Automobile battery fault prediction method and device

By performing feature decomposition and feature selection of battery voltage data, combined with long and short-term memory network model, the rapid and accurate prediction of battery failures is achieved, and the misjudgment and missed detection problems caused by environmental changes and sensor accuracy in the existing technology are solved. Improves the reliability and stability of fault detection.

CN119025975AInactive Publication Date: 2024-11-26GAC AION NEW ENERGY AUTOMOBILE CO LTD

Patent Information

Application Number
CN202411507203.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-28
Publication Date
2024-11-26
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art is susceptible to environmental changes and sensor accuracy in battery failure prediction, resulting in misjudgment or missed detection, and insufficient reliability and stability of fault detection.

Method used

The fault prediction method based on battery voltage data is adopted, and the feature decomposition and feature selection of timing voltage sampling data is constructed to construct a long and short-term memory network model, fault detection and classification are carried out, and the dependence on sensor accuracy and environmental changes is reduced.

Benefits of technology

It realizes fast and accurate prediction of battery failures, improves the reliability and stability of fault detection, and improves the accuracy and robustness of battery status diagnosis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119025975A_ABST
    Figure CN119025975A_ABST
Patent Text Reader

Abstract

The invention provides an automobile battery fault prediction method and device. The method comprises the following steps: collecting time sequence voltage sampling data and battery state sampling data; performing characteristic decomposition and characteristic selection on the time sequence voltage sampling data to obtain strong characterization characteristic data; constructing a feature data set according to the strong representation feature data and the battery state sampling data; constructing an original fault detection classification model based on a long short-term memory network; training the original fault detection classification model according to the feature data set to obtain a trained fault diagnosis classification model; acquiring a real-time battery voltage value of the to-be-diagnosed vehicle; predicting whether the battery state has a fault according to the real-time battery voltage value and a fault diagnosis classification model; and if yes, outputting battery fault early warning information. According to the method and the device, the battery fault can be quickly and accurately predicted in real time based on the battery voltage data, the dependence on the sensor precision and the environmental change is reduced, and the reliability and the stability of fault detection are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of battery management technology, and in particular to a method and device for predicting automobile battery failure. Background Art

[0002] With the global emphasis on environmental protection and sustainable development, new energy vehicles have become the first choice to replace traditional fuel vehicles. Among them, battery technology, as the core of new energy vehicles, is crucial to vehicle endurance, performance and safety. At present, fault prediction and status monitoring of new energy vehicle batteries are key research areas, aiming to prevent battery failures, safety hazards, health damage and shortened life. However, existing technologies rely on specific parameters or sensor measurements to determine battery status, but are easily affected by environmental changes or sensor accuracy, resulting in misjudgment or missed detection. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a method and device for predicting automobile battery faults, which can quickly and accurately predict battery faults in real time based on battery voltage data, reduce dependence on sensor accuracy and environmental changes, and improve the reliability and stability of fault detection.

[0004] The first aspect of the present application provides a method for predicting automobile battery failure, comprising: Collecting a voltage data sample set; wherein the voltage data sample set includes time-series voltage sampling data and battery status sampling data corresponding to the time-series voltage sampling data; Performing characteristic decomposition on the time series voltage sampling data to obtain frequency characteristic components; Perform feature selection according to the frequency characteristic component to obtain strong characterization feature data; Constructing a feature data set according to the strong characterization feature data and the battery state sampling data; Building an original fault detection classification model based on long short-term memory network; Training the original fault detection classification model according to the feature data set to obtain a trained fault diagnosis classification model; Collect the real-time battery voltage value of the vehicle to be diagnosed; Predicting whether a battery state is faulty according to the real-time battery voltage value and the fault diagnosis classification model; If yes, output battery failure warning information.

[0005] In the above implementation process, the method can quickly and accurately predict battery faults in real time based on battery voltage data, reduce dependence on sensor accuracy and environmental changes, and improve the reliability and stability of fault detection.

[0006] Furthermore, the collecting of voltage data sample sets includes: Collecting battery voltage continuous measurement values ​​fed back by a battery management system and battery status information corresponding to the battery voltage continuous measurement values; Sampling the battery voltage continuous measurement value to obtain time series voltage sampling data; Sampling the battery status information to obtain status sampling data corresponding to the time-series voltage sampling data; The time-series voltage sampling data and the state sampling data are aggregated to obtain a voltage data sample set.

[0007] Furthermore, the characteristic decomposition of the time series voltage sampling data to obtain frequency characteristic components includes: An empirical mode decomposition algorithm is used to perform frequency domain decomposition on the time series voltage sampling data to obtain multiple frequency characteristic components.

