Battery current fault detection method and device, medium and equipment

Abstract

The invention belongs to the technical field of battery management, and provides a battery current fault detection method and device, a medium and equipment, and the method comprises the steps: obtaining a current signal and current characteristics of a to-be-detected battery; determining a wavelet basis function and a decomposition level based on the current signal through a first neural network model, performing wavelet transformation on the current signal based on the determined wavelet basis function and the decomposition level, extracting a time-frequency domain feature, and obtaining a depth feature through a second neural network model based on the time-frequency domain feature; and based on the current feature, the time-frequency domain feature and the depth feature, obtaining a fault category through a classifier. The method is helpful for detecting transient fault features, effectively improves the flexibility and accuracy of time-frequency domain feature extraction, and can efficiently and accurately recognize the current fault of the battery.

Description

Technical Field

[0001] The present invention belongs to the technical field of battery management, and in particular relates to a battery current fault detection method, device, medium and equipment. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Abnormal changes in battery current may lead to performance degradation or even safety issues. Therefore, battery current fault detection is crucial for the battery management system (BMS). In the prior art, relying solely on cloud or edge computing for fault detection has the problem of insufficient real-time performance or limited computing resources. At this stage, the battery cells in the industry are developing rapidly, and current fault detection and processing are particularly important.

[0004] More importantly, existing battery current fault detection methods all use deep learning models to directly extract current signals, which makes it difficult to focus on transient anomalies in current signals. Although wavelet transform can extract local frequency changes in current signals, its wavelet basis function and decomposition scale rely on manual selection and cannot be applied to complex and changeable current signals. Summary of the invention

[0005] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a battery current fault detection method, device, medium and equipment, which can adaptively select wavelet basis functions and decomposition scales instead of relying on manual selection, and can more accurately analyze the local frequency changes of current signals, which is helpful to detect short-term fault characteristics, effectively improves the flexibility and accuracy of time-frequency domain feature extraction, and can efficiently and accurately identify battery current faults.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A first aspect of the present invention provides a battery current fault detection method.

[0008] A battery current fault detection method, comprising:

[0009] Obtaining the current signal and current characteristics of the battery to be tested;

[0010] Based on the current signal, after determining the wavelet basis function and the decomposition level through the first neural network model, the current signal is subjected to wavelet transform based on the determined wavelet basis function and the decomposition level to extract the time-frequency domain features, and based on the time-frequency domain features, the second neural network model is used to obtain the deep features;

[0011] Based on current features, time-frequency domain features and depth features, the fault category is obtained through the classifier.

[0012] Furthermore, the current characteristics include current mean, current standard deviation, current change rate, skewness and peak value.

[0013] Furthermore, the training process of the first neural network model, the second neural network model and the classifier includes:

[0014] Acquire a training set, wherein the training set records the current signal and the actual fault category of each sample;

[0015] For each current signal in the training set, extract the current feature;

[0016] For each sample, based on the current signal, after the wavelet basis function and decomposition level are determined through the first neural network model, the current signal is subjected to wavelet transform based on the determined wavelet basis function and decomposition level to extract the time-frequency domain features, and based on the time-frequency domain features, the second neural network model is used to obtain the deep features; based on the current features, time-frequency domain features and deep features, the fault category is obtained through the classifier;

[0017] Based on the actual fault category and the fault category obtained by the classifier, after the loss is calculated, it is determined whether the termination condition is met; if so, the trained first neural network model, the second neural network model and the classifier are output; otherwise, the parameters of the first neural network model, the second neural network model and the classifier are updated, and the wavelet basis function and the decomposition level are returned to be re-determined.

[0018] Furthermore, the current signal is collected by a battery management system; the current feature is extracted by an edge computing node deployed on the vehicle side; the first neural network model, the second neural network model and the classifier are all deployed on a cloud server.

[0019] Furthermore, the fault categories include normal, overcurrent fault, abnormal current fluctuation and continuous current drop.

[0020] A second aspect of the present invention provides a battery current fault detection device.

[0021] A battery current fault detection device, comprising:

[0022] A data acquisition module, which is configured to: acquire a current signal and current characteristics of a battery to be detected;

[0023] The feature extraction module is configured to: determine the wavelet basis function and the decomposition level through the first neural network model based on the current signal, perform wavelet transform on the current signal based on the determined wavelet basis function and the decomposition level, extract time-frequency domain features, and obtain deep features through the second neural network model based on the time-frequency domain features;

[0024] The fault detection module is configured to obtain the fault category through a classifier based on current features, time-frequency domain features and depth features.

