Electric vehicle battery fault detection method and system, electronic equipment and storage medium

Abstract

The invention discloses an electric vehicle battery fault detection method and system, an electronic device and a storage medium, which consider electric heating data of multiple dimensions of a battery, can more comprehensively and accurately reflect the working state of the battery, thereby improving the accuracy of electric vehicle battery fault diagnosis, and improving the reliability of the electric vehicle battery fault diagnosis. Dimension reduction processing is carried out on collected high-dimensional electric heating data, the data processing amount is greatly reduced while electric heating data characteristics are reserved, diagnosis efficiency is improved, meanwhile, the improved Minkowski algorithm is used for calculating the distance between electric heating data characteristic sequences, distance measurement is not affected by data scales of all dimensions, and the accuracy of diagnosis is improved. And the accuracy of the distance calculation result is improved, so that the accuracy of electric vehicle battery fault diagnosis is further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery fault detection, and in particular to an electric vehicle battery fault detection method and system, electronic equipment, and a computer-readable storage medium. Background Art

[0002] With the rapid construction of charging facilities, electric vehicles have received great attention and strong support as one of the effective ways to ensure the security of national energy strategies and solve environmental crises. As the "heart" of electric vehicles, batteries have been rapidly developed and widely used. However, safety accidents caused by battery failures occur frequently, posing a serious threat to the personal safety of users. Therefore, a fast and accurate electric vehicle battery fault detection method is urgently needed to maintain the safe operation of electric vehicles.

[0003] At present, there are three main methods for battery fault detection. The first is a model-based method, which compares the residual signal of the model output and the measured value, or compares the calculation using different model parameters. However, this method requires an accurate battery model, and the establishment of the battery model is often complex and depends on specific working conditions. For unknown or complex fault modes, the model cannot accurately predict; the second is a knowledge-based method, such as neural networks, expert systems, and fault trees. This method has poor real-time performance, while fault detection requires timely discovery. Once the knowledge base is established, its rules and logic are usually fixed, and the flexibility and scalability are poor; the third is a data-driven method, such as comparing battery consistency or capturing abnormal fluctuations and other characteristics of measured values. This method does not rely on a specific battery model and has good real-time performance, meeting the requirements of fault detection. However, the existing data-driven method detects battery faults based only on measured voltage. For example, patent CN116840725A discloses a battery pack fault detection method, which identifies faulty single cells based on the voltage characteristics of single cells, but the voltage characteristics of some faults are very similar and difficult to distinguish, resulting in poor detection accuracy. Summary of the invention

[0004] The present invention provides an electric vehicle battery fault detection method and system, electronic equipment, and a computer-readable storage medium, which can improve the accuracy of electric vehicle battery fault diagnosis and can also improve the diagnosis efficiency.

[0005] According to one aspect of the present invention, a method for detecting a battery failure in an electric vehicle is provided, comprising the following contents:

[0006] Collecting characteristic data of each battery cell in the power battery in real time, constructing a characteristic data sequence of each battery cell, and preprocessing and normalizing the collected characteristic data, wherein the characteristic data includes a plurality of electrical characteristic data and a plurality of thermal characteristic data;

[0007] Perform dimensionality reduction processing on the normalized feature data sequence of each battery cell to extract the electric and thermal data feature sequence of each battery cell;

[0008] The improved Minkowski algorithm is used to calculate the distance between the electrothermal data feature sequences of any two battery cells. If the distance between any two electrothermal data feature sequences is less than a preset threshold, the battery state is determined to be normal, otherwise the battery is determined to be faulty.

[0009] Furthermore, the process of performing dimensionality reduction processing on the normalized feature data sequence of each battery cell to extract the electric and thermal data feature sequence of each battery cell includes the following:

[0010] Divide all data points in the normalized feature data sequence into two data point sets;

[0011] Using multiple query data points to search in two data point sets, each data point set is divided into multiple local data sets, each local data set includes (k+1) nearest neighbor points of each query data point;

[0012] Find the local data set where each data point in the feature data sequence is located, and calculate the weight between the data point and all other data points in the local data set to construct a weight vector;

[0013] Embed each data point in the feature data sequence from the high-dimensional data space into the low-dimensional data space to obtain an embedded data sequence;

[0014] Based on the weight vector and the embedded data sequence, the embedding error of the feature data sequence is calculated. If the embedding error is less than a threshold, the embedded data sequence is used as the electrothermal data feature sequence.

[0015] Furthermore, the process of dividing the normalized feature data sequence into two data point sets includes the following:

[0016] Calculate the average value of all data points in the feature data sequence in each feature data dimension, and construct the average data point based on the average value of all feature data dimensions;

[0017] Calculate the similarity between all data points and the average data point;

[0018] If the similarity between a data point and the average data point is greater than a preset similarity threshold, the data point is divided into the first data point set, otherwise the data point is divided into the second data point set.

[0019] Furthermore, the similarity between any data point and the average data point is calculated based on the following formula:

[0020]

[0021] Among them, d i represents the similarity between the i-th data point and the average data point in the feature data sequence, n represents the feature data dimension, t im represents the value of the feature data of the i-th data point in the m-th dimension, C m Represents the value of the m-th dimension feature data of the average data point.

