Charging state monitoring method and system of electric vehicle, electronic equipment and storage medium

Abstract

The invention discloses a charging state monitoring method and system of an electric vehicle, electronic equipment and a storage medium, and the method comprises the steps: building a charging state probabilistic prediction model, and quantifying the prediction uncertainty of the model in real time; according to the method and the system, the charging state is subjected to fault monitoring by adopting a dual verification mechanism fusing prediction deviation and quantitative uncertainty, so that the charging state can be accurately predicted, and an alarm is given only when the model has high confidence coefficient for abnormal prediction through the uncertainty of quantitative prediction, so that a real charging fault can be accurately identified, and the reliability of the system is improved. False alarms caused by data noise or rare working conditions are greatly inhibited, and more reliable and robust charging state abnormity early warning is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric vehicle charging, in particular, to a charging state monitoring method and system for electric vehicles, an electronic device and a computer readable storage medium. BACKGROUND

[0002] With the wide application of electric vehicles, the construction and use of charging facilities are increasing, and how to ensure the efficiency, safety and reliability of the electric vehicle charging process has become a problem to be solved. The charging state (also known as state of charge, SOC for short) in the electric vehicle charging process is a key indicator for measuring the remaining battery capacity. Accurate SOC monitoring is crucial for optimizing charging efficiency, prolonging battery life and ensuring the normal operation of charging facilities. At present, the traditional SOC monitoring method is to monitor the voltage, current and temperature of the battery through the sensors inside the battery, and then estimate the SOC based on the above physical quantities by the battery management system. However, this method based on real-time monitoring data of sensors has a lag, which is difficult to meet the real-time requirements of SOC monitoring. Moreover, due to the effects of battery aging, environmental changes, measurement noise, etc., the monitoring accuracy of the traditional method is poor. In order to improve the real-time performance of the charging state monitoring, the existing technology applies deep learning technology to SOC monitoring. For example, a deep learning model (such as an LSTM model) is used to predict the charging data of the electric vehicle, and based on the predicted charging data, charging fault diagnosis and identification are performed, and the SOC can also be estimated in real time based on the predicted charging data, which can effectively solve the problem of lag. However, since the traditional deep learning model is a deterministic model, it cannot distinguish between real charging fault signs and prediction deviations caused by its own cognitive limitations, and there is a false alarm caused by data noise or rare working conditions. In addition, due to the lack of training data, the prediction accuracy of the model is poor, which leads to poor accuracy of the charging state fault monitoring. SUMMARY

[0003] The present application provides a low-voltage area electric energy meter time correction method and system based on voltage transition events, an electronic device and a computer readable storage medium, which can accurately identify real charging faults, greatly suppress false alarms caused by data noise or rare working conditions, and realize more reliable and robust charging state anomaly early warning.

[0004] According to one aspect of the present application, a charging state monitoring method for an electric vehicle is provided, comprising the following contents: Collecting historical charging state data of the battery and constructing a sample data set; Constructing a charging state probabilistic prediction model for outputting predicted values of the charging state data and their quantitative uncertainties, and training the model with the sample data set until the model converges; Real-time measurement of the battery's current state of charge data is input into a trained probabilistic prediction model of the state of charge to obtain the predicted value of the state of charge data and its quantified uncertainty. Based on the prediction deviation and quantification uncertainty between the predicted value and the corresponding measured value of the charging status data, fault monitoring of the charging status is carried out.

[0005] Furthermore, the loss function of the probabilistic prediction model for the state of charging is: ; ; ; in, Indicates the overall loss. Indicates the predicted loss. Indicates the complexity loss. This represents the weight hyperparameters used to balance the two loss terms. Indicates the number of training samples. Indicates the first i Adaptive weight coefficients for each training sample. and They represent the first i The model's predicted values ​​and actual values ​​for each training sample. and This represents the gain hyperparameter. Indicates the first i The voltage values ​​of each training sample. This represents the incremental capacity function. Indicates the first i The incremental capacity corresponding to the voltage values ​​of each training sample. This represents the incremental capacity change function. .

