Battery internal short circuit fault diagnosis method, equipment, medium and product

Abstract

The invention discloses a battery internal short circuit fault diagnosis method and device, a medium and a product, and relates to the technical field of battery fault diagnosis, and the method comprises the steps: obtaining healthy battery data, and carrying out the preprocessing, and obtaining a training set; establishing a neural network model for fault signal anomaly detection; performing offline training on the neural network model by adopting the training set to obtain an anomaly detection model and setting a safety threshold; performing online diagnosis on real-time battery data by adopting the anomaly detection model to obtain an anomaly score; and comparing the abnormal score with the safety threshold, and judging whether the battery has an internal short-circuit fault or not. According to the battery internal short-circuit fault diagnosis method and device, the medium and the product provided by the invention, the battery internal short-circuit fault can be quickly and accurately diagnosed.

Description

Technical Field

[0001] The present application relates to the technical field of battery fault diagnosis, and in particular to a method, device, medium and product for diagnosing internal short circuit fault of a battery. Background Art

[0002] With the continuous development of electric vehicles and energy storage technology, various types of batteries have a core position in the field of energy storage. Among them, lithium-ion batteries have become the most widely used type of batteries due to their high energy density and high power density. However, with the continuous increase in battery installed capacity, the battery safety that comes with it is the biggest obstacle to its further development. In recent years, battery safety problems have frequently occurred due to failures or other factors. Battery safety is one of the issues that people are most concerned about.

[0003] Thermal runaway of lithium-ion batteries is mainly caused by two aspects: on the one hand, there are problems with the materials and production processes of lithium-ion batteries, and on the other hand, there are problems in the use of lithium-ion batteries. There are many reasons why lithium-ion batteries form thermal runaway during use, such as internal and external short circuits, excessive charging and discharging, high-rate charging and discharging, high and low temperature environments, cycle aging, extrusion deformation, etc. Current research shows that internal short circuits in lithium-ion batteries are the most common cause of thermal runaway. In order to maximize the performance of lithium-ion batteries, improve the safety of lithium-ion batteries and extend their service life, it is necessary to monitor the lithium-ion battery status online and perform internal short circuit diagnosis of lithium-ion batteries at the same time. Internal short circuit diagnosis of lithium-ion batteries is to monitor the temperature, voltage, current and other status information of lithium-ion batteries in real time, and to achieve early warning of internal short circuits in lithium-ion batteries through certain model algorithms.

[0004] At present, researchers have proposed some methods for detecting and diagnosing internal short circuits in lithium-ion batteries from different perspectives. For example: (1) Statistical methods based on battery voltage and current signals; (2) Statistical methods based on battery pack consistency; (3) Methods based on thermoelectric characteristic threshold diagnosis; (4) Methods based on external auxiliary measurement circuits; (5) Fault warning methods based on battery models, etc. Among them, the statistical method based on battery voltage and current signals is simple and easy to apply. It only collects the input excitation signal and response signal of the battery, and can achieve early warning of internal short circuit faults in the battery through algorithms. It is widely used in engineering. Up to now, with the improvement of computing power and storage capacity, methods based on machine learning are also widely used in the field of fault diagnosis. However, most of the current methods have the problems of low diagnostic accuracy and long diagnostic time. Summary of the invention

[0005] The purpose of this application is to provide a battery internal short circuit fault diagnosis method, equipment, medium and product, which can realize rapid and accurate diagnosis of battery internal short circuit fault.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a method for diagnosing a battery internal short circuit fault, comprising:

[0008] Obtain healthy battery data and preprocess it to obtain a training set;

[0009] Establish a neural network model for fault signal anomaly detection;

[0010] Using the training set to perform offline training on the neural network model, obtain an anomaly detection model and set a safety threshold;

[0011] The anomaly detection model is used to perform online diagnosis on real-time battery data to obtain an anomaly score;

[0012] The abnormal score is compared with the safety threshold to determine whether the battery has an internal short circuit fault.

[0013] Optionally, healthy battery data is obtained and preprocessed to obtain a training set, specifically including:

[0014] Acquire healthy battery data; the battery data includes: battery voltage and current signals;

[0015] The healthy battery data is preprocessed using a Z-score standardization method, and a training set is constructed.

[0016] Optionally, the neural network model is an Auto-Encoder model; the Auto-Encoder model includes: an encoder and a decoder; the encoder includes: several layers of LSTM networks; the encoder is used to encode the input data, and map the input data to a latent space to obtain a latent code; the decoder is used to decode the latent code, and map the latent code to an actual space to obtain reconstructed data; the mean square error between the input data and the reconstructed data is the reconstruction residual.