[0022] Furthermore, the embedding error is calculated based on the following formula:

[0023]

[0024] Where E(T) represents the embedding error, N represents the total number of data points in the feature data sequence, and J i represents the set of k nearest neighbors of the ith data point, w ij represents the weight between the i-th data point and its j-th nearest neighbor, t i represents the embedded data of the i-th data point, t j Represents the embedded data of the jth nearest neighbor.

[0025] Further, the process of searching in two data point sets using multiple query data points to divide each data point set into multiple local data sets includes the following:

[0026] Set a query data point, calculate the similarity between the query data point and the average data point, if the similarity between the two is greater than a preset threshold, the query data point is classified into the first data point set, otherwise it is classified into the second data point set, and the (k+1) data point with the greatest similarity to the current query data point is screened out from the classified data point set, and the (k+1) data point is divided into a local data set, and the (k+1) data point is removed from the current data point set; then, the query data point is reset, and the above process is repeated, and it is iterated continuously until each data point set is divided into multiple local data sets, each local data set contains (k+1) data points.

[0027] Furthermore, the distance between the electrothermal data feature sequences of any two battery cells is calculated based on the following formula:

[0028]

[0029] Among them, x and y represent the characteristic sequences of the electric and thermal data of the two battery cells, p is a parameter with a value range of [1, 2, ∞], and x i andi represents the ith data feature value of the electrothermal data feature sequence of battery cell x and battery cell y, respectively. The superscript T represents transposition, and S -1 () represents the inverse matrix of the covariance matrix.

[0030] In addition, the present invention also provides an electric vehicle battery fault detection system, which adopts the electric vehicle battery fault detection method as described above, and comprises:

[0031] The data acquisition and processing module collects the characteristic data of each battery cell in the power battery in real time, constructs a characteristic data sequence of each battery cell, and pre-processes and normalizes the collected characteristic data, wherein the characteristic data includes a plurality of electrical characteristic data and a plurality of thermal characteristic data;

[0032] A data dimension reduction processing module is used to perform dimension reduction processing on the normalized feature data sequence of each battery cell to extract the electric and thermal data feature sequence of each battery cell;

[0033] The battery fault diagnosis module is used to calculate the distance between the electrothermal data feature sequences of any two battery cells using the improved Minkowski algorithm. If the distance between any two electrothermal data feature sequences is less than a preset threshold, the battery state is determined to be normal, otherwise the battery is determined to be faulty.

[0034] In addition, the present invention also provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the processor executes the steps of the above method by calling the computer program stored in the memory.

[0035] In addition, the present invention also provides a computer-readable storage medium for storing a computer program for performing electric vehicle battery fault detection, wherein the computer program executes the steps of the method described above when running on a computer.

[0036] The present invention has the following beneficial effects:

[0037] The electric vehicle battery fault detection method of the present invention first collects the electrical characteristic data and thermal characteristic data of each battery cell in real time and performs normalization processing, then performs dimensionality reduction processing on the characteristic data sequence after normalization processing of each battery cell, extracts the electric thermal data characteristic sequence of each battery cell, and finally, uses the improved Minkowski algorithm to calculate the distance between the electric thermal data characteristic sequences of any two battery cells. If the distance between any two electric thermal data characteristic sequences is less than a preset threshold, it is determined that the battery state is normal, otherwise it is determined that the battery fails. The electric vehicle battery fault detection method of the present invention takes into account the electric thermal data of multiple dimensions of the battery, and can more comprehensively and accurately reflect the working state of the battery, thereby improving the accuracy of the electric vehicle battery fault diagnosis, and through the dimensionality reduction processing of the collected high-dimensional electric thermal data, the data processing amount is greatly reduced while retaining the characteristics of the electric thermal data, and the diagnostic efficiency is improved. At the same time, the distance between the electric thermal data characteristic sequences is calculated using the improved Minkowski algorithm, so that the distance metric is not affected by the scale of the data in each dimension, and the accuracy of the distance calculation result is improved, thereby further improving the accuracy of the electric vehicle battery fault diagnosis.

[0038] In addition, the electric vehicle battery fault detection system of the present invention also has the above advantages.

[0039] In addition to the above-described purposes, features and advantages, the present invention has other purposes, features and advantages. The present invention will be further described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings constituting a part of this application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0041] Figure 1 It is a flow chart of an electric vehicle battery fault detection method of a current transformer according to a preferred embodiment of the present application.

[0042] Figure 2 yes Figure 1 Schematic diagram of the sub-process of step S2 in FIG.

[0043] Figure 3 yes Figure 2 Schematic diagram of the sub-process of step S21.

[0044] Figure 4 It is another flow chart of the electric vehicle battery fault detection method of the current transformer according to the preferred embodiment of the present application.

[0045] Figure 5It is another flow chart of the electric vehicle battery fault detection method of the current transformer according to the preferred embodiment of the present application.

