from each charging segment data to obtain the vehicle battery feature parameters and charging pile feature parameters.

[0051] Specifically, the safety monitoring terminal extracts feature parameters from the data of each charging segment, and mines key indicators that can characterize the operating status of the battery and charging pile at that stage from each local charging segment, thus obtaining characteristic parameters of the vehicle battery and the charging pile. For the characteristic parameters of the vehicle battery, the voltage mean is obtained by calculating the average value of all voltage data within the charging segment, which reflects the overall level of the battery voltage within that segment; the current variance is obtained by calculating the average of the sum of squares of the deviations of the current data from the mean, which reflects the degree of current fluctuation, and the larger the variance, the more unstable the current; the temperature slope is obtained by linearly fitting the temperature change trend over time within the segment and calculating the slope of the fitted line, which reflects the rate of temperature change, and a positive slope indicates a temperature increase, and a larger absolute value of the slope indicates a faster temperature change. For the characteristic parameters of charging piles, output voltage fluctuation is calculated by the difference between the maximum and minimum voltage values; a larger difference indicates poorer voltage stability. Output current stability is calculated by the coefficient of variation (the ratio of standard deviation to mean) of the current data; a smaller coefficient of variation indicates more stable current. The rate of change of insulation resistance is calculated by dividing the difference between the insulation resistance at the start and end of the segment by the insulation resistance at the start, reflecting the trend of insulation performance change. A negative rate of change with a large absolute value indicates a significant decline in insulation performance. These characteristic parameters can accurately depict the operating status of the vehicle and charging pile within each charging segment.

[0052] In an optional embodiment, based on the predicted value of the power battery voltage and the power battery voltage, fault diagnosis of the vehicle's power battery is performed to obtain fault diagnosis results, including the following steps: S41, use the following formula to calculate the predicted value of the power battery voltage and the residual value of the power battery voltage to obtain the residual sequence: ; in, For the first The residual value at each time step. For the first The power battery voltage at each time step For the first The predicted value of the power battery voltage at each time step.

[0053] Specifically, the safety monitoring terminal calculates the predicted value of the power battery voltage and the residual value of the power battery voltage according to a given formula, forming a residual sequence. The residual value is the core indicator for measuring the degree of deviation between the predicted value and the actual value. In the above formula for calculating the residual value, For the first The residual value at each time step. For the first The power battery voltage at each time step For the first The predicted voltage of the power battery at each time step is calculated. By calculating the difference between the two values ​​at each time step, the continuous deviation information is transformed into an ordered residual sequence. The residual sequence can intuitively reflect the error distribution law of the prediction model. If the residual sequence shows random fluctuations and a stable range, it indicates that the model prediction effect is good. If there is a systematic deviation or abrupt change, there may be an anomaly in the battery or the model needs to be optimized.

[0054] S42, use the following formulas to obtain the residual mean and residual standard deviation of the residual series: ; in, The mean of the residuals, The standard deviation of the residuals. For window size, For the first in the window Each residual value.

[0055] Specifically, the safety monitoring terminal calculates the residual mean and residual standard deviation of the residual sequence according to formulas. In the formulas for calculating the residual mean and residual standard deviation, The mean of the residuals is used to reflect the overall trend of the deviation. If the mean deviates significantly from zero, there may be systematic errors in the prediction model. The standard deviation of the residual measures the degree of dispersion of the residual value from the mean and reflects the fluctuation range of the error. The larger the standard deviation, the more dispersed the error distribution. For window size, For the first in the window Each residual value is calculated. The window size needs to balance the continuity of the time series with the ability to capture local features. By using a sliding window to segment the residual series, the mean and standard deviation corresponding to each window are obtained, which dynamically reflects the error characteristics in different time periods and provides a quantitative basis for determining the subsequent dynamic threshold.

[0056] S43. Based on the residual mean and residual standard deviation, the upper limit and lower limit of the dynamic threshold are obtained using the following formula: ; in, This is the upper limit of the dynamic threshold. This is the lower limit of the dynamic threshold. The mean of the residuals, This represents the standard deviation of the residuals.

