Lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis

Abstract

The invention relates to a lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis. The lithium battery internal short circuit fault diagnosis method comprises the following steps that a sliding window collects voltage data; performing continuous variation mode decomposition to extract trend components; performing trend component standardization processing, and calculating a covariance matrix; calculating a feature quantity and a feature vector, and calculating a square prediction error statistic control limit and a square prediction error statistic; and outputting the fault battery number and the fault occurrence time. According to the lithium battery internal short circuit fault diagnosis method, the early-stage tiny internal short circuit fault can be effectively identified on line, and the fault battery can be accurately positioned.

Description

Technical Field

[0001] The present invention relates to the field of lithium battery internal short circuit fault diagnosis, and in particular to a lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis. Background Art

[0002] Under the dual influence of environmental pollution and energy crisis, traditional fuel vehicles are facing increasingly severe challenges. Electric vehicles have become an important driving direction for the development of the contemporary automobile industry due to their advantages such as low environmental pollution and high energy conversion rate. In recent years, the production and research and development technology of lithium-ion batteries has become increasingly mature, providing an important breakthrough for the development of electric vehicles. Behind the booming development of lithium-ion batteries and electric vehicles, there are also many hidden dangers. In particular, the frequent safety accidents of electric vehicle batteries have brought great threats to people's property and life safety. The occurrence of electric vehicle fire accidents is actually caused by thermal runaway caused by lithium-ion battery failure. There are many types of lithium-ion battery failure inducements, and the failure inducements that can eventually lead to thermal runaway can be roughly divided into: mechanical abuse, thermal abuse and electrical abuse. Internal short circuit is the common link under the three inducing factors of thermal runaway of lithium-ion batteries. Therefore, timely and accurate diagnosis of internal short circuit faults can effectively prevent safety accidents in battery systems and ensure personal and property safety.

[0003] Existing short-circuit fault diagnosis methods for lithium-ion batteries can be mainly divided into the following three categories: battery model-based diagnosis methods, machine learning-based diagnosis methods, and sensor signal-based diagnosis methods. The accuracy of battery model-based diagnosis methods is highly dependent on the accuracy of the model used, and due to the large amount of model calculation, it is difficult to apply to real-time fault diagnosis; machine learning-based diagnosis methods do not require accurate battery models and can achieve automatic fault classification, but require a large amount of data to train the model and have high requirements for data quality; signal processing-based diagnosis methods do not require accurate battery models and can achieve online detection of the internal state of the battery, but have high requirements for sensor accuracy, high complexity of signal processing methods, and large amount of calculation. Summary of the invention

[0004] In order to solve the technical problems existing in the prior art, the present invention provides a lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis, which can effectively identify early-stage minor internal short circuit faults online and accurately locate faulty batteries.

[0005] To achieve the above object, the technical solution of the present invention is as follows:

[0006] A lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis includes the following steps:

[0007] S1: Sliding window to collect voltage data;

[0008] S2: continuous variational mode decomposition to extract trend components;

[0009] S3: Standardization of trend component and calculation of covariance matrix;

[0010] S4: Calculate the characteristic quantity and characteristic vector, calculate the Hotelling statistic and the squared prediction error statistic and their corresponding control limits;

[0011] S5: Output the faulty battery number and the time when the fault occurred.

[0012] As a preferred technical solution, step S2 includes: adding constraints to the variational modal decomposition process; performing constrained continuous variational modal decomposition on the collected voltage data to obtain a trend component.

[0013] As a preferred technical solution, in step S2, the formula for adding constraints to the variational mode decomposition process is:

[0014]

[0015] In the formula, ω L is the center frequency of the Lth mode, β i (t), β L (t) is the impulse response of the filter.

[0016] As a preferred technical solution, in step S3, the trend component standardization process is to assume that the original data X' is N samples containing M-dimensional measurement variables and standardize them:

[0017]

[0018] In the formula, μ is X i ', σ is the mean of X i ', the standardized data is X=[X1,...,X M ], the covariance matrix of X after standardization is approximately expressed as:

[0019] As a preferred technical solution, in step S4, the eigenvalue corresponding to the selected principal element is used to calculate the Hotelling statistic control limit Sum of squared prediction error statistics control limits

[0020]

[0021]

[0022] In the formula, F α(k,nk) refers to the F distribution with the first degree of freedom k and the second degree of freedom nk at the α test level (usually 95%); c α Represents the limit of the standard normal distribution at the α test level.

[0023] As a preferred technical solution, after the control limit is obtained by principal component analysis modeling of historical normal data, the data to be tested X t ' est After standardization, X test Substitute into the principal component analysis model and calculate the Hotelling statistic and the squared prediction error statistic:

[0024]

[0025] Where I is an M*M dimensional unit vector.

[0026] As a preferred technical solution, in step S5, the contribution of each battery to the squared prediction error statistic is expressed as: SPE i =(X test (IP pc P pc T )) 2 , introduce the single variable limit value η i , if satisfied It is considered that the fault is caused by this variable, and the specific number and time of the faulty battery are located through the square prediction error statistics and the statistical contribution exceeding the limit value.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] (1) The lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis of the present invention realizes adaptive selection of variational mode decomposition K value by adding decomposition constraints, improves the extraction effect of voltage low-frequency trend component, and improves the traditional variational mode algorithm for online modeling and diagnosis integration, thus solving the problem that the variational mode online fault diagnosis model may not be suitable.

