A vehicle power battery fault diagnosis method
By collecting voltage data from vehicle power batteries and combining Shannon entropy and Z-Score models with long short-term memory networks, accurate location and type identification of battery faults are achieved. This solves the problems of accuracy and timeliness in power battery fault diagnosis in existing technologies, and enables rapid response and high-precision fault prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- DALIAN INSTITUTE OF CHEMICAL PHYSICS CHINESE ACADEMY OF SCIENCES
- Filing Date
- 2026-05-11
- Publication Date
- 2026-06-09
AI Technical Summary
Existing technologies struggle to ensure accurate prediction of the entire lifespan of power batteries while simultaneously enabling immediate response and blocking of sudden faults. In particular, under the pressure of network transmission delays and massive data concurrent processing, traditional vehicle-side systems are unable to meet the accuracy and timeliness requirements for battery faults.
By collecting individual cell voltage data, calculating the Shannon entropy distribution, and using the Z-Score anomaly coefficient model to identify anomalies, combined with the Long Short-Term Memory Network to predict voltage drop trends, the system achieves accurate fault location and type identification.
It enables rapid anomaly detection and real-time alarms on edge devices, and utilizes cloud computing power for accurate fault prediction, improving the accuracy and timeliness of fault diagnosis, and can respond to sudden battery failures in milliseconds.
Smart Images
Figure CN122172035A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power battery fault diagnosis technology, and in particular to a method for diagnosing vehicle power battery faults. Background Technology
[0002] With the explosive growth of the new energy vehicle industry, the aftermarket service system for power batteries is facing severe data and technological challenges. Battery fault diagnosis technology is caught in a dilemma between accuracy and timeliness: traditional vehicle-side battery management systems are limited by hardware computing power and cannot run complex mechanism models to ensure diagnostic accuracy. While relying solely on cloud-based big data platforms has sufficient computing power, it is limited by network transmission latency and the pressure of processing massive amounts of concurrent data, and cannot meet the need for millisecond-level real-time early warning for sudden safety risks such as battery thermal runaway. As a result, existing technologies cannot achieve immediate response and prevention of sudden faults while ensuring the accuracy of full life cycle life prediction.
[0003] Therefore, a vehicle power battery fault diagnosis method that combines accuracy and timeliness is needed. Summary of the Invention
[0004] In view of this, the present invention provides a method for diagnosing vehicle power battery faults, which facilitates data acquisition and anomaly coefficient calculation by the equipment to determine the location of battery abnormalities, and uses a cloud-based long short-term memory network to predict voltage drop trends and determine the fault type, thus enabling vehicle power battery fault diagnosis to be both accurate and timely.
[0005] Therefore, the present invention provides the following technical solution:
[0006] A method for diagnosing vehicle power battery faults, comprising: Collect voltage data of all individual cells in the vehicle's power battery; The Shannon entropy distribution of each individual cell was calculated based on the voltage data. Using the Shannon entropy distribution as input, the anomaly coefficient value of each individual cell is obtained through the Z-Score anomaly coefficient model; Determine whether each individual cell has an abnormality based on the aforementioned abnormality coefficient value; If an anomaly occurs, the voltage drop trend is obtained through a long short-term memory network using voltage, current, and state of charge data as input. The abnormality type of the individual battery cell is determined based on the voltage drop trend.
[0007] Furthermore, using the Shannon entropy distribution as input, the anomaly coefficient value of each individual cell is obtained through the Z-Score anomaly coefficient model, including: The Z-Score of each cell is calculated based on the Shannon entropy of each cell. The abnormality coefficient value of each individual cell is determined based on the Z-Score value combined with a preset threshold.
[0008] Furthermore, the Z-Score value of each individual cell is calculated based on the Shannon entropy of each individual cell:
[0009] in, For single cell batteries Shannon entropy; for The mean; for Standard deviation ; For single cell batteries The Z-Score value.
[0010] Furthermore, based on the Z-Score value combined with a preset threshold, the anomaly coefficient value of each individual cell is determined, including:
[0011] in, For single cell batteries The abnormal coefficient value; This is a preset threshold.