[0057] Specifically, the safety monitoring terminal calculates the upper and lower limits of the dynamic threshold using a formula based on the residual mean and residual standard deviation. The setting of the dynamic threshold follows statistical principles. The principle is that normal data is likely distributed within the range of the mean plus or minus... Data outside the range of one-times the standard deviation is considered outlier. In the above formulas for calculating the upper and lower limits of the dynamic threshold, The mean of the residuals, The standard deviation of the residuals. The upper limit of the dynamic threshold is Add 3 times , The lower limit of the dynamic threshold is Reduce by 3 times This setting method allows the threshold to be dynamically adjusted according to the mean and fluctuation characteristics of the residual sequence. Compared with a fixed threshold, it is more adaptable to changes in the battery state during charging. The residual distribution characteristics may be different in the early and late stages of charging. The dynamic threshold can accurately match the error characteristics of different stages and improve the accuracy of anomaly detection.

[0058] S44. Based on the upper limit and lower limit of the dynamic threshold, the dynamic threshold range is obtained.

[0059] For example, the security monitoring terminal defines the range between the upper and lower limits of the dynamic threshold as the dynamic threshold range. The dynamic threshold range serves as the benchmark for determining whether residual values ​​are normal and is updated in real time as the time window moves. The width of the dynamic threshold range is determined by the standard deviation of the residuals; the greater the error fluctuation, the wider the range; the smaller the error fluctuation, the narrower the range. By setting this range, the distribution interval of normal residuals is clearly defined, providing a clear criterion for subsequent anomaly detection. This ensures that only residual values ​​that significantly deviate from the normal pattern are identified as abnormal, reducing the probability of false positives.

[0060] S45, when the residual value exceeds the dynamic threshold range, an abnormal fault result is obtained.

[0061] For example, the safety monitoring terminal compares the residual value with the dynamic threshold range. When the residual value exceeds the upper limit of the dynamic threshold or falls below the lower limit, an anomaly is identified, resulting in a fault anomaly result. The anomaly time point is the specific time step in which the residual value exceeds the threshold range; the time of occurrence can be precisely located using a timestamp. The anomaly type is initially determined based on the direction and magnitude of the residual exceeding the threshold. For example, a residual value significantly higher than the upper limit of the dynamic threshold may correspond to a voltage surge anomaly, while a residual value significantly lower than the lower limit may correspond to a voltage drop anomaly. By monitoring the relationship between the residual and the threshold in real time, rapid identification of abnormal battery states is achieved, providing initial anomaly information for subsequent fault diagnosis.

[0062] S46. Based on the fault anomaly results, the fault diagnosis results are obtained.

[0063] The safety monitoring terminal integrates and analyzes fault and anomaly results to obtain fault diagnosis results. These results include fault type, frequency of occurrence, and duration. The terminal summarizes the abnormal time points and types, and combines this information with other characteristic parameters such as battery current and temperature to determine whether the anomaly is caused by a battery fault, eliminating false anomalies due to data acquisition errors. It then identifies the fault type, such as voltage runaway or individual cell anomaly; counts the frequency of occurrence (the number of abnormal events per unit time); and records the duration of the fault (the time interval from the occurrence of the anomaly to the return to normal). By comprehensively analyzing this information, a complete fault diagnosis result is generated, including fault type, frequency of occurrence, and duration.

[0064] In an optional embodiment, based on the fault anomaly result, a fault diagnosis result is obtained, including the following steps: S51, perform cluster analysis on the abnormal time points in the fault abnormal results to obtain abnormal event clusters; Specifically, the safety monitoring terminal performs cluster analysis on abnormal time points in fault anomaly results, grouping time-related anomalies into the same abnormal event cluster to identify complete fault events. A single abnormal time point may only be an instantaneous disturbance, while consecutive or densely occurring anomalies often constitute a complete fault event. The safety monitoring terminal uses clustering algorithms, such as Density-Based Spatial Clustering of Applications with Noise (DBSCAN, a clustering algorithm based on data point density), to group anomalies based on the time interval between them: when the interval between two anomalies is less than a preset threshold, they are determined to be consecutive anomalies of the same event; when the interval is greater than the threshold, they are classified as different events. The abnormal event clusters obtained through clustering can accurately reflect the complete temporal distribution of fault events, avoiding misjudging multiple anomalies of the same fault as multiple independent faults.

