A battery thermal runaway early warning method based on a deep convolutional long short-term memory network

Abstract

The application discloses a battery thermal runaway early warning method based on a deep convolutional long short-term memory network, S1: obtaining historical battery data field data of a power battery from a data platform and performing pretreatment; S2: selecting battery monomer voltage and probe temperature data from the pretreated data, respectively extracting statistical features thereof, and respectively extracting time-frequency features thereof by using time-frequency analysis technology; S3: constructing and training a prediction model by using CLSTM, and the training input being the results of S1 and S2; S4: obtaining real-time data of the power battery and inputting the prediction model to obtain monomer battery voltage and probe temperature prediction values in a prediction time window; S5: calculating improved variance and average extreme value entropy of the two groups of prediction values respectively, comparing the results with a set threshold value, and if the value of any one of them is greater than the set threshold value, a thermal runaway early warning is issued. The model used in the application can capture the non-stationary characteristics of the battery data, and improve the prediction accuracy of the battery data.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power battery safety, and in particular to a battery thermal runaway early warning method based on a deep convolution long short-term memory network (CLSTM). BACKGROUND

[0002] Lithium-ion power batteries are widely used in the field of electric vehicles due to their high specific energy and long cycle life, but the safe and efficient management of lithium-ion power batteries is still a major technical challenge for the high-quality growth of electric vehicles. Batteries must be strictly managed to prevent overcharging, overdischarging, and overtemperature. In addition, due to its complex internal structure, the charging and discharging process of lithium batteries is a highly coupled physical and chemical reaction. Different charging and discharging rates, environmental temperatures, and other factors can have a significant impact on the degree of battery degradation and service life, and the electrochemical parameters inside the battery will gradually change as the battery is used. This requires the battery management system (BMS) to be able to identify changes in parameters and take appropriate control strategies to ensure its safety. When the battery is about to cross the safety boundary into a fault state, the fault diagnosis system should give timely warnings, determine the fault based on existing engineering application experience, and provide reasonable countermeasures.

[0003] Recurrent neural networks are used for time series prediction because they can capture long dependencies in time periods. However, when using a long short-term memory (LSTM) network for battery data prediction, the statistical features extracted from the battery signal cannot well represent the non-stationary information. In actual applications, non-stationary characteristics are crucial to battery data prediction. Therefore, it is necessary to add the extraction of non-stationary information features to the traditional prediction algorithm to improve the accuracy of battery data prediction and also facilitate the extraction of battery thermal runaway information. SUMMARY

[0004] To address the shortcomings of existing time series algorithms that cannot well extract battery non-stationary information, the present application proposes a battery thermal runaway early warning method based on a deep convolution long short-term memory network (CLSTM) that adds the extraction of non-stationary information features, uses the CLSTM algorithm to predict the voltage and temperature of the battery for a certain period of time in the future, and determines whether the battery is abnormal by predicting the improved variance and average extreme value entropy of the voltage and temperature, thereby improving the accuracy of battery data prediction and also facilitating the extraction of battery thermal runaway information.

[0005] The specific technical solutions are as follows:

[0006] A battery thermal runaway early warning method based on a deep convolutional long short-term memory network, comprising the following steps:

[0007] S1: obtaining historical data of the power battery from the data platform and preprocessing the same; the historical data includes battery data field data;

[0008] S2: selecting battery monomer voltage and probe temperature data from the preprocessed data, extracting statistical features of the selected two groups of data, and extracting time-frequency features of the selected two groups of data using time-frequency analysis technology;

[0009] S3: constructing and training a prediction model using a deep convolutional long short-term memory network, the input data for training being the preprocessed battery data field data obtained in S1, and the statistical features and time-frequency features of the battery monomer voltage and probe temperature data obtained in S2; the trained prediction model is used to realize prediction of the battery monomer voltage and probe temperature;

[0010] S4: obtaining real-time data of the power battery, the real-time data including battery data field data; inputting the real-time data into the trained prediction model to obtain monomer battery voltage prediction values and probe temperature prediction values in a prediction time window;

[0011] S5: calculating the improved variance and average extreme value entropy of the monomer battery voltage prediction values and probe temperature prediction values respectively, comparing the obtained monomer battery voltage improved variance, monomer battery voltage average extreme value entropy, probe temperature improved variance, and probe temperature average extreme value entropy with their respective set thresholds, and if any of the values is greater than the set threshold, determining that the battery has a thermal runaway risk and issuing a thermal runaway early warning; the set threshold is obtained by statistical analysis of historical data.