[0046] Figure 6 It is a schematic diagram of the module structure of an electric vehicle battery fault detection system of a current transformer according to another embodiment of the present application. DETAILED DESCRIPTION

[0047] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0048] Reference Figure 1 The preferred embodiment of the present application provides a method for detecting a battery failure in an electric vehicle, comprising the following contents:

[0049] Step S1: collecting characteristic data of each battery cell in the power battery in real time, constructing a characteristic data sequence of each battery cell, and preprocessing and normalizing the collected characteristic data, wherein the characteristic data includes a plurality of electrical characteristic data and a plurality of thermal characteristic data;

[0050] Step S2: performing dimensionality reduction processing on the normalized feature data sequence of each battery cell to extract the electric and thermal data feature sequence of each battery cell;

[0051] Step S3: using the improved Minkowski algorithm to calculate the distance between the electrothermal data feature sequences of any two battery cells, if the distance between any two electrothermal data feature sequences is less than a preset threshold, the battery state is determined to be normal, otherwise the battery is determined to be faulty.

[0052] It can be understood that the electric vehicle battery fault detection method of the present invention first collects the electrical characteristic data and thermal characteristic data of each battery cell in real time and performs normalization processing, then performs dimensionality reduction processing on the characteristic data sequence after normalization processing of each battery cell, extracts the electric and thermal data characteristic sequence of each battery cell, and finally, uses the improved Minkowski algorithm to calculate the distance between the electric and thermal data characteristic sequences of any two battery cells. If the distance between any two electric and thermal data characteristic sequences is less than the preset threshold, the battery state is determined to be normal, otherwise the battery is determined to be faulty. The electric vehicle battery fault detection method of the present invention takes into account the electric and thermal data of multiple dimensions of the battery, which can more comprehensively and accurately reflect the working state of the battery, thereby improving the accuracy of the electric vehicle battery fault diagnosis, and through the dimensionality reduction processing of the collected high-dimensional electric and thermal data, the data processing amount is greatly reduced while retaining the characteristics of the electric and thermal data, and the processing efficiency is improved. At the same time, the distance between the electric and thermal data characteristic sequences is calculated using the improved Minkowski algorithm, so that the distance metric is not affected by the scale of the data in each dimension, and the accuracy of the distance calculation result is improved, thereby further improving the accuracy of the electric vehicle battery fault diagnosis.

[0053] It can be understood that in step S1, the characteristic data of each battery cell in the power battery is collected in real time, wherein the characteristic data includes multiple electrical characteristic data and multiple thermal characteristic data, specifically, the electrical characteristic data includes voltage, current, peak voltage and state of charge (SOC), and the thermal characteristic data includes temperature, thermal resistance, thermal distribution uniformity and health state (SOH), so that the characteristic data sequence of each battery cell can be constructed. In addition, the collected characteristic data needs to be preprocessed, such as cleaning the data, removing noise and outliers, and then normalizing the preprocessed characteristic data to facilitate subsequent data analysis and processing.

[0054] It is understandable that Figure 2 As shown, in the step S2, the process of performing dimensionality reduction processing on the normalized feature data sequence of each battery cell to extract the electric and thermal data feature sequence of each battery cell includes the following:

[0055] Step S21: Divide all data points in the normalized feature data sequence into two data point sets;

[0056] Step S22: using multiple query data points to search in two data point sets, so as to divide each data point set into multiple local data sets, each local data set including the (k+1) nearest neighbor points of each query data point;

[0057] Step S23: Find the local data set where each data point in the feature data sequence is located, and calculate the weight between the data point and all other data points in the local data set to construct a weight vector;

[0058] Step S24: embed each data point in the feature data sequence from the high-dimensional data space into the low-dimensional data space to obtain an embedded data sequence;

[0059] Step S25: Based on the weight vector and the embedded data sequence, the embedding error of the feature data sequence is calculated. If the embedding error is less than the threshold, the embedded data sequence is used as the electrothermal data feature sequence.

[0060] Specifically, first, all data points in the normalized feature data sequence of each battery cell are divided into two data point sets based on similarity, and the feature data sequence is preliminarily divided. Figure 3 As shown, in the step S21, the process of dividing the normalized feature data sequence into two data point sets includes the following contents:

[0061] Step S211: Calculate the average value of all data points in the feature data sequence in each feature data dimension, and construct an average data point based on the average values ​​of all feature data dimensions;

[0062] Step S212: Calculate the similarity between all data points and the average data point;

[0063] Step S213: if the similarity between a certain data point and the average data point is greater than a preset similarity threshold, the data point is divided into the first data point set; otherwise, the data point is divided into the second data point set.

[0064] Specifically, for each battery cell's normalized feature data sequence, first calculate the average value of all data points in the feature data sequence in each feature data dimension, and then construct the average data point based on the average value of all feature data dimensions. Each data point is an eight-dimensional vector. For example, first calculate the average values ​​of all data points in the voltage dimension, current dimension, peak voltage dimension, charging state dimension, temperature dimension, thermal resistance dimension, thermal distribution uniformity dimension, and health state dimension, and then construct the average data point based on the eight average values, that is, the average data point includes the average values ​​of the eight feature data dimensions. Then, calculate the similarity between each data point and the average data point, and specifically calculate the similarity between any data point and the average data point based on the following formula:

[0065]

[0066] Among them, d irepresents the similarity between the ith data point in the feature data sequence and the average data point, n represents the feature data dimension, for example, n = 8, t im represents the value of the feature data of the i-th data point in the m-th dimension, C m Represents the value of the m-th dimension feature data of the average data point.