[0065] S52, extract features from the abnormal event clusters to obtain fault frequency features and fault duration.

[0066] The fault frequency feature characterizes the number of times a fault occurs within a preset time period; the fault duration characterizes the time period during which the abnormal event occurs. Specifically, the safety monitoring terminal extracts features from abnormal event clusters to obtain the fault frequency feature and fault duration. The fault frequency feature is obtained by counting the number of abnormal event clusters within the preset time period, reflecting the density of fault occurrences; a higher frequency indicates poorer battery stability. The fault duration is obtained by calculating the time difference between the first and last abnormal time points in each abnormal event cluster, reflecting the duration of the fault event; a longer duration indicates a more severe fault impact.

[0067] S53. Based on the fault frequency characteristics and fault duration, the fault type classification results are obtained.

[0068] For example, the safety monitoring terminal obtains a fault type classification result based on the fault frequency characteristics and fault duration, combined with preset fault type classification rules. Different types of battery faults have different time characteristics. Specifically, cell aging faults may manifest as low-frequency, short-duration voltage anomalies; internal short-circuit faults may manifest as high-frequency, long-duration voltage drops and temperature rises occurring simultaneously. The safety monitoring terminal compares the extracted fault frequency and duration characteristics with preset fault feature templates, determines the most likely fault type through similarity matching, and the classification result clearly indicates the specific attributes of the fault, providing a basis for accurate diagnosis of battery problems.

[0069] S54. Based on the fault type classification results, the fault diagnosis results are obtained.

[0070] Specifically, the safety monitoring terminal integrates information such as fault frequency characteristics and fault duration based on the fault type classification results to form the final fault diagnosis result. The fault diagnosis result not only includes the specific type of fault but also covers the temporal pattern (frequency) and impact range (duration) of the fault, providing a comprehensive characterization of the battery fault state. For example, the diagnosis result might be a cell aging fault that occurred twice in the past hour, each lasting approximately 30 seconds. This structured result clearly identifies the nature of the fault and provides quantifiable fault characteristics, offering detailed and accurate input data for subsequent safety risk assessments, ensuring the scientific rigor and relevance of the risk assessment.

[0071] In an optional embodiment, based on the fault diagnosis results, a comprehensive evaluation of the charging safety risks of the vehicle is performed to obtain a safety risk evaluation result, including the following steps: S61. Based on the fault diagnosis results, risk assessment indicators are obtained.

[0072] For example, the safety monitoring terminal extracts key information from fault diagnosis results and combines it with the characteristic parameters of the vehicle battery and the charging pile to determine risk assessment indicators. These indicators cover multi-dimensional state information of the battery and the charging pile. Vehicle battery characteristic parameters include voltage stability, current fluctuation, temperature change rate, state of charge, and capacity degradation rate, directly reflecting the battery's health status and safety boundaries. Charging pile characteristic parameters include output voltage stability, output current stability, temperature, insulation resistance, charging efficiency, and fault frequency, reflecting the impact of the charging pile's operating status on charging safety. Specifically, the terminal integrates these indicators to construct a charging safety risk assessment system containing target, project, and indicator layers. It then determines the weights of the safety risk assessment indicators, uses the analytic hierarchy process (AHP) and entropy weighting method to determine the combined weights of each indicator, and calculates the charging safety risk indicators for each vehicle based on the constructed risk assessment system. The indicators are then normalized to obtain a safety risk assessment score for each vehicle. The superiority of this method is verified by comparing the results with the AHP analysis, thus achieving charging safety risk assessment.

[0073] S62, use the following formula to calculate the overall weight of each risk assessment indicator: ; in, For the first The comprehensive weight of each risk assessment indicator, To determine the matrix element, the first element is represented as the first element. The risk assessment indicator is relative to the first The importance of each risk assessment indicator This represents the total number of risk assessment indicators. To determine the order of a matrix.