[0029] (2) The lithium battery internal short-circuit fault diagnosis method of the present invention based on improved variational mode decomposition and principal component analysis, the proposed lithium-ion battery early micro-internal short-circuit fault diagnosis method can resist Gaussian white noise with a signal-to-noise ratio greater than or equal to 50dB, effectively identify early micro-internal short-circuit faults online, and accurately locate the faulty battery.

[0030] (3) Compared with the internal short-circuit fault diagnosis method of the prior art based on the correlation coefficient method, the lithium battery internal short-circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis of the present invention has the advantages of wider application range, higher accuracy and stronger robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a flow chart of the lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis of the present invention;

[0032] Figure 2 It is a fault diagnosis flow chart of the improved VMD-PCA method in the lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis of the present invention. DETAILED DESCRIPTION

[0033] The technical solution of the present invention is further described below in conjunction with specific implementation methods:

[0034] like Figure 1 As shown, a lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis includes the following steps:

[0035] S1: Sliding window to collect voltage data;

[0036] In this embodiment, the voltage data of the battery pack is collected through the battery management system, and a sliding window for real-time updating of the voltage data is introduced. The value of the sliding window N is set to 80, and the sliding window moves every 5 seconds, updating 5 data points each time.

[0037] S2: continuous variational mode decomposition to extract trend components;

[0038] Specifically include:

[0039] (1) Add constraints to the variational mode decomposition process:

[0040]

[0041] In the formula, ω L is the center frequency of the Lth mode, β i (t), β L (t) is the impulse response of the filter.

[0042] (2) The collected voltage data is subjected to constrained continuous variational mode decomposition to obtain the trend component.

[0043] S3: Standardization of trend component, calculation of covariance matrix; specifically, standardization of trend component is to assume that the original data X' is N samples containing M-dimensional measurement variables and standardize them:

[0044]

[0045] In the formula, μ is X i ', σ is the mean of X i ', the standardized data is X=[X1,...,X M ], the covariance matrix of X after standardization can be approximately expressed as:

[0046] S4: Calculate the characteristic quantity and characteristic vector, calculate the Hotelling statistic and the squared prediction error statistic and their corresponding control limits; calculate the Hotelling statistic control limits using the eigenvalues ​​corresponding to the selected principal element Sum of squared prediction error statistics control limits

[0047]

[0048] In the formula, F α (k,nk) refers to the F distribution with the first degree of freedom k and the second degree of freedom nk at the α test level (usually 95%); c α Represents the limit of the standard normal distribution at the α test level.

[0049] After the control limit is obtained by principal component analysis modeling of historical normal data, the data to be tested X t ' est After standardization, X test Substitute into the principal component analysis model and calculate the Hotelling statistic and the squared prediction error statistic:

[0050]

[0051] Where I is an M*M dimensional unit vector.

[0052] S5: Output the fault battery number and the time of fault occurrence. Specifically, the contribution of each battery to the square prediction error statistic is expressed as: SPE i =(X test (IP pc P pc T )) 2 , introduce the single variable limit value ηi, if it satisfies The fault is considered to be caused by this variable.

[0053] The specific number and time of the faulty battery are located through the squared prediction error statistics and the statistical contribution exceeding the limit.

[0054] This embodiment is only a further explanation of the present invention, not a limitation of the present invention. After reading this specification, those skilled in the art may make non-creative modifications to this embodiment as needed, but as long as it is within the scope of the claims of the present invention, it will be protected by the patent law.

Claims

1. A lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis, characterized in that: The following steps are involved: S1: Sliding window to collect voltage data; S2: continuous variational mode decomposition to extract trend components; S3: Standardization of trend component and calculation of covariance matrix; S4: Calculate the characteristic quantity and characteristic vector, calculate the Hotelling statistic and the squared prediction error statistic and their corresponding control limits; S5: Output the faulty battery number and the time when the fault occurred.

2. The lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis according to claim 1 is characterized in that: The step S2 includes: adding constraints to the variational mode decomposition process; performing constrained continuous variational mode decomposition on the collected voltage data to obtain a trend component.

3. The lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis according to claim 2 is characterized in that: In step S2, the formula for adding constraints to the variational mode decomposition process is: In the formula, ω L is the center frequency of the Lth mode, β i (t), β L (t) is the impulse response of the filter.

4. The lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis according to claim 1 is characterized in that: In step S3, the trend component standardization process is to assume that the original data X' is N samples containing M-dimensional measurement variables and standardize them: In the formula, μ is X i ', σ is the mean of X i ', the standardized data is X=[X1,...,X M ], the covariance matrix of X after standardization is approximately expressed as:

5. The lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis according to claim 1 is characterized in that: In step S4, the Hotelling statistic control limit is calculated by using the eigenvalue corresponding to the selected principal component. Sum of squared prediction error statistics control limits In the formula, F α (k,nk) refers to the F distribution with the first degree of freedom k and the second degree of freedom nk at the α test level (usually 95%); c α Represents the limit of the standard normal distribution at the α test level.

6. The lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis according to claim 1 is characterized in that: After the control limit is obtained by principal component analysis modeling of historical normal data, the data to be tested X t ' est After standardization, X test Substitute into the principal component analysis model and calculate the Hotelling statistic and the squared prediction error statistic: Where I is an M*M dimensional unit vector.

7. A lithium battery internal short circuit fault diagnosis method based on improved variational mode decomposition and principal component analysis according to claim 1, characterized in that: In step S5, the contribution of each battery to the squared prediction error statistic is expressed as: SPE i =(X test (IP pc P pc T )) 2 , introduce the single variable limit value η i , if satisfied It is considered that the fault is caused by this variable, and the specific number and time of the faulty battery are located through the square prediction error statistics and the statistical contribution exceeding the limit value.

Images