Method, system and equipment for diagnosing short-circuit fault in series lithium battery pack and medium

Abstract

The invention discloses a series lithium battery pack internal short circuit fault diagnosis method, system and device and a medium, and the method comprises the steps: employing the historical voltage data of a single cell of a lithium battery pack as the input of a kernel principal component analysis model, and employing a first square prediction error in a residual subspace as an internal short circuit detection parameter; according to the first square prediction error, a self-adaptive square prediction error threshold value based on kernel density estimation is obtained, and the self-adaptive square prediction error threshold value serves as an internal short circuit detection threshold value; and inputting the obtained voltage data of the single battery cell into the kernel principal component analysis model, obtaining a second square prediction error, comparing the second square prediction error with an adaptive square prediction error threshold, and judging whether an internal short circuit fault occurs or not. According to the method, voltage data feature mining is taken as a core, the algorithm structure is simple, the calculated amount is small, a complex mechanism model does not need to be constructed, the diagnosis time is short, the precision is high, and real-time identification and early warning of the short-circuit fault in the battery cell can be realized.

Description

Technical Field

[0001] This invention belongs to the field of lithium battery system safety and condition monitoring technology, specifically relating to a method, system, device and medium for diagnosing short circuit faults in a series lithium battery pack. Background Technology

[0002] With the large-scale deployment of lithium-ion battery energy storage systems in grid-side and industrial / commercial applications, these systems exhibit characteristics such as high energy density, compact structure, numerous cells, and long operating cycles. Consequently, their safety risks are becoming more systemic and amplified. Among various failure modes, internal short-circuit faults are typically induced by factors such as separator defects, metallic foreign objects, electrode burrs, and dendrite growth. This manifests as undesigned electrical connections forming within the cell, leading to continuous self-discharge and localized Joule heating. During the evolution of internal short circuits from soft to hard short circuits, early externally observable signals such as slight voltage drops, weak temperature rises, and equivalent parameter drifts are often masked by noise and inconsistency fluctuations, easily leading to missed detections. Once the system enters the runaway stage, it may trigger accelerated side reactions and positive feedback heating, further inducing thermal runaway. This runaway can then propagate to the module / cluster / module level through thermal diffusion and electrical coupling, causing shutdowns, fires, explosions, and significant economic losses.

[0003] Existing internal short-circuit diagnostic technologies include rule-based criteria based on voltage and temperature thresholds, residual diagnosis based on equivalent models and parameter identification, detection based on impedance and thermal characteristics, and data-driven anomaly detection and fault classification. However, these technologies generally face limitations in applicability and engineering implementation in energy storage applications. Soft short-circuit signals are weak and non-specific, easily confused with factors such as ambient temperature fluctuations, sensor drift, control strategy switching, and equalization actions, leading to both false alarms and missed alarms. Furthermore, engineering observations are limited, making it difficult to fully capture local thermo-electric evolution and achieve long-term online, low-cost, and interpretable diagnosis and early warning. Therefore, there is an urgent need for a method to achieve early, reliable, and interpretable diagnosis of internal short-circuit faults under complex operating conditions and limited observation, in order to reduce the probability of internal short circuits causing thermal runaway and system-level accidents, and improve the safety operation and maintenance level and risk management capabilities of energy storage systems throughout their entire lifecycle. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system, device and medium for diagnosing short-circuit faults in series lithium battery packs, so as to solve the problems of low detection accuracy and limited application scenarios caused by existing technologies that rely only on single-point voltage thresholds and have simple diagnostic rules.

[0005] This invention is achieved through the following technical solution: A method for diagnosing short-circuit faults within a series-connected lithium battery pack includes the following steps: S01. Historical voltage data of individual cells in lithium battery packs are used as input to the core principal component analysis model, and the first square prediction error in the residual subspace is used as the internal short circuit detection parameter. S02. Based on the first squared prediction error in step S01, obtain the adaptive squared prediction error threshold based on kernel density estimation, and use the adaptive squared prediction error threshold as the internal short circuit detection threshold. S03. Input the acquired single cell voltage data into the principal component analysis model to obtain the second squared prediction error. Compare the second squared prediction error with the adaptive squared prediction error threshold to determine whether an internal short circuit fault has occurred.

[0006] In some embodiments, historical voltage data of individual cells in a lithium battery pack is obtained, represented as... , , where n represents the number of battery packs and m represents the number of individual battery cells in the battery pack.