[0006] Furthermore, the expression for the complexity loss is: ; in, M This represents the number of weight parameters in the probabilistic prediction model of the state of charge. and They represent the first i The mean and standard deviation of each weight parameter are obtained after learning.

[0007] Furthermore, the probabilistic prediction model for the state of charge includes a Bayesian convolutional network and a Bayesian bidirectional gated recurrent unit. The Bayesian convolutional network is used to extract a sequence of charging state features containing uncertainty from the charging state data, and the Bayesian bidirectional gated recurrent unit is used to make predictions based on the extracted charging state feature sequences, outputting the predicted value of the charging state data and its quantified uncertainty.

[0008] Furthermore, the convolution operation of the Bayesian convolutional network is represented as follows: ; ; ; in, This represents the result extracted after the convolution operation. t Constant charging status characteristics express t Real-time charging status data, Indicates weight, Indicates bias. and Let represent the mean and standard deviation of the weight distribution, respectively. and Let represent the mean and standard deviation of the biased distribution, respectively. and Let these represent the weighted random noise and the biased random noise sampled from the standard normal distribution, respectively. This represents the activation function for the convolution operation.

[0009] Furthermore, the hidden state update formula for the Bayesian bidirectional gated recurrent unit is: ; ; ; ; ; ; in, express t The hidden state of the Bayesian bidirectional gated loop unit at any given time. and They represent t The forward hidden state of the forward-gated loop unit and the backward hidden state of the backward-gated loop unit at any given time. and They represent t The forward hidden state of the forward-gated loop unit and the backward hidden state of the backward-gated loop unit at time -1. denotes an activation function, denotes t denotes a charging state feature at time and denote t denotes the output weight of the forward and backward gated recurrent unit at time denotes t denotes a bias at time and denotes the mean of the corresponding weight distribution, and denotes the standard deviation of the corresponding weight distribution, and denote the mean and standard deviation of the bias distribution, respectively, and denotes a weight random noise sampled from a standard normal distribution, denotes a bias random noise sampled from a standard normal distribution.

[0010] Further, when the prediction bias exceeds the prediction bias upper threshold and the quantified uncertainty is lower than the uncertainty upper threshold, it is determined that a high-confidence charging state anomaly occurs.

[0011] In addition, the present application also provides a charging state monitoring system of an electric vehicle, comprising: a sample data acquisition module, configured to acquire historical charging state data of a battery and construct a sample data set; a model training module, configured to construct a charging state probabilistic prediction model for outputting a predicted value of the charging state data and its quantified uncertainty, and train the model using the sample data set until the model converges; a model prediction module, configured to measure current charging state data of the battery in real time, input the data into the trained charging state probabilistic prediction model, and obtain the predicted value of the charging state data and its quantified uncertainty; a fault monitoring module, configured to perform charging state fault monitoring based on the prediction bias and quantified uncertainty between the predicted value and the corresponding measured value of the charging state data.

[0012] In addition, the present application also provides an electronic device comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to execute the steps of the method described above by invoking the computer program stored in the memory.

[0013] In addition, the present application also provides a computer-readable storage medium for storing a computer program for monitoring the charging state of an electric vehicle, wherein the computer program executes the steps of the method described above when running on a computer.

[0014] The application has the following beneficial effects: The electric vehicle charging state monitoring method of the application can quantize the uncertainty of the model's own prediction in real time by constructing a charging state probabilistic prediction model, and can perform charging state fault monitoring by using a double verification mechanism that fuses the prediction deviation and the quantized uncertainty. The method can not only accurately predict the charging state, but also can only issue an alarm when the model has a high confidence in abnormal prediction by quantizing the uncertainty of the prediction, so as to accurately identify the real charging fault and greatly inhibit the false alarm caused by data noise or rare working conditions, thereby realizing more reliable and robust charging state abnormal early warning.