[0012] Further, in S5, the improved variance S 2 is calculated according to the following expression:

[0013]

[0014] In the formula, M is the total number of battery monomers in the battery pack, avg represents the mean value of M battery monomer data sequences K, and the battery monomer data sequence K represents the battery monomer voltage E or the probe temperature T, wherein K(1) represents the first battery monomer data sequence; K(i,j) represents the i-th battery monomer data sequence at j time, i=1,2,…,M, j=1,2,…,N, and N is the prediction length corresponding to the prediction time window.

[0015] Further, in S5, the average extreme value entropy is calculated according to the following expression:

[0016]

[0017] In the formula, N is the prediction duration corresponding to the prediction time window, K max represents the maximum value in the battery monomer data sequence at time t, K min represents the minimum value in the battery monomer data sequence at time t.

[0018] Further, in the S5, the threshold is set in the following manner: the improved variance and average extreme value entropy of the battery monomer voltage in the historical data are calculated, the improved variance and average extreme value entropy of the probe temperature are calculated, the normal data and the thermal runaway data of the battery are analyzed by box plot, the improved variance S E The value of the upper quartile of the box plot analysis is taken as the set threshold of the improved variance S E The value of the upper quartile of the box plot analysis is taken as the set threshold of the average extreme value entropy of the monomer battery voltage S The value of the upper quartile of the box plot analysis is taken as the set threshold of the improved variance S The value of the upper quartile of the box plot analysis is taken as the set threshold of the improved variance S T The value of the upper quartile of the box plot analysis is taken as the set threshold of the average extreme value entropy of the monomer battery voltage S T The value of the upper quartile of the box plot analysis is taken as the set threshold of the average extreme value entropy of the monomer battery voltage S The value of the upper quartile of the box plot analysis is taken as the set threshold of the average extreme value entropy of the monomer battery voltage S The value of the upper quartile of the box plot analysis is taken as the set threshold of the average extreme value entropy of the monomer battery voltage S.

[0019] Further, in the S1, the historical data of the power battery includes: battery monomer voltage, probe temperature, battery pack voltage, current, SOC, SOH, mileage.

[0020] Further, in the S1, the pre-processing manner is specifically: first, data cleaning is performed, then data normalization is performed, and finally standardization processing is performed.

[0021] Further, in the S2, the statistical characteristics of the pre-processed battery monomer voltage include the mean, standard deviation, variance, maximum value and minimum value of the voltage; and the statistical characteristics of the pre-processed temperature data include the mean, standard deviation, variance, maximum value and minimum value of the temperature.

[0022] Further, in the S2, the time-frequency analysis technique includes: short-time Fourier transform, wavelet transform, empirical mode decomposition, and ensemble empirical mode decomposition.

[0023] Further, in the S3, the CLSTM model includes: a convolution layer, an LSTM layer, and a full connection layer; the convolution layer is used to capture local dependency, and the LSTM layer is used to perform sequence prediction.

[0024] The present application has the following advantages:

[0025] (1) The application uses time-frequency analysis technology to extract time-frequency characteristics, and uses a convolutional neural network algorithm to improve the accuracy of battery data prediction; the CLSTM algorithm used can process the time domain and frequency domain characteristics of time series data through convolution operation, and can capture the signal changes for the non-stationary characteristics of battery voltage and temperature signals; moreover, CLSTM can be constructed into a deep network structure, and each layer can learn different levels of abstract features. This enables CLSTM to learn more complex patterns from raw data, making the algorithm more sensitive to subtle changes in battery performance.

[0026] (2) The application uses improved variance and average extreme entropy of power battery monomer voltage and probe temperature in historical data to set the thermal runaway alarm threshold, calculates the improved variance and average extreme entropy of real-time monomer battery voltage prediction value and probe temperature prediction value, and compares them with the set threshold to determine whether the thermal runaway alarm condition is reached. This combination of machine learning and mechanism analysis can accurately determine whether the power battery will thermal runaway in the future for a long period of time, is accurate and efficient, and has certain application prospect. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is a flowchart of the battery thermal runaway early warning method based on the deep convolutional long short-term memory network according to the embodiment of the application.