[0007] In some embodiments, in step S01, the kernel function of the kernel principal component analysis model is set as follows: ,in, and This represents two sample points in the data space. It is a hyperparameter that controls the width or stretching of the kernel function. yes and The Euclidean distance between them; The mapping process satisfies ,in, and This represents two sample points in the data space. and The representative will and Nonlinear mapping functions that map to higher-dimensional spaces. represent The transpose of the matrix; Squared prediction error ,in, and Statistics representing the eigenvalues ​​of the residual subspace. It is the confidence limit of the standard normal distribution. This represents the shape correction parameter.

[0008] In some embodiments, step S02 includes: S021. Define the density function for kernel density estimation, expressed as: Where q represents the data sequence length and h represents the bandwidth. This represents the kernel function, where x is the input data. Represents the i-th sample point; S022. Using the first squared prediction error as input, calculate its probability density, expressed as follows: ,in, This represents the squared prediction error threshold. It is the squared prediction error coefficient. Represents the confidence limit. This represents the first-squared prediction error; S023. Perform kernel density estimation on the first squared prediction error to obtain the distribution range. Based on the probability density curve, the adaptive squared prediction error threshold is calculated sequentially using the infinitesimal method and the composite trapezoidal rule; among which, h min This represents the minimum value within the distribution interval. h max This represents the maximum value within the distribution range.

[0009] In some embodiments, in step S023, the infinitesimal element method and the composite trapezoidal rule are used sequentially to calculate until the following formula is satisfied: ; in, This represents the area under the probability density curve. This represents the area of ​​probability accumulated from the right side of the probability density curve as we search for the squared prediction error threshold. represents the confidence limit, and k represents the number of equally divided intervals within the distribution range. The step size for each interval is... , Representing the i The step size of each interval, Represents from the right end h max Start to With step size The number of intervals accumulated segment by segment, This represents the squared prediction error threshold. j Represents from the right end h max Start to The corresponding between The number.

[0010] In some embodiments, in step S03, the voltage data of individual cells are sequentially input into the principal component analysis model using a sliding time window to obtain the second squared prediction error corresponding to each sliding time window.

[0011] In some embodiments, in step S03, if the second squared prediction error corresponding to the sliding time window exceeds the adaptive squared prediction error threshold, it is determined that an internal short circuit fault has occurred; otherwise, it is determined that no internal short circuit fault has occurred.

[0012] On the other hand, the present invention also provides a short-circuit fault diagnosis system for series lithium battery packs, which uses a short-circuit fault diagnosis method for series lithium battery packs to diagnose internal short-circuit faults, including: The data acquisition module is used to collect voltage data of individual cells in the lithium battery pack. The internal short-circuit detection parameter acquisition module is used to obtain the first squared prediction error in the residual subspace based on the historical voltage data of individual cells in the lithium battery pack and the core principal component analysis model. The adaptive squared prediction error threshold acquisition module is used to obtain the adaptive squared prediction error threshold based on kernel density estimation by obtaining the squared prediction error obtained by the internal short-circuit detection parameter acquisition module. The internal short-circuit fault diagnosis module is used to obtain the second squared prediction error based on the individual cell voltage data and the core principal component analysis model, and to determine whether an internal short-circuit fault has occurred based on the second squared prediction error and the adaptive squared prediction error threshold.

[0013] On the other hand, the present invention also provides an electronic device, comprising: Processor; and Memory for storing the executable instructions of the processor; The processor is configured to execute the short-circuit fault diagnosis method within a series lithium battery pack by executing the executable instructions.

[0014] On the other hand, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the aforementioned method for diagnosing short-circuit faults within a series lithium battery pack.

[0015] Compared with the prior art, the present invention has the following advantages and beneficial effects: This invention focuses on voltage data feature mining. The algorithm has a simple structure and low computational cost. It does not require the construction of complex mechanism models, has a short diagnosis time and high accuracy, and can realize real-time identification and early warning of short circuit faults in battery cells.

[0016] The system employs an adaptive squared prediction error threshold based on kernel density estimation for dynamic detection of internal short-circuit faults. It does not rely on fixed absolute model parameters or specific calibration conditions, enabling it to be applied to power battery packs of different specifications and aging stages, and exhibiting good cross-model applicability and scenario versatility. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the short-circuit fault detection process within the battery pack in an embodiment of the present invention.

[0019] Figure 2 This is a probability density map based on kernel density estimation in an embodiment of the present invention.