[0067] Next, the data set is divided, and the calculated similarity is compared with the preset similarity threshold. If the similarity is greater than the preset similarity threshold, the data point is divided into the first data point set, and if the similarity is not greater than the preset similarity threshold, the data point is divided into the second data point set. This process is repeated until all data points in the feature data sequence are divided into the first data point set and the second data point set.

[0068] Then, the query data point is used to search in the two data point sets divided in step S21 to find the (k+1) data points that are the nearest neighbors of each query data point, and the (k+1) data points are constructed as a local data set. The process of using multiple query data points to search in the two data point sets to divide each data point set into multiple local data sets includes the following:

[0069] Set a query data point, calculate the similarity between the query data point and the average data point, if the similarity between the two is greater than a preset threshold, the query data point is classified into the first data point set, otherwise it is classified into the second data point set, and the (k+1) data point with the greatest similarity to the current query data point is screened out from the classified data point set, and the (k+1) data point is divided into a local data set, and the (k+1) data point is removed from the current data point set; then, the query data point is reset, and the above process is repeated, and it is iterated continuously until each data point set is divided into multiple local data sets, each local data set contains (k+1) data points.

[0070] Specifically, the query data point is also an n-dimensional vector, for example, including characteristic data of eight dimensions, namely, voltage, current, peak voltage, charging state, temperature, thermal resistance, heat distribution uniformity and health state. The similarity between the query data point and the average data point is calculated using the aforementioned similarity calculation formula. If the similarity between the two is greater than the preset threshold, the query data point is classified into the first data point set, otherwise it is classified into the second data point set. Then, assuming that the query data point is classified into the first data point set, the similarity between the query data point and all data points is calculated again in the first data point set, and the (k+1) data points with the largest similarity are selected, and the (k+1) data points are divided into a local data set. It can be considered that the similarity between any two data points in the (k+1) data points is large, and the (k+1) data point is removed from the current data point set. Then, the query data point is reset, and the above process is repeated, and it is continuously iterated until the first data point set is divided into multiple local data sets, each of which contains (k+1) data points. In addition, when some data points still remain in the first data point set after iterative division, they can be completed by interpolation to form a local data set, or the remaining data points can be directly eliminated.

[0071] Next, find the local data set where each data point in the feature data sequence is located. Since the first data point set and the second data point set have been divided into multiple local data sets in step S22, the local data set where each data point is located can be directly found. Moreover, since the similarity between the (k+1) data points in the local data set and the query data point is relatively large, it can be considered that the remaining k data points in the local data set are the k nearest neighbor points of the data point. Then, the weights between the data point and all the remaining data points in the local data set are calculated respectively to construct a weight vector, where the weight vector can be expressed as: W i =[w i1 ,w i2 ,...,w ik ],W i represents the weight vector of the i-th data point, w ik Represents the weight between the ith data point and its kth nearest neighbor, and the weight is calculated using a Gaussian function.

[0072] Then, the data dimension of each data point is embedded from the high-dimensional data space to the low-dimensional data space, for example, the data dimension of each data point is embedded from the eight-dimensional data space to the one-dimensional or two-dimensional data space to obtain an embedded data sequence, where the embedded data sequence can be expressed as: T = [t1, t2, ..., t N ], T represents the embedded data sequence, N represents the total number of data points, t NRepresents the embedded data of the Nth data point. The specific data embedding process belongs to the prior art and will not be described here. For example, the dimension of the low-dimensional space is first determined, and then the embedding matrix (i.e., the embedded data sequence) is randomly initialized, and the embedding matrix is ​​optimized using an optimization algorithm. For the objective function of the optimization algorithm, the gradient of the embedding matrix needs to be calculated, and then the value of the embedding matrix is ​​updated using the gradient descent method, and finally the embedding matrix is ​​obtained through the optimization process.

[0073] Finally, based on the weight vector and the embedded data sequence, the embedding error of the feature data sequence is calculated. If the embedding error is less than the threshold, the embedded data sequence is used as the electrothermal data feature sequence. If the embedding error is not less than the threshold, it is considered that the data embedding does not meet the requirements and does not retain the local characteristics of the data well. It is necessary to readjust the similarity threshold or adjust the embedding parameters until the embedding error is less than the threshold. The embedding error is calculated based on the following formula:

[0074]

[0075] Where E(T) represents the embedding error, N represents the total number of data points in the feature data sequence, and J i represents the set of k nearest neighbors of the ith data point, w ij represents the weight between the i-th data point and its j-th nearest neighbor, t i represents the embedded data of the i-th data point, t j Represents the embedded data of the jth nearest neighbor.