[0028] Figure 2 is a flowchart of the threshold for determining whether the battery has a thermal runaway risk according to the embodiment of the application. DETAILED DESCRIPTION

[0029] The application will be described in detail below according to the drawings and preferred embodiments, and the purposes and effects of the application will become more apparent. The application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application and do not limit the application.

[0030] As shown in Figure 1 , a battery thermal runaway early warning method based on a deep convolutional long short-term memory network specifically comprises the following steps:

[0031] S1: Obtain the historical data of the power battery from the data platform and preprocess it. The historical data of the power battery includes: battery monomer voltage, probe temperature, battery pack voltage, current, SOC, SOH, mileage and other battery data fields. The preprocessing method is as follows: first, data cleaning, then data normalization, and finally standardization.

[0032] S2: Select the battery cell voltage and probe temperature data from the preprocessed data, extract the statistical features of the two groups of data (i.e. the battery cell voltage and probe temperature data after preprocessing in S1) respectively, and apply time-frequency analysis techniques to extract the time-frequency features of the two groups of data respectively, the purpose is to better capture the non-stationary characteristics of the battery data. The data obtained in S2 is: the statistical features and time-frequency features of the battery cell voltage, the statistical features and time-frequency features of the probe temperature data. Among them, the statistical features of the preprocessed battery cell voltage include the mean, standard deviation, variance, maximum and minimum of the voltage; the statistical features of the preprocessed temperature data include the mean, standard deviation, variance, maximum and minimum of the temperature. Time-frequency analysis techniques include: Short-Time Fourier Transform (STFT), Wavelet Transform (WT), Empirical Mode Decomposition (EMD), Ensemble Empirical Mode Decomposition (EEMD) and other algorithms; in this embodiment, the time-frequency analysis technique selected is the STFT algorithm.

[0033] S3: Build a deep convolutional long short-term memory network (CLSTM) model, which includes a convolutional layer, an LSTM layer, and a fully connected layer. The convolutional layer is used to capture local dependencies, which is beneficial for capturing local fluctuations in battery voltage and temperature; the LSTM layer is used for sequence prediction. Use the S1 preprocessed battery data field data (excluding battery cell voltage and probe temperature) and the S2 feature extracted data (including statistical features and time-frequency features of battery cell voltage, statistical features and time-frequency features of probe temperature data) to train the CLSTM model, and obtain a prediction model for predicting battery cell voltage and probe temperature.

[0034] S4: Obtain real-time data of the power battery (including the battery data fields selected in S1, i.e. "battery cell voltage, probe temperature, battery pack voltage, current, SOC, SOH, range"), and input these real-time data into the prediction model to obtain the predicted values of the power battery data in the prediction time window, i.e. the predicted values of the battery cell voltage and the probe temperature.

[0035] S5: Calculate the improved variance and average extreme value entropy of the battery cell voltage and probe temperature predicted values obtained in S4, and use the obtained battery cell voltage improved variance, battery cell voltage average extreme value entropy, probe temperature improved variance, and probe temperature average extreme value entropy as the judgment indexes of thermal runaway. The specific calculation process is as follows:

[0036] 1) Improved variance S 2 :

[0037]

[0038]

[0039] In the formula, M is the total number of battery cells in the battery pack, avg represents the mean of the M battery cell data sequences K, and the battery cell data sequence K represents the battery cell voltage E or the probe temperature T (that is, when calculating the improved variance of the battery cell voltage, K in both formula (1) and formula (2) represents the battery cell voltage E, and when calculating the improved variance of the probe temperature, K in both formula (1) and formula (2) represents the probe temperature T), where K(1) represents the data sequence of the first battery cell. K(i,j) represents the data sequence of the i-th battery cell at time j, i=1,2,…,M, j=1,2,…,N, where N is the prediction duration corresponding to the prediction time window.