[0025] Furthermore, it also includes a model training module, which is configured to:

[0026] Acquire a training set, wherein the training set records the current signal and the actual fault category of each sample;

[0027] For each current signal in the training set, extract the current feature;

[0028] For each sample, based on the current signal, after the wavelet basis function and decomposition level are determined through the first neural network model, the current signal is subjected to wavelet transform based on the determined wavelet basis function and decomposition level to extract the time-frequency domain features, and based on the time-frequency domain features, the second neural network model is used to obtain the deep features; based on the current features, time-frequency domain features and deep features, the fault category is obtained through the classifier;

[0029] Based on the actual fault category and the fault category obtained by the classifier, after the loss is calculated, it is determined whether the termination condition is met; if so, the trained first neural network model, the second neural network model and the classifier are output; otherwise, the parameters of the first neural network model, the second neural network model and the classifier are updated, and the wavelet basis function and the decomposition level are returned to be re-determined.

[0030] Furthermore, the current signal is collected by a battery management system; the current feature is extracted by an edge computing node deployed on the vehicle side; and the feature extraction module and the fault detection module are both deployed on a cloud server.

[0031] A third aspect of the present invention provides a computer-readable storage medium.

[0032] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in a battery current fault detection method as described in the first aspect above.

[0033] A fourth aspect of the present invention provides a computer device.

[0034] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the battery current fault detection method as described in the first aspect above are implemented.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] The present invention can adaptively select wavelet basis functions and decomposition scales instead of relying on manual selection, can more accurately analyze local frequency changes in current signals, helps detect transient fault features, effectively improves the flexibility and accuracy of time-frequency domain feature extraction, and can efficiently and accurately identify battery current faults.

[0037] The present invention further extracts higher-level features from the time-frequency domain features through a neural network model, and fuses them with current features and time-frequency domain features to perform fault identification, thereby further optimizing the performance of fault detection or signal classification tasks.

[0038] The cloud-edge collaborative computing model of the present invention optimizes the utilization of computing resources and realizes efficient data processing and intelligent analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of 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 creative work.

[0040] Figure 1 is a flow chart of a battery current fault detection method shown in Embodiment 1 of the present invention;

[0041] Figure 2 1 is a schematic diagram of the process of selecting wavelet basis functions and their parameters shown in the first embodiment of the present invention;

[0042] Figure 3 is a schematic diagram of wavelet transform shown in the first embodiment of the present invention;

[0043] Figure 4 is a schematic diagram of deep feature extraction shown in the first embodiment of the present invention;

[0044] Figure 5 is a schematic diagram of an end-to-end deep learning model shown in Embodiment 1 of the present invention;

[0045] Figure 6 It is a structural diagram of a computer device shown in Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0046] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0047] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0048] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0049] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the methods and devices according to various embodiments of the present invention. It should be noted that each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the module, the program segment, or a part of the code may include one or more executable instructions for implementing the logical functions specified in the various embodiments. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the flowchart and / or block diagram, and the combination of boxes in the flowchart and / or block diagram can be implemented using a dedicated hardware-based device that performs a specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0050] Embodiment 1

[0051] This embodiment provides a battery current fault detection method.

[0052] A battery current fault detection method provided in this embodiment aims to achieve efficient current fault detection and processing for new power battery packs through cloud-edge collaborative computing and a support vector machine (SVM) algorithm, so as to improve the real-time performance and reliability of the battery management system, and at the same time propose fault handling measures for a certain battery pack.

[0053] This embodiment provides a battery current fault detection method, such as Figure 1 As shown, the following steps are included:

[0054] Step 1: Building or training a feature extraction model and a fault classification model.

[0055] Step 101: Obtain a labeled training data set, and filter out the abnormal numbers in the training data set through Z value centering and normalization processing.

[0056] In this embodiment, data labeling includes: selecting battery data under known conditions from historical data, labeling normal and fault conditions, and forming a training data set; for each data sample, recording its current data and corresponding labels (normal, excessive current fault, abnormal current fluctuation, and continuous current decline).

[0057] Step 102: extracting current characteristics such as current mean, current standard deviation, current change rate, skewness and peak value from the preprocessed current data.

