Fault diagnosis method for off-grid solar storage battery

Abstract

The invention relates to storage battery fault diagnosis, in particular to an off-grid solar storage battery fault diagnosis method, which comprises the following steps: acquiring voltage data of a target storage battery, performing wavelet packet decomposition on the voltage data, and constructing a wavelet packet coefficient matrix; performing singular value decomposition on the wavelet packet coefficient matrix to obtain a singular value matrix, and determining a feature vector curve corresponding to the fault feature value; the similarity between the feature vector curve and a reference feature vector curve is calculated, whether the target storage battery breaks down or not and the fault type are judged, and a first fault diagnosis result is obtained; when the target storage battery has a fault, constructing a feature sequence of the target storage battery according to the voltage data; respectively inputting the feature sequence into a plurality of isolated forest models to obtain corresponding binary tree features, and respectively calculating the abnormal probability of the feature sequence under each isolated forest model according to the binary tree features; according to the technical scheme provided by the invention, the defect of difficulty in accurate fault diagnosis of the storage battery in the prior art can be overcome.

Description

Technical Field

[0001] The invention relates to battery fault diagnosis, and in particular to a fault diagnosis method for an off-grid solar battery. Background Art

[0002] In recent years, with the widespread popularity of new energy vehicles, charging stations, battery swap stations, and battery swap cabinets are also being built in an all-round way. Due to the low cost of solar photovoltaic power generation, it has been widely applied and promoted, and there are more and more solar energy storage systems using solar photovoltaic power generation technology. In solar energy storage systems, batteries are widely used, and battery failure may lead to battery performance degradation, capacity loss, slower charging speed, and even battery failure. Therefore, developing an accurate battery fault diagnosis method is crucial to improving the reliability and service life of batteries.

[0003] At present, some battery fault diagnosis methods have been proposed. One common method is to monitor and analyze the internal parameters of the battery and evaluate the health of the battery in combination with the charging and discharging process. However, this method has the problem of inaccurate diagnosis of fault types, resulting in the need to improve the accuracy and reliability of the diagnosis results.

[0004] Another common method is to monitor and analyze the external characteristics of the battery (such as shell temperature, changes in battery appearance, etc.), combined with abnormal phenomena during use, to preliminarily determine whether the battery is faulty. However, this method can only provide rough fault diagnosis results. Summary of the invention

[0005] 1. Technical issues to be solved

[0006] In view of the above-mentioned shortcomings of the prior art, the present invention provides a fault diagnosis method for an off-grid solar battery, which can effectively overcome the defect of the prior art that it is difficult to accurately diagnose the fault of the battery.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] A fault diagnosis method for an off-grid solar battery comprises the following steps:

[0010] S1. Collect voltage data of the target battery, perform wavelet packet decomposition on the voltage data, and construct a wavelet packet coefficient matrix;

[0011] S2. Perform singular value decomposition on the wavelet packet coefficient matrix to obtain a singular value matrix, and determine the eigenvector curve corresponding to the fault eigenvalue;

[0012] S3, calculating the similarity between the characteristic vector curve and the reference characteristic vector curve, determining whether the target battery has a fault and the fault type, and obtaining a first fault diagnosis result;

[0013] S4. When a target battery fails, construct a characteristic sequence of the target battery according to the voltage data;

[0014] S5, inputting the feature sequence into multiple isolation forest models respectively, obtaining corresponding binary tree features, and calculating the abnormal probability of the feature sequence under each isolation forest model according to the binary tree features;

[0015] S6, repeat S5 until the preset number of cycles is reached, calculate the average abnormal probability of the feature sequence under each isolation forest model, and determine the fault type of the target battery to obtain a second fault diagnosis result;

[0016] S7. Combine the first fault diagnosis result and the second fault diagnosis result to perform comprehensive fault diagnosis on the target battery to obtain a final fault diagnosis result.

[0017] Preferably, in S1, voltage data of the target battery is collected, wavelet packet decomposition is performed on the voltage data, and a wavelet packet coefficient matrix is ​​constructed, including:

[0018] Collect voltage data of the target battery, perform wavelet packet decomposition on the voltage data, and obtain low-frequency signal wavelet coefficients and high-frequency signal wavelet coefficients;

[0019] The wavelet packet coefficient matrix is ​​constructed according to the wavelet coefficients of the low-frequency signal and the high-frequency signal.

