Lithium battery module multi-working-condition multi-fault diagnosis method and system

Abstract

The invention discloses a lithium battery module multi-working-condition multi-fault diagnosis method and system, and belongs to the field of battery pack fault diagnosis. Comprising the steps that under different working conditions, voltage data of each single battery in a battery pack under different faults are collected and preprocessed; calculating the voltage range of a single battery in the battery pack, and extracting the voltage range with a fixed length as an input feature data set through a moving window method; dividing the input feature data set into a training set and a test set, and training an LSTM model by using the training set; inputting the test set into the trained LSTM model, and determining a fault and a fault type according to a model output result; and according to the fault type and the change rule of the maximum voltage battery number and the minimum voltage battery number in the feature window, positioning the fault battery. According to the method, accurate classification diagnosis of multiple faults of the battery pack is realized through the voltage range in combination with the LSTM model, the position of the fault battery is determined, and the safety and reliability of operation of the battery pack are improved.

Description

Technical Field

[0001] The present invention belongs to the field of battery pack fault diagnosis. Specifically, the present invention relates to a method and system for diagnosing multiple faults under multiple working conditions of a lithium battery module. Background Art

[0002] With the wide application of electric vehicles (EVs) and hybrid electric vehicles (HEVs), the importance of lithium batteries as their core power sources has become increasingly prominent. The safety and reliability of lithium battery packs are directly related to the performance of the vehicles and the user experience. However, during actual use, various faults may occur in the battery pack, which not only affect the performance and service life of the batteries but may also cause serious safety problems such as thermal runaway and fires. Therefore, developing effective battery pack fault diagnosis technologies to detect and locate faults in a timely manner is of great significance for ensuring the safe operation of electric vehicles.

[0003] Existing battery pack fault diagnosis technologies mainly focus on model-based methods, data-driven methods, and signal processing-based methods. Model-based methods detect faults by establishing a mathematical model of the battery pack and using the difference between the output of the model and the actual measured values. For example, the equivalent circuit model (ECM) identifies fault states by simulating the voltage and current responses of the battery. However, these methods usually rely on accurate model parameters and have high requirements for the complexity and computational cost of the model, making it difficult to adapt to complex working condition changes in actual applications.

[0004] Therefore, the present invention proposes a method and system for diagnosing multiple faults under multiple working conditions of a lithium battery module. Summary of the Invention

[0005] The present invention aims to overcome the deficiencies of the prior art and proposes a method and system for diagnosing multiple faults under multiple working conditions of a lithium battery module to achieve the following objectives: By calculating the voltage range and combining with the LSTM model, accurate classification diagnosis of multiple faults in the battery pack is realized, and the positions of the faulty batteries are determined, improving the safety and reliability of the battery pack operation.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is: A method for diagnosing multiple faults under multiple working conditions of a lithium battery module, the method comprising the following steps:

[0007] Step S1, under different working conditions, collect the voltage data of each single battery in the battery pack under different faults and perform preprocessing, where the faults include normal state, connection fault, and short circuit fault;

[0008] Step S2, calculate the voltage range of the single batteries in the battery pack, and extract multiple voltage ranges of a fixed length as the input feature dataset by the moving window method;

[0009] Step S3: Divide the input feature dataset into a training set and a test set; construct an LSTM model and train it using the training set;

[0010] Step S4: Input the test set data into the trained LSTM model, and determine whether there is a fault and the type of the fault according to the model output result;

[0011] Step S5: According to the type of the fault, analyze the variation law of the battery numbers with the maximum and minimum voltages within the feature window, and locate the faulty battery.

[0012] Preferably, in step S1, data collection is carried out by constructing an experimental platform; the experimental platform consists of a Shenzhen Yakeyuan BTS60-100-4CH battery pack test system, ternary lithium batteries, a YKYTECH408L high and low temperature test chamber and a host computer. The battery pack is composed of four ternary lithium batteries connected in series, with a rated voltage of 3.7V, a rated capacity of 40Ah, a charging cut-off voltage of 4.2V, and a discharging cut-off voltage of 2.7V; in the experiment, the battery pack is charged and discharged under two working conditions of UDDS and NEDC. Among them, the measurement accuracy of the battery voltage is 0.1%, and the voltage acquisition frequency is 1Hz.