[0076] It can be understood that the data dimensionality reduction process of the present invention first divides all data points into two data point sets based on the similarity algorithm, realizes the coarse clustering of data points, and then uses the query data point to search in the two data point sets to divide each data point set into multiple local data sets, thereby realizing the fine clustering of data points, so that the k nearest neighbor points of each data point can be quickly found, which greatly improves the efficiency of dimensionality reduction, especially when performing data dimensionality reduction on large-scale high-dimensional data, the efficiency is more obvious, and the local characteristics of the feature data sequence are also retained. In addition, the accuracy of data dimensionality reduction is evaluated by the constructed embedded error algorithm, which ensures that the data after dimensionality reduction retains the local characteristics of the original data, which is conducive to improving the accuracy of fault diagnosis. Compared with the existing dimensionality reduction algorithm, it has better local structure optimization ability, interpretability and robustness, and can better handle nonlinear data. For example, compared with the existing local linear embedding algorithm, the present invention can better process nonlinear data and retain the local structure of the data; compared with the existing T-SNE algorithm, the present invention can generate a clearer low-dimensional representation and can better display the relationship between data; compared with the existing LLE algorithm, the LLE algorithm is not effective when processing large-scale data and is very sensitive to the number of selected neighboring points. The present invention is more efficient and more robust when processing large-scale high-dimensional data by first constructing a combination of two data points and then dividing them into different local data sets.

[0077] It can be understood that in step S3, the improved Minkowski algorithm is used to calculate the distance between the electrothermal data feature sequences of any two battery cells. If the distance between any two electrothermal data feature sequences is less than a preset threshold, the battery state is determined to be normal, otherwise the battery is determined to be faulty. It can be understood that the electrothermal feature data between multiple battery cells in normal operation will show a certain similarity in the data space. If the similarity between the electrothermal feature data of two battery cells exceeds the deviation range, it is determined that the battery has a certain fault or is in an abnormal operating state. Therefore, the present invention uses an improved Minkowski algorithm to calculate the distance between the electrothermal data feature sequences of any two battery cells. If the distance between the electrothermal data feature sequences of any two battery cells is less than a preset threshold, the battery is determined to be normal, otherwise the battery is determined to be faulty. Among them, the expression of the improved Minkowski algorithm is:

[0078]

[0079] Among them, x and y represent the characteristic sequences of the electric and thermal data of the two battery cells, p is a parameter with a value range of [1, 2, ∞], and x i and i represents the ith data feature value of the electrothermal data feature sequence of battery cell x and battery cell y, respectively. The superscript T represents transposition, and S -1() represents the inverse matrix of the covariance matrix. The value of p varies with the data type. When the data is very sensitive to incremental changes in each dimension, p takes 1. When each dimension is equally important and independent of each other, p takes 2. When the data is very sensitive to extreme values, p takes ∞. For example, a small change in temperature sometimes indicates that the system is approaching its thermal limit, so when calculating the temperature between different battery cells, p takes 1.

[0080] It can be understood that the covariance structure of the data is introduced into the improved Minkowski algorithm of the present invention, so that the distance metric is not affected by the scale of the characteristic data of each dimension, can better reflect the inherent structure of the data, and improve the accuracy of the distance calculation results, thereby further improving the accuracy of electric vehicle battery fault diagnosis.

[0081] Optionally, in order to further improve the accuracy of battery fault diagnosis, taking into account that the data characteristics of the battery in the charging state and the battery discharging state are quite different, the present invention performs the above-mentioned data collection, dimensionality reduction, and distance calculation processes in the battery charging state and the battery discharging state respectively. Since the distance measurement judgment thresholds in the charging stage and the discharging stage are different, a new threshold is obtained by assigning weights to the two thresholds, and then the new threshold is used to determine whether a fault occurs.

[0082] Alternatively, if Figure 4 As shown, the electric vehicle battery fault detection method also includes the following contents:

[0083] Step S4: Evaluate the fault detection result.

[0084] Specifically, the F1-score method is used to evaluate the detection results to verify the effectiveness and accuracy of battery fault detection. The F1-score method includes four indicators, namely accuracy, precision, recall rate and F1 score. A represents accuracy, P represents precision, R represents recall, F1 represents F1 score, TT represents diagnosed as faulty and actually faulty, TF represents diagnosed as faulty and actually normal, FT represents diagnosed as normal and actually faulty, and FF represents diagnosed as normal and actually normal. Then, the effectiveness and accuracy of the fault detection results are evaluated based on the four indicators, wherein the specific evaluation logic belongs to the prior art and will not be repeated here. In addition, when P is higher but R is lower, a normal battery may be misdiagnosed as a faulty battery, proving that the probability of misjudging a normal battery as a fault is higher, which means that the distance threshold is set low and the threshold should be increased; if R is higher but P is lower, it means that the detection is too conservative and some battery faults cannot be identified, which means that the distance threshold is set high and the threshold should be lowered.

[0085] Alternatively, if Figure 5As shown, the electric vehicle battery fault detection method also includes the following contents:

[0086] Step S5: construct a correlation analysis model based on the collected electrothermal characteristic data, and use the improved Pearson correlation coefficient to analyze the correlation between the characteristic data, and determine the fault type according to the correlation result.

