Power battery abnormality identification method based on greedy-relaxation hyperparameter optimization

Abstract

The application discloses a battery abnormality identification method based on a greedy-relaxation hyperparameter optimization, which comprises the following steps: step 1, preprocessing vehicle operation data; step 2, selecting data with strong correlation as input items of a model; step 3, determining a hyperparameter category to be optimized, and adopting a greedy-relaxation hyperparameter optimization method to optimize the hyperparameters and obtain comprehensive optimal hyperparameters; step 4, applying the comprehensive optimal hyperparameters to a neural network prediction model to obtain an optimized neural network prediction model; inputting the data with strong correlation into the optimized neural network prediction model to obtain output values, i.e., predicted values; and step 5, calculating the residual error of the predicted values and actual values of the neural network prediction model, and regarding the abnormality when the difference between the actual values and the predicted values is greater than an abnormality threshold value, otherwise, regarding the normality. The application seeks a balance between model performance and calculation efficiency, realizes the rapid and accurate construction of a battery abnormality identification neural network model, and improves the efficiency of the model.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power battery safety, and relates to a battery abnormality identification method, in particular to a power battery abnormality identification method based on greedy-relaxation hyperparameter optimization. BACKGROUND

[0002] In recent years, power battery technology has developed rapidly with the rapid growth of the number of electric vehicles. The power battery not only provides energy supply for the vehicle, but also relates to the safety of the vehicle. Before serious failure, abnormal changes will occur in battery parameters such as voltage and temperature, so accurate identification of battery abnormalities is of great significance to ensure the safe operation of the vehicle.

[0003] Neural network models have been widely used in power battery anomaly detection, and hyperparameters have a very important influence on neural networks, to some extent, affecting the robustness, stability and generalization ability of the network. Therefore, selecting appropriate hyperparameters is of great significance to improve the performance of the network.

[0004] The current commonly used hyperparameter optimization algorithm cannot balance the optimization time and optimization performance, resulting in certain instability of the neural network model in large-scale application. With the continuous and rapid growth of the number of electric vehicles, how to obtain a hyperparameter combination with better accuracy in a short time is of great significance to the use of the neural network model under the big data framework. SUMMARY

[0005] The purpose of the application is to provide a battery abnormality identification method to solve the problem of hyperparameter optimization in battery abnormality identification of neural network models under the big data framework, and to realize fast optimization of the hyperparameters of the battery abnormality identification model. The method seeks a balance between model performance and computational efficiency based on vehicle cloud platform monitoring data, realizes the rapid and accurate construction of a battery abnormality identification neural network model, and thus improves the efficiency of the model.

[0006] In order to achieve the above purpose, the application adopts the following technical solutions to realize it:

[0007] A battery abnormality identification method based on greedy-relaxation hyperparameter optimization, specifically comprising the following steps:

[0008] Step 1: Preprocessing the vehicle operation data to obtain preprocessed vehicle operation data;

[0009] Step 2: Correlation analysis of different feature items and output features, and selection of data with strong correlation as input items of the model;

[0010] Step 3, determine the hyperparameter categories that need to be optimized, use the greedy-relaxation hyperparameter optimization method to optimize the hyperparameters, and obtain the comprehensive optimal hyperparameters; the hyperparameters that need to be optimized include four categories: learning rate lr, batch size batch_size, number of neural network layers, and number of neurons; the loss function needs to be selected as the evaluation index; the specific steps are as follows:

[0011] Step 31, hyperparameter initialization: determine the multiple hyperparameter categories that need to be optimized according to the task requirements, list the possible hyperparameter values for each hyperparameter category, and randomly select a set as the initial hyperparameters;

[0012] Step 32, hyperparameter pre-adjustment: for each hyperparameter category in the initial hyperparameters, iterate through all possible values while keeping the values of other categories unchanged, train the model on this combination of hyperparameters and calculate the loss obtained in each iteration, and obtain the loss of the model for each hyperparameter category at all possible values;

[0013] Step 33, hyperparameter category importance evaluation: calculate the range of the loss of the model for the same hyperparameter category at all possible values, which represents the importance of the hyperparameter category;