[0017] Optionally, the neural network model is trained offline using the training set to obtain an anomaly detection model and set a safety threshold, specifically including:

[0018] The healthy battery data in the training set is input into the neural network model for encoding and decoding, and the reconstruction residual during offline training is calculated; the reconstruction residual during offline training is the reconstruction residual corresponding to the healthy battery data;

[0019] Using the neural network model when the reconstructed residual during the offline training converges as an anomaly detection model;

[0020] The maximum value of the reconstructed residual during offline training after convergence of a set multiple is set as a safety threshold; the set multiple is 2-3 times.

[0021] Optionally, the anomaly detection model is used to perform online diagnosis on real-time battery data to obtain an anomaly score, specifically including:

[0022] The preprocessed real-time battery data is input into the abnormality detection model for encoding and decoding, and the reconstruction residual during online diagnosis is calculated; the reconstruction residual during online diagnosis is the reconstruction residual corresponding to the real-time battery data;

[0023] The reconstructed residual during the online diagnosis is used as the abnormality score.

[0024] Optionally, comparing the abnormality score with the safety threshold to determine whether the battery has an internal short circuit fault specifically includes:

[0025] comparing the anomaly score to the safety threshold;

[0026] If the abnormal score exceeds the safety threshold, it is determined that the battery has an internal short circuit fault;

[0027] If the abnormal score does not exceed the safety threshold, it is determined that the battery does not have an internal short circuit fault.

[0028] Optionally, the battery internal short circuit fault diagnosis method further includes:

[0029] If the battery has an internal short circuit fault, an early warning will be issued;

[0030] If the battery does not have an internal short circuit fault, continue monitoring.

[0031] In a second aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the battery internal short circuit fault diagnosis method.

[0032] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the battery internal short circuit fault diagnosis method when executed by a processor.

[0033] In a fourth aspect, the present application provides a computer program product, including a computer program, which implements the battery internal short circuit fault diagnosis method when executed by a processor.

[0034] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0035] The present application provides a battery internal short circuit fault diagnosis method, equipment, medium and product. By using healthy battery data to train the neural network model offline, an abnormal detection model is obtained and a safety threshold is set, so that the abnormal detection model can be directly used to perform online diagnosis on real-time battery data to obtain an abnormal score, and then combined with the safety threshold to determine whether the battery has an internal short circuit fault. Compared with the prior art, the present application only requires healthy battery data as a training set without abnormal data, and the data acquisition difficulty is low, which is more in line with the actual situation; and the present application uses an abnormal detection model and a safety threshold to achieve rapid and accurate diagnosis of battery internal short circuit faults and early warning of faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 A flow chart of a battery internal short circuit fault diagnosis method provided in this application;

[0038] Figure 2 A specific flow chart of the battery internal short circuit fault diagnosis method provided in this application;

[0039] Figure 3 A schematic diagram of the pre-processing process provided for this application;

[0040] Figure 4 The structure diagram of the Auto-Encoder model provided for this application;

[0041] Figure 5 The structural diagram of the LSTM network provided for this application. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0043] At present, most battery internal short-circuit diagnosis methods have problems such as low diagnostic accuracy, long diagnosis time, high false alarm rate, complex algorithm, and difficulty in practical application. In order to solve the above problems, the present application provides a battery internal short-circuit fault diagnosis method, equipment, medium and product, which is based on data-driven machine learning method, has high accuracy, fast diagnosis speed, good practicality, and can perform early abnormal detection and safety warning of battery short-circuit faults.

[0044] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0045] In an exemplary embodiment, the present application provides a method for diagnosing a battery internal short circuit fault, such as Figure 1 and Figure 2 As shown, the method includes the following steps 1 to 5.

[0046] Step 1: Obtain healthy battery data and preprocess it to obtain a training set.

[0047] The battery data includes: voltage and current signals of the battery. A large amount of offline healthy battery data is preprocessed to make the input data conform to the input structure required by the model.

[0048] Because the networks used for training are linear networks and LSTM (Long Short-Term Memory) networks, which require that the input data must be normalized, the Z-score normalization method is used to preprocess the healthy battery data to meet the requirements of the model, as shown in the following formula:

[0049] Z=(XE[X] / σ(X)

[0050] Where Z represents the detrended data, X represents the original data, E[X] and σ(X) represent the mean and standard deviation of the original data, respectively. The preprocessing process and results are shown in Figure 3 shown.

[0051] Step 2: Establish a neural network model for fault signal anomaly detection.