[0020] Preferably, in S2, singular value decomposition is performed on the wavelet packet coefficient matrix to obtain a singular value matrix, and a characteristic vector curve corresponding to the fault characteristic value is determined, including:

[0021] Perform singular value decomposition on the wavelet packet coefficient matrix to obtain a singular value matrix;

[0022] The non-zero singular values ​​of the wavelet packet coefficient matrix are determined according to the singular value matrix, and the mean of the non-zero singular values ​​is calculated to obtain the fault characteristic value, and the characteristic vector curve corresponding to the fault characteristic value is determined.

[0023] Preferably, calculating the similarity between the characteristic vector curve and the reference characteristic vector curve in S3, determining whether the target battery has a fault and the type of fault, and obtaining a first fault diagnosis result includes:

[0024] Calculating the similarity between the characteristic vector curve and the reference characteristic vector curve, and determining whether the similarity is greater than a first preset distance threshold;

[0025] If the similarity is greater than a first preset distance threshold, it is determined that the target battery has a fault, otherwise it is determined that the target battery has not a fault;

[0026] When it is determined that a fault occurs in the target battery, a corresponding fault type is determined to obtain a first fault diagnosis result.

[0027] Preferably, when it is determined that a target battery fails, determining a corresponding fault type and obtaining a first fault diagnosis result includes:

[0028] Calculate the similarity between the feature vector curve and all sampling points in the reference feature vector curve, and select the target sampling points of the feature vector curve whose similarity is greater than a second preset distance threshold;

[0029] Extract the non-zero singular values ​​of the target sampling points to obtain the fault characteristic values;

[0030] The corresponding fault type is determined according to the fault characteristic value to obtain a first fault diagnosis result.

[0031] Preferably, the method for generating the reference characteristic vector curve comprises:

[0032] Sort all non-zero singular values ​​in the characteristic vector curve of each battery from small to large, and obtain the median of all non-zero singular values ​​after sorting;

[0033] A reference eigenvector curve is generated based on the median of all non-zero singular values ​​extracted from the eigenvector curve of each battery.

[0034] Preferably, when the target battery fails in S4, a characteristic sequence of the target battery is constructed according to the voltage data, including:

[0035] When a target battery fails, the voltage data is cleaned, and a characteristic sequence of the target battery is constructed based on the cleaned voltage data;

[0036] Among them, the K nearest neighbor method is used to supplement the missing values ​​and abnormal values ​​in the voltage data.

[0037] Preferably, in S5, the feature sequence is input into multiple isolation forest models respectively to obtain corresponding binary tree features, and the abnormal probability of the feature sequence under each isolation forest model is calculated according to the binary tree features, including:

[0038] The feature sequence is input into multiple isolation forest models respectively, and the binary tree algorithm is used to repeatedly search for the feature sequence in each isolation forest model until the preset termination condition is reached, and the binary tree features corresponding to the feature sequence in each isolation forest model are obtained;

[0039] According to the binary tree features corresponding to the feature sequence in each isolation forest model, the abnormal probability of the feature sequence under each isolation forest model is calculated respectively;

[0040] Among them, each isolation forest model corresponds to a different fault type.

[0041] Preferably, the step of calculating the abnormal probability of the feature sequence under each isolation forest model according to the binary tree features corresponding to the feature sequence in each isolation forest model comprises:

[0042] According to the binary tree characteristics, the abnormality and normality of the feature sequence under each isolation forest model are calculated;

[0043] According to the abnormality and normality of the feature sequence under each isolation forest model, the abnormal probability of the feature sequence under each isolation forest model is calculated respectively.

[0044] Preferably, in S6, the average abnormal probability of the feature sequence under each isolation forest model is calculated respectively, and the fault type of the target battery is determined to obtain a second fault diagnosis result, including:

[0045] The average abnormal probability of the feature sequence under each isolation forest model is calculated respectively, and the fault type corresponding to the isolation forest model with the largest average abnormal probability is taken as the fault type of the target battery to obtain the second fault diagnosis result.