[0087] Specifically, for short-circuit faults, the correlation analysis model constructed is: R1 = [F(V1′, V2′)F(V2′, V3′)...F(V N ′,V1′)], V1′ represents the voltage sum of the first battery cell and the second battery cell, V i ′ represents the voltage sum of the ith battery cell and the i+1th battery cell, and F() represents the correlation calculated using the improved Pearson correlation coefficient. When the ith battery cell has a short circuit fault, the correlation coefficient F(V i ' -2 ,V i ' -1 ) and F(V i ′,V i ' +1 ) will be abnormal, and F(V i ' -1 ,V i ′) is still normal, where the abnormal correlation coefficient means that it deviates from the preset value range, and normal means that it is within the preset value range. Therefore, by performing correlation analysis on the voltage data of the battery cell, the short circuit fault can be accurately diagnosed. Further, for the connection fault, the constructed correlation analysis model is: R2=[F(ΔV 1,2 +m,I) F(ΔV 2,3 +m,I) ... F(ΔV N,1 +m,I)], where ΔV 1,2 =|V1′-V′2|, m represents the correction value, and I represents the battery current. When the connection between the i-th battery cell and the i+1-th battery cell fails, the correlation coefficient F(V i-2 ,V i-1 ) and F(V i ,V i+1 ) will be swapped and will not approach 0, while F(ΔV i-1，i +m,I) and F(ΔV i,i+1+m,I) will be close to 1. Therefore, based on the correlation analysis of the voltages of different battery cells, the correlation analysis of the voltage difference and current of different battery cells can be performed to diagnose the connection fault. In addition, the voltage sensor fault is also based on the correlation analysis of the voltages of different battery cells. By performing the correlation analysis of the voltage difference and current of different batteries, when the voltage sensor corresponding to the i-th battery cell fails, the correlation coefficient F(V i-2 ,V i-1 ) and F(V i ,V i+1 ) will be swapped, and no large correlation value will appear in R2, that is, all correlation coefficients in R2 are smaller than the preset threshold.

[0088] In addition, for the heat dissipation coefficient failure, the correlation analysis model constructed is: R3 = [F(T1,θ1) F(T2,θ2) ...F(T N ,θ N )], R4=[F(T1′,T′2) F(T′2,T′3) ... F(T′ N ,T′1)], T1 and θ1 represent the temperature and thermal resistance of the first battery cell respectively, and T1′ represents the temperature and thermal resistance of the first battery cell and the second battery cell. When the cooling system of the i-th battery cell fails, the correlation coefficient F(T i ,θ i )、F(T′ i-2 ,T′ i-1 )、F(T′ i ,T′ i+1 ) is abnormal, and F(T′ i-1 ,T′ i ) is still normal. Therefore, by performing correlation analysis between the thermal resistance and temperature of each battery cell, and the temperature between different battery cells, it is possible to diagnose heat dissipation system failures.

[0089] In addition, for thermal runaway failure, the correlation analysis model constructed is: R5 = [F(V1, T1)F(V2, T2)...F(V N ,T N )], R6=[F(I1,T1)F(I2,T2)...F(I N ,T N )], V1, I1 and T1 represent the voltage, current and temperature of the first battery cell respectively. When the i-th battery cell has a thermal runaway failure, its temperature will rise sharply, its voltage will drop sharply, and its current will rise sharply. The correlation coefficient F(V i ,T i ) will decrease abnormally, while F(I i ,T i) shows an abnormal increase. Therefore, by analyzing the correlation between the voltage and temperature, and the current and temperature of each battery cell, thermal runaway faults can be diagnosed.

[0090] In addition, for battery insulation failure, the correlation analysis model constructed is: R7 = [F(θ1,θ2) F(θ2,θ3) ...F(θ N ,θ1)], when the insulation failure occurs in the i-th battery cell, its thermal resistance will decrease, and the thermal resistance of different battery cells will increase inconsistently, and the correlation coefficient F(θ i-1 ,θ i )、F(θ i ,θ i+1 ) will cause an exception.

[0091] In addition, for low capacity and low SOC faults, the constructed correlation analysis model is: It represents the ratio of the voltage of the ith battery cell to the average battery voltage of the normal battery pack. If the voltage ratio shows obvious abnormal fluctuations and is accompanied by abnormal temperature rise, when the ith battery cell has a low capacity and low SOC fault, An exception will occur.

[0092] In addition, for internal resistance fault, the correlation analysis model constructed is: R9 = [F (V 1,max ,SOC1) F(V 2,max ,SOC2) ... F(V N,max ,SOC 1N )],V 1,max and SOC1 represent the peak voltage and SOC state of the first battery cell respectively. When an internal resistance fault occurs, the peak voltage of the battery drops abnormally, and the SOC state and peak voltage will present a nonlinear state. Therefore, when an internal resistance fault occurs in the i-th battery cell, F(V i,max ,SOC i ) will cause an exception.

[0093] In addition, for different correlation analysis models, the process of analyzing the correlation between feature data using the improved Pearson correlation coefficient is as follows:

[0094] First, construct the input data matrix X based on the collected feature data, X∈R N×T×F , N represents the number of samples, T represents the time step, and F represents the number of features, that is, the feature dimension, where the features include the voltage V of each battery cell i 、Current I i , peak voltage V i，max 、Thermal resistanceθ i and temperature T i , and the voltage of two adjacent battery cells and V i′、voltage difference ΔV i-1,i , battery current I, the ratio of the voltage of the battery cell to the average battery voltage of a normal battery pack Let y be the label vector, where y∈{0,1} N , 0 means normal, 1 means fault.