[0058] The current feature extraction in this embodiment extracts fault-related features from the current signal, including the current mean, current standard deviation, current change rate, etc.

[0059] (1) Current average:

[0060]

[0061] (2) Current standard deviation:

[0062]

[0063] (3) Current change rate:

[0064]

[0065] (4) Skewness: It reflects the asymmetry of signal distribution. The current signal may show asymmetric changes when a fault occurs, especially in short-term abnormalities, when the skewness change may be significant.

[0066] (5) Kurtosis: It is used to measure the degree of spikes in the signal. Faults may cause extreme peaks in the current signal. Changes in kurtosis can be used to detect these transient anomalies.

[0067] Among them, the extraction of skewness and kurtosis can be done using the following code:

[0068] import numpy as np

[0069] from scipy.stats import skew,kurtosis

[0070] #Assume curr_signal is the current signal

[0071] curr_signal = np.random.randn(1000) # Example signal

[0072] # Calculate skewness and kurtosis

[0073] curr_skewness=skew(curr_signal)

[0074] curr_kurtosis=kurtosis(curr_signal)

[0075] print(f"Skewness: {curr_skewness}, Kurtosis: {curr_kurtosis}").

[0076] Step 103: extract frequency domain features such as spectrum energy and harmonic components from the preprocessed current data.

[0077] In this embodiment, the frequency domain features are extracted using wavelet transform (WT). Wavelet transform is a time-frequency analysis method that is more suitable for processing non-stationary signals than fast Fourier transform (FFT). By using wavelet transform, the local frequency changes of the current signal can be analyzed more carefully, which is helpful to detect short-term fault characteristics.

[0078] In this embodiment, deep learning is combined with adaptive selection of wavelet transform. By training the neural network model, the wavelet basis function and its corresponding scale (decomposition level) that best suits the input signal can be automatically selected to better extract features. Suppose there is a task of fault signal detection. Normally, appropriate wavelet basis functions (such as Daubechies, Symlet, Coiflet, etc.) may be manually selected for wavelet transform to extract features. However, in different signal scenarios, fixed wavelet basis functions and decomposition levels may not always be optimal. Through deep learning, the neural network model can adaptively select the best wavelet basis function and its decomposition scale.

[0079] In this embodiment, a deep neural network is designed, combining the adaptive selection of wavelet transform with the feature extraction capability of convolutional neural network (CNN). The specific process is as follows:

[0080] (1) Signal input: Input the original time domain signal (time series signal), that is, the preprocessed current signal.

[0081] Assume that X = [x_1, x_2, ..., x_n], which represents a fault detection signal.

[0082] (2) Automatic wavelet basis function selection:

[0083] A neural network model (the first neural network model) is used to predict the wavelet basis function and decomposition level that best suits the input signal. This neural network model receives the input signal and learns to output the optimal wavelet basis function and its parameters.

[0084] In this embodiment, a simple fully connected neural network or convolutional neural network is established, the input signal X is output, and the selection parameters and decomposition scale of the wavelet basis function are output. Assuming that a simple fully connected network is used, the selection process of the wavelet basis function and its parameters is as follows: Figure 2 As shown; alternatively, the wavelet basis function can be selected through a classification model, or the decomposition scale can be determined through a regression model.

[0085] (3) Wavelet transform layer: According to the wavelet basis function and decomposition scale predicted by the neural network model, the input current signal is transformed by wavelet transform to extract the time-frequency domain features.

[0086] The wavelet transform step in step (3) is as follows: Figure 3 As shown, including:

[0087] (301) Obtain the wavelet basis function (such as Morlet, Daubechies, Coiflet, etc.) obtained in step (2).

[0088] (302) Decomposition: The current signal is decomposed into different scales and frequencies through discrete wavelet transform (DWT), which can analyze the local characteristics of the current signal in different frequency bands.

[0089] (303) Feature extraction: Wavelet coefficients are extracted at different scales, and then features such as energy and entropy are calculated based on these wavelet coefficients. These features can be used as input for fault detection models.

[0090] Step 104: Input the frequency domain features of the signal after wavelet transformation into a convolutional neural network (CNN) to further extract higher-level features (deep features), such as Figure 4 shown.

[0091] Step 105: Use a classifier to output a fault category or other target label based on the current features, the time-frequency domain features, and the depth features.

[0092] Among them, the classifier can use a fully connected layer and Softmax, etc.; an SVM model can also be used.