[0015] In addition, the electric vehicle charging state monitoring system of the application also has the above advantages.

[0016] In addition to the purposes, features and advantages described above, the application has other purposes, features and advantages. The application will be further described below with reference to the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0017] The accompanying drawings, which form a part of the present application, are included to provide a further understanding of the application, and are incorporated herein for explanation by reference. In the drawings: Figure 1 FIG. 1 is a flowchart of an electric vehicle charging state monitoring method according to a preferred embodiment of the application; Figure 2 FIG. 2 is a schematic diagram of the module structure of an electric vehicle charging state monitoring system according to another embodiment of the application. DETAILED DESCRIPTION

[0018] It should be noted that the embodiments and features in the application can be combined with each other without conflict. The application will be described in detail below with reference to the drawings and in combination with the embodiments.

[0019] With reference to Figure 1 The preferred embodiment of the application provides an electric vehicle charging state monitoring method, which includes the following contents: Step S1: Collecting historical charging state data of the battery and constructing a sample data set; Step S2: Constructing a charging state probabilistic prediction model for outputting the predicted value of the charging state data and its quantized uncertainty, and training the model with the sample data set until the model converges; Step S3: Real-time measurement of the current charging state data of the battery, inputting the data into the trained charging state probabilistic prediction model to obtain the predicted value of the charging state data and its quantized uncertainty; Step S4: based on the prediction deviation between the predicted value and the corresponding measured value of the state of charge data and the quantified uncertainty, performing fault monitoring of the state of charge.

[0020] It can be understood that the state of charge monitoring method of the electric vehicle in the embodiment can quantize the uncertainty of the model itself prediction in real time by constructing a state of charge probabilistic prediction model, and perform fault monitoring of the state of charge by using a double verification mechanism that fuses the prediction deviation and the quantified uncertainty. Not only can the state of charge be accurately predicted, but also by quantizing the uncertainty of the prediction, an alarm is only sent when the model has a high confidence in the abnormal prediction, so that the real charging fault can be accurately identified, the false alarm caused by data noise or rare working conditions is greatly inhibited, and more reliable and robust charging state abnormal early warning is achieved.

[0021] In the step S1, the historical state of charge data of the battery is collected by the battery management system of the electric vehicle, wherein the state of charge data specifically includes charging voltage, charging current and charging temperature, etc., and a sample data set is constructed based on the historical state of charge data. In addition, the sample data set can be divided into a training set, a test set and a validation set according to a preset ratio, and the specific division ratio can be set according to actual needs, for example, the ratio is 7:2:1.

[0022] Optionally, after collecting the historical state of charge data, the missing data can be filled in, and then the data is normalized to obtain a data set with a value interval of 0 to 1. Specifically, an improved forward filling algorithm is used to fill in the missing values, and the calculation formula is: wherein, represents the missing value at time t, t , , , , , and represent the weights of the corresponding effective observation values, and the closer to the missing value, the greater the weight coefficient, so as to improve the accuracy of data filling. Then, the Min-Max normalization method is used to map the state of charge data to the interval [0, 1], and the mapping formula is: wherein, represents the normalized state of charge data, represents the original state of charge data at time t, and represent the maximum and minimum values of the original state of charge data, respectively.

[0023] In addition, in the step S2, the state-of-charge probabilistic prediction model comprises a Bayesian convolutional network and a Bayesian bidirectional gated recurrent unit, the Bayesian convolutional network is used to extract a state-of-charge feature sequence containing uncertainty from state-of-charge data, and the Bayesian bidirectional gated recurrent unit is used to make a prediction based on the extracted state-of-charge feature sequence and output a predicted value of the state-of-charge data and quantified uncertainty thereof.