[0008] Furthermore, the feature selection is performed according to the frequency characteristic component to obtain strong characterization feature data, including: Calculate the total correlation coefficient between each of the frequency characteristic components and other frequency characteristic components; Screening the frequency characteristic components according to the total correlation coefficient to obtain strong characterization frequency characteristic components; Obtaining the decomposition number corresponding to the strong characterization frequency feature component; The strong characterization frequency feature components are concatenated according to the number of decompositions to obtain strong characterization feature data.

[0009] Furthermore, constructing a feature data set according to the strong characterization feature data and the battery status sampling data includes: Acquire the battery state corresponding to the strong characterization feature data according to the state sampling data; A feature data set is constructed according to the strong characterization feature data and the battery state.

[0010] Furthermore, the fault diagnosis classification model includes a first LSTM layer, a second LSTM layer and a fully connected layer, the second LSTM layer is connected to the fully connected layer, the activation function between the second LSTM layer and the fully connected layer is a tanh function, and a softmax activation function is used after the fully connected layer.

[0011] Further, the predicting whether a battery state is faulty according to the real-time battery voltage value and the fault diagnosis classification model includes: Sampling the real-time battery voltage value to obtain a real-time time series voltage signal; Extracting features of the real-time time series voltage signal to obtain real-time time series voltage features; Performing feature selection on the real-time time series voltage feature to obtain real-time strong characterization feature data; Inputting the real-time strong characterization feature data into the fault diagnosis classification model to perform battery state prediction to obtain a prediction result; Determining whether a battery failure is predicted according to the prediction result; If yes, determining the battery fault type according to the prediction result, and outputting the battery fault warning information; If not, the step of collecting the real-time battery voltage value of the vehicle to be diagnosed is performed.

[0012] A second aspect of the present application provides a vehicle battery failure prediction device, the vehicle battery failure prediction device comprising: A collecting unit, configured to collect a voltage data sample set; wherein the voltage data sample set includes time-series voltage sampling data and battery status sampling data corresponding to the time-series voltage sampling data; A characteristic decomposition unit, used for performing characteristic decomposition on the time series voltage sampling data to obtain a frequency characteristic component; A feature selection unit, used for performing feature selection according to the frequency characteristic component to obtain strong characterization feature data; A first construction unit, configured to construct a feature data set according to the strong characterization feature data and the battery state sampling data; A second construction unit is used to construct an original fault detection classification model based on a long short-term memory network; A training unit, used to train the original fault detection classification model according to the feature data set to obtain a trained fault diagnosis classification model; A real-time acquisition unit, used to acquire the real-time battery voltage value of the vehicle to be diagnosed; A fault prediction unit, used to predict whether a battery state is faulty according to the real-time battery voltage value and the fault diagnosis classification model; The early warning unit is used to output battery failure early warning information when a battery failure is predicted.

[0013] Furthermore, the collection unit comprises: A collecting subunit, used to collect the battery voltage continuous measurement values ​​fed back by the battery management system and the battery status information corresponding to the battery voltage continuous measurement values; A first sampling subunit is used to sample and process the battery voltage continuous measurement value to obtain time series voltage sampling data; The first sampling subunit is further used to perform sampling processing on the battery status information to obtain status sampling data corresponding to the time-series voltage sampling data; The summarizing subunit is used to summarize the time-series voltage sampling data and the state sampling data to obtain a voltage data sample set.

[0014] Furthermore, the characteristic decomposition unit is specifically used to perform frequency domain decomposition on the time series voltage sampling data by using an empirical mode decomposition algorithm to obtain multiple frequency characteristic components.

[0015] Furthermore, the feature selection unit comprises: A calculation subunit, used for calculating the total correlation coefficient between each of the frequency characteristic components and other frequency characteristic components; A screening subunit, used for screening the frequency characteristic components according to the total correlation coefficient to obtain strong characterization frequency characteristic components; A first acquisition subunit is used to acquire the decomposition times corresponding to the strong frequency characteristic component; The splicing subunit is used to splice the strong characterization frequency feature components according to the decomposition times to obtain strong characterization feature data.

[0016] Furthermore, the first construction unit includes: A second acquisition subunit, configured to acquire a battery state corresponding to the strong characterization feature data according to the state sampling data; A construction subunit is used to construct a feature data set according to the strong characterization feature data and the battery state.

[0017] Furthermore, the fault diagnosis classification model includes a first LSTM layer, a second LSTM layer and a fully connected layer, the second LSTM layer is connected to the fully connected layer, the activation function between the second LSTM layer and the fully connected layer is a tanh function, and a softmax activation function is used after the fully connected layer.

[0018] Furthermore, the fault prediction unit includes: A second sampling subunit is used to sample the real-time battery voltage value to obtain a real-time time series voltage signal; An extraction subunit, used for extracting features from the real-time time series voltage signal to obtain real-time time series voltage features; A selection subunit is used to perform feature selection on the real-time sequential voltage feature to obtain real-time strong characterization feature data; A prediction subunit, used for inputting the real-time strong characterization feature data into the fault diagnosis classification model to perform battery state prediction and obtain a prediction result; The judgment subunit is used to judge whether a battery failure is predicted based on the prediction result; and when a battery failure occurs, determine the battery failure type based on the prediction result, and trigger the warning unit to output battery failure warning information; or when the battery does not fail, trigger the real-time acquisition unit to collect the real-time battery voltage value of the vehicle to be diagnosed.