[0013] Preferably, in step S1, the preprocessing includes denoising the originally collected voltage data by using the moving filtering method. The specific formula of the moving filtering method is:

[0014]

[0015] where V filtered (t) represents the filtered voltage at time t, and V(i) represents the originally collected voltage data; W represents the size of the filtering window.

[0016] Preferably, in step S2, the voltage range difference represents the difference between the maximum voltage and the minimum voltage among all the single-cell voltages of the battery pack, that is:

[0017] V r =V max -V min

[0018] where V r is the voltage range difference, V max represents the maximum voltage among all the single-cell voltages, and V min represents the minimum voltage among all the single-cell voltages.

[0019] Preferably, in step S2, the moving window method includes: dividing the voltage range difference sequence into sliding window segments according to a preset window size, and taking the voltage range difference within each window as an input feature.

[0020] Preferably, in the step S3, the LSTM model includes an input layer, an LSTM layer, a Relu activation layer, a Dropout layer, a fully connected layer, a Softmax layer, and an output layer connected in sequence, where:

[0021] The input layer is used to input the extracted voltage difference data;

[0022] The LSTM layer includes multiple neurons and is used to capture the long-term dependencies in the input voltage difference data;

[0023] The Relu activation layer is used to enhance the expression ability of the model by introducing a non-linear activation function;

[0024] The Dropout layer is used to prevent overfitting by randomly discarding neurons;

[0025] The fully connected layer includes multiple output units, corresponding to different fault states respectively, including normal state, connection fault, and short-circuit fault;

[0026] The Softmax layer is used to convert the output result of the fully connected layer into a probability distribution and output the fault type;

[0027] The output layer is used to convert the probability distribution output by the Softmax layer into the final output value.

[0028] Preferably, in the step S4, according to the value output by the LSTM model, it is judged whether there is a fault and the fault type. Among them, an output of 1 represents the normal state; an output of 2 represents a connection fault; an output of 3 represents a short-circuit fault.

[0029] Preferably, in the step S5, the method for determining the position of the faulty battery includes:

[0030] (1) In the case of a connection fault, analyze the change diagram of the maximum voltage battery number and the minimum voltage battery number. If the time when a certain battery number appears in the change diagram is greater than the preset first time threshold, then this battery is the faulty battery;

[0031] (2) In the case of a short-circuit fault, analyze the change diagram of the minimum voltage battery number. If the time when a certain battery number appears in the change diagram is greater than the preset second time threshold, then this battery is the faulty battery.

[0032] (3) In the normal state, the maximum or minimum voltage battery number fluctuates randomly.

[0033] Meanwhile, the present application also proposes a multi-condition and multi-fault diagnosis system for a lithium battery module. The system includes a data acquisition module, a data analysis module, and a fault location module, where:

[0034] The data acquisition module is used to collect the voltage data of each single battery in the battery pack under different faults in different working conditions and perform preprocessing;

[0035] The data analysis module is used to calculate the voltage range and construct an LSTM model to analyze the collected voltage data to determine whether the battery is faulty and the type of fault;

[0036] The fault location module is used to locate the faulty battery according to the type of fault.

[0037] The technical effects of the present invention are as follows:

[0038] (1) The present invention uses the moving filtering method to denoise the original voltage data, reduces the influence of noise on the data, improves the reliability of the data, and helps to improve the fault diagnosis accuracy of the LSTM model;

[0039] (2) The present invention inputs the extracted voltage range characteristics into the LSTM model for training and testing, realizing accurate classification of multiple faults of the battery pack; when a fault is determined to occur, the position of the faulty battery is also determined according to the change of the battery number with the maximum or minimum voltage in the type of fault, with high accuracy.

[0040] (3) The method of the present invention can simultaneously identify multiple types of faults, adapt to the changes between different working conditions, has high fault diagnosis accuracy and model generalization ability, and can effectively improve the safety and reliability of the operation of the battery pack. Description of the Drawings

[0041] Figure 1 It is a flowchart of a method for diagnosing multiple working conditions and multiple faults of a lithium battery module provided by an embodiment of the present invention;

[0042] Figure 2 It is the battery number with the maximum or minimum voltage in case of connection fault provided by an embodiment of the present invention: among them, Figure 2 (a) is a change diagram of the battery number with the maximum voltage in case of connection fault; Figure 2 (b) is a change diagram of the battery number with the minimum voltage.