[0024] In order to extract local features containing uncertainty in the state-of-charge data, the application introduces a Bayesian on the basis of a convolutional neural network, and specifically adopts a Bayesian improved one-dimensional convolutional neural network for feature extraction, each weight coefficient and a bias coefficient is described by an independent, learnable and optimized Gaussian distribution Before each forward propagation convolution operation is performed, the weight and bias are first sampled from the learned probability distribution, and the sampling process is usually realized through a reparameterization technique, and the calculation is as follows: 、 Then, the sampled weight and are used to perform convolution operation, and the calculation is as follows: , wherein t represents the state-of-charge feature at the moment t, represents the state-of-charge data at the moment t, t represents the convolution operation, represents the weight, represents the bias, and represent the mean and standard deviation of the weight distribution, respectively, and represent the mean and standard deviation of the bias distribution, respectively, and represent the weight random noise and bias random noise sampled from the standard normal distribution, respectively, represents the activation function of the convolution operation, which can be a SELU activation function, and is represented as: , and are constants.

[0025] In addition, after the state-of-charge feature sequence is extracted, a bidirectional network composed of a Bayesian improved bidirectional gated recurrent unit (BiGRU for short) is adopted, all weight matrices and bias terms used for calculating update gates, reset gates and candidate hidden states in the network are described by Gaussian distributions ​The description includes the following: The hidden state update formula for the Bayesian bidirectional gated recurrent unit is: ; ; ; ; ; ; in, express t The hidden state of the Bayesian bidirectional gated loop unit at any given time. and They represent t The forward hidden state of the forward-gated loop unit and the backward hidden state of the backward-gated loop unit at any given time. and They represent t The forward hidden state of the forward-gated loop unit and the backward hidden state of the backward-gated loop unit at time -1. The activation function can be represented by the tanh function. express t The charging state characteristics at any given time. and They represent t The output weights of the forward-gated loop unit and the backward-gated loop unit at each time step. express t Time offset, and This represents the mean of the corresponding weight distribution. and This represents the standard deviation of the corresponding weight distribution. and Let represent the mean and standard deviation of the biased distribution, respectively. and This represents weighted random noise sampled from a standard normal distribution. This represents biased random noise sampled from a standard normal distribution at each time step. t Before the operation, the internal weights of the GRU unit need to be randomly sampled from these probability distributions.

[0026] It is understood that the basic architecture of the probabilistic prediction model for charging state of the present invention adopts a combination of CNN and BiGRU, which has a powerful spatiotemporal feature extraction capability and can accurately capture the complex dynamic changes of key parameters such as voltage, current and temperature during the charging process. While ensuring high-precision prediction, it also introduces an uncertainty quantification mechanism through Bayesian method, which can quantify the uncertainty of the model's own prediction in real time, thereby improving the reliability and robustness of the prediction results.

[0027] Optionally, this invention also designs a loss function tailored to the characteristics of different stages of the charging process. By training the model, the distribution characteristics of normal data are learned, and by optimizing the loss function, the optimal parameters of all weight probability distributions in the network are learned, providing a regularization method that is more in line with Bayesian principles. The loss function of the probabilistic prediction model for the charging state is: ; ; ; in, Indicates the overall loss. Indicates the predicted loss. Indicates the complexity loss. This represents the weight hyperparameter used to balance the two loss terms, typically set to 0.2. Indicates the number of training samples. Indicates the first i Adaptive weight coefficients for each training sample. and They represent the first i The model's predicted values ​​and actual values ​​for each training sample. and This represents the gain hyperparameter. Indicates the first i The voltage values ​​of each training sample. This represents the incremental capacity function, obtained by performing incremental capacity analysis on a segment of the charging curve. Indicates the first i The incremental capacity corresponding to the voltage values ​​of each training sample. This represents the incremental capacity change function. .