[0040] 2) Average extreme value entropy:

[0041] The extreme value entropy is defined as the ratio of the maximum value to the minimum value in the battery cell data sequence at a certain moment:

[0042] Average extreme value entropy This is the average extreme value entropy at each moment within the prediction time window. Considering the possibility of outlier fluctuations, the average extreme value entropy is used. As an indicator of thermal runaway, its calculation formula is as follows:

[0043]

[0044] The calculated improved variance S of the single cell voltage E Average extreme entropy of single cell voltage Improved variance of probe temperature S T Probe temperature average extreme entropy Each is compared with its set threshold; if S exists... E , S T , If a certain value in the battery exceeds a set threshold, it is determined that the battery is at risk of thermal runaway, and a thermal runaway warning is issued.

[0045] like Figure 2 As shown, the threshold is set by statistical analysis of historical data. Specifically, firstly, the improved variance and average extreme value entropy of the battery cell voltage, and the improved variance and average extreme value entropy of the probe temperature in the historical data are calculated according to equations (1) to (3). Then, the normal battery data and thermal runaway data are analyzed by box plot analysis, and S is...E the upper quartile of the box plot analysis as S E a set threshold value, the upper quartile of the box plot analysis as S a set threshold value, T the upper quartile of the box plot analysis as S T a set threshold value, the upper quartile of the box plot analysis as S a set threshold value.

[0046] Specifically, the improved variance of the monomer voltage and the temperature and the average extreme value entropy exceeding the threshold value include the following cases, and any one of them can trigger the thermal runaway early warning:

[0047] a) the improved variance S E of the monomer battery voltage exceeds the threshold value. In this embodiment, the set threshold value of the improved variance S E of the monomer battery voltage is 1.0, that is, S E > 1.0, it is judged that the power battery has a thermal runaway risk.

[0048] b) the improved variance S T of the probe temperature exceeds the threshold value. In this embodiment, the set threshold value of the improved variance S T of the probe temperature is 0.5, that is, S T > 0.5, it is judged that the power battery has a thermal runaway risk.

[0049] c) the average extreme value entropy S of the monomer battery voltage exceeds the threshold value. In this embodiment, the set threshold value of the average extreme value entropy S of the monomer battery voltage is 1.1, that is, S it is judged that the power battery has a thermal runaway risk.

[0050] d) the average extreme value entropy S of the probe temperature exceeds the normal range. In this embodiment, the set threshold value of the average extreme value entropy S of the probe temperature is 1.05, that is, S it is judged that the power battery has a thermal runaway risk.

[0051] It should be noted that the above-mentioned 1.0, 0.5, 1.1, 1.05 and the like are for illustration only, and do not limit the protection scope of the present application.

[0052] The application uses a convolutional neural network algorithm to improve the accuracy of battery data prediction, and the CLSTM algorithm used can process the time domain and frequency domain characteristics of time series data through convolution operation, and can well capture the signal changes for the non-stationary characteristics of signals such as battery voltage and probe temperature. Moreover, the CLSTM can be constructed into a deep network structure, and each layer can learn different levels of abstract features, which makes the CLSTM learn more complex patterns from the original data, so that the algorithm is more sensitive to the subtle changes of battery performance. At the same time, the application uses the improved variance and average extreme entropy of the power battery monomer voltage and probe temperature in the historical data to set the threshold of the thermal runaway early warning, calculates the real-time monomer battery voltage prediction value and probe temperature prediction value, and compares it with the set threshold to judge whether the thermal runaway early warning condition is reached. This method combining machine learning and mechanism analysis can accurately judge whether the power battery will thermal runaway in a long period of time in the future, and is accurate and efficient, and has certain application prospect.

[0053] Those skilled in the art can understand that the above description is only preferred examples of the application and is not used to limit the application, although the application has been described in detail with reference to the foregoing examples, those skilled in the art can still modify the technical solutions recorded in the foregoing examples or make equivalent replacement for part of the technical features. Any modification, equivalent replacement, etc. within the spirit and principles of the application shall be included in the protection scope of the application.