[0019] A third aspect of the present application provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the automotive battery failure prediction method described in any one of the first aspects of the present application.

[0020] A fourth aspect of the present application provides a computer-readable storage medium storing computer program instructions. When the computer program instructions are read and executed by a processor, the automobile battery fault prediction method described in any one of the first aspect of the present application is executed.

[0021] The beneficial effects of the present application are: using empirical mode decomposition (EMD) to decompose the voltage signal in the frequency domain, and filtering the features through Pearson correlation analysis, so as to extract more representative and reliable features, thereby improving the accuracy and stability of electric vehicle battery fault detection; A fault detection and classification model based on a long short-term memory network (LSTM) is used to conduct a comprehensive and efficient analysis of time series data. Compared with traditional methods, it can better capture the dynamic changes of the battery system and achieve accurate classification and diagnosis. The loss function design combining Euclidean distance and cross entropy principle can make the fault detection model have stronger learning ability for abnormal data, thereby improving the reliability of the model for electric vehicle battery fault detection, making the model more robust, and thus better adapting to complex working environments and data changes; The twin training mechanism enables the network to better capture the data distribution between battery signals, which is beneficial to improving the flexibility and generalization ability of the detection model, making the model more robust and reliable when facing complex electric vehicle scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0023] Figure 1 A schematic diagram of a process flow of a method for predicting vehicle battery failure provided in an embodiment of the present application; Figure 2 A schematic diagram of a flow chart of another method for predicting vehicle battery failures provided in an embodiment of the present application; Figure 3 A schematic diagram of a process flow of voltage data collection and preprocessing provided in an embodiment of the present application; Figure 4 A schematic diagram of the main process of EMD decomposition provided in an embodiment of the present application; Figure 5 An overall block diagram of implementing feature selection based on Pearson analysis provided in an embodiment of the present application; Figure 6 A schematic diagram of a fault detection classification network model based on LSTM provided in an embodiment of the present application; Figure 7 A real-time battery fault detection block diagram provided in an embodiment of the present application; Figure 8 A schematic diagram of the structure of a vehicle battery failure prediction device provided in an embodiment of the present application; Fig. 9 A schematic diagram of the structure of another vehicle battery failure prediction device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0025] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0026] Example 1 Please see Figure 1 , Figure 1 A schematic diagram of a flow chart of a method for predicting a vehicle battery failure provided in this embodiment. The method for predicting a vehicle battery failure includes: S101. Collect a voltage data sample set.

[0027] In this embodiment, the voltage data sample set includes time-series voltage sampling data and battery status sampling data corresponding to the time-series voltage sampling data.

[0028] In this embodiment, the method may preferentially collect and preprocess voltage data based on a battery management system (BMS).

[0029] S102, performing feature decomposition on the time series voltage sampling data to obtain frequency feature components.

[0030] S103, performing feature selection according to the frequency characteristic component to obtain strong characterization feature data.

[0031] In this embodiment, the method can construct a voltage-based feature decomposition and feature selection module, wherein the feature decomposition uses empirical mode decomposition (EMD) and the feature selection is based on Pearson correlation analysis.

[0032] S104: construct a feature data set according to the strong characterization feature data and the battery status sampling data.

[0033] In this embodiment, the method can use feature extraction and feature selection modules to obtain a feature data set with strong characterization capabilities for the preprocessed time series data set.

[0034] S105. Construct an original fault detection classification model based on the long short-term memory network.

[0035] S106. Train the original fault detection classification model according to the feature data set to obtain a trained fault diagnosis classification model.

[0036] In this embodiment, the method can establish a fault detection classification model based on LSTM, and design a corresponding loss function and model training method to train the fault diagnosis analysis model.

[0037] S107: Collect the real-time battery voltage value of the vehicle to be diagnosed.

[0038] S108, predicting whether the battery state is faulty based on the real-time battery voltage value and the fault diagnosis classification model, if so, executing step S109; if not, ending this process.

[0039] In this embodiment, the method can perform feature extraction and feature selection on new data collected by the sensor to obtain corresponding feature data, and then use the fault detection classification model to perform state detection.

[0040] S109: Output battery failure warning information.

[0041] In this embodiment, the execution subject of the method may be a computing device such as a computer or a server, and no limitation is made in this embodiment.

[0042] In this embodiment, the execution subject of the method may also be a smart device such as a smart phone, a tablet computer, etc., which is not limited in this embodiment.