[0046] (III) Beneficial effects

[0047] Compared with the prior art, the fault diagnosis method of an off-grid solar battery provided by the present invention has the following beneficial effects:

[0048] 1) Collect voltage data of the target battery, perform wavelet packet decomposition on the voltage data, construct a wavelet packet coefficient matrix, perform singular value decomposition on the wavelet packet coefficient matrix to obtain a singular value matrix, and determine the eigenvector curve corresponding to the fault eigenvalue, and be able to obtain the eigenvector curve based on the voltage data of the target battery, and by calculating the similarity between the eigenvector curve and the reference eigenvector curve, determine whether the target battery has a fault and the type of fault;

[0049] 2) When a target battery fails, a feature sequence of the target battery is constructed according to the voltage data, and the feature sequences are respectively input into a plurality of isolation forest models to obtain corresponding binary tree features, and the abnormal probability of the feature sequence under each isolation forest model is respectively calculated according to the binary tree features, and the average abnormal probability of the feature sequence under each isolation forest model is respectively calculated, and the fault type of the target battery is judged to obtain a second fault diagnosis result, and the fault type of the target battery is judged by inputting the feature sequence into a plurality of isolation forest models respectively, and calculating the average abnormal probability of the feature sequence under each isolation forest model;

[0050] 3) Combining the first fault diagnosis result and the second fault diagnosis result, a comprehensive fault diagnosis is performed on the target battery to obtain a final fault diagnosis result, thereby enabling accurate fault diagnosis of the battery and effectively improving the accuracy and reliability of the diagnosis result. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0052] Figure 1 It is a schematic diagram of the process of the present invention;

[0053] Figure 2 A schematic diagram of a flow chart of using a characteristic vector curve to determine whether a target battery has a fault and the type of fault in the present invention;

[0054] Figure 3 It is a flow chart of using the isolation forest model to judge the fault type of the target battery in the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] A fault diagnosis method for an off-grid solar battery, such as Figure 1 and Figure 2 As shown, ① collect the voltage data of the target battery, perform wavelet packet decomposition on the voltage data, and construct a wavelet packet coefficient matrix, which specifically includes:

[0057] Collect voltage data of the target battery, perform wavelet packet decomposition on the voltage data, and obtain low-frequency signal wavelet coefficients and high-frequency signal wavelet coefficients;

[0058] The wavelet packet coefficient matrix is ​​constructed according to the wavelet coefficients of the low-frequency signal and the high-frequency signal.

[0059] ② Perform singular value decomposition on the wavelet packet coefficient matrix to obtain the singular value matrix and determine the eigenvector curve corresponding to the fault eigenvalue, including:

[0060] Perform singular value decomposition on the wavelet packet coefficient matrix to obtain a singular value matrix;

[0061] The non-zero singular values ​​of the wavelet packet coefficient matrix are determined according to the singular value matrix, and the mean of the non-zero singular values ​​is calculated to obtain the fault characteristic value, and the characteristic vector curve corresponding to the fault characteristic value is determined.

[0062] ③ Calculate the similarity between the characteristic vector curve and the reference characteristic vector curve, determine whether the target battery has a fault, and the type of fault, and obtain a first fault diagnosis result, which specifically includes:

[0063] Calculating the similarity between the characteristic vector curve and the reference characteristic vector curve, and determining whether the similarity is greater than a first preset distance threshold;

[0064] If the similarity is greater than a first preset distance threshold, it is determined that the target battery has a fault, otherwise it is determined that the target battery has not a fault;

[0065] When it is determined that a fault occurs in the target battery, a corresponding fault type is determined to obtain a first fault diagnosis result.

[0066] Specifically, when it is determined that a target battery fails, a corresponding fault type is determined to obtain a first fault diagnosis result, including:

[0067] Calculate the similarity between the feature vector curve and all sampling points in the reference feature vector curve, and select the target sampling points of the feature vector curve whose similarity is greater than a second preset distance threshold;

[0068] Extract the non-zero singular values ​​of the target sampling points to obtain the fault characteristic values;

[0069] The corresponding fault type is determined according to the fault characteristic value to obtain a first fault diagnosis result.

[0070] In the technical solution of the present application, the method for generating a reference characteristic vector curve includes:

[0071] Sort all non-zero singular values ​​in the characteristic vector curve of each battery from small to large, and obtain the median of all non-zero singular values ​​after sorting;

[0072] A reference eigenvector curve is generated based on the median of all non-zero singular values ​​extracted from the eigenvector curve of each battery.

[0073] The above technical solution collects voltage data of the target battery, performs wavelet packet decomposition on the voltage data, constructs a wavelet packet coefficient matrix, performs singular value decomposition on the wavelet packet coefficient matrix, obtains a singular value matrix, and determines the eigenvector curve corresponding to the fault eigenvalue. It is possible to obtain the eigenvector curve based on the voltage data of the target battery, and determine whether the target battery has a fault and the type of fault by calculating the similarity between the eigenvector curve and the reference eigenvector curve.