[0014] Step 34, search for local optimal hyperparameters: based on the initial hyperparameters obtained in step 31, first optimize the hyperparameter category with the highest importance obtained in step 33;

[0015] Step 35, search for comprehensive optimal hyperparameters: based on the local optimal candidate hyperparameters of the current hyperparameter category, fix the parameter values of the optimized parameter categories unchanged, and search for the local optimal hyperparameters of the most important category among the remaining hyperparameter categories; when the optimization of the least important hyperparameter category is completed, select the hyperparameters with the smallest loss as the comprehensive optimal hyperparameters;

[0016] Step 4, apply the comprehensive optimal hyperparameters to the neural network prediction model to obtain the optimized neural network prediction model; input the data with strong correlation obtained in step 2 into the optimized neural network prediction model to obtain the output value, i.e. the predicted value; the model predicted parameter is identified as the normal parameter of the battery;

[0017] Step 5, calculate the residual error between the predicted value and the true value of the neural network prediction model obtained in step 4; when the difference between the true value and the predicted value is greater than the abnormal threshold, it is identified as abnormal, otherwise it is normal.

[0018] Further, in step 1, the preprocessing principles are as follows:

[0019] 1) Sort the vehicle operation data by time and perform de-duplication operation;

[0020] 2) If the current data frame has null values, the average of the two frames of data adjacent to the current data frame is used as the value of the current frame data;

[0021] 3) Z-score method is used to realize data standardization.

[0022] Further, in step 2, the Pearson correlation coefficient is used to evaluate the correlation between the data.

[0023] Further, in step 5, the box plot method is used to set the abnormal threshold, and the calculation formula is as follows:

[0024] Thre=Q3+1.5*IQR

[0025] In the formula, Q3 is the third quartile of the residual of the predicted value and the true value, and IQR is the interquartile range.

[0026] Compared with the prior art, the present application focuses on the power battery abnormality identification method of electric vehicles, and based on the real vehicle operation big data on the electric vehicles, the power battery abnormality identification method based on the greedy-relaxation hyperparameter optimization is proposed for the instability of the model efficiency and performance caused by the selection of hyperparameters in the neural network model in the vehicle big data. The greediness ensures that the search of the hyperparameters is in the direction of loss reduction, and the relaxation reduces the probability of falling into local optimum. The difference between the model predicted value and the true value is used for battery abnormality identification, which can provide early warning for vehicles with potential fault risk and heat runaway, thereby improving the driving safety. The method of the present application is based on the vehicle cloud platform monitoring data, seeks the balance between the model performance and the calculation efficiency, realizes the rapid and accurate construction of the battery abnormality identification neural network model, and thus improves the efficiency of the model. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is the method flowchart of the present application; DETAILED DESCRIPTION

[0028] The present application will be described in detail below in combination with the drawings and specific embodiments.

[0029] As Figure 1 shown, the battery abnormality identification method based on the greedy-relaxation hyperparameter optimization provided by the present application specifically includes the following steps:

[0030] Step 1, the vehicle operation data is preprocessed to obtain the preprocessed vehicle operation data to ensure that the model needs are met.

[0031] In this embodiment, according to the historical data of the electric vehicle uploaded according to the national standard GB / T 32960 "Technical Specification for Electric Vehicle Remote Service and Management System", the data needs to be preprocessed to meet the needs of the model, and the specific preprocessing principles are as follows:

[0032] 1) The vehicle operation data is sorted by time and de-duplicated to overcome the problems of vehicle operation data time disorder and data duplication.

[0033] 2) If the current data frame has a null value, the average of the two adjacent data frames of the current data frame is used as the value of the current frame data;

[0034] 3) Since different data items in the national standard GB / T 32960 have different effective ranges, the Z-score method is used to realize data standardization in this embodiment to ensure that each feature has similar scale and range, and the calculation formula is as follows:

[0035]

[0036] In the formula, x is the data to be standardized, μ is the average value of the data corresponding to the feature item of x, σ is the standard deviation of all data under the data item corresponding to x, and Z is the standardized score.