[0043] Figure 3 It is the battery number with the maximum or minimum voltage in case of short circuit fault provided by an embodiment of the present invention: among them, Figure 3 (a) is a change diagram of the battery number with the maximum voltage in case of connection fault; Figure 3 (b) is a change diagram of the battery number with the minimum voltage.

[0044] Figure 4 It is the battery number with the maximum or minimum voltage in the normal state provided by an embodiment of the present invention: among them, Figure 4 (a) is a change diagram of the battery number with the maximum voltage in case of connection fault; Figure 4(b) is the variation diagram of the minimum voltage battery number. Detailed implementation manner

[0045] The following is a more detailed description of the specific implementation manner of the present invention by describing the embodiments with reference to the accompanying drawings, aiming to help those skilled in the art have a more complete, accurate and in-depth understanding of the inventive concept and technical solution of the present invention, and contribute to its implementation. It should be noted that the terms "first", "second", etc. described in this application are only for the convenience of describing the technical solution to distinguish different components, and do not limit this application. To make the technical solution of the present invention clearer, the present invention is explained and illustrated through the following embodiments.

[0046] This embodiment provides a method for diagnosing multiple working conditions and multiple faults of a lithium battery module, as Figure 1 shown, the method includes the following steps:

[0047] Step S1: Under different working conditions, collect the voltage data of each single battery in the battery pack under different faults and perform preprocessing. The faults include normal state, connection fault and short circuit fault;

[0048] Step S2: Calculate the voltage range of the single battery in the battery pack, and extract multiple voltage ranges of a fixed length as the input feature dataset by the moving window method;

[0049] Step S3: Divide the input feature dataset into a training set and a test set; construct an LSTM model and train it using the training set;

[0050] Step S4: Input the test set data into the trained LSTM (Long Short-Term Memory Network) model, and determine whether there is a fault and the fault type according to the model output result;

[0051] Step S5: According to the fault type, analyze the change rules of the maximum and minimum voltage battery numbers within the feature window to locate the faulty battery.

[0052] Specifically, in step S1 of this embodiment, an experimental platform is first constructed. The experimental platform includes equipment such as a battery pack test system, a battery pack, a high and low temperature test chamber, and a host computer, so that charge and discharge experiments of the battery under different working conditions can be carried out, and data during the experiment can be collected.

[0053] The experimental platform of this embodiment consists of a Shenzhen Yakeyuan BTS60-100-4CH battery pack test system, a ternary lithium battery, a YKYTECH408L high and low temperature test chamber, and a host computer. The battery pack is composed of four ternary lithium batteries connected in series, with a rated voltage of 3.7V, a rated capacity of 40Ah, a charging cut-off voltage of 4.2V, and a discharging cut-off voltage of 2.7V. In the experiment, two working conditions, UDDS and NEDC, are used to charge and discharge the battery pack. Among them, the measurement accuracy of the battery voltage is 0.1%, and the voltage acquisition frequency is 1Hz.

[0054] The original voltage data collected in the experiment contains noise, which will reduce the accuracy of battery fault diagnosis. Therefore, in order to reduce the influence of noise on the data, preprocessing is also required. The preprocessing of this embodiment includes using the moving filtering method to denoise the originally collected voltage data. The specific formula of the moving filtering method is:

[0055]

[0056] Where V filtered (t) represents the filtered voltage at time t, and V(i) represents the originally collected voltage data; W represents the size of the filtering window. In this embodiment, W is selected as 5. Specifically in implementation, it can be flexibly selected according to the actual situation.

[0057] The battery pack is composed of multiple single cells, and there are certain differences in the performance of each single cell. When performing battery fault diagnosis, it is necessary to select representative data features as the basis for diagnosis. The voltage range can intuitively reflect the voltage difference degree between single cells and reflect the consistency of the battery pack. If the voltage range of each single cell in the battery pack is large, it means that the consistency of the battery pack is poor, and there may be performance degradation or faults in some batteries, which will affect the overall performance and life of the battery pack. Therefore, in step S2 of this embodiment, the voltage range is extracted from the data as the characteristic data.

[0058] The voltage range represents the difference between the maximum voltage and the minimum voltage among the voltages of all single cells in the battery pack, that is:

[0059] V r =V max -V min

[0060] Where V r is the voltage range, V max represents the maximum voltage among the voltages of all single cells, and V min represents the minimum voltage among the voltages of all single cells.