[0043] It can be seen that the implementation of the automobile battery fault prediction method described in this embodiment can quickly and accurately predict battery faults in real time based on battery voltage data, reduce dependence on sensor accuracy and environmental changes, and improve the reliability and stability of fault detection.

[0044] Example 2 Please see Figure 2 , Figure 2 A schematic diagram of a flow chart of a method for predicting a vehicle battery failure provided in this embodiment. The method for predicting a vehicle battery failure includes: S201. Collecting battery voltage continuous measurement values ​​fed back by a battery management system and battery status information corresponding to the battery voltage continuous measurement values.

[0045] S202: Sampling the battery voltage continuous measurement values ​​to obtain time series voltage sampling data.

[0046] S203: Sampling the battery status information to obtain status sampling data corresponding to the time-series voltage sampling data.

[0047] S204 , summarizing the time-series voltage sampling data and the state sampling data to obtain a voltage data sample set.

[0048] In this embodiment, the voltage data sample set includes time-series voltage sampling data and battery status sampling data corresponding to the time-series voltage sampling data.

[0049] In this embodiment, the method aims to diagnose the fault of the battery system by using the battery voltage measurement value.

[0050] In this embodiment, the battery management system (BMS) can measure the battery status of the electric vehicle in real time and provide continuous feedback of the battery voltage results. However, during driving, the voltage may fluctuate slightly due to external factors. Therefore, only the voltage value at a single moment cannot provide enough characteristic data for battery fault detection. For the above reasons, this method uses time series voltage data for battery fault detection. The voltage data collection and preprocessing process is as follows: Figure 3 As shown, the specific steps include: ① Collect the battery voltage continuous measurement value fed back by the BMS system in units of 30s, and collect the corresponding battery status information S out , including normal operation, battery degradation failure, battery cell failure, internal short circuit, external short circuit and electrolyte leakage, etc., a total of N groups of collection results were obtained; ② For the continuous voltage measurement value, a set of time series result data is obtained with a sampling frequency of 1Hz ; ③ All N groups of collected results are subjected to the sampling process of step ②; ④Get the battery voltage data and corresponding status information, as shown below: (1) (2) S205 , using an empirical mode decomposition algorithm to perform frequency domain decomposition on the time series voltage sampling data to obtain multiple frequency characteristic components.

[0051] As an optional implementation, an empirical mode decomposition algorithm is used to perform frequency domain decomposition on the time series voltage sampling data to obtain multiple frequency characteristic components, including: Determine the local maximum and local minimum of each time series voltage sampling data; According to the local maximum and local minimum, the cubic spline interpolation algorithm is used to calculate the upper and lower envelopes of each time series voltage sampling data; Calculate the signal average value of each time-series voltage sampling data according to the upper and lower envelopes; Calculate the component signal of each time-series voltage sampling data according to the time-series voltage sampling data and the signal average value; When the component signal satisfies a preset condition, the component signal is determined as a decomposed signal of the corresponding time-series voltage sampling data; Calculate the residual signal of each time-series voltage sampling data according to the decomposed signal and the time-series voltage sampling data; The residual signal is decomposed multiple times to obtain multiple frequency characteristic components of each time series voltage sampling data; the multiple frequency characteristic components include IMF component signals from high frequency to low frequency.

[0052] In this embodiment, there are multiple noises in the initial time series voltage data. Directly performing fault detection based on the collected time series characteristics may affect the detection accuracy. Therefore, this method uses the empirical mode decomposition (EMD) method to perform frequency domain decomposition on the battery voltage signal to obtain data of multiple IMF frequency characteristic components. The main process of EMD decomposition is as follows: Figure 4 shown.

[0053] For example, the EMD analysis process may include the following steps: ① Determine the decomposed input timing signal as , where t∈[1,30]; ②Determine the input timing signal All local maxima and minima of , and use the cubic spline interpolation method to get the upper and lower envelopes of the signal as V max and V min , calculate the signal average value V m (t) is as follows: (3) ③According to the timing signal and average value calculation V m (t) component h1(t): (4) ④ Determine whether h1(t) meets the following conditions: the number of extreme values ​​and zero crossing points of the signal h1(t) is equal or differs by at most 1; the average value of the upper and lower envelopes of h1(t) is zero.

[0054] ⑤If h1(t) satisfies the two conditions, h1(t) is used as the first decomposition signal of IMF and the residual signal r1(t) is calculated: (5) ⑥ Use the residual signal r1(t) as the input time series signal again, repeat steps ② to ⑤ to perform multiple signal decompositions, and gradually obtain the IMF component signals from high frequency to low frequency; ⑦ Since the high-frequency component signal in IMF can reflect voltage fluctuations and capture fault characteristics during operation, while the fluctuation response of low-frequency signals is slower, that is, it contains less characteristic information, this method only decomposes the data of frequency characteristic components from IMF1 to IMF6.

[0055] S206. Calculate the total correlation coefficient between each frequency characteristic component and other frequency characteristic components.