[0037] Step 2, in order to obtain better model performance, correlation analysis needs to be performed on different feature items and output features. Strongly correlated data is selected as the input item of the model. In this embodiment, the Pearson correlation coefficient is used to evaluate the correlation between data. Strong correlation represents greater contribution to model performance. Generally, when the correlation coefficient reaches 0.6 or more, it is considered to be strongly correlated and can be used as the input item of the model.

[0038] The calculation formula of the Pearson correlation coefficient is as follows:

[0039]

[0040] In the formula, n is the number of samples included in the measured data item, x i and y i represent the ith value of the measured data item and the target data item, respectively, and are the average values of the model candidate input item and the target feature item data, respectively, and coef is the Pearson correlation coefficient.

[0041] Step 3, determine the type of hyperparameters to be optimized, use the greedy-relaxation hyperparameter optimization method to optimize the hyperparameters, and get the comprehensive optimal hyperparameters. Specifically, the hyperparameters to be optimized include four categories: learning rate lr, batch size batch_size, number of neural network layers, and number of neurons. At the same time, in order to evaluate the performance of the model, the loss function needs to be selected as the evaluation index, and the specific steps are as follows:

[0042] Step 31, hyperparameter initialization: determine the categories of hyperparameters that need to be optimized according to the task requirements, list the possible values of each category of hyperparameters, and randomly select a set as the initial hyperparameters;

[0043] Step 32, hyperparameter pre-adjustment: for each category of hyperparameters in the initial hyperparameters, iterate through all possible values while keeping the values of hyperparameters in other categories unchanged, train the model on this combination of hyperparameters and calculate the loss (i.e. the value of the loss function) obtained in each iteration, and obtain the loss of the model for each category of hyperparameters at all possible values.

[0044] Step 33, hyperparameter category importance evaluation: calculate the range (difference between the maximum and minimum values of the loss) of the loss of the model for the same category of hyperparameters at all possible values, which is used to represent the importance of this category of hyperparameters. The larger the range, the more significant the impact of this category of hyperparameters on the performance of the model, and the higher the importance of this category of hyperparameters.

[0045] According to the above method, the range of all hyperparameter categories is calculated to quantitatively evaluate the importance of the hyperparameter categories.

[0046] Step 34, search for local optimal hyperparameters: based on the initial hyperparameters obtained in step 31, first optimize the hyperparameter category with the highest importance (i.e. the largest range) obtained in step 33. Specifically, keep the remaining hyperparameter categories unchanged except the optimized hyperparameter category, iterate through all possible values of the optimized hyperparameter category, and obtain the model loss for different combinations of hyperparameters in this category. Assuming there are n different hyperparameter values in the current hyperparameter category (i.e. the optimized category), keep the round(n / ln(n)) corresponding hyperparameter values with the smallest loss as the local optimal candidate hyperparameters for the current hyperparameter category, where round(·) represents rounding and n represents the number of samples.

[0047] Step 35, search for comprehensive optimal hyperparameters:

[0048] According to the local optimal candidate hyperparameters of the current hyperparameter category, fix the parameter values in the optimized parameter category unchanged, search for the local optimal hyperparameters in the most important category among the remaining hyperparameter categories; when the optimization of the least important hyperparameter category is completed, select the hyperparameter with the smallest loss as the comprehensive optimal hyperparameter.

[0049] Step 4, apply the comprehensive optimal hyperparameters to the neural network prediction model to obtain the optimized neural network prediction model; input the data with strong correlation obtained in step 2 into the optimized neural network prediction model to obtain the output value, i.e. the predicted value.

[0050] Since the prediction result of the neural network prediction model tends to be inclined to the majority class sample, and in the vehicle operation data, the normal data accounts for the vast majority, therefore the model prediction parameter can be identified as the normal parameter of the battery.

[0051] Step 5, calculate the residual of the prediction value and the true value of the neural network prediction model obtained in step 4, when the difference between the true value and the prediction value is greater than the abnormal threshold, it is identified as abnormal, otherwise it is normal.

[0052] Since the model prediction parameter is identified as the normal parameter of the battery, when the true value and the prediction value produce a large deviation, it is undoubtedly an abnormal performance. In order to accurately distinguish the normal and abnormal battery parameters, based on the vehicle historical data, the box plot method is used to set the abnormal threshold, and the calculation formula is as follows:

[0053] Thre=Q3+1.5*IQR

[0054] In the formula, Q3 is the third quartile of the residual of the prediction value and the true value, and IQR is the interquartile range.