[0074] Specifically, the safety monitoring terminal uses a given formula to calculate the comprehensive weight of each risk assessment indicator. The weight reflects the importance of different indicators in the risk assessment. The formula for calculating the comprehensive weight is as follows: For the first The comprehensive weight of each risk assessment indicator, To determine the first element in the matrix The first indicator is relative to the first The importance of each indicator is scored using a judgment matrix constructed through the Analytic Hierarchy Process (AHP, a method for decomposing complex decision problems into hierarchical structures for analysis), quantifying the relative importance among the indicators. For example, the importance of fault type may be higher than that of fault frequency, reflected in... The values ​​in the matrix are larger. By normalizing the elements of the judgment matrix, i.e., the numerator is the sum of the importance scores of a certain indicator and all other indicators, and the denominator is the sum of the importance scores of all indicators among themselves, the initial comprehensive weight of each indicator is obtained, providing a weight basis for subsequent risk calculation.

[0075] S63, the comprehensive weights of each risk assessment indicator are adjusted to obtain the adjusted comprehensive weights of each risk assessment indicator.

[0076] Specifically, the safety monitoring terminal uses the entropy weight method to correct the comprehensive weights of each risk assessment indicator, resulting in a more objective and reasonable corrected weight. The entropy weight method is based on information entropy theory, determining weights by analyzing the dispersion of each indicator's data. The greater the data dispersion (i.e., the more significant the change in indicator value), the greater the amount of information it carries, and the higher the weight; conversely, the lower the weight, the lower the weight. Correction using the entropy weight method effectively compensates for potential subjective biases in the analytic hierarchy process (AHP). For example, when the actual data for the battery insulation resistance indicator fluctuates significantly (indicating a significant impact on safety), the entropy weight method will increase its weight. During the correction process, the terminal first calculates the information entropy of each indicator, then calculates the entropy weight based on the information entropy, and finally combines the initial weights obtained from the AHP with the entropy weights to obtain the corrected comprehensive weight, ensuring that the weight allocation reflects both expert experience and actual data characteristics.

[0077] S64. Based on the revised comprehensive weights of each risk assessment indicator and the individual risk assessment indicators, the safety risk assessment score is obtained using the following formula: ; in, As a safety risk assessment score, For the revised first The comprehensive weight of each risk assessment indicator, For the first The quantitative value of each risk assessment indicator. This represents the total number of risk assessment indicators.

[0078] Specifically, the security monitoring terminal is based on the corrected comprehensive weight. and the quantitative values ​​of each risk assessment indicator The safety risk assessment score is calculated using the formula described above. Quantified value This involves quantifying risk indicators, assigning values ​​to fault types based on risk levels, fault frequencies based on frequency ranges, and fault durations based on duration ranges. The formula uses a weighted sum to integrate the quantified values ​​of each indicator into a single score; indicators with higher weights have a greater impact on the total score. Multiplying the quantified value of a high-risk fault type by its higher weight will significantly increase the risk score. The safety risk assessment score is used to comprehensively reflect the overall level of charging safety risk, providing a quantitative basis for risk level classification.

[0079] S65, based on the safety risk assessment score, the safety risk level is obtained; the safety risk level includes the first safety level, the second safety level, and the third safety level; Specifically, the safety monitoring terminal determines the corresponding safety risk level based on the range of the safety risk assessment score, including Level 1 (low risk), Level 2 (medium risk), and Level 3 (high risk). The threshold ranges for level classification are set based on historical fault data and safety standards. For example, a score below a preset lower limit is Level 1, indicating that the charging process is safe and controllable; a score between the lower and upper limits is Level 2, indicating potential risks requiring close monitoring; and a score above the upper limit is Level 3, indicating a high risk requiring immediate intervention. Different levels correspond to different safety states, and the level classification transforms the abstract risk score into an intuitive risk status indicator, facilitating a quick understanding of the risk level.

[0080] S66, based on the safety risk level, yields a safety risk assessment result; the safety risk assessment result is used to indicate the safety risks of the power battery when the vehicle is charging.