[0056] S207. Screen the frequency characteristic components according to the total correlation coefficient to obtain strong characterization frequency characteristic components.

[0057] S208, obtaining the decomposition times corresponding to the strongly representative frequency characteristic components.

[0058] S209, splicing the strong characterization frequency feature components according to the number of decompositions to obtain strong characterization feature data.

[0059] In this embodiment, the method can filter out IMF component data with high correlation through Pearson similarity analysis, while retaining more representative feature data. Finally, this method is used to eliminate noise interference in the voltage signal sampling process. In addition, since there may be component data with high correlation in the frequency feature component, that is, redundant component features. Therefore, in order to improve the fault detection speed and ensure the detection performance, the method uses Pearson correlation analysis to filter redundant components. The overall block diagram of the feature selection based on Pearson analysis is as follows: Figure 5 shown.

[0060] For example, the specific implementation process is as follows: Calculate the total correlation coefficient p of each component in IMF1 to IMF6 with the other five components i : (6) in, and Represents the component IMFi and IMF j The average value of |·| indicates the calculation of the Euclidean distance between two data points; ABS(·) indicates the calculation of the absolute value; Formula (6) is used to obtain the total correlation coefficient of each component [p1, p2, …, p6], where the larger the total correlation coefficient, the greater the correlation between the component and other components, that is, it contains more redundant information. Therefore, the four smallest total correlation coefficients are selected, and finally the EMD decomposition number corresponding to the component with the strongest representation ability is obtained (IMF1, IMF2, IMF3, IMF5 are retained, and the corresponding EMD decomposition numbers are 1, 2, 3, 5).

[0061] Therefore, the feature decomposition and feature selection module converts the time series voltage signal Decompose into frequency signals: IMF1, IMF2, IMF3, IMF5; splice the frequency signals to obtain a signal F1(t) with stronger representation ability: (7) S210: Obtain a battery state corresponding to the strong characterization feature data according to the state sampling data.

[0062] S211. Construct a feature data set according to the strong characterization feature data and the battery status.

[0063] For example, the process of obtaining a feature data set using feature extraction and feature selection modules may include: ① For the preprocessed N groups of time series data sets , for each set of data Perform feature extraction and feature selection to obtain the corresponding signal with strong characterization ability .but Corresponding battery status information Construct a dataset: (8) in, To input feature information, To output battery status information.

[0064] ② Repeat step ① to obtain a complete feature data set: (9) S212. Construct an original fault detection classification model based on long short-term memory network.

[0065] S213. Train the original fault detection classification model according to the feature data set to obtain a trained fault diagnosis classification model.

[0066] In this embodiment, the fault diagnosis classification model includes a first LSTM layer, a second LSTM layer and a fully connected layer. The second LSTM layer is connected to the fully connected layer. The activation function between the second LSTM layer and the fully connected layer is the tanh function. The softmax activation function is used after the fully connected layer.

[0067] For example, the process of establishing and training a fault detection classification model based on LSTM can be as follows: ① Use the long short-term memory network (LSTM) layer to build an overall fault detection classification model. The network architecture includes two LSTM layers and one fully connected layer. The activation function between the LSTM layer and the fully connected layer uses the tanh function. And add a softmax activation function after the fully connected layer for output classification. Set the input data time step of the LSTM layer to 30, the number of features to 4, and the output dimension of the fully connected layer to 6 (there are 6 battery states in total). The LSTM-based fault detection classification network model is as follows Figure 6 As shown: ② Design a custom loss function for the network model, which combines the Euclidean distance and cross entropy principles to guide model learning and training. Based on the output of the softmax layer, the model's prediction probability for fault category i is defined as , the true probability is y i ; The corresponding probability prediction error function is: (10) Among them, w i is the category weight for fault category i. Since the purpose of this method is to detect battery fault status, the battery fault category has a higher weight to enhance the learning attention to abnormal (fault) feature data during the training process.

[0068] Based on this, combined with information entropy and Euclidean distance calculation, the final loss function is: (11) in, is the one-hot vector of the predicted output, and y is the true high temperature fault category vector.

[0069] ③ Use the twin training mechanism to train the fault detection classification network model. The process includes: a. Initialize two LSTM-based fault detection classification network models. The two detection networks share the same network parameters and weights. b. In each training step, two similar but different training samples of the same category are input into the two sub-networks respectively, and their losses are calculated; c. Use the Adam optimizer to optimize the two sub-network models respectively, calculate the update changes of the network model parameters respectively, and then average the two update changes and use them to update the parameters of the two sub-networks respectively; d. Repeat steps b and c until the preset number of iterations is reached; e. The iteratively updated network model parameters are used in the fault detection classification network model, and finally a fully trained fault detection model is obtained.

[0070] S214: Collect the real-time battery voltage value of the vehicle to be diagnosed.