[0052] The neural network model is an Auto-Encoder model built based on pytorch. It takes the voltage and current of the battery at a certain moment after preprocessing as input, and takes the reconstructed residual corresponding to the moment, that is, the abnormal score, as output. Its structure is as follows: Figure 4 As shown. The Auto-Encoder model includes: an encoder and a decoder. The encoder includes: several layers of LSTM networks. The encoder is used to input data, that is, from the actual time series signal input Encode, map the input data to the latent space to obtain the latent code ε(x); the decoder is used to decode the latent code, map the latent code to the actual space to obtain the reconstructed data Where G() and ε() represent the mapping relationship between the decoder and the encoder in different spaces. At this time, the dimension of the decoded data is consistent with the actual input dimension. The mean square error between the input data and the reconstructed data is the reconstruction residual.

[0053] Assuming x is the input of time space, the output of the encoder ε() is:

[0054] z1=ε(x1)

[0055] in, It represents the hidden code distribution of the latent space. The size and dimension of the encoded z1 are different from the original input x1, but it contains the information of the original input. In fact, the encoder ε() is equivalent to two space mappings, that is, the mapping relationship from the actual space to the hidden space. Similarly, for the decoder G(), it represents the mapping relationship from the latent space to the actual space:

[0056] x2=g(z2)

[0057] Among them, z2 represents the latent space input, and x2 represents the decoded output. For a time series x in the actual space, the encoding and decoding processes can be expressed as:

[0058]

[0059] in, Represents the reconstructed data, which is generated by encoding and decoding the input data, and has the same size and dimension as the input data. The residual of the two, i.e. the reconstructed residual Res, is defined as follows:

[0060]

[0061] Among them, MSE represents mean square error. It is worth noting that the principle of an effective Auto-Encoder for anomaly monitoring is: when abnormal data is input into the model, some information will be lost when it is encoded, so the decoded and reconstructed data will be inconsistent with the original input at this moment. At this time, the residual of the two can be used as a basis for judging whether a fault has occurred. Therefore, two anomaly scores S are defined train With S det , which are equal to the reconstruction residual Res during offline training and online diagnosis respectively. train Represents the reconstructed residual during offline training. The residual Res after model training convergence is low, that is, S train is also low, and the threshold τ can be set at this time:

[0062] τ=K·max{S train}

[0063] Among them, K can generally be 2 to 3. Finally, in the online monitoring process, once there is abnormal data, the residual Res at this time, that is, the abnormal score S det It will increase rapidly and exceed the threshold, at which point it is determined to be an abnormal fault. Figure 2 .

[0064] Since the input data can be considered as a time series signal, the LSTM network is used as the basic network composition in the encoder, and its structure is as follows Figure 5 The specific operation process is described as follows:

[0065]

[0066] Among them, σ represents the sigmoid function and tanh represents the hyperbolic tangent function: f t 、i t and t Respectively represent the forget gate, input gate and output gate at the tth sampling moment. t represents the long-term memory at the t-th sampling moment, represents the short-term memory at the t-th sampling moment, h t represents the hidden layer at the tth sampling moment. Usually, a single-layer LSTM only outputs h at the last step. t V and W represent the weight parameter matrices to be trained, b represents the bias parameter, Indicates the multiplication relationship between matrix elements. In addition, subscripts f, i, o, and c represent the forget gate, input gate, output gate, and memory unit, respectively. Subscript t represents the tth sampling moment, and subscript m represents the length of the entire LSTM input time series.

[0067] Step 3: Use the training set to perform offline training on the neural network model to obtain an anomaly detection model and set a safety threshold.

[0068] Specifically, the healthy battery data in the training set is input into the neural network model for encoding and decoding, and the reconstruction residual during offline training is calculated; the reconstruction residual during offline training is the reconstruction residual corresponding to the healthy battery data; the neural network model when the reconstruction residual during offline training converges is used as an abnormality detection model; the maximum value of the reconstruction residual during offline training after convergence of a set multiple is set as a safety threshold; the set multiple is 2-3 times.

[0069] Step 4: Use the anomaly detection model to perform online diagnosis on the real-time battery data to obtain an anomaly score.

[0070] Specifically, the preprocessed real-time battery data is input into the anomaly detection model for encoding and decoding, and the reconstruction residual during online diagnosis is calculated; the reconstruction residual during online diagnosis is the reconstruction residual corresponding to the real-time battery data; the reconstruction residual during online diagnosis is used as the anomaly score.

[0071] The above steps 3 and 4 are performed under offline and online conditions respectively, that is, the anomaly detection model is pre-trained offline, and the real-time battery data (online battery data) is directly used as the input of the model without training online. The model output value (anomaly score) during offline training is considered to be the safety threshold.

[0072] Step 5: Compare the abnormal score with the safety threshold to determine whether the battery has an internal short circuit fault.

[0073] Specifically, the abnormality score is compared with the safety threshold; if the abnormality score exceeds the safety threshold, it is determined that the battery has an internal short circuit fault (it is considered that the battery has a potential safety fault or has already failed); if the abnormality score does not exceed the safety threshold, it is determined that the battery does not have an internal short circuit fault. Further, if the battery has an internal short circuit fault, an early warning is issued; if the battery does not have an internal short circuit fault, monitoring continues.