[0028] It is understandable that traditional models predicting losses typically use mean squared error (MSE). MSE treats all time points equally and is insensitive to early, subtle anomalies. However, 90% of charging failures occur within the first 1-3 charging cycles, resulting in a small shift of 0.5%-2% in the peak region of the incremental capacity curve. Traditional MSE is slow to respond to such localized, subtle changes, making early warning difficult. Incremental capacity (…) The curve can amplify the subtle changes in the voltage platform region, and abnormal charging behavior will preferentially appear as peak position deviation, peak height change or new peak in these regions, therefore, the application introduces an adaptive weight coefficient based on incremental capacity analysis , so that the model can dynamically focus on the key charging stage with more fault diagnosis information during training according to the dynamic characteristics implied in the data, The peak region corresponds to the stage with the most intense electrochemical reaction inside the battery, reflecting the intensity of the electrochemical phase change process, the application fuses the information of the amplitude and slope of the incremental capacity change in two dimensions, and sets the adaptive weight coefficient deeply bound to the inherent physical dynamic process of the battery, while considering the intensity of the phase change reaction and the sensitivity of capturing dynamic changes, the adaptive weight mechanism has physical self-consistency and lifelong adaptability, for new batteries, the peak is sharp and concentrated, and the high value region of the adaptive weight coefficient is concentrated in a narrow voltage range, while with the aging of the battery, the peak gradually broadens and shifts, and the weight distribution automatically adjusts, without the need for recalibration or manual intervention, realizing an adaptive weight distribution mechanism that evolves synchronously with the battery health state, which can adapt to the aging and working condition changes of the battery, thereby improving the prediction accuracy of the model.

[0029] Optionally, the expression of the complexity loss is: ; wherein, M represents the number of weight parameters in the state-of-charge probabilistic prediction model, and respectively represent the mean and standard deviation of the first i weight parameter after learning.

[0030] It can be understood that the complexity loss term of the application gives the difference between the weight posterior distribution learned by the model and the standard normal prior distribution , which punishes the model complexity from the information theory perspective, when the learned distribution deviates from the standard normal distribution (i.e. is not 0 or is not 1), the loss value will increase, thereby effectively preventing the model from overfitting, especially when the available sample data set is small, it can still maintain high accuracy and stability, ensuring that the model has strong generalization ability.

[0031] Furthermore, after constructing the probabilistic prediction model for the state of charge, the model is trained using the training set until it converges, for example, when the overall loss is less than a preset threshold or the maximum number of iterations is reached. Then, the model's performance is tested using the test set. The root mean square error (RMSE) and mean absolute percentage error (MAPE) can be used to reflect the overall accuracy of the prediction data, and can be expressed as: , ,in, and They represent the first i The actual and predicted values ​​of each test sample n This indicates the number of samples in the test set. and These represent the root mean square error and the mean absolute percentage error, respectively. and The smaller the value, the more accurate the model's predictions. Finally, the model's performance is validated using a validation set. During performance validation, a judgment threshold is set based on the model's performance on the validation set, which can be expressed as: , This indicates the warning threshold for a certain type of rating (e.g., normal charging voltage). and Let these represent the mean and standard deviation of this type of rating on the validation set, respectively. k This represents the warning coefficient, which can be set according to actual needs. For example, when... k When the value is 3, it is called the 3-sigma rule. After inputting the validation set into the model, the absolute value of the difference between the predicted value and the actual value of all samples is calculated, and the mean of the deviation sequence is obtained. and standard deviation Therefore, the upper limit threshold for prediction deviation can be set as follows: Similarly, the mean of the uncertainty sequence is obtained by taking all the quantified uncertainties (i.e., variances) of the model output on the statistical validation set. and standard deviation Therefore, the upper limit threshold for uncertainty can be set as follows: The subsequent steps are based on the upper limit threshold of the prediction deviation. and uncertainty upper limit threshold To perform fault diagnosis of the charging status.