[0095] Then, using X as input, use f1 filters of size k1 to obtain the feature map C1: C1 = ReLU(X*W1+b1), where * represents the convolution operation, W1 and b1 are the weight and bias of the convolution kernel respectively, ReLU is the activation function, and ReLU(x) = max(0,x).

[0096] Next, perform maximum pooling on C1 to obtain P1, which is the output of the feature extraction layer. Then use f2 filters of size k2 to obtain the feature map C2: C2 = ReLU (P1*W2+b2)

[0097] Then perform maximum pooling on C2 to obtain P2, which is the output of the feature optimization layer. Then flatten P2 into a feature vector F to obtain the improved Pearson correlation coefficient, which is expressed as:

[0098]

[0099] Among them, ρ i Represents the Pearson correlation coefficient vector of the i-th dimension feature, that is, it includes multiple F(), f i,j represents the eigenvalue of the jth sample on the i-th dimension feature, f i,j ∈F, and Represents the i-th dimension feature f i and the mean of the label vector y, represents the weight coefficient of the i-th dimension feature, Represents the weight coefficient of the i-th label.

[0100] It can be understood that the present invention extracts nonlinear features by introducing some CNN mechanisms to better capture the complex relationship between variables, so that the entire network including feature extraction and correlation calculation is trainable, and optimizes the problem that the traditional person coefficient cannot be updated and the data is real-time.

[0101] In addition, if Figure 6 As shown, another embodiment of the present invention further provides an electric vehicle battery fault detection system, preferably using the electric vehicle battery fault detection method as described above, including:

[0102] The data acquisition and processing module collects the characteristic data of each battery cell in the power battery in real time, constructs a characteristic data sequence of each battery cell, and pre-processes and normalizes the collected characteristic data, wherein the characteristic data includes a plurality of electrical characteristic data and a plurality of thermal characteristic data;

[0103] A data dimension reduction processing module is used to perform dimension reduction processing on the normalized feature data sequence of each battery cell to extract the electric and thermal data feature sequence of each battery cell;

[0104] The battery fault diagnosis module is used to calculate the distance between the electrothermal data feature sequences of any two battery cells using the improved Minkowski algorithm. If the distance between any two electrothermal data feature sequences is less than a preset threshold, the battery state is determined to be normal, otherwise the battery is determined to be faulty.

[0105] It can be understood that the electric vehicle battery fault detection system of the present invention first collects the electrical characteristic data and thermal characteristic data of each battery cell in real time and performs normalization processing, then performs dimensionality reduction processing on the characteristic data sequence after normalization processing of each battery cell, extracts the electric and thermal data characteristic sequence of each battery cell, and finally, uses the improved Minkowski algorithm to calculate the distance between the electric and thermal data characteristic sequences of any two battery cells. If the distance between any two electric and thermal data characteristic sequences is less than the preset threshold, the battery state is determined to be normal, otherwise the battery is determined to be faulty. The electric vehicle battery fault detection method of the present invention takes into account the electric and thermal data of multiple dimensions of the battery, which can more comprehensively and accurately reflect the working state of the battery, thereby improving the accuracy of the electric vehicle battery fault diagnosis, and through the dimensionality reduction processing of the collected high-dimensional electric and thermal data, the data processing amount is greatly reduced while retaining the characteristics of the electric and thermal data, and the processing efficiency is improved. At the same time, the distance between the electric and thermal data characteristic sequences is calculated using the improved Minkowski algorithm, so that the distance metric is not affected by the scale of the data in each dimension, and the accuracy of the distance calculation result is improved, thereby further improving the accuracy of the electric vehicle battery fault diagnosis.

[0106] In addition, the electric vehicle battery fault detection system also includes:

[0107] The fault detection result evaluation module is used to evaluate the fault detection result.

[0108] In addition, the electric vehicle battery fault detection system also includes:

[0109] The fault type diagnosis and identification module is used to build a correlation analysis model based on the collected electrothermal characteristic data, and use the improved Pearson correlation coefficient to analyze the correlation between the characteristic data, and determine the fault type according to the correlation results.

[0110] Among them, each module of the system embodiment corresponds to each step of the above method embodiment, so the specific working process and principle of each module will not be repeated here, and the corresponding reference is made to the above method embodiment.

[0111] In addition, another embodiment of the present invention further provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the processor executes the steps of the above method by calling the computer program stored in the memory.

[0112] In addition, another embodiment of the present invention further provides a computer-readable storage medium for storing a computer program for performing electric vehicle battery fault detection, wherein the computer program executes the steps of the method described above when running on a computer.

[0113] The general form of computer readable storage media includes: floppy disk, flexible disk, hard disk, magnetic tape, any other magnetic medium, CD-ROM, any other optical medium, punch cards, paper tape, any other physical medium with a pattern of holes, random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), flash erasable programmable read-only memory (FLASH-EPROM), any other memory chip or cartridge, or any other medium that can be read by a computer. The instructions can further be transmitted or received by a transmission medium. The term transmission medium can include any tangible or intangible medium that can be used to store, encode or carry instructions for execution by a machine, and includes digital or analog communication signals or other intangible media that facilitate the communication of the above instructions. Transmission media include coaxial cables, copper wires and optical fibers, which include the wires of a bus used to transmit a computer data signal.