[0061] To further extract features, in step S2 of this embodiment, the moving window method is used to divide the voltage range difference into multiple input features. The moving window method includes: sliding window segmentation of the voltage range difference sequence according to a preset window size, and the voltage range difference within each window is used as an input feature, and finally a data set with the voltage range difference as the input feature is obtained. In this embodiment, the window size is initially selected as 50 - 250, and after a large number of experimental verifications, the window size is set to 200 to ensure the reliability of the model input feature data.

[0062] Next, in step S3 of this embodiment, the extracted feature data set is divided into a training set and a test set according to a ratio of 7:3. Among them, the training set is used to train the LSTM model, and the test set is used to verify the performance of the model. The LSTM model constructed in this embodiment includes an input layer, an LSTM layer, a Relu activation layer, a Dropout layer, a fully connected layer, a Softmax layer, and an output layer connected in sequence, where:

[0063] The input layer is used to input the extracted voltage range difference data; the feature length is 200, and the feature dimension is 1;

[0064] The LSTM layer includes multiple neurons, which are set to 100 in this embodiment. The input voltage range difference data in this embodiment is a type of time series data, and the LSTM layer can be used to capture long-term dependencies in time series data with high accuracy;

[0065] The Relu activation layer is used to enhance the expression ability of the model by introducing a non-linear activation function;

[0066] The Dropout layer is used to prevent overfitting by randomly discarding neurons. In this embodiment, 20% of the neurons are randomly discarded;

[0067] The fully connected layer includes multiple output units, corresponding to different fault states respectively; 3 output units are set in this embodiment, corresponding to the normal state, connection fault, and short circuit fault respectively;

[0068] The Softmax layer is used to convert the output result of the fully connected layer into a probability distribution and output the fault type;

[0069] The output layer is used to convert the probability distribution output by the Softmax layer into the final output value.

[0070] Each neuron in the LSTM layer is an LSTM cell. With its unique structures of forget gate, input gate, and output gate, the LSTM cell realizes the effective update and interaction of long-term and short-term memories. Compared with the traditional RNN (Recurrent Neural Network), the LSTM network shows higher accuracy in extracting the features of sequence signals, and thus can achieve higher recognition accuracy and better classification performance, thereby improving the accuracy of fault recognition in this application.

[0071] The first part of the internal unit of the LSTM is the forget gate, which performs a sigmoid function operation based on the hidden state h at the previous moment t-1 and the input X at the current moment t . The specific formula is:

[0072] f1 = sigmoid(w1[h t-1 , X t ′ + b1)

[0073] where w1 is the weight, b1 is the bias vector, and the output range of f1 is [0, 1]. After multiplying f1 by the storage unit at the previous moment, the content in the storage unit can be selectively deleted.

[0074] The second part is the input gate, which performs sigmoid function and tanh function operations based on the hidden state h at the previous moment t-1 and the input state X at the current moment t . The specific formula is:

[0075]

[0076] where the value range of the tanh function is [-1, 1], which sorts and summarizes the data; w2, are the weights, and b2, are the bias vectors. After adding f2 to the storage unit C at the previous moment after deletion, new content can be added, which is specifically expressed as: t-1 O t

[0077] t = σ(w0[h t-1 , X t +b0)

[0078] h t = O t tanh(C t )

[0079] where h t represents the hidden state at the current moment, w0 is the weight, b0 is the bias vector, and C t represents the storage unit at the current moment.

[0080] In step S4 of this embodiment, the test set data is input into the LSTM model, and the LSTM model will output a value. Based on the value output by the LSTM model, it can be determined whether there is a fault and the type of the fault. Among them, an output of 1 indicates a normal state; an output of 2 indicates a connection fault; an output of 3 indicates a short-circuit fault.

[0081] Next, after obtaining the fault type, in step S5 of this embodiment, the position of the faulty battery is determined according to the change of the battery number with the maximum or minimum voltage in the fault characteristics. The specific method includes:

[0082] (1) In the case of a connection fault, the battery that appears for a long time in the change diagram of the battery number with the maximum or minimum voltage is the faulty battery. Specifically, in the case of a connection fault, analyze the change diagrams of the battery numbers with the maximum voltage and the minimum voltage. If the time when a certain battery number appears in the change diagram is greater than a preset first time threshold, then this battery is the faulty battery;

[0083] (2) In the case of a short-circuit fault, the battery that appears most of the time in the change diagram of the minimum voltage is the faulty battery. Specifically, in the case of a short-circuit fault, analyze the change diagram of the battery number with the minimum voltage. If the time when a certain battery number appears in the change diagram is greater than a preset second time threshold, then this battery is the faulty battery.