[0081] Specifically, the safety monitoring terminal generates a complete safety risk assessment result based on the safety risk level and the specific circumstances of the risk assessment indicators. This result includes not only the risk level but also the key factors contributing to that level, along with targeted safety recommendations. For example, a key factor might be at the third safety level, primarily due to frequent internal battery short-circuit faults and a decrease in the insulation resistance of the charging pile. Safety recommendations might include immediately stopping charging and inspecting the battery's internal condition and the charging pile's insulation performance. The safety risk assessment result provides clear decision-making support for users, charging operators, or vehicle control systems, enabling effective management of charging safety risks and ensuring the safety and reliability of the new energy vehicle charging process.

[0082] The aforementioned method for monitoring the charging safety of new energy vehicles acquires multi-dimensional data from vehicle batteries and charging piles. This data is then standardized to eliminate the impact of data discrepancies and fused to form vehicle-pile integrated data, breaking down data silos. Subsequently, feature parameters are extracted and input into a voltage prediction model containing structures such as convolutional layers and bidirectional gate control loop units. Combined with actual voltage data, fault diagnosis is achieved, and finally, a safety risk assessment is completed based on the diagnostic results. This method effectively addresses issues such as significant differences between laboratory and actual charging environments and noise interference with real-vehicle data. It fully integrates vehicle-pile data resources and fuses safety indicators, considering their coupling effect. This avoids the limitations of traditional monitoring methods, improves the accuracy and applicability of charging safety monitoring, meets the safety monitoring needs of actual charging scenarios, and provides reliable protection for the charging safety of new energy vehicles.

[0083] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0084] Based on the same inventive concept, this application also provides a new energy vehicle charging safety monitoring system for implementing the aforementioned new energy vehicle charging safety monitoring method. The solution provided by this system is similar to the implementation scheme described in the above method; therefore, the specific limitations of one or more embodiments of the new energy vehicle charging safety monitoring system provided below can be found in the limitations of the new energy vehicle charging safety monitoring method described above, and will not be repeated here.

[0085] In one exemplary embodiment, such as Figure 2 As shown, a new energy vehicle charging safety monitoring system 200 is provided, including: The data acquisition module 201 is used to acquire battery data of the vehicle and charging pile data; the battery data of the vehicle includes power battery voltage, current and temperature parameters; the charging pile data includes charging pile output voltage, charging pile current, charging pile temperature and insulation resistance value. The data standardization processing module 202 is used to standardize the battery data of the vehicle and the charging pile data to obtain standardized vehicle battery data and standardized charging pile data. The vehicle-charging pile data fusion module 203 is used to fuse standardized vehicle battery data and standardized charging pile data to obtain vehicle-charging pile fusion data; The feature parameter extraction module 204 is used to extract features from the vehicle-charging pile fusion data to obtain vehicle battery feature parameters and charging pile feature parameters; The battery voltage prediction module 205 is used to input the characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile into the preset power battery voltage prediction model to obtain the predicted value of the power battery voltage. The power battery voltage prediction model includes a convolutional layer, a bidirectional gate control loop unit, an attention mechanism layer and a fully connected layer. The fault diagnosis module 206 is used to perform fault diagnosis on the vehicle's power battery based on the predicted value of the power battery voltage and the power battery voltage, and obtain the fault diagnosis result. The safety evaluation module 207 is used to comprehensively evaluate the charging safety risks of the vehicle based on the fault diagnosis results, and obtain the safety risk evaluation results, which include a safety risk evaluation score.

[0086] Furthermore, the battery voltage prediction module 205 is also used for: The characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile are concatenated to obtain a joint feature vector; the joint feature vector is used as the input of the power battery voltage prediction model. Local features are extracted from the joint feature vector through convolutional layers, and a convolutional feature map is output. Temporal features are extracted from the convolutional feature map by a bidirectional gate-controlled recurrent unit, resulting in a temporal feature vector. The temporal feature vector includes a forward hidden state and a backward hidden state. The temporal feature vector is weighted by an attention mechanism layer to obtain a weighted feature vector; The weighted feature vector is input into the fully connected layer to obtain the predicted value of the power battery voltage.