[0093] The SVM model selects the RBF kernel function and optimizes the hyperparameters C and γ. The SVM optimization problem formula is:

[0094]

[0095] y i (w T φ(x i )+b)≥1-ξ i

[0096] ξi ≥0

[0097] Among them, w is the weight vector, b is the bias, ξi is the slack variable, C is the penalty parameter, and φ(xi) is the characteristic function mapped to the high-dimensional space.

[0098] Step 106, calculate the loss function based on the actual fault category and the fault category obtained in step 105; determine whether the end condition is met (such as the number of iterations reaches the set value), if so, output the trained neural network model, convolutional neural network and classifier; otherwise, update the parameters of the neural network model, convolutional neural network and classifier, and return to step 103.

[0099] If the classifier is an SVM model, its parameters are hyperparameters C and γ, and its classification decision function is:

[0100]

[0101] Among them, α is the Lagrange multiplier, K is the kernel function, and b is the bias.

[0102] The parameters of the neural network model, convolutional neural network and classifier are updated through back propagation. The update of the neural network model can optimize the wavelet basis function selection and decomposition scale parameters, so that the neural network model can automatically select the optimal wavelet parameters in a specific task and obtain the best classification or prediction results.

[0103] It can be seen that this embodiment combines automatic wavelet selection and CNN to build an end-to-end deep learning model. Figure 5 shown.

[0104] In this way, the model can adaptively select the wavelet basis function and decomposition scale instead of relying on manual selection, effectively improving the flexibility and accuracy of feature extraction. In practical applications, this method can capture different features of the signal and further optimize the performance of fault detection or signal classification tasks through deep learning.

[0105] Step 2: Deploy on the cloud.

[0106] The edge computing node is deployed on the vehicle side to perform preliminary data processing and current feature extraction.

[0107] The cloud big data database is configured with a trained first neural network model, a convolutional neural network (second neural network model) and a classifier.

[0108] The feature data is transmitted to the vehicle T-box via the CAN network, and then uploaded to the cloud database by the T-box. The cloud server runs the trained first neural network model, convolutional neural network (second neural network model) and classifier in real time to perform fault detection on the uploaded data and output the predicted label and fault probability.

[0109] Step 3: Data collection and preprocessing.

[0110] A new type of battery pack is installed on an electric car, and the voltage, current, temperature, SOC and SOH data are collected through the BMS (battery management system). The data is collected once a second and processed through Z value centering and normalization to filter out abnormal data.

[0111] In this embodiment, a new type of battery for a PHEV model is used, and the following are the basic parameters of the battery: voltage range: 2.5V-4.2V; current range: -100A-100A; temperature range: -20℃-60℃; SOC range: 0%-100%; SOH range: 60%-100%.

[0112] In this embodiment, the battery management software in the vehicle-side controller collects battery data in real time, including voltage signals, current signals, SOC, SOH, etc.; the data acquisition module is used to monitor the battery to ensure the continuity and real-time nature of the data.

[0113] In this embodiment, an anomaly detection algorithm is applied to filter abnormal data in real time, and Z value centering and normalization are performed; the centering formula is: z=(x-μ) / σ, where x is the original data, μ is the data mean, and σ is the data standard deviation.

[0114] Step 4: Real-time detection.

[0115] Step 401, edge computing node: deploy edge computing nodes on the vehicle side to perform preliminary data processing and current feature extraction to reduce the burden of data transmission. At the same time, the vehicle side uses the traditional ON-OFF algorithm to handle faults.

[0116] Step 402, data transmission and storage: The processed battery signal and current characteristic data are transmitted to the vehicle T-box via the CAN network, and then uploaded to the cloud big database by the T-box.

[0117] (3) Real-time detection:

[0118] The cloud server runs the trained first neural network model, convolutional neural network (second neural network model) and classifier in real time, performs fault detection on the uploaded data, and outputs the detection results.

[0119] The output results of the classifier include: predicted label (normal or faulty), fault probability (such as the probability of a fault occurring is 95%).

[0120] Step 5: Troubleshooting and feedback.

[0121] After detecting a current fault, a processing plan is automatically generated and sent to the vehicle end for execution, including power reduction operation, system restart, etc. The corresponding processing plan is generated according to the fault type and sent to the vehicle end for execution: in the case of an overcurrent fault (such as a current greater than 300A), the power is reduced to reduce the maximum output power by 50%; in the case of abnormal current fluctuations (such as the current standard deviation exceeds the set threshold), the system enters the protection mode, stops the battery charging and discharging operation, and issues a warning to the user; in the case of a continuous decrease in current (such as a continuous negative current change rate that exceeds the set time), the BMS is automatically restarted and self-checked.