[0020] Figure 3 This is experimental data on the voltage of individual battery cells tested in the battery pack embodiment of the present invention.

[0021] Figure 4 This is a schematic diagram of internal short-circuit fault diagnosis in an embodiment of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, specific embodiments of this application will be described in further detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely for explaining this application and not for limiting it. It should also be noted that, for ease of description, only the parts relevant to this application are shown in the drawings, not all of them. Before discussing exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe operations (or steps) as sequential processes, many of these operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but may also have additional steps not included in the drawings. The process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0023] In this invention, an internal short circuit fault in a battery refers to a fault in which an electrical connection occurs between the positive and negative terminals inside the battery.

[0024] In some embodiments of the present invention, reference is made to Figure 1 A method for diagnosing short-circuit faults within a series-connected lithium battery pack includes the following steps: S1. Obtain historical cycle voltage data of the series lithium battery pack to train the squared prediction error in the core principal component analysis model.

[0025] S2. Train the kernel principal component analysis model using historical voltage data of the battery pack, and use the squared prediction error in the residual subspace as the internal short circuit detection parameter to obtain the first squared prediction error.

[0026] Since the calculation of the squared prediction error threshold through the kernel principal component analysis model needs to follow a normal distribution, but the input voltage data does not completely follow a normal distribution, kernel density estimation is used to achieve a normal distribution.

[0027] S3. Obtain the adaptive squared prediction error threshold based on kernel density estimation as the internal short circuit detection threshold. The squared prediction error threshold obtained by mapping it completely conforms to the normal distribution condition.

[0028] S4. Using a sliding time window, the voltage data of individual cells are sequentially input into the principal component analysis model to calculate the squared prediction error, obtain the second squared prediction error, and compare it with the adaptive squared prediction error threshold to determine the internal short circuit fault.

[0029] This invention uses kernel density estimation to achieve a normal distribution, so that the threshold of the squared prediction error statistic obtained by mapping it conforms to the normal distribution condition, and obtains the squared prediction error threshold for fault diagnosis, realizing the diagnosis of internal short circuit faults directly based on the voltage data of individual cells; using the voltage data of individual cells as input, direct diagnosis can be achieved when an internal short circuit fault exists, with low platform computing power requirements.

[0030] Using a battery pack consisting of four ternary lithium batteries connected in series as the test object, a resistor is connected in parallel across the two ends of the battery cell to simulate an internal short-circuit fault scenario. The detection method of this invention is used to analyze and determine the internal short-circuit fault of the battery pack, so as to verify the effectiveness and applicability of the method of this invention.

[0031] In this embodiment, as Figure 1 As shown, the procedure for detecting short circuit faults within the battery pack is as follows: S1. Obtain historical cycle voltage data of the battery pack, which adopts dynamic stress test conditions and has a sampling period of 1 second. Specifically: S1.1 Obtain historical voltage data of individual battery cells in the battery pack. , Where n represents the number of battery packs and m represents the number of individual battery cells in the battery pack.

[0032] S2. Train the kernel principal component analysis model using historical battery pack data, and use the squared prediction error in the residual subspace as the internal short-circuit detection parameter to obtain the first squared prediction error; specifically: S2.1, Define the kernel function of the kernel principal component analysis model as follows: ,in, and This represents two sample points in the data space. It is a hyperparameter that controls the width or stretching of the kernel function. yes and The Euclidean distance between them; The mapping process satisfies ,in, and This represents two sample points in the data space. and The representative will and Nonlinear mapping functions that map to higher-dimensional spaces. represent The transpose of the matrix; Squared prediction error ,in, and Statistics representing the eigenvalues ​​of the residual subspace. It is the confidence limit of the standard normal distribution. Represents shape correction parameters; S2.2, Historical voltage data of individual battery cells in the battery pack , As input to the kernel principal component analysis module, the squared prediction error is calculated to obtain the first squared prediction error.

[0033] In this process, firstly, a kernel function is used to map the voltage data to a high-dimensional feature space, and then principal component analysis is used to extract the main feature components, thereby representing the data in a low-dimensional space. Next, these principal components are used to reconstruct the original data, obtaining the reconstructed voltage values. The squared prediction error is calculated as the squared difference between the original value and the reconstructed value for each data point, used to measure the reconstruction error.