[0071] S215: Sampling the real-time battery voltage value to obtain a real-time sequential voltage signal.

[0072] S216: Extract features of the real-time time series voltage signal to obtain real-time time series voltage features.

[0073] S217, performing feature selection on the real-time time series voltage feature to obtain real-time strong characterization feature data.

[0074] S218, inputting the real-time strong characterization feature data into the fault diagnosis classification model to predict the battery status and obtain a prediction result.

[0075] S219: Determine whether a battery failure is predicted based on the prediction result. If so, execute step S220; if not, execute step S214.

[0076] S220: Determine the battery failure type according to the prediction result, and output battery failure warning information.

[0077] For example, a voltage-based fault detection application flow might include: ① Collect voltage values ​​in real time based on BMS, and use the above method to obtain the timing voltage signal Test in (t); ②Use the above method to obtain Test in (t) is a strong representation feature data F in (t); ③Use the fault detection model trained by the above method to transform the feature data F in (t) Input model and predict the corresponding battery state S o (normal operation, battery degradation failure, battery cell failure, internal short circuit, external short circuit or electrolyte leakage); ④ Repeat steps ①, ②, and ③ to implement real-time detection of battery status information, provide timely feedback when a battery failure is predicted, and return the fault category.

[0078] The real-time battery fault detection block diagram is as follows: Figure 7 shown.

[0079] In summary, the innovations of this method are as follows: ①By applying empirical mode decomposition (EMD) to decompose the voltage signal in the frequency domain and using Pearson correlation analysis to filter the features, more representative features can be extracted and the representativeness of the features can be enhanced.

[0080] ②Introducing a fault detection classification model based on the long short-term memory network (LSTM) to achieve higher diagnostic accuracy, which helps to analyze time series data more comprehensively and effectively and improve the intelligence level of fault detection.

[0081] ③ The loss function design combining Euclidean distance and cross entropy principles enables the fault detection model to have stronger learning ability for abnormal data, thereby improving the reliability of the model for electric vehicle battery fault detection.

[0082] ④ Using the twin training mechanism to train the fault detection classification network model can enable the network to better capture the similarities and differences between battery signal data and improve the flexibility and generalization ability of the detection model.

[0083] In this embodiment, the execution subject of the method may be a computing device such as a computer or a server, and no limitation is made in this embodiment.

[0084] In this embodiment, the execution subject of the method may also be a smart device such as a smart phone, a tablet computer, etc., which is not limited in this embodiment.

[0085] It can be seen that the implementation of the automobile battery fault prediction method described in this embodiment can quickly and accurately predict battery faults in real time based on battery voltage data, reduce dependence on sensor accuracy and environmental changes, and improve the reliability and stability of fault detection.

[0086] Example 3 Please see Figure 8 , Figure 8 This is a schematic diagram of the structure of a vehicle battery fault prediction device provided in this embodiment. Figure 8 As shown, the automobile battery failure prediction device includes: The collecting unit 310 is used to collect a voltage data sample set; wherein the voltage data sample set includes time-series voltage sampling data and battery status sampling data corresponding to the time-series voltage sampling data; The characteristic decomposition unit 320 is used to perform characteristic decomposition on the time series voltage sampling data to obtain frequency characteristic components; A feature selection unit 330 is used to perform feature selection according to the frequency characteristic component to obtain strong characterization feature data; A first constructing unit 340 is used to construct a feature data set according to the strong characterization feature data and the battery state sampling data; A second construction unit 350 is used to construct an original fault detection classification model based on a long short-term memory network; A training unit 360 is used to train the original fault detection classification model according to the feature data set to obtain a trained fault diagnosis classification model; A real-time acquisition unit 370 is used to acquire the real-time battery voltage value of the vehicle to be diagnosed; A fault prediction unit 380 is used to predict whether a battery state is faulty based on the real-time battery voltage value and the fault diagnosis classification model; The early warning unit 390 is used to output battery failure early warning information when a battery failure is predicted.

[0087] In this embodiment, the explanation of the vehicle battery fault prediction device can refer to the description in Embodiment 1 or Embodiment 2, and will not be further elaborated in this embodiment.

[0088] It can be seen that the automobile battery fault prediction device described in this embodiment can quickly and accurately predict battery faults in real time based on battery voltage data, reduce dependence on sensor accuracy and environmental changes, and improve the reliability and stability of fault detection.