[0122] Example of troubleshooting solution:

[0123] (1) Overcurrent fault: P new =0.5×P max ;

[0124] (2) Abnormal current fluctuation: stop charging and discharging, and issue a warning;

[0125] (3) The current continues to decrease: Restart the BMS and the self-test process begins;

[0126] User feedback: Provides real-time information to users through the vehicle display or mobile application to provide feedback on detection results and processing suggestions.

[0127] The user interface displays the following: current battery status (normal or faulty); fault type and handling suggestions; real-time current, voltage, SOC and SOH data.

[0128] User feedback example: Fault type: Excessive current; Handling suggestion: The power has been reduced, please go to the service center for inspection as soon as possible.

[0129] For example, after detecting an overcurrent fault (current greater than 300A), the system automatically reduces power to reduce the maximum output power by 50%, and provides feedback on the fault type and handling suggestions to the user through the on-board display.

[0130] In actual applications, the cloud server uses the real-time current data collected by the edge device to perform real-time fault detection and alarm through a pre-trained model. Taking a certain electric car as an example, suppose that during a driving process, the cloud server detects abnormal current data and triggers the alarm mechanism in time to notify the user to check. Subsequently, the cloud server conducts in-depth analysis of the fault data and provides detailed fault cause analysis and maintenance suggestions in combination with the vehicle's historical data.

[0131] Compared with the traditional ON-OFF algorithm, the traditional method cannot handle faults such as small current fluctuations. The method in this embodiment can solve and predict future current faults. The method of the present invention can efficiently and accurately identify the current fault of a certain battery pack and handle it in time, which significantly improves the real-time and reliability of the battery management system and ensures the safe operation and stable performance of the battery. At the same time, the cloud-edge collaborative computing mode optimizes the utilization of computing resources and realizes efficient data processing and intelligent analysis.

[0132] Embodiment 2

[0133] This embodiment provides a battery current fault detection device.

[0134] A battery current fault detection device, comprising:

[0135] The model training module is configured as follows: obtaining a training set, in which the current signal and the actual fault category of each sample are recorded; for each current signal in the training set, extracting current features; for each sample, based on the current signal, after determining the wavelet basis function and the decomposition level through the first neural network model, performing wavelet transform on the current signal based on the determined wavelet basis function and the decomposition level, extracting time-frequency domain features, and based on the time-frequency domain features, obtaining deep features through the second neural network model; obtaining the fault category through the classifier based on the current features, the time-frequency domain features and the deep features; after calculating the loss based on the actual fault category and the fault category obtained by the classifier, judging whether the termination condition is met; if so, outputting the trained first neural network model, the second neural network model and the classifier; otherwise, updating the parameters of the first neural network model, the second neural network model and the classifier, and returning to re-determine the wavelet basis function and the decomposition level.

[0136] A data acquisition module is configured to: acquire a current signal and current characteristics of a battery to be detected;

[0137] The feature extraction module is configured to: determine the wavelet basis function and the decomposition level through the first neural network model based on the current signal, perform wavelet transform on the current signal based on the determined wavelet basis function and the decomposition level, extract the time-frequency domain features, and obtain the deep features through the second neural network model based on the time-frequency domain features;

[0138] The fault detection module is configured to obtain the fault category through a classifier based on current features, time-frequency domain features and depth features.

[0139] Among them, the collection of current signals is completed by the battery management system; the extraction of current characteristics is completed by the edge computing node deployed on the vehicle side; the feature extraction module and the fault detection module are both deployed on the cloud server.

[0140] It should be noted that: the battery current fault detection device provided in the above embodiment is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the electronic device is divided into different functional modules to complete all or part of the functions described above. In addition, the battery current fault detection device provided in the above embodiment and the battery current fault detection method embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0141] Embodiment 3

[0142] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the battery current fault detection method described in the first embodiment are implemented.

[0143] Embodiment 4

[0144] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the battery current fault detection method described in the first embodiment are implemented.

[0145] Figure 6 The schematic diagram of the structure of a computer device provided by an exemplary embodiment of the present application is shown. The computer device includes: a processor and a memory.