[0074] like Figure 1 and Figure 3 As shown, ④ when the target battery fails, a characteristic sequence of the target battery is constructed according to the voltage data, specifically including:

[0075] When a target battery fails, the voltage data is cleaned, and a characteristic sequence of the target battery is constructed based on the cleaned voltage data;

[0076] Among them, the K nearest neighbor method is used to supplement the missing values ​​and abnormal values ​​in the voltage data.

[0077] ⑤ Input the feature sequence into multiple isolation forest models respectively to obtain the corresponding binary tree features, and calculate the abnormal probability of the feature sequence under each isolation forest model according to the binary tree features, including:

[0078] The feature sequence is input into multiple isolation forest models respectively, and the binary tree algorithm is used to repeatedly search for the feature sequence in each isolation forest model until the preset termination condition is reached, and the binary tree features corresponding to the feature sequence in each isolation forest model are obtained;

[0079] According to the binary tree features corresponding to the feature sequence in each isolation forest model, the abnormal probability of the feature sequence under each isolation forest model is calculated respectively;

[0080] Among them, each isolation forest model corresponds to a different fault type.

[0081] Specifically, according to the binary tree features corresponding to the feature sequence in each isolation forest model, the abnormal probability of the feature sequence under each isolation forest model is calculated respectively, including:

[0082] According to the binary tree characteristics, the abnormality and normality of the feature sequence under each isolation forest model are calculated;

[0083] According to the abnormality and normality of the feature sequence under each isolation forest model, the abnormal probability of the feature sequence under each isolation forest model is calculated respectively.

[0084] ⑥ Repeat ⑤ until the preset number of cycles is reached, calculate the average abnormal probability of the feature sequence under each isolation forest model, and determine the fault type of the target battery to obtain the second fault diagnosis result, which specifically includes:

[0085] The average abnormal probability of the feature sequence under each isolation forest model is calculated respectively, and the fault type corresponding to the isolation forest model with the largest average abnormal probability is taken as the fault type of the target battery to obtain the second fault diagnosis result.

[0086] In the above technical scheme, when a target battery fails, a feature sequence of the target battery is constructed based on voltage data, and the feature sequences are respectively input into multiple isolation forest models to obtain corresponding binary tree features, and the abnormal probability of the feature sequence under each isolation forest model is calculated based on the binary tree features, and the average abnormal probability of the feature sequence under each isolation forest model is calculated, and the fault type of the target battery is judged to obtain a second fault diagnosis result, and the fault type of the target battery is judged by inputting the feature sequence into multiple isolation forest models and calculating the average abnormal probability of the feature sequence under each isolation forest model.

[0087] ⑦ Combine the first fault diagnosis result and the second fault diagnosis result to perform comprehensive fault diagnosis on the target battery and obtain the final fault diagnosis result.

[0088] The above technical solution combines the first fault diagnosis result and the second fault diagnosis result to perform comprehensive fault diagnosis on the target battery and obtain a final fault diagnosis result, thereby being able to perform accurate fault diagnosis on the battery and effectively improving the accuracy and reliability of the diagnosis result.

[0089] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fault diagnosis method for an off-grid solar battery, characterized in that: The following steps are involved: S1. Collect voltage data of the target battery, perform wavelet packet decomposition on the voltage data, and construct a wavelet packet coefficient matrix; S2. Perform singular value decomposition on the wavelet packet coefficient matrix to obtain a singular value matrix, and determine the eigenvector curve corresponding to the fault eigenvalue; S3, calculating the similarity between the characteristic vector curve and the reference characteristic vector curve, determining whether the target battery has a fault and the fault type, and obtaining a first fault diagnosis result; S4. When a target battery fails, construct a characteristic sequence of the target battery according to the voltage data; S5, inputting the feature sequence into multiple isolation forest models respectively, obtaining corresponding binary tree features, and calculating the abnormal probability of the feature sequence under each isolation forest model according to the binary tree features; S6, repeat S5 until the preset number of cycles is reached, calculate the average abnormal probability of the feature sequence under each isolation forest model, and determine the fault type of the target battery to obtain a second fault diagnosis result; S7. Combine the first fault diagnosis result and the second fault diagnosis result to perform comprehensive fault diagnosis on the target battery to obtain a final fault diagnosis result.

2. The off-grid solar battery fault diagnosis method according to claim 1, characterized in that: In S1, the voltage data of the target battery is collected, the voltage data is decomposed by wavelet packets, and a wavelet packet coefficient matrix is ​​constructed, including: Collect voltage data of the target battery, perform wavelet packet decomposition on the voltage data, and obtain low-frequency signal wavelet coefficients and high-frequency signal wavelet coefficients; The wavelet packet coefficient matrix is ​​constructed according to the wavelet coefficients of the low-frequency signal and the high-frequency signal.