[0089] Example 4 Please see Fig. 9 , Fig. 9 This is a schematic diagram of the structure of a vehicle battery fault prediction device provided in this embodiment. Fig. 9 As shown, the automobile battery failure prediction device includes: The collecting unit 310 is used to collect a voltage data sample set; wherein the voltage data sample set includes time-series voltage sampling data and battery status sampling data corresponding to the time-series voltage sampling data; The characteristic decomposition unit 320 is used to perform characteristic decomposition on the time series voltage sampling data to obtain frequency characteristic components; A feature selection unit 330 is used to perform feature selection according to the frequency characteristic component to obtain strong characterization feature data; A first constructing unit 340 is used to construct a feature data set according to the strong characterization feature data and the battery state sampling data; A second construction unit 350 is used to construct an original fault detection classification model based on a long short-term memory network; A training unit 360 is used to train the original fault detection classification model according to the feature data set to obtain a trained fault diagnosis classification model; A real-time acquisition unit 370 is used to acquire the real-time battery voltage value of the vehicle to be diagnosed; A fault prediction unit 380 is used to predict whether a battery state is faulty based on the real-time battery voltage value and the fault diagnosis classification model; The early warning unit 390 is used to output battery failure early warning information when a battery failure is predicted.

[0090] As an optional implementation, the collecting unit 310 includes: The collecting subunit 311 is used to collect the battery voltage continuous measurement value fed back by the battery management system and the battery status information corresponding to the battery voltage continuous measurement value; The first sampling subunit 312 is used to sample and process the continuous measurement value of the battery voltage to obtain time-series voltage sampling data; The first sampling subunit 312 is further used to sample and process the battery status information to obtain status sampling data corresponding to the time-series voltage sampling data; The summarizing subunit 313 is used to summarize the time-series voltage sampling data and the state sampling data to obtain a voltage data sample set.

[0091] As an optional implementation manner, the characteristic decomposition unit 320 is specifically configured to perform frequency domain decomposition on the time series voltage sampling data by using an empirical mode decomposition algorithm to obtain a plurality of frequency characteristic components.

[0092] As an optional implementation, the feature selection unit 330 includes: A calculation subunit 331 is used to calculate the total correlation coefficient between each frequency characteristic component and other frequency characteristic components; A screening subunit 332 is used to screen the frequency characteristic components according to the total correlation coefficient to obtain the strong characterization frequency characteristic components; The first acquisition subunit 333 is used to acquire the decomposition times corresponding to the strong frequency characteristic component; The concatenation subunit 334 is used to concatenate the strong characterization frequency feature components according to the number of decompositions to obtain strong characterization feature data.

[0093] As an optional implementation, the first building unit 340 includes: The second acquisition subunit 341 is used to acquire the battery state corresponding to the strong characterization feature data according to the state sampling data; The construction subunit 342 is used to construct a feature data set according to the strong characterization feature data and the battery status.

[0094] In this embodiment, the fault diagnosis classification model includes a first LSTM layer, a second LSTM layer and a fully connected layer. The second LSTM layer is connected to the fully connected layer. The activation function between the second LSTM layer and the fully connected layer is the tanh function. The softmax activation function is used after the fully connected layer.

[0095] As an optional implementation, the fault prediction unit 380 includes: The second sampling subunit 381 is used to sample the real-time battery voltage value to obtain a real-time sequential voltage signal; The extraction subunit 382 is used to extract the characteristics of the real-time time series voltage signal to obtain the real-time time series voltage characteristics; A selection subunit 383 is used to perform feature selection on the real-time sequential voltage feature to obtain real-time strong characterization feature data; The prediction subunit 384 is used to input the real-time strong characterization feature data into the fault diagnosis classification model to predict the battery state and obtain the prediction result; The judgment subunit 385 is used to judge whether a battery failure is predicted based on the prediction result; and when a battery failure occurs, determine the battery failure type based on the prediction result, and trigger the warning unit 390 to output battery failure warning information; or when the battery does not fail, trigger the real-time acquisition unit 370 to collect the real-time battery voltage value of the vehicle to be diagnosed.

[0096] In this embodiment, the explanation of the vehicle battery fault prediction device can refer to the description in Embodiment 1 or Embodiment 2, and will not be further elaborated in this embodiment.

[0097] It can be seen that the automobile battery fault prediction device described in this embodiment can quickly and accurately predict battery faults in real time based on battery voltage data, reduce dependence on sensor accuracy and environmental changes, and improve the reliability and stability of fault detection.

[0098] An embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the automobile battery fault prediction method in Embodiment 1 or Embodiment 2 of the present application.

[0099] An embodiment of the present application provides a computer-readable storage medium storing computer program instructions. When the computer program instructions are read and executed by a processor, the automobile battery fault prediction method in Embodiment 1 or Embodiment 2 of the present application is executed.

[0100] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0101] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0102] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0103] The above description is only an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.

[0104] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0105] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

Claims

1. A method for predicting automobile battery failure, characterized in that: include: Collecting a voltage data sample set; wherein the voltage data sample set includes time-series voltage sampling data and battery status sampling data corresponding to the time-series voltage sampling data; Performing characteristic decomposition on the time series voltage sampling data to obtain frequency characteristic components; Perform feature selection according to the frequency characteristic component to obtain strong characterization feature data; Constructing a feature data set according to the strong characterization feature data and the battery state sampling data; Building an original fault detection classification model based on long short-term memory network; Training the original fault detection classification model according to the feature data set to obtain a trained fault diagnosis classification model; Collect the real-time battery voltage value of the vehicle to be diagnosed; Predicting whether a battery state is faulty according to the real-time battery voltage value and the fault diagnosis classification model; If yes, output battery failure warning information.