[0032] Furthermore, in step S3, after confirming that the model possesses high-precision predictive capabilities, it is deployed in a real-time monitoring task. The battery management system collects the battery's current state of charge data in real time and inputs it into the trained probabilistic prediction model for the state of charge to obtain predicted values ​​of the state of charge data and their quantified uncertainties. Specifically, prediction is made based on data within the current sliding window by processing the Bayesian network model...T a prediction set containing T different prediction values , based on which the final comprehensive prediction value at time t can be calculated as: , , wherein, represents the final comprehensive prediction value at time t, t represents the quantitative uncertainty of the current prediction obtained by calculating the variance at time t. t In addition, in the step S4, the prediction deviation between the final comprehensive prediction value obtained by the prediction in the step S3 and the actual measurement value at time t is calculated as:

[0033] If the prediction deviation exceeds the upper threshold value of the prediction deviation and the quantitative uncertainty is lower than the upper threshold value of the uncertainty, it means that the state of charge data deviates from the normal mode, and a high-confidence state of charge anomaly is determined to occur, and a warning signal is issued. In addition, the SOC can also be estimated in real time based on the prediction value of the state of charge data, wherein the specific estimation principle belongs to the prior art and will not be described here. t

[0034] In addition, in order to verify the effectiveness and superiority of the state of charge probabilistic prediction model of the present application, the present application also carries out a comparative test with the traditional neural network, and the root mean square error RMSE and the mean absolute percentage error MAPE are used as evaluation indexes, and the comparative test results are shown in Table 1.

[0035] Table 1, state of charge prediction results of different neural network models ; In addition, the present application also carries out a comparative test with the traditional neural network on the real data set collected at a new energy vehicle charging station, and the comparative test results are shown in Table 2.

[0036] Table 2, state of charge anomaly detection results of different neural network models ; It can be seen that the state of charge probabilistic prediction model of the present application shows lower error values in the prediction of charging current, charging voltage and temperature, and is far superior to the traditional LSTM, GRU and BiLSTM neural network models, effectively improving the prediction accuracy, robustness and reliability of the state of charge.

[0037] In addition, as Figure 2 ​​​​​​​​​As shown, another embodiment of the present application also provides a charging state monitoring system of an electric vehicle, preferably using the charging state monitoring method of an electric vehicle as described above, comprising: a sample data collection module for collecting historical charging state data of the battery and constructing a sample data set; a model training module for constructing a charging state probabilistic prediction model for outputting predicted values of the charging state data and their quantified uncertainty, and training the model using the sample data set until the model converges; a model prediction module for measuring the current charging state data of the battery in real time, inputting it into the trained charging state probabilistic prediction model, and obtaining the predicted values of the charging state data and their quantified uncertainty; a fault monitoring module for monitoring the charging state based on the prediction bias and quantified uncertainty between the predicted values and corresponding measured values of the charging state data.

[0038] It can be understood that the charging state monitoring system of the electric vehicle of the present embodiment can quantize the uncertainty of the model's own prediction in real time by constructing a charging state probabilistic prediction model, and can monitor the charging state based on the prediction bias and quantized uncertainty of the double verification mechanism, not only accurately predicting the charging state, but also quantizing the uncertainty of the prediction, only issuing an alarm when the model has a high confidence in abnormal prediction, so as to accurately identify real charging faults, greatly inhibit false alarms caused by data noise or rare working conditions, and achieve more reliable and robust charging state abnormal early warning.

[0039] In addition, another embodiment of the present application also provides an electronic device comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to execute the steps of the method as described above by invoking the computer program stored in the memory.

[0040] In addition, another embodiment of the present application also provides a computer-readable storage medium for storing a computer program for monitoring the charging state of an electric vehicle, wherein the computer program executes the steps of the method as described above when running on a computer.

[0041] The computer readable storage medium can be a machine-readable storage medium, including but not limited to diskette, floppy disk, hard disk, global positioning system, read-only memory, random access memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, flash memory, compact disc read-only memory, tape, cassette, cassette tape, magnetic tape, punch cards, paper tape, any other physical medium with patterns of holes, any other medium that can be used to store information suitable for use with a computer, or any suitable combination of the foregoing. The computer program product can also be a transmission medium that carries or communicates program code in the form of instructions or data arranged to be transferred, used, or processed by a machine, such as a computer, a processor, or an embedded processor, for example. Such a transmission medium can be for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or any other combination of the foregoing. Examples of a transmission medium include a copper wire, an optical fiber, a carrier wave, and any other medium suitable for storing or transmitting information.