[0087] Furthermore, the vehicle-charging station data fusion module 203 is also used for: Obtain vehicle charging order data; charging order data includes vehicle identification number, order creation time, charging start time, and charging end time; Based on the vehicle identification number and order creation time, standardized vehicle battery data is matched with charging order data to obtain preliminary matching data; Based on the charging start time and charging end time, the preliminary matching data is matched by time window to obtain time window matching data; Data alignment and stitching are performed on the time window matching data to obtain preliminary vehicle-pile fusion data; The initial vehicle-pile fusion data is denoised to obtain the vehicle-pile fusion data.

[0088] Furthermore, the feature parameter extraction module 204 is also used for: The vehicle-charging station fusion data is segmented using a sliding window to obtain multiple charging segment data. Feature parameters are extracted from the data of each charging segment to obtain the characteristic parameters of the vehicle battery and the charging pile. The characteristic parameters of the vehicle battery include the mean voltage, the variance of the current, and the temperature slope. The characteristic parameters of the charging pile include the output voltage fluctuation, the stability of the output current, and the rate of change of insulation resistance.

[0089] Furthermore, the fault diagnosis module 206 is also used for: The predicted value of the power battery voltage and the residual value of the power battery voltage are calculated using the following formula to obtain the residual sequence: ; in, For the first The residual value at each time step. For the first The power battery voltage at each time step For the first Predicted power battery voltage at each time step; The residual mean and residual standard deviation of the residual series can be obtained using the following formulas: ; in, The mean of the residuals, The standard deviation of the residuals. For window size, For the first in the window One residual value; Based on the residual mean and residual standard deviation, the upper limit and lower limit of the dynamic threshold are obtained using the following formula: ; in, This is the upper limit of the dynamic threshold. This is the lower limit of the dynamic threshold. The mean of the residuals, The standard deviation of the residuals; The dynamic threshold range is obtained based on the upper limit and lower limit of the dynamic threshold. When the residual value exceeds the dynamic threshold range, a fault anomaly result is obtained; the fault anomaly result includes the abnormal time point and the anomaly type. Based on the abnormal fault results, fault diagnosis results are obtained; the fault diagnosis results include fault type, fault occurrence frequency and fault duration.

[0090] Furthermore, the fault diagnosis module 206 is also used for: Cluster analysis is performed on abnormal time points in the fault and abnormal results to obtain abnormal event clusters; Feature extraction is performed on the cluster of abnormal events to obtain fault frequency features and fault duration; the fault frequency features are used to characterize the number of times a fault occurs within a preset time period; the fault duration is used to characterize the time period during which the abnormal event occurs. Based on the fault frequency characteristics and fault duration, the fault type classification results are obtained; the fault type classification results are used to indicate the type of fault. Based on the fault type classification results, the fault diagnosis results are obtained.

[0091] Furthermore, the safety evaluation module 207 is also used for: Based on the fault diagnosis results, risk assessment indicators are obtained; the risk assessment indicators include vehicle battery characteristic parameters and charging pile characteristic parameters. Use the following formula to calculate the overall weight of each risk assessment indicator: ; in, For the first The comprehensive weight of each risk assessment indicator, To determine the matrix element, the first element is represented as the first element. The risk assessment indicator is relative to the first The importance of each risk assessment indicator This represents the total number of risk assessment indicators. To determine the order of a matrix; The overall weights of each risk assessment indicator are adjusted to obtain the adjusted overall weights of each risk assessment indicator. Based on the revised comprehensive weights of each risk assessment indicator and the individual risk assessment indicators, the safety risk assessment score is obtained using the following formula: ; in, As a safety risk assessment score, For the revised first The comprehensive weight of each risk assessment indicator, For the first The quantitative value of each risk assessment indicator. This represents the total number of risk assessment indicators. Based on the safety risk assessment score, a safety risk level is obtained; the safety risk level includes the first safety level, the second safety level, and the third safety level. Based on the safety risk level, a safety risk assessment result is obtained; the safety risk assessment result is used to indicate the safety risk of the power battery when the vehicle is charging.

[0092] In one embodiment, such as Figure 3 A computer device 300 is provided, comprising: At least one processor 301, and at least one memory 302 communicatively connected to the processor 301; the memory stores application code executable by at least one processor, the application code being executed by at least one processor to enable at least one processor to perform the steps of the new energy vehicle charging safety monitoring method as described above; The computer device may also include: sensor 303; The processor 301, memory 302, and sensor 303 can be connected via bus 304 or other means. The figure shows an example of connection via bus 304. Figure 3 The character is represented by a single thick line, but this does not mean that there is only one bus or a type of bus.