[0012] Further, determining whether each individual cell exhibits an anomaly based on the anomaly coefficient value includes: If a single cell If the anomaly coefficient value is 1 within N consecutive windows, then the single cell is judged to be... An abnormality is detected, and a buzzer alarm is triggered.
[0013] Furthermore, the step of obtaining the voltage drop trend through a long short-term memory network using voltage, current, and state of charge data as input includes: Based on the historical voltage, current, and state of charge data of the individual cells that exhibited abnormalities within a preset time window, time-series input features are constructed. The time series input features are fed into a pre-trained long short-term memory network model, which outputs a voltage sequence within a preset future time period. The voltage sequence is fitted with linear regression to calculate the slope of voltage change as the voltage drop trend.
[0014] Furthermore, based on the voltage drop trend, the abnormality type of the individual battery cell exhibiting the abnormality is determined, including: If the voltage drop trend is less than a preset threshold and the duration exceeds a preset duration, the abnormality type is determined to be an internal short circuit. If the voltage drop trend is greater than or equal to a preset threshold, the abnormality type is determined to be a sensor failure. If the voltage drop trend is less than the preset threshold but does not last for the preset duration, the abnormality type is determined to be a sensor failure.
[0015] A vehicle power battery fault diagnosis system, comprising: The module consists of a data acquisition module, a detection module, and a recognition module. The acquisition module collects voltage data of all individual cells in the vehicle's power battery. The detection module calculates the Shannon entropy distribution of each individual cell based on the voltage data; uses the Shannon entropy distribution as input to obtain the anomaly coefficient value of each individual cell through the Z-Score anomaly coefficient model; and determines whether each individual cell has an anomaly based on the anomaly coefficient value. When an anomaly occurs, the identification module uses voltage, current, and state of charge data as input to obtain the voltage drop trend through a long short-term memory network; based on the voltage drop trend, it determines the anomaly type of the individual battery cell that has an anomaly.
[0016] Furthermore: The acquisition module and the detection module are deployed on edge devices; the recognition module is deployed in the cloud.
[0017] Advantages and positive effects of the present invention: This invention uses an anomaly coefficient model combined with preset thresholds to identify anomalies in individual battery cells, achieving precise fault location. By using a long short-term memory network to predict voltage drop trends based on voltage, current, and charge state, the accuracy of voltage prediction is improved. The anomaly type is determined based on the voltage trend, ruling out sensor faults and further enhancing the accuracy of fault diagnosis. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 Flowchart of vehicle power battery fault diagnosis method; Figure 2 This is a framework diagram of a vehicle power battery fault diagnosis system. Detailed Implementation
[0020] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0021] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0022] This invention provides a method for diagnosing faults in vehicle power batteries. The method involves collecting voltage data from all individual cells in the vehicle power battery; calculating the Shannon entropy distribution of each individual cell based on the voltage data; obtaining the anomaly coefficient value for each individual cell using the Shannon entropy distribution as input through a Z-Score anomaly coefficient model; determining whether each individual cell exhibits an anomaly based on the anomaly coefficient value; if an anomaly is found, obtaining the voltage drop trend using voltage, current, and state of charge data through a long short-term memory network; and determining the anomaly type of the individual cell exhibiting the anomaly based on the voltage drop trend.
[0023] Combination Figure 1 As shown, the present invention provides a method for diagnosing vehicle power battery faults, including: S1. Collect voltage data of all individual cells in the vehicle's power battery; S2. Calculate the Shannon entropy distribution of each individual cell based on voltage data; S3. Using the Shannon entropy distribution as input, the anomaly coefficient value of each individual cell is obtained through the Z-Score anomaly coefficient model; S4. Determine whether each individual cell has an abnormality based on the abnormality coefficient value; S5. If an anomaly occurs, the voltage drop trend is obtained through a long short-term memory network using voltage, current, and state of charge data as input. S6. Determine the anomaly type of the individual cell based on the voltage drop trend.
[0024] Example 1 A method for diagnosing vehicle power battery faults, comprising: S1. Collect voltage data of all individual cells in the vehicle's power battery; Data is collected using an active polling communication strategy (by sending a specific request frame to trigger the ECU / BMS). Time sliding window: Set a time window of length L. The window slides forward with time t, new data enters and old data is removed.