[0074] In summary, this application first pre-processes a large amount of normal healthy battery data (no faults), and then builds a neural network model for fault signal anomaly detection based on a data-driven method, and uses healthy battery data to train the model offline. Finally, the online real-time battery data (unknown faults) is input into the pre-trained model for detection, and it is promptly determined whether it is potential abnormal fault data, and screened. The battery internal short circuit fault diagnosis method provided in this application is an effective and fast battery early warning safety strategy, which can capture abnormal voltages in the early stages of the fault method and provide timely safety warnings, and has high potential application value in the future.

[0075] Compared with the prior art, this application has the following advantages:

[0076] 1) The neural network used in this application has a simple structure and low data requirements during training. It only requires normal battery data as a training set without abnormal data, which is in line with the actual situation.

[0077] 2) This application can respond to potential faults quickly and accurately, and can provide early warning of faults. At the same time, this application conforms to the current development trend of artificial intelligence, combines with reality, and has great potential application value in the future.

[0078] In an exemplary embodiment, the present application further provides a computer device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0079] In an exemplary embodiment, the present application further provides a computer-readable storage medium storing a computer program, which implements the steps in the above-mentioned method embodiments when executed by a processor.

[0080] In an exemplary embodiment, the present application further provides a computer program product, including a computer program, which implements the steps in the above-mentioned method embodiments when executed by a processor.

[0081] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant legal provisions.

[0082] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0083] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0084] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0085] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for diagnosing a battery internal short circuit fault, characterized in that: include: Obtain healthy battery data and preprocess it to obtain a training set; Establish a neural network model for fault signal anomaly detection; Using the training set to perform offline training on the neural network model, obtain an anomaly detection model and set a safety threshold; The anomaly detection model is used to perform online diagnosis on real-time battery data to obtain an anomaly score; The abnormal score is compared with the safety threshold to determine whether the battery has an internal short circuit fault.

2. The battery internal short circuit fault diagnosis method according to claim 1, characterized in that: Obtain healthy battery data and preprocess it to obtain a training set, including: Acquire healthy battery data; the battery data includes: battery voltage and current signals; The healthy battery data is preprocessed using a Z-score standardization method, and a training set is constructed.

3. The battery internal short circuit fault diagnosis method according to claim 1, characterized in that: The neural network model is an Auto-Encoder model; the Auto-Encoder model includes: an encoder and a decoder; the encoder includes: several layers of LSTM networks; the encoder is used to encode the input data, and map the input data to a latent space to obtain a latent code; the decoder is used to decode the latent code, and map the latent code to an actual space to obtain reconstructed data; the mean square error between the input data and the reconstructed data is a reconstruction residual.

4. The battery internal short circuit fault diagnosis method according to claim 3, characterized in that: The training set is used to perform offline training on the neural network model to obtain an anomaly detection model and set a safety threshold, specifically including: The healthy battery data in the training set is input into the neural network model for encoding and decoding, and the reconstruction residual during offline training is calculated; the reconstruction residual during offline training is the reconstruction residual corresponding to the healthy battery data; Using the neural network model when the reconstructed residual during the offline training converges as an anomaly detection model; The maximum value of the reconstructed residual during offline training after convergence of a set multiple is set as a safety threshold; the set multiple is 2-3 times.

5. The battery internal short circuit fault diagnosis method according to claim 3, characterized in that: The anomaly detection model is used to perform online diagnosis on real-time battery data to obtain an anomaly score, which specifically includes: The preprocessed real-time battery data is input into the abnormality detection model for encoding and decoding, and the reconstruction residual during online diagnosis is calculated; the reconstruction residual during online diagnosis is the reconstruction residual corresponding to the real-time battery data; The reconstructed residual during the online diagnosis is used as the abnormality score.

6. The battery internal short circuit fault diagnosis method according to claim 1, characterized in that: Comparing the abnormal score with the safety threshold to determine whether the battery has an internal short circuit fault specifically includes: comparing the anomaly score with the safety threshold; If the abnormal score exceeds the safety threshold, it is determined that the battery has an internal short circuit fault; If the abnormal score does not exceed the safety threshold, it is determined that the battery does not have an internal short circuit fault.

7. The battery internal short circuit fault diagnosis method according to claim 1, characterized in that: Also includes: If the battery has an internal short circuit fault, an early warning will be issued; If the battery does not have an internal short circuit fault, continue monitoring.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the battery internal short circuit fault diagnosis method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the battery internal short circuit fault diagnosis method described in any one of claims 1 to 7 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the battery internal short circuit fault diagnosis method described in any one of claims 1 to 7 is implemented.

Images