[0114] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiments of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0115] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0116] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0117] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0118] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0119] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

[0120] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for detecting battery failure in an electric vehicle, characterized in that: Includes the following: Collecting characteristic data of each battery cell in the power battery in real time, constructing a characteristic data sequence of each battery cell, and preprocessing and normalizing the collected characteristic data, wherein the characteristic data includes a plurality of electrical characteristic data and a plurality of thermal characteristic data; Perform dimensionality reduction processing on the normalized feature data sequence of each battery cell to extract the electric and thermal data feature sequence of each battery cell; The improved Minkowski algorithm is used to calculate the distance between the electrothermal data feature sequences of any two battery cells. If the distance between any two electrothermal data feature sequences is less than a preset threshold, the battery state is determined to be normal, otherwise the battery is determined to be faulty.

2. The electric vehicle battery fault detection method according to claim 1, characterized in that: The process of performing dimensionality reduction processing on the normalized feature data sequence of each battery cell to extract the electric and thermal data feature sequence of each battery cell includes the following: Divide all data points in the normalized feature data sequence into two data point sets; Using multiple query data points to search in two data point sets, each data point set is divided into multiple local data sets, each local data set includes (k+1) nearest neighbor points of each query data point; Find the local data set where each data point in the feature data sequence is located, and calculate the weight between the data point and all other data points in the local data set to construct a weight vector; Embed each data point in the feature data sequence from the high-dimensional data space into the low-dimensional data space to obtain an embedded data sequence; Based on the weight vector and the embedded data sequence, the embedding error of the feature data sequence is calculated. If the embedding error is less than a threshold, the embedded data sequence is used as the electrothermal data feature sequence.

3. The electric vehicle battery fault detection method according to claim 2, characterized in that: The process of dividing the normalized feature data sequence into two data point sets includes the following: Calculate the average value of all data points in the feature data sequence in each feature data dimension, and construct the average data point based on the average value of all feature data dimensions; Calculate the similarity between all data points and the average data point; If the similarity between a data point and the average data point is greater than a preset similarity threshold, the data point is divided into the first data point set, otherwise the data point is divided into the second data point set.

4. The electric vehicle battery fault detection method according to claim 3, characterized in that: The similarity between any data point and the average data point is calculated based on the following formula: Among them, d i represents the similarity between the i-th data point and the average data point in the feature data sequence, n represents the feature data dimension, t im represents the value of the feature data of the i-th data point in the m-th dimension, C m Represents the value of the m-th dimension feature data of the average data point.

5. The electric vehicle battery fault detection method according to claim 4, characterized in that: The embedding error is calculated based on the following formula: Where E(T) represents the embedding error, N represents the total number of data points in the feature data sequence, and J i represents the set of k nearest neighbors of the ith data point, w ij represents the weight between the i-th data point and its j-th nearest neighbor, t i represents the embedded data of the i-th data point, t j Represents the embedded data of the jth nearest neighbor.

6. The electric vehicle battery fault detection method according to claim 3, characterized in that: The process of searching in two data point sets using multiple query data points to divide each data point set into multiple local data sets includes the following: Set a query data point, calculate the similarity between the query data point and the average data point, if the similarity between the two is greater than a preset threshold, the query data point is classified into the first data point set, otherwise it is classified into the second data point set, and the (k+1) data point with the greatest similarity to the current query data point is screened out from the classified data point set, and the (k+1) data point is divided into a local data set, and the (k+1) data point is removed from the current data point set; then, the query data point is reset, and the above process is repeated, and it is iterated continuously until each data point set is divided into multiple local data sets, each local data set contains (k+1) data points.

7. The electric vehicle battery fault detection method according to claim 1, characterized in that: The distance between the electrothermal data feature sequences of any two battery cells is calculated based on the following formula: Among them, x and y represent the characteristic sequences of the electric and thermal data of the two battery cells, p is a parameter with a value range of [1, 2, ∞], and x i and i represents the ith data feature value of the electrothermal data feature sequence of battery cell x and battery cell y, respectively. The superscript T represents transposition, and S -1 () represents the inverse matrix of the covariance matrix.

8. An electric vehicle battery fault detection system, using the electric vehicle battery fault detection method according to any one of claims 1 to 7, characterized in that: include: The data acquisition and processing module collects the characteristic data of each battery cell in the power battery in real time, constructs a characteristic data sequence of each battery cell, and pre-processes and normalizes the collected characteristic data, wherein the characteristic data includes a plurality of electrical characteristic data and a plurality of thermal characteristic data; A data dimension reduction processing module is used to perform dimension reduction processing on the normalized feature data sequence of each battery cell to extract the electric and thermal data feature sequence of each battery cell; The battery fault diagnosis module is used to calculate the distance between the electrothermal data feature sequences of any two battery cells using the improved Minkowski algorithm. If the distance between any two electrothermal data feature sequences is less than a preset threshold, the battery state is determined to be normal, otherwise the battery is determined to be faulty.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a computer program, and the processor executes the steps of the method according to 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 performing electric vehicle battery fault detection, characterized in that: When the computer program is run on a computer, the steps of the method according to any one of claims 1 to 7 are executed.

Images