[0025] Edge devices maintain an N in memory The matrix L, where N is the number of individual cells, is expressed by the formula: V=
[0026] in, This is a voltage data matrix; Indicates the number of individual battery cells in the power battery pack; This indicates the length of the sliding time window and the corresponding number of sampling points; Indicates the first Cell battery in time Voltage values collected at all times; Indicates the first The individual battery cells move forward at the current moment Voltage value per sampling period, Indicates the current sampling time.
[0027] Denoising: Wavelet transform was used to denoise the original voltage signal. The Daubechies (db3) wavelet basis was selected for three-level decomposition to extract low-frequency approximation coefficients and reconstruct the signal, thereby filtering out high-frequency noise caused by electromagnetic interference.
[0028] S2. Calculate the Shannon entropy distribution of each individual cell based on voltage data: 1) Discretize the continuous voltage data into a probability distribution.
[0029] 2) Find the maximum value of all voltage values within the current window. and minimum value .
[0030] 3) Divide the interval Divided into A number of equal-width subintervals.
[0031] The choice of affects the sensitivity; in this embodiment, the value is between 10 and 50, or it can be dynamically calculated according to the Freedman-Diaconis criterion.
[0032] 4) Count the number of occurrences in each sub-interval. Number of voltage data points within Calculate the probability of its occurrence. The formula is expressed as:
[0033] 5) The voltage distribution entropy within the current time window is expressed by the formula:
[0034] The unit is bits. If all battery voltages are exactly the same (falling into the same bin), then... =0; if the voltage is extremely discrete, then Approaching the maximum value .
[0035] S3. Using the Shannon entropy distribution as input, the anomaly coefficient value of each individual cell is obtained through the Z-Score anomaly coefficient model; The Z-Score of each individual cell is calculated based on the Shannon entropy of each cell:
[0036] in, For single cell batteries Shannon entropy; for The mean; for Standard deviation ; For single cell batteries The Z-Score value.
[0037] The anomaly coefficient value of each individual cell is determined based on the Z-Score value combined with a preset threshold.
[0038] in, For single cell batteries The abnormal coefficient value; The preset threshold is 3 in this embodiment, which corresponds to a confidence level of over 99%.
[0039] S4. Determine whether each individual cell has an anomaly based on the anomaly coefficient value; If a single cell If the anomaly coefficient value is 1 within N consecutive windows, then the single cell is judged to be... An abnormality is detected, and a buzzer alarm is triggered.
[0040] In this embodiment, N is 5.
[0041] S5. If an anomaly occurs, the voltage drop trend is obtained through a long short-term memory network using voltage, current, and state of charge data as input. Based on the historical voltage, current, and state of charge data of the individual cells that exhibited abnormalities within a preset time window, time-series input features are constructed. The input features are standardized to eliminate the impact of dimensional differences on model training; The time series input features are fed into a pre-trained long short-term memory network model, which outputs the voltage sequence within a preset future time period. The slope of voltage change is calculated by performing linear regression fitting on the voltage sequence as the voltage drop trend.
[0042] In this embodiment, the Long Short-Term Memory (LSTM) network includes at least two LSTM layers and a fully connected output layer; the first LSTM layer extracts short-term dynamic features; the second LSTM layer extracts long-term dependency features; and the output layer predicts the voltage values at the next H sampling times.
[0043] S6. Determine the anomaly type of the individual cell based on the voltage drop trend.
[0044] If the voltage drop trend is less than the preset threshold and the duration exceeds the preset duration, the abnormality type is determined to be an internal short circuit. If the voltage drop trend is greater than or equal to the preset threshold, the abnormality type is determined to be a sensor failure. If the voltage drop trend is less than the preset threshold but does not continue for the preset duration, the abnormality type is determined to be a sensor failure.
[0045] Example 2 like Figure 2 As shown, a vehicle power battery fault diagnosis system includes: The module consists of a data acquisition module, a detection module, and a recognition module. The data acquisition module collects voltage data from all individual cells in the vehicle's power battery. The detection module calculates the Shannon entropy distribution of each individual cell based on voltage data; using the Shannon entropy distribution as input, it obtains the anomaly coefficient value of each individual cell through the Z-Score anomaly coefficient model; and determines whether each individual cell has an anomaly based on the anomaly coefficient value. When an anomaly occurs, the identification module uses voltage, current, and state of charge data as input to obtain the voltage drop trend through a long short-term memory network; based on the voltage drop trend, it determines the anomaly type of the individual cell that is experiencing the anomaly.