[0093] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0094] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0095] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for monitoring the charging safety of new energy vehicles, characterized in that, The method includes: Acquire vehicle battery data and charging pile data; the vehicle battery data includes power battery voltage, current and temperature parameters; the charging pile data includes charging pile output voltage, charging pile current, charging pile temperature and insulation resistance value; The battery data of the vehicle and the charging pile data are standardized to obtain standardized vehicle battery data and standardized charging pile data. The standardized vehicle battery data and the standardized charging pile data are fused to obtain vehicle-charging pile fused data; Feature extraction is performed on the vehicle-charging pile fusion data to obtain vehicle battery feature parameters and charging pile feature parameters; The characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile are input into a preset power battery voltage prediction model to obtain the predicted value of the power battery voltage; the power battery voltage prediction model includes a convolutional layer, a bidirectional gate control loop unit, an attention mechanism layer and a fully connected layer; Based on the predicted value of the power battery voltage and the power battery voltage, the power battery of the vehicle is diagnosed for fault, and the fault diagnosis result is obtained. Based on the fault diagnosis results, a comprehensive evaluation of the charging safety risks of the vehicle is conducted to obtain a safety risk evaluation result; the safety risk evaluation result includes a safety risk evaluation score.

2. The method according to claim 1, characterized in that, The step of inputting the characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile into a preset power battery voltage prediction model to obtain the predicted power battery voltage value includes: The characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile are concatenated to obtain a joint feature vector; the joint feature vector is used as the input of the power battery voltage prediction model. The convolutional layer extracts local features from the joint feature vector and outputs a convolutional feature map. Temporal features are extracted from the convolutional feature map by the bidirectional gate-controlled loop unit to obtain a temporal feature vector; the temporal feature vector includes a forward hidden state and a backward hidden state; The attention mechanism layer is used to weight the temporal feature vector to obtain a weighted feature vector; The weighted feature vector is input into the fully connected layer to obtain the predicted value of the power battery voltage.

3. The method according to claim 1, characterized in that, The process of fusing the standardized vehicle battery data and the standardized charging pile data to obtain vehicle-charging pile fused data includes: Obtain the charging order data of the vehicle; the charging order data includes the vehicle identification number, order creation time, charging start time, and charging end time; Based on the vehicle identification number and the order creation time, the standardized vehicle battery data is matched with the charging order data to obtain preliminary matching data; Based on the charging start time and the charging end time, the preliminary matching data is subjected to time window matching to obtain time window matching data; The time window matching data is aligned and stitched together to obtain preliminary vehicle-pile fusion data; The preliminary vehicle-pile fusion data is denoised to obtain the vehicle-pile fusion data.

4. The method according to claim 1, characterized in that, The feature extraction of the vehicle-charging pile fusion data yields vehicle battery feature parameters and charging pile feature parameters, including: The vehicle-charging station fusion data is segmented using a sliding window to obtain multiple charging segment data. Feature parameters are extracted from the data of each charging segment to obtain the characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile; the characteristic parameters of the vehicle battery include the average voltage, the variance of the current, and the temperature slope; the characteristic parameters of the charging pile include the output voltage fluctuation, the stability of the output current, and the rate of change of insulation resistance.

5. The method according to claim 1, characterized in that, The step of performing fault diagnosis on the vehicle's power battery based on the predicted power battery voltage and the power battery voltage, and obtaining fault diagnosis results, includes: The predicted value of the power battery voltage and the residual value of the power battery voltage are calculated using the following formula to obtain the residual sequence: ； in, For the first The residual value at each time step. For the first The voltage of the power battery at each time step, For the first The predicted value of the power battery voltage at each time step; The residual mean and residual standard deviation of the residual sequence are obtained using the following formulas: ； in, The mean of the residuals, The standard deviation of the residual is... For window size, For the first in the window One residual value; Based on the mean and standard deviation of the residuals, the upper limit and lower limit of the dynamic threshold are obtained using the following formula: ； in, This is the upper limit of the dynamic threshold. This is the lower limit of the dynamic threshold. The mean of the residuals, The standard deviation of the residual; The dynamic threshold range is obtained based on the upper limit and the lower limit of the dynamic threshold. When the residual value exceeds the dynamic threshold range, a fault anomaly result is obtained; the fault anomaly result includes the anomaly time point and the anomaly type; Based on the abnormal fault results, a fault diagnosis result is obtained; the fault diagnosis result includes the fault type, fault occurrence frequency, and fault duration.