[0055] With the increasing number of electric vehicles and the gradual maturity of cloud computing technology, it has gradually become a mainstream trend to monitor the safety of electric vehicles through the cloud platform. In view of the important role of the neural network model in cloud computing, the present application proposes a greedy-relaxation hyperparameter optimization method, which effectively balances the efficiency and accuracy of the model. At the same time, according to the characteristics of vehicle big data, a power battery abnormality identification method based on greedy-relaxation hyperparameter optimization is proposed, which accurately predicts the normal parameters of the battery to detect battery abnormalities and provides support for real-time safety risk monitoring of running vehicles.

Claims

1. A battery anomaly identification method based on greedy-relaxation hyperparameter optimization, characterized in that, Specifically, the steps include the following: Step 1: Preprocess the vehicle operation data to obtain preprocessed vehicle operation data; Step 2: Perform correlation analysis on different feature items and output features, and select data with strong correlation as input items for the model; Step 3: Determine the categories of hyperparameters to be optimized, and use the greedy-relaxation hyperparameter optimization method to optimize the hyperparameters to obtain the overall optimal hyperparameters. The hyperparameters to be optimized include four categories: learning rate (lr), batch size (batch_size), number of neural network layers, and number of neurons. The loss function needs to be selected as the evaluation metric. The specific steps are as follows: Step 31, Hyperparameter initialization: Determine the multiple categories of hyperparameters that need to be optimized according to the task requirements, list the possible hyperparameter values ​​for each category, and randomly select a set as the initial hyperparameters; Step 32, Hyperparameter pre-tuning: For each hyperparameter category in the initial hyperparameters, iterate through all possible values ​​while keeping the hyperparameter values ​​of other categories unchanged. Train the model on such hyperparameter combinations and calculate the loss obtained in each iteration to obtain the model loss for each hyperparameter category under all possible values. Step 33, Hyperparameter category importance assessment: Calculate the range of the model loss for the same hyperparameter category under all possible values, which is used to characterize the importance of the hyperparameter category; Step 34, Search for local optimal hyperparameters: Based on the initial hyperparameters obtained in step 31, first optimize the hyperparameter category with the highest importance obtained in step 33; Step 35, Search for the overall optimal hyperparameter: Based on the local optimal candidate hyperparameters of the current hyperparameter category, keep the parameter values ​​of the optimized parameter categories unchanged, and search for the local optimal hyperparameter of the most important category among the remaining hyperparameter categories; when the hyperparameter category with the lowest importance is optimized, select the hyperparameter with the smallest loss as the overall optimal hyperparameter. Step 4: Apply the comprehensive optimal hyperparameters to the neural network prediction model to obtain the optimized neural network prediction model; input the highly correlated data obtained in Step 2 into the optimized neural network prediction model to obtain the output value, i.e., the predicted value; the model prediction parameters are identified as the normal parameters of the battery. Step 5: Calculate the residual between the predicted value and the actual value of the neural network prediction model obtained in Step 4. If the difference between the actual value and the predicted value is greater than the abnormal threshold, it is considered abnormal; otherwise, it is considered normal.

2. The battery anomaly identification method based on greedy-relaxation hyperparameter optimization as described in claim 1, characterized in that, In step 1, the preprocessing principles are as follows: 1) Sort vehicle operation data by time and perform redundancy removal; 2) If the current data frame contains null values, the average of the two adjacent data frames is used as the value of the current data frame. 3) Use the Z-score method to achieve data standardization.

3. The battery anomaly identification method based on greedy-relaxation hyperparameter optimization as described in claim 1, characterized in that, In step 2, the Pearson correlation coefficient is used to evaluate the correlation between the data.

4. The battery anomaly identification method based on greedy-relaxation hyperparameter optimization as described in claim 1, characterized in that, In step 5, the box plot method is used to set the anomaly threshold, and the calculation formula is as follows: Thre = Q³ + 1.5 * IQR In the formula, Q3 is the third quartile of the residual between the predicted and actual values, and IQR is the interquartile range.

Images