2. The method for predicting automobile battery failure according to claim 1, characterized in that: The collecting of voltage data sample sets includes: Collecting battery voltage continuous measurement values ​​fed back by a battery management system and battery status information corresponding to the battery voltage continuous measurement values; Sampling the battery voltage continuous measurement value to obtain time series voltage sampling data; Sampling the battery status information to obtain status sampling data corresponding to the time-series voltage sampling data; The time-series voltage sampling data and the state sampling data are aggregated to obtain a voltage data sample set.

3. The method for predicting automobile battery failure according to claim 1, characterized in that: The characteristic decomposition of the time series voltage sampling data to obtain frequency characteristic components includes: An empirical mode decomposition algorithm is used to perform frequency domain decomposition on the time series voltage sampling data to obtain multiple frequency characteristic components.

4. The method for predicting automobile battery failure according to claim 1, characterized in that: The step of performing feature selection according to the frequency characteristic component to obtain strong characterization feature data includes: Calculate the total correlation coefficient between each of the frequency characteristic components and other frequency characteristic components; Screening the frequency characteristic components according to the total correlation coefficient to obtain strong characterization frequency characteristic components; Obtaining the decomposition number corresponding to the strong characterization frequency feature component; The strong characterization frequency feature components are concatenated according to the number of decompositions to obtain strong characterization feature data.

5. The automobile battery failure prediction method according to claim 1, characterized in that: The constructing a feature data set according to the strong characterization feature data and the battery state sampling data includes: Acquire the battery state corresponding to the strong characterization feature data according to the state sampling data; A feature data set is constructed according to the strong characterization feature data and the battery state.

6. The method for predicting automobile battery failure according to claim 1, characterized in that: The fault diagnosis classification model includes a first LSTM layer, a second LSTM layer and a fully connected layer, the second LSTM layer is connected to the fully connected layer, the activation function between the second LSTM layer and the fully connected layer is a tanh function, and a softmax activation function is used after the fully connected layer.

7. The automobile battery failure prediction method according to claim 1, characterized in that: The predicting whether a battery state is faulty according to the real-time battery voltage value and the fault diagnosis classification model includes: Sampling the real-time battery voltage value to obtain a real-time time series voltage signal; Extracting features of the real-time time series voltage signal to obtain real-time time series voltage features; Performing feature selection on the real-time time series voltage feature to obtain real-time strong characterization feature data; Inputting the real-time strong characterization feature data into the fault diagnosis classification model to perform battery state prediction to obtain a prediction result; Determining whether a battery failure is predicted according to the prediction result; If yes, determining the battery fault type according to the prediction result, and outputting the battery fault warning information; If not, the step of collecting the real-time battery voltage value of the vehicle to be diagnosed is performed.

8. An automobile battery failure prediction device, characterized in that: The automobile battery failure prediction device comprises: A collecting unit, configured to collect a voltage data sample set; wherein the voltage data sample set includes time-series voltage sampling data and battery status sampling data corresponding to the time-series voltage sampling data; A characteristic decomposition unit, used for performing characteristic decomposition on the time series voltage sampling data to obtain a frequency characteristic component; A feature selection unit, used for performing feature selection according to the frequency characteristic component to obtain strong characterization feature data; A first construction unit, configured to construct a feature data set according to the strong characterization feature data and the battery state sampling data; A second construction unit is used to construct an original fault detection classification model based on a long short-term memory network; A training unit, used to train the original fault detection classification model according to the feature data set to obtain a trained fault diagnosis classification model; A real-time acquisition unit, used to acquire the real-time battery voltage value of the vehicle to be diagnosed; A fault prediction unit, used to predict whether a battery state is faulty according to the real-time battery voltage value and the fault diagnosis classification model; The early warning unit is used to output battery failure early warning information when a battery failure is predicted.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the automobile battery failure prediction method according to any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium stores computer program instructions, and when the computer program instructions are read and executed by a processor, the automobile battery fault prediction method according to any one of claims 1 to 7 is executed.

Citation Information

Patent Citations

  • Fault diagnosis method and system for fuel cell device and storage medium

    CN114048772A

  • New energy automobile remote diagnosis method and system based on cloud computing

    CN118331239A

  • Multivariate empirical mode decomposition-based top oil temperature prediction method for transformer

    WO2022247241A1

Cited By

  • Automobile emergency starting power supply control system and method thereof

    CN119812523A

  • Battery fault diagnosis method based on graph structure and LSTM

    CN120387091A

  • Battery abnormity type determination method and device and electronic equipment

    CN120703603A