6. The method according to claim 5, characterized in that, The process of obtaining fault diagnosis results based on the fault anomaly results includes: Cluster analysis is performed on the abnormal time points in the fault anomaly results to obtain abnormal event clusters; Feature extraction is performed on the abnormal event cluster to obtain fault frequency features and fault duration; the fault frequency features are used to characterize the number of times the fault occurs within a preset time period; the fault duration is used to characterize the time period during which the abnormal event occurs. Based on the fault frequency characteristics and the fault duration, a fault type classification result is obtained; the fault type classification result is used to indicate the type of fault. The fault diagnosis result is obtained based on the fault type classification result.

7. The method according to claim 1, characterized in that, Based on the fault diagnosis results, a comprehensive evaluation of the charging safety risks of the vehicle is conducted to obtain a safety risk evaluation result, including: Based on the fault diagnosis results, risk assessment indicators are obtained; the risk assessment indicators include the characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile. Calculate the overall weight of each of the aforementioned risk assessment indicators using the following formula: ； in, For the first The comprehensive weight of each risk assessment indicator, To determine the matrix element, the first element is represented as the first element. The risk assessment indicator is relative to the first The importance of each risk assessment indicator This represents the total number of risk assessment indicators. To determine the order of a matrix; The comprehensive weights of each of the aforementioned risk assessment indicators are adjusted to obtain the adjusted comprehensive weights of each of the aforementioned risk assessment indicators. Based on the revised comprehensive weights of each of the aforementioned risk assessment indicators and the individual risk assessment indicators, the safety risk assessment score is obtained using the following formula: ； in, The safety risk assessment score is... For the revised first The comprehensive weight of each risk assessment indicator, For the first The quantitative value of each risk assessment indicator. This represents the total number of risk assessment indicators. Based on the security risk assessment score, a security risk level is obtained; the security risk level includes a first security level, a second security level, and a third security level. Based on the safety risk level, a safety risk assessment result is obtained; the safety risk assessment result is used to indicate the safety risk of the power battery of the vehicle during charging.

8. A new energy vehicle charging safety monitoring system, characterized in that, The system includes: The data acquisition module is used to acquire battery data of the vehicle and charging pile data; the battery data of the vehicle includes power battery voltage, current and temperature parameters; the charging pile data includes charging pile output voltage, charging pile current, charging pile temperature and insulation resistance value. The data standardization processing module is used to standardize the battery data of the vehicle and the charging pile data to obtain standardized vehicle battery data and standardized charging pile data. The vehicle-charging pile data fusion module is used to fuse the standardized vehicle battery data and the standardized charging pile data to obtain vehicle-charging pile fused data; The feature parameter extraction module is used to extract features from the vehicle-charging pile fusion data to obtain vehicle battery feature parameters and charging pile feature parameters. The battery voltage prediction module is used to input the characteristic parameters of the vehicle battery and the characteristic parameters of the charging pile into a preset power battery voltage prediction model to obtain the predicted value of the power battery voltage; the power battery voltage prediction model includes a convolutional layer, a bidirectional gate control loop unit, an attention mechanism layer and a fully connected layer; The fault diagnosis module is used to perform fault diagnosis on the power battery of the vehicle based on the predicted value of the power battery voltage and the power battery voltage, and obtain the fault diagnosis result. The safety evaluation module is used to comprehensively evaluate the charging safety risks of the vehicle based on the fault diagnosis results, and obtain a safety risk evaluation result; the safety risk evaluation result includes a safety risk evaluation score.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images