3. The fault diagnosis method for off-grid solar battery according to claim 2, characterized in that: In S2, the singular value decomposition of the wavelet packet coefficient matrix is ​​performed to obtain the singular value matrix, and the eigenvector curve corresponding to the fault eigenvalue is determined, including: Perform singular value decomposition on the wavelet packet coefficient matrix to obtain a singular value matrix; The non-zero singular values ​​of the wavelet packet coefficient matrix are determined according to the singular value matrix, and the mean of the non-zero singular values ​​is calculated to obtain the fault characteristic value, and the characteristic vector curve corresponding to the fault characteristic value is determined.

4. The off-grid solar battery fault diagnosis method according to claim 3, characterized in that: In S3, the similarity between the characteristic vector curve and the reference characteristic vector curve is calculated to determine whether the target battery has a fault and the type of fault, and obtain a first fault diagnosis result, including: Calculating the similarity between the characteristic vector curve and the reference characteristic vector curve, and determining whether the similarity is greater than a first preset distance threshold; If the similarity is greater than the first preset distance threshold, it is determined that the target battery has a fault, otherwise it is determined that the target battery has not a fault; When it is determined that a fault occurs in the target battery, a corresponding fault type is determined to obtain a first fault diagnosis result.

5. The off-grid solar battery fault diagnosis method according to claim 4, characterized in that: When it is determined that the target battery fails, the corresponding fault type is determined to obtain a first fault diagnosis result, including: Calculate the similarity between the feature vector curve and all sampling points in the reference feature vector curve, and select the target sampling points of the feature vector curve whose similarity is greater than a second preset distance threshold; Extract the non-zero singular values ​​of the target sampling points to obtain the fault characteristic values; The corresponding fault type is determined according to the fault characteristic value to obtain a first fault diagnosis result.

6. The off-grid solar battery fault diagnosis method according to claim 4 or 5, characterized in that: The method for generating the reference characteristic vector curve comprises: Sort all non-zero singular values ​​in the characteristic vector curve of each battery from small to large, and obtain the median of all non-zero singular values ​​after sorting; A reference eigenvector curve is generated based on the median of all non-zero singular values ​​extracted from the eigenvector curve of each battery.

7. The off-grid solar battery fault diagnosis method according to claim 1, characterized in that: In S4, when a target battery fails, a characteristic sequence of the target battery is constructed according to the voltage data, including: When a target battery fails, the voltage data is cleaned, and a characteristic sequence of the target battery is constructed based on the cleaned voltage data; Among them, the K nearest neighbor method is used to supplement the missing values ​​and abnormal values ​​in the voltage data.

8. The off-grid solar battery fault diagnosis method according to claim 1, characterized in that: In S5, the feature sequence is input into multiple isolation forest models respectively to obtain the corresponding binary tree features, and the abnormal probability of the feature sequence under each isolation forest model is calculated according to the binary tree features, including: The feature sequence is input into multiple isolation forest models respectively, and the binary tree algorithm is used to repeatedly search for the feature sequence in each isolation forest model until the preset termination condition is reached, and the binary tree features corresponding to the feature sequence in each isolation forest model are obtained; According to the binary tree features corresponding to the feature sequence in each isolation forest model, the abnormal probability of the feature sequence under each isolation forest model is calculated respectively; Among them, each isolation forest model corresponds to a different fault type.

9. The off-grid solar battery fault diagnosis method according to claim 8, characterized in that: The method of calculating the abnormal probability of the feature sequence under each isolation forest model according to the binary tree features corresponding to the feature sequence in each isolation forest model includes: According to the binary tree characteristics, the abnormality and normality of the feature sequence under each isolation forest model are calculated; According to the abnormality and normality of the feature sequence under each isolation forest model, the abnormal probability of the feature sequence under each isolation forest model is calculated respectively.

10. The off-grid solar battery fault diagnosis method according to claim 8, characterized in that: In S6, the average abnormal probability of the feature sequence under each isolation forest model is calculated respectively, and the fault type of the target battery is determined to obtain the second fault diagnosis result, including: The average abnormal probability of the feature sequence under each isolation forest model is calculated respectively, and the fault type corresponding to the isolation forest model with the largest average abnormal probability is taken as the fault type of the target battery to obtain the second fault diagnosis result.

Images