[0146] The processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, and the like. The processor may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field Programmable Gate Array), and PLA (Programmable Logic Array). The processor may also include a main processor and a coprocessor. The main processor is a processor for processing data in an awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in a standby state. In some embodiments, the processor may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.

[0147] The memory may include one or more computer-readable storage media, which may be non-transitory. The memory may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory is used to store at least one computer program, which is used to be used by the processor to implement the vehicle occupant detection method provided in the method embodiment of the present application.

[0148] Those skilled in the art will appreciate that the illustrated structure does not constitute a limitation on the computer device, and may include more or fewer components than illustrated, or combine certain components, or adopt a different component arrangement.

[0149] 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 battery current fault detection method, characterized in that: include: Obtaining the current signal and current characteristics of the battery to be tested; Based on the current signal, after determining the wavelet basis function and the decomposition level through the first neural network model, the current signal is subjected to wavelet transform based on the determined wavelet basis function and the decomposition level to extract the time-frequency domain features, and based on the time-frequency domain features, the second neural network model is used to obtain the deep features; Based on current features, time-frequency domain features and depth features, the fault category is obtained through the classifier.

2. A battery current fault detection method according to claim 1, characterized in that: The current characteristics include current mean, current standard deviation, current change rate, skewness and peak value.

3. A battery current fault detection method according to claim 1, characterized in that: The training process of the first neural network model, the second neural network model and the classifier includes: Acquire a training set, wherein the training set records the current signal and the actual fault category of each sample; For each current signal in the training set, extract the current feature; For each sample, based on the current signal, after the wavelet basis function and decomposition level are determined through the first neural network model, the current signal is subjected to wavelet transform based on the determined wavelet basis function and decomposition level to extract the time-frequency domain features, and based on the time-frequency domain features, the second neural network model is used to obtain the deep features; based on the current features, time-frequency domain features and deep features, the fault category is obtained through the classifier; Based on the actual fault category and the fault category obtained by the classifier, after the loss is calculated, it is determined whether the termination condition is met; if so, the trained first neural network model, the second neural network model and the classifier are output; otherwise, the parameters of the first neural network model, the second neural network model and the classifier are updated, and the wavelet basis function and the decomposition level are returned to be re-determined.

4. A battery current fault detection method according to claim 1, characterized in that: The current signal is collected by the battery management system; the current feature is extracted by the edge computing node deployed on the vehicle side; the first neural network model, the second neural network model and the classifier are all deployed on the cloud server.

5. A battery current fault detection method according to claim 1, characterized in that: The fault categories include normal, overcurrent fault, abnormal current fluctuation and continuous current drop.

6. A battery current fault detection device, characterized in that: include: A data acquisition module, which is configured to: acquire a current signal and current characteristics of a battery to be detected; The feature extraction module is configured to: determine the wavelet basis function and the decomposition level through the first neural network model based on the current signal, perform wavelet transform on the current signal based on the determined wavelet basis function and the decomposition level, extract time-frequency domain features, and obtain deep features through the second neural network model based on the time-frequency domain features; The fault detection module is configured to obtain the fault category through a classifier based on current features, time-frequency domain features and depth features.

7. A battery current fault detection device according to claim 6, characterized in that: Also included is a model training module, which is configured to: Acquire a training set, wherein the training set records the current signal and the actual fault category of each sample; For each current signal in the training set, extract the current feature; For each sample, based on the current signal, after the wavelet basis function and the decomposition level are determined through the first neural network model, the current signal is subjected to wavelet transform based on the determined wavelet basis function and the decomposition level to extract the time-frequency domain features, and based on the time-frequency domain features, the second neural network model is used to obtain the deep features; Based on current features, time-frequency domain features and deep features, the fault category is obtained through the classifier; Based on the actual fault category and the fault category obtained by the classifier, after calculating the loss, determine whether the termination condition is met; If so, output the trained first neural network model, second neural network model and classifier; otherwise, update the parameters of the first neural network model, second neural network model and classifier, and return to re-determine the wavelet basis function and decomposition level.

8. A battery current fault detection device according to claim 6, characterized in that: The current signal is collected by the battery management system; the current feature is extracted by the edge computing node deployed on the vehicle side; The feature extraction module and the fault detection module are both deployed on a cloud server.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in a battery current fault detection method as described in any one of claims 1 to 5 are implemented.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the battery current fault detection method according to any one of claims 1 to 5 are implemented.

Images