[0042] Those skilled in the art will appreciate that embodiments of the present application can be readily used as a method, apparatus, or computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-readable storage media (including, but not limited to, disk memory, CD-ROMs, optical storage devices, etc.) embodying computer readable program code thereon. Embodiments of the present application can be implemented in various computer languages, including object-oriented programming languages such as Java and JavaScript.

[0043] The present application is described in reference to the flowchart illustrations and / or block diagrams according to the embodiments of the application. It should be understood that each flow and / or block in the flowchart illustrations and / or block diagrams, and combinations of flows and / or blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 The means for carrying out the functions specified in the flowchart illustrations and / or block diagrams can be embodied in one or more of the following: Figure 1 The means for carrying out the functions specified in the flowchart illustrations and / or block diagrams can be embodied in one or more of the following:

[0044] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks

[0045] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks

[0046] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the preferred embodiments by those of skill in the art once they have the benefit of the present disclosure without departing from the spirit and scope of the application. Accordingly, the appended claims are intended to encompass within their scope all such variations and modifications as are within the scope of the application.

[0047] It will be apparent to those skilled in the art that various modifications and variations can be made to the present application without departing from the spirit or scope of the application. Thus, it is intended that the present application cover modifications and variations of this application provided they come within the scope of the appended claims and their equivalents.

[0048] The preferred embodiments of the application described hereinabove are therefore to be considered in all respects as illustrative and not restrictive, and all changes coming within the meaning and equivalency range of the appended claims are therefore intended to be embraced therein.

[0045] The preferred embodiments of the application described hereinabove are therefore to be considered in all respects as illustrative and not restrictive, and all changes coming within the meaning and equivalency range of the appended claims are therefore intended to be embraced therein. Figure 1 Figure 1 The preferred embodiments of the application described hereinabove are therefore to be considered in all respects as illustrative and not restrictive, and all changes coming within the meaning and equivalency range of the appended claims are therefore intended to be embraced therein.

[0046] The preferred embodiments of the application described hereinabove are therefore to be considered in all respects as illustrative and not restrictive, and all changes coming within the meaning and equivalency range of the appended claims are therefore intended to be embraced therein.

[0047] The preferred embodiments of the application described hereinabove are therefore to be considered in all respects as illustrative and not restrictive, and all changes coming within the meaning and equivalency range of the appended claims are therefore intended to be embraced therein.

[0048] The preferred embodiments of the application described hereinabove are therefore to be considered in all respects as illustrative and not restrictive, and all changes coming within the meaning and equivalency range of the appended claims are therefore intended to be embraced therein.

Claims

1. A method for monitoring the charging status of an electric vehicle, characterized in that, Includes the following: Collect historical charging state data of the battery and construct a sample dataset; A probabilistic prediction model for the state of charge is constructed to output the predicted values ​​of the state of charge data and its quantified uncertainty. The model is then trained using a sample dataset until it converges. Real-time measurement of the battery's current state of charge data is input into a trained probabilistic prediction model of the state of charge to obtain the predicted value of the state of charge data and its quantified uncertainty. Based on the prediction deviation and quantification uncertainty between the predicted value and the corresponding measured value of the charging status data, fault monitoring of the charging status is carried out.

2. The method for monitoring the charging status of an electric vehicle as described in claim 1, characterized in that, The loss function of the probabilistic prediction model for the state of charge is: ； ； ； in, Indicates the overall loss. Indicates the predicted loss. Indicates the complexity loss. This represents the weight hyperparameters used to balance the two loss terms. Indicates the number of training samples. Indicates the first i Adaptive weight coefficients for each training sample. and They represent the first i The model's predicted values ​​and actual values ​​for each training sample. and This represents the gain hyperparameter. Indicates the first i The voltage values ​​of each training sample. This represents the incremental capacity function. Indicates the first i The incremental capacity corresponding to the voltage values ​​of each training sample. This represents the incremental capacity change function. .