[0084] (3) In the normal state, the battery numbers with the maximum or minimum voltage show random fluctuations. According to the above method, the faulty battery can be determined. After that, the user can perform subsequent operations such as recording and repairing the fault.

[0085] In this embodiment, as Figure 2 shown, Figure 2 (a) and Figure 2 (b) are respectively the change diagrams of the battery number with the maximum voltage and the minimum voltage in the case of a connection fault. It can be seen from the figures that in the change diagrams of the battery number with the maximum voltage and the minimum voltage, the battery that appears for a long time contains battery No. 1. Therefore, when it is determined that a connection fault has occurred, the faulty battery can be determined according to the battery number that appears for a long time in the change diagram of the maximum or minimum voltage battery number.

[0086] As Figure 3 shown, Figure 3 (a) and Figure 3 (b) are respectively the change diagrams of the battery number with the maximum voltage and the minimum voltage in the case of a short-circuit fault. It can be seen from the figures that most of the time, battery No. 1 appears in the change diagram of the minimum voltage. Therefore, when it is determined that a short-circuit fault has occurred, the faulty battery can be determined according to the battery that exists most of the time in the change diagram of the battery number with the minimum voltage.

[0087] As shown Figure 4 in Figure 4 (a) and Figure 4 (b) are the variation diagrams of the maximum or minimum voltage battery numbers in the normal state. Due to the inconsistency between battery packs, their variations are chaotic.

[0088] Meanwhile, the present application also proposes a multi - condition and multi - fault diagnosis system for a lithium - battery module. The system includes a data acquisition module, a data analysis module, and a fault location module, where:

[0089] The data acquisition module is used to collect the voltage data of each single battery in the battery pack under different faults in different working conditions and perform pre - processing;

[0090] The data analysis module is used to calculate the voltage range and construct an LSTM model to analyze the collected voltage data to judge whether the battery is faulty and the type of fault;

[0091] The fault location module is used to locate the faulty battery according to the type of fault.

[0092] In summary, the method of the present invention first collects the single - cell voltage data of a series battery pack under different fault states and different working conditions, and uses the moving average filter method to denoise the data. Then, calculate the difference between the maximum voltage and the minimum voltage of the single cells in the battery pack, that is, the voltage range, and use the moving window method to divide the voltage range into multiple input features. Then, input the extracted features into the LSTM model for training and testing to achieve accurate classification of multiple faults of the battery pack. Finally, when a fault is determined to occur, determine the position of the faulty battery according to the change of the battery number with the maximum or minimum voltage in the fault characteristics. The present invention realizes accurate classification diagnosis of multiple faults of the battery pack and determines the position of the faulty battery through the voltage range and in combination with the LSTM model, improving the safety and reliability of the operation of the battery pack. The method of the present invention supports cross - condition transfer learning. Through mutual verification of the data of two working conditions, UDDS and NEDC, the fault diagnosis accuracy rate of the LSTM model under different working conditions is not less than 80%.

[0093] The above has described the present invention by way of example with reference to the accompanying drawings. Obviously, the specific implementation of the present invention is not limited by the above - mentioned manner. As long as various non - substantial improvements are made by adopting the method concept and technical solution of the present invention; or without improvement, directly applying the above - mentioned concept and technical solution of the present invention to other occasions, they are all within the protection scope of the present invention.

Claims

1. A multi-operating-condition and multi-fault diagnosis method for a lithium battery module, characterized in that: The method comprises the following steps: Step S1, under different working conditions, collecting and preprocessing the voltage data of each single cell in the battery pack under different faults, the faults including normal state, connection fault and short circuit fault; Step S2, calculating the voltage range difference of the single cells in the battery pack, and extracting multiple fixed-length voltage range differences as input feature data sets by using a moving window method; Step S3, dividing the input feature data set into a training set and a test set; constructing an LSTM model and using the training set for training; Step S4: input the test set data into the trained LSTM model, and determine whether there is a fault and the fault type according to the model output result; Step S5: Analyze the variation pattern of the maximum and minimum voltage battery numbers in the characteristic window according to the fault type, and locate the faulty battery.