[0034] S3. Design an adaptive squared prediction error threshold based on kernel density estimation as the internal short-circuit detection threshold, specifically as follows: S3.1. Define the density function for kernel density estimation, expressed as: Where q represents the data sequence length and h represents the bandwidth. This represents the kernel function, where x is the input data. Represents the i-th sample point; S3.2. Using the first squared prediction error as input, calculate its probability density, expressed as: ,in, This represents the squared prediction error threshold. It is the squared prediction error coefficient. Represents the confidence limit. Represents the squared prediction error; S3.3. Perform kernel density estimation on the first squared prediction error to obtain the distribution range. and probability density curves, where, h minThis represents the minimum value within the distribution interval. h max Indicates the maximum value within the distribution interval; such as Figure 2 As shown, the following formula is calculated sequentially using the infinitesimal element method and the composite trapezoidal rule: ; in, This represents the area under the probability density curve. This represents the area of ​​probability accumulated from the right side of the probability density curve as we search for the squared prediction error threshold. represents the confidence limit, and k represents the number of equally divided intervals within the distribution range. The step size for each interval is... , Representing the i The step size of each interval, Represents from the right end h max Start to With step size The number of intervals accumulated segment by segment, This represents the squared prediction error threshold. j Represents from the right end h max Start to The corresponding between The number.

[0035] Here, historical data is used for calculation, and the squared prediction error threshold is found to be 10.909. Figure 4 As shown by the red dashed line in the image.

[0036] S4. Using a sliding time window, sequentially input the test cell data into the core principal component analysis model to calculate the squared prediction error, obtaining the second squared prediction error. The test cell data is as follows: Figure 3 As shown; The second squared prediction error calculated in this step is compared with the adaptive squared prediction error threshold to determine the internal short-circuit fault. Specifically: S4.1, will Figure 3 The voltage data shown Divide the time window into segments of size 500 and iterate through the entire battery voltage data in sequence. S4.2 Input the voltage sliding time window data into the kernel principal component analysis model to calculate the squared prediction error and obtain the second squared prediction error; S4.3 In this sliding time window, the internal short circuit fault in the test data occurs at approximately 320s; Fault detection based on an adaptive squared prediction error threshold using kernel density estimation also detected an internal short-circuit fault at 320s. Figure 4As shown, the continuous detection and alarm after the fault occurs demonstrates that the detection method of the present invention has high detection accuracy.

[0037] Furthermore, the voltage fluctuation at 278s did not affect the diagnostic results, indicating that the detection method of the present invention has good reliability and there is no false alarm.

[0038] Compared with the traditional diagnostic method that sets fixed upper and lower thresholds based on voltage data and only triggers an alarm when the voltage of a single cell has significantly exceeded the limit, the internal short circuit fault diagnosis method of the series lithium battery pack of the present invention realizes the diagnosis of internal short circuit faults of the battery pack. It has the characteristics of fast response speed, high identification accuracy and wide applicability, which significantly enhances the safety management and control capabilities of lithium battery systems and effectively improves the overall efficiency of operation and maintenance.

[0039] On the other hand, the present invention also provides a short-circuit fault diagnosis system for series lithium battery packs, which uses a short-circuit fault diagnosis method for series lithium battery packs to diagnose internal short-circuit faults, including: The data acquisition module is used to collect voltage data of individual cells in the lithium battery pack. The internal short-circuit detection parameter acquisition module is used to obtain the first squared prediction error in the residual subspace based on the historical voltage data of individual cells in the lithium battery pack and the core principal component analysis model. The adaptive squared prediction error threshold acquisition module is used to obtain the adaptive squared prediction error threshold based on kernel density estimation by obtaining the squared prediction error obtained by the internal short-circuit detection parameter acquisition module. The internal short-circuit fault diagnosis module is used to obtain the second squared prediction error based on the individual cell voltage data and the core principal component analysis model, and to determine whether an internal short-circuit fault has occurred based on the second squared prediction error and the adaptive squared prediction error threshold.

[0040] On the other hand, the present invention also provides an electronic device, comprising: Processor; and Memory for storing the executable instructions of the processor; The processor is configured to execute the short-circuit fault diagnosis method within a series lithium battery pack by executing the executable instructions.

[0041] On the other hand, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the aforementioned method for diagnosing short-circuit faults within a series lithium battery pack.