3. The method for monitoring the charging status of an electric vehicle as described in claim 2, characterized in that, The expression for the complexity loss is: ； in, M This represents the number of weight parameters in the probabilistic prediction model of the state of charge. and They represent the first i The mean and standard deviation of each weight parameter are obtained after learning.

4. The method for monitoring the charging status of an electric vehicle as described in claim 1, characterized in that, The probabilistic prediction model for the state of charge includes a Bayesian convolutional network and a Bayesian bidirectional gated recurrent unit. The Bayesian convolutional network is used to extract a sequence of charging state features containing uncertainty from the charging state data. The Bayesian bidirectional gated recurrent unit is used to make predictions based on the extracted charging state feature sequences and output the predicted value of the charging state data and its quantified uncertainty.

5. The method for monitoring the charging status of an electric vehicle as described in claim 4, characterized in that, The convolution operation of the Bayesian convolutional network is represented as follows: ； ； ； in, This represents the result extracted after the convolution operation. t Constant charging status characteristics express t Real-time charging status data, Indicates weight, Indicates bias. and Let represent the mean and standard deviation of the weight distribution, respectively. and Let represent the mean and standard deviation of the biased distribution, respectively. and Let these represent the weighted random noise and the biased random noise sampled from the standard normal distribution, respectively. This represents the activation function for the convolution operation.

6. The method for monitoring the charging status of an electric vehicle as described in claim 4, characterized in that, The hidden state update formula for the Bayesian bidirectional gated recurrent unit is: ； ； ； ； ； ； in, express t The hidden state of the Bayesian bidirectional gated loop unit at any given time. and They represent t The forward hidden state of the forward-gated loop unit and the backward hidden state of the backward-gated loop unit at any given time. and They represent t The forward hidden state of the forward-gated loop unit and the backward hidden state of the backward-gated loop unit at time -1. This represents the activation function. express t The charging state characteristics at any given time. and They represent t The output weights of the forward-gated loop unit and the backward-gated loop unit at each time step. express t Time offset, and This represents the mean of the corresponding weight distribution. and This represents the standard deviation of the corresponding weight distribution. and Let represent the mean and standard deviation of the biased distribution, respectively. and This represents weighted random noise sampled from a standard normal distribution. This represents biased random noise sampled from a standard normal distribution.

7. The method for monitoring the charging status of an electric vehicle as described in claim 1, characterized in that, When the prediction deviation exceeds the upper limit threshold of prediction deviation and the quantification uncertainty is lower than the upper limit threshold of uncertainty, a high-confidence charging state anomaly is determined to have occurred.

8. A charging status monitoring system for electric vehicles, characterized in that, include: The sample data acquisition module is used to collect historical charging state data of the battery and build a sample dataset. The model training module is used to build a probabilistic prediction model for the state of charge, output the predicted value of the state of charge data and its quantified uncertainty, and train the model using the sample dataset until the model converges. The model prediction module is used to measure the current state of charge of the battery in real time, input it into the trained probabilistic prediction model of the state of charge, and obtain the predicted value of the state of charge data and its quantified uncertainty. The fault monitoring module is used to monitor charging status faults based on the prediction deviation and quantification uncertainty between the predicted value and the corresponding measured value of the charging status data.

9. An electronic device, characterized in that, The method includes a processor and a memory, wherein the memory stores a computer program, and the processor executes the steps of the method as described in any one of claims 1 to 7 by calling the computer program stored in the memory.

10. A computer-readable storage medium for storing a computer program for monitoring the charging status of an electric vehicle, characterized in that, The computer program, when run on a computer, performs the steps of the method as described in any one of claims 1 to 7.

Images