2. A multi-operating-condition and multi-fault diagnosis method for a lithium battery module according to claim 1, characterized in that: In the step S1, data collection is performed by constructing an experimental platform; the experimental platform consists of a Shenzhen Yakeyuan BTS60-100-4CH battery pack test system, a ternary lithium battery, a YKYTECH408L high and low temperature test chamber and a host computer. The battery pack consists of four ternary lithium batteries connected in series, with a rated voltage of 3.7V, a rated capacity of 40Ah, a charging cut-off voltage of 4.2V, and a discharging cut-off voltage of 2.7V. In the experiment, the battery pack is charged and discharged under two working conditions, UDDS and NEDC, wherein the battery voltage measurement accuracy is 0.1% and the voltage acquisition frequency is 1Hz.

3. A multi-operating-condition and multi-fault diagnosis method for a lithium battery module according to claim 1 or 2, characterized in that: In step S1, the preprocessing includes performing denoising on the original collected voltage data using a moving filter method. The specific formula of the moving filter method is: Among them, V filtered (t) represents the filtered voltage at time t, V(i) represents the original collected voltage data; W represents the filter window size.

4. A lithium battery module multi-operating condition multi-fault diagnosis method according to claim 1, characterized in that: In step S2, the voltage range difference represents the difference between the maximum voltage and the minimum voltage of all single cell voltages of the battery pack, that is: V r =V max -V min Among them, V r is the voltage difference, V max Indicates the maximum voltage among all single battery voltages, V min Indicates the minimum voltage among all single battery voltages.

5. A multi-operating-condition and multi-fault diagnosis method for a lithium battery module according to claim 1 or 4, characterized in that: In step S2, the moving window method includes: dividing the voltage range sequence into sliding windows according to a preset window size, and taking the voltage range in each window as an input feature.

6. A multi-operating-condition and multi-fault diagnosis method for a lithium battery module according to claim 1, characterized in that: In step S3, the LSTM model includes an input layer, an LSTM layer, a Relu activation layer, a Dropout layer, a fully connected layer, a Softmax layer and an output layer connected in sequence, wherein: The input layer is used to input the extracted voltage range data; The LSTM layer includes a plurality of neurons and is used to capture long-term dependencies in the input voltage range data; The Relu activation layer is used to enhance the expressiveness of the model by introducing a nonlinear activation function; The Dropout layer is used to prevent overfitting by randomly dropping neurons; The fully connected layer includes a plurality of output units corresponding to different fault states, including a normal state, a connection fault, and a short circuit fault; The Softmax layer is used to convert the output result of the fully connected layer into a probability distribution and output the fault type; The output layer is used to convert the probability distribution output by the Softmax layer into a final output value.

7. A lithium battery module multi-operating condition multi-fault diagnosis method according to claim 1 or 6, characterized in that: In step S4, whether there is a fault and the type of fault are determined based on the value output by the LSTM model, wherein an output of 1 indicates a normal state; an output of 2 indicates a connection fault; and an output of 3 indicates a short circuit fault.

8. A multi-operating-condition and multi-fault diagnosis method for a lithium battery module according to claim 1, characterized in that: In step S5, the method for determining the location of the faulty battery includes: (1) In the connection fault state, the change graph of the maximum voltage battery number and the minimum voltage battery number is analyzed. If the time that a certain battery number appears in the change graph is greater than a preset first time threshold, the battery is a faulty battery; (2) Under the short-circuit fault state, the variation graph of the minimum voltage battery number is analyzed. If the time that a certain battery number appears in the variation graph is greater than a preset second time threshold, the battery is a faulty battery. (3) Under normal conditions, the maximum or minimum voltage battery number fluctuates randomly.

9. A lithium battery module multi-operating condition and multi-fault diagnosis system according to any one of claims 1 to 8, characterized in that: The system includes a data acquisition module, a data analysis module, and a fault location module, wherein: The data acquisition module is used to collect voltage data of each single cell of the battery pack under different faults under different working conditions and perform preprocessing; The data analysis module is used to calculate the voltage range and build an LSTM model to analyze the collected voltage data to determine whether the battery is faulty and the type of fault; The fault locating module is used to locate the faulty battery according to the fault type.

Images