[0042] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any simple modifications or equivalent changes made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for diagnosing a short-circuit fault in a series lithium battery pack, characterized by, The method comprises the following steps: S01, using the historical voltage data of the lithium battery monomer cell as the input of the kernel principal component analysis model, and taking the first square prediction error in the residual subspace as the internal short circuit detection parameter; S02, obtaining the adaptive square prediction error threshold based on the kernel density estimation according to the first square prediction error in step S01, and taking the adaptive square prediction error threshold as the internal short circuit detection threshold; S03, inputting the obtained monomer cell voltage data into the kernel principal component analysis model, obtaining the second square prediction error, and comparing the second square prediction error with the adaptive square prediction error threshold to determine whether an internal short circuit fault occurs.

2. The method of diagnosing a short circuit fault in series connection lithium battery pack according to claim 1, characterized by, Historical voltage data of lithium battery pack single cells are acquired, represented as , wherein n represents the number of battery packs, and m represents the number of single cells in the battery pack.

3. The method of claim 1, wherein the method further comprises: In step S01, a kernel function of a kernel principal component analysis model is set as wherein, and denote two sample points in a data space, is a hyperparameter that controls the width or extent of the kernel function, is and the Euclidean distance between and The mapping process satisfies wherein, and denote two sample points in the data space, and represent a nonlinear mapping function that maps and to a higher dimensional space, represent the transpose of ; squared prediction error wherein and represents a statistic of the eigenvalues of the residual subspace, is a confidence limit of a standard normal distribution, represents a shape correction parameter.

4. The method of claim 1, wherein the method further comprises: Step S02 comprises: S021、Set the density function of kernel density estimation, denoted as where q represents the length of data sequence, h represents the bandwidth, represents the kernel function, x is the input data, represents the i-th sample point; S022、adopting the first square prediction error as input, calculating its probability density, denoted as wherein, represents a square prediction error threshold, is a square prediction error coefficient, represents a confidence limit, represents the first square prediction error; S023、performing kernel density estimation on the first square prediction error to obtain a distribution interval range and a probability density curve, and the adaptive square prediction error threshold is sequentially calculated by using the micro-element method and the composite trapezoidal rule; wherein, h min represents the minimum value of the distribution interval, h max represents the maximum value of the distribution interval.

5. The method of diagnosing a short circuit fault in series connection lithium battery pack according to claim 4, characterized by, In step S023, the micro-element method and the composite trapezoidal rule are used to calculate sequentially until the following formula is satisfied: ； wherein, represents the area of the probability density curve, represents the probability area cumulated from the right of the area of the probability density curve to find the squared prediction error threshold, represents the confidence limit, k represents the number of intervals that equally divide the range of the distribution interval, is the step size for each interval, i.e. , represents the step size of the i i th interval, represents the number of intervals cumulated from the right end h max to with step size , represents the squared prediction error threshold, j represents the number of the i h th interval corresponding to the range between the right end max and .

6. The method of claim 1, wherein the method further comprises: In step S03, the monomer cell voltage data is input into the kernel principal component analysis model using a sliding time window to obtain the second square prediction error corresponding to each sliding time window.

7. The method of diagnosing a short circuit fault in series connection lithium battery pack according to claim 6, characterized by, In step S03, when the second square prediction error corresponding to the sliding time window exceeds the adaptive square prediction error threshold, it is determined that an internal short circuit fault occurs, otherwise it is determined that no internal short circuit fault occurs.

8. A short circuit fault diagnosis system in series lithium battery pack characterized by, The internal short circuit fault diagnosis method of the series lithium battery pack according to any one of claims 1-7 is used for internal short circuit fault diagnosis, comprising: a data acquisition module for acquiring lithium battery monomer cell voltage data; an internal short circuit detection parameter acquisition module for obtaining the first square prediction error in the residual subspace according to the lithium battery monomer cell historical voltage data and the kernel principal component analysis model; an adaptive square prediction error threshold acquisition module for obtaining the adaptive square prediction error threshold based on the kernel density estimation according to the square prediction error obtained by the internal short circuit detection parameter acquisition module; an internal short circuit fault diagnosis module for obtaining the second square prediction error according to the monomer cell voltage data and the kernel principal component analysis model, and determining whether an internal short circuit fault occurs according to the second square prediction error and the adaptive square prediction error threshold.

9. An electronic device, characterized by comprise: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the series lithium battery pack internal short circuit fault diagnosis method according to any one of claims 1-7 by executing the executable instructions.

10. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the series lithium battery pack internal short circuit fault diagnosis method according to any one of claims 1-7.

Images