[0046] The acquisition module and the detection module are deployed on edge devices; the recognition module is deployed in the cloud. The detection module is deployed on edge devices to achieve millisecond-level anomaly detection and real-time alarm; the recognition module is deployed on a cloud server to utilize cloud computing power to complete model inference and trend prediction of the long short-term memory network, thereby reducing the computing burden on the vehicle terminal.
[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for diagnosing vehicle power battery faults, characterized in that, include: Collect voltage data of all individual cells in the vehicle's power battery; The Shannon entropy distribution of each individual cell was calculated based on the voltage data. Using the Shannon entropy distribution as input, the anomaly coefficient value of each individual cell is obtained through the Z-Score anomaly coefficient model; Determine whether each individual cell has an abnormality based on the aforementioned abnormality coefficient value; If an anomaly occurs, the voltage drop trend is obtained through a long short-term memory network using voltage, current, and state of charge data as input. The abnormality type of the individual battery cell is determined based on the voltage drop trend.
2. The method according to claim 1, characterized in that, Using the Shannon entropy distribution as input, the anomaly coefficient value of each individual cell is obtained through the Z-Score anomaly coefficient model, including: The Z-Score of each cell is calculated based on the Shannon entropy of each cell. The abnormality coefficient value of each individual cell is determined based on the Z-Score value combined with a preset threshold.
3. The method according to claim 2, characterized in that, The Z-Score value of each individual cell is calculated based on the Shannon entropy of each individual cell: in, For single cell batteries Shannon entropy; for The mean; for Standard deviation ; For single cell batteries The Z-Score value.
4. The method according to claim 2, characterized in that, The abnormality coefficient value of each individual cell is determined based on the Z-Score value combined with a preset threshold, including: in, For single cell batteries The abnormal coefficient value; This is a preset threshold.
5. The method according to claim 4, characterized in that, Determining whether each individual cell exhibits an anomaly based on the aforementioned anomaly coefficient value includes: If a single cell If the anomaly coefficient value is 1 within N consecutive windows, then the single cell is judged to be... An abnormality is detected, and a buzzer alarm is triggered.
6. The method according to claim 1, characterized in that, The method of obtaining voltage drop trends through a long short-term memory network using voltage, current, and state of charge data as input includes: Based on the historical voltage, current, and state of charge data of the individual cells that exhibited abnormalities within a preset time window, time-series input features are constructed. The time series input features are fed into a pre-trained long short-term memory network model, which outputs a voltage sequence within a preset future time period. The voltage sequence is fitted with linear regression to calculate the slope of voltage change as the voltage drop trend.
7. The method according to claim 6, characterized in that, Based on the voltage drop trend, the abnormality type of the individual battery cell is determined, including: If the voltage drop trend is less than a preset threshold and the duration exceeds a preset duration, the abnormality type is determined to be an internal short circuit. If the voltage drop trend is greater than or equal to a preset threshold, the abnormality type is determined to be a sensor failure. If the voltage drop trend is less than the preset threshold but does not last for the preset duration, the abnormality type is determined to be a sensor failure.
8. A vehicle power battery fault diagnosis system, characterized in that, include: The module includes a data acquisition module, a detection module, and a recognition module. The acquisition module collects voltage data of all individual cells in the vehicle's power battery. The detection module calculates the Shannon entropy distribution of each individual cell based on the voltage data; and obtains the anomaly coefficient value of each individual cell by using the Shannon entropy distribution as input through the Z-Score anomaly coefficient model. Determine whether each individual cell has an abnormality based on the aforementioned abnormality coefficient value; When an anomaly occurs, the identification module uses voltage, current, and state of charge data as input to obtain the voltage drop trend through a long short-term memory network. The abnormality type of the individual battery cell is determined based on the voltage drop trend.
9. The system according to claim 8, characterized in that: The acquisition module and the detection module are deployed on edge devices; the recognition module is deployed in the cloud.