Claims

1.A battery thermal runaway early warning method based on a deep convolutional long short-term memory network, characterized in that, The method comprises the following steps: S1: obtaining historical data of the power battery from a data platform and preprocessing the historical data; the historical data comprises battery data field data; S2: selecting battery monomer voltage and probe temperature data from the preprocessed data, extracting statistical features of the two groups of selected data respectively, and extracting time-frequency features of the two groups of selected data respectively using time-frequency analysis technology; S3: constructing and training a prediction model using a deep convolutional long short-term memory network, the input data for training being the preprocessed battery data field data obtained in S1, and the statistical features and time-frequency features of the battery monomer voltage and probe temperature data obtained in S2; the trained prediction model is used to realize prediction of the battery monomer voltage and probe temperature; S4: obtaining real-time data of the power battery, the real-time data comprising battery data field data; inputting the real-time data into the trained prediction model to obtain monomer battery voltage prediction values and probe temperature prediction values in a prediction time window; S5: calculating the improved variance and average extreme value entropy of the monomer battery voltage prediction values and the probe temperature prediction values respectively, comparing the obtained monomer battery voltage improved variance, monomer battery voltage average extreme value entropy, probe temperature improved variance, and probe temperature average extreme value entropy with their respective set thresholds, and if the value of any one of them is greater than the set threshold, determining that the battery has a risk of thermal runaway and issuing a thermal runaway warning; the set threshold is obtained by statistical analysis of historical data. 2.The battery thermal runaway early warning method based on deep convolutional long short-term memory network according to claim 1, wherein, In the S5, the improved variance S 2 The calculation expression is as follows: In the formula, M is the total number of battery monomers in the battery pack, avg represents the mean value of M battery monomer data sequences K, the battery monomer data sequence K represents the battery monomer voltage E or the probe temperature T, wherein K(1) represents the first battery monomer data sequence; K(i,j) represents the i-th battery monomer data sequence at j time, i=1,2,…,M, j=1,2,…,N, and N is the prediction length corresponding to the prediction time window. 3.The battery thermal runaway pre-warning method based on deep convolutional long short-term memory network according to claim 1, wherein, In S5, the average extreme entropy The calculation expression is as follows: In the formula, N is a prediction time length corresponding to a prediction time window, K max represents a maximum value in the battery monomer data sequence at time t, K min represents a minimum value in the battery monomer data sequence at time t. 4.The battery thermal runaway early warning method based on deep convolutional long short-term memory network according to claim 1, wherein, In S5, the threshold value is set by calculating the improved variance and average extreme entropy of the battery cell voltage in the historical data, the improved variance and average extreme entropy of the probe temperature, analyzing the battery normal data and thermal runaway data by box plot, and taking the improved variance S E The value of the upper quartile of the box plot analysis as the set threshold value of S E The value of the upper quartile of the box plot analysis as the set threshold value of S The value of the upper quartile of the box plot analysis as the set threshold value of S The value of the upper quartile of the box plot analysis as the set threshold value of S T The value of the upper quartile of the box plot analysis as the set threshold value of S T The value of the upper quartile of the box plot analysis as the set threshold value of S The value of the upper quartile of the box plot analysis as the set threshold value of S The value of the upper quartile of the box plot analysis as the set threshold value of S 5.The battery thermal runaway pre-warning method based on deep convolutional long short-term memory network according to claim 1, wherein, In S1, the historical data of the power battery comprises battery monomer voltage, probe temperature, battery pack voltage, current, SOC, SOH, and mileage. 6.The battery thermal runaway pre-warning method based on deep convolutional long short-term memory network according to claim 1, wherein, In S1, the preprocessing method specifically comprises the following steps: first, data cleaning, then data normalization, and finally standardization processing. 7.The battery thermal runaway early warning method based on deep convolutional long short-term memory network according to claim 1, wherein, In S2, the statistical features of the preprocessed battery monomer voltage include the mean value, standard deviation, variance, maximum value, and minimum value of the voltage; The statistical features of the preprocessed temperature data include the mean value, standard deviation, variance, maximum value, and minimum value of the temperature. 8.The battery thermal runaway pre-warning method based on deep convolutional long short-term memory network according to claim 1, wherein, In S2, the time-frequency analysis technology includes short-time Fourier transform, wavelet transform, empirical mode decomposition, and ensemble empirical mode decomposition. 9.The battery thermal runaway pre-warning method based on deep convolutional long short-term memory network according to claim 1, wherein, In S3, the CLSTM model comprises a convolutional layer, an LSTM layer, and a fully connected layer; the convolutional layer is used to capture local dependency, and the LSTM layer is used to perform sequence prediction.

Images