CN117171588A - Method for detecting gradient utilization faults of power battery - Google Patents
Method for detecting gradient utilization faults of power battery Download PDFInfo
- Publication number
- CN117171588A CN117171588A CN202311442766.0A CN202311442766A CN117171588A CN 117171588 A CN117171588 A CN 117171588A CN 202311442766 A CN202311442766 A CN 202311442766A CN 117171588 A CN117171588 A CN 117171588A
- Authority
- CN
- China
- Prior art keywords
- feature
- fault type
- fault
- vector
- feature vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims description 47
- 239000013598 vector Substances 0.000 claims abstract description 242
- 239000011159 matrix material Substances 0.000 claims abstract description 28
- 238000001514 detection method Methods 0.000 claims abstract description 23
- 238000000354 decomposition reaction Methods 0.000 claims description 7
- 238000000513 principal component analysis Methods 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000007599 discharging Methods 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
Landscapes
- Charge And Discharge Circuits For Batteries Or The Like (AREA)
Abstract
The invention relates to the technical field of data processing, and provides a power battery echelon utilization fault detection method, which comprises the following steps: collecting related data of different echelons of a plurality of power batteries and fault types; respectively constructing initial matrixes for related data of different echelons of the same fault type, acquiring feature vectors for each initial matrix to obtain a plurality of feature chains of each fault type, and acquiring a plurality of enhanced feature vectors of each fault type; according to the feature vector matching result of each enhanced feature vector of each fault type in other fault types, obtaining the matching feature value combination of each enhanced feature vector, obtaining a larger sequence of each fault type, and obtaining the fault vector and the fault threshold of each fault type; and performing fault detection on the power battery according to the fault vector and the fault threshold value to finish the fault detection of the power battery cascade utilization. The invention aims to solve the problem that similar characteristic changes in different fault modes affect fault detection results.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a gradient utilization fault detection method for a power battery.
Background
The gradient utilization of the power battery means that after the battery of the electric vehicle is used for a period of time, the capacity of the battery is reduced, and the gradient utilization of the power battery is realized by secondarily utilizing the battery to other fields; as the battery capacity decreases, the power battery used in cascade may generate various faults, such as faults of battery capacity attenuation, internal resistance increase, etc.; in the prior art, faults generated by gradient utilization of a power battery are usually detected by a neural network training method according to related data generated when the battery is used; however, the neural network training has larger requirements on data quantity, the accuracy of the data is also higher, and meanwhile, the characteristic changes of some fault modes on part of related data are similar, so that an accurate fault detection result cannot be obtained by the related data alone; therefore, the change characteristics of the relevant data of the battery in different echelon use processes need to be analyzed, and further, the characteristics capable of singly reflecting the fault mode are extracted, so that the fault detection of the echelon utilization of the power battery is realized, and the fault detection accuracy is improved.
Disclosure of Invention
The invention provides a power battery echelon utilization fault detection method, which is used for solving the problem that similar characteristic changes in different existing fault modes influence fault detection results, and adopts the following technical scheme:
one embodiment of the present invention provides a method for power cell cascade utilization fault detection, comprising the steps of:
collecting related data of different echelons of a plurality of power batteries, and obtaining fault types of each power battery;
acquiring a plurality of enhanced feature vectors of each fault type according to a feature chain formed by feature vectors of different echelon initial matrixes of each fault type;
according to the feature vector matching result of each enhanced feature vector of each fault type in other fault types, obtaining the fault vector and the fault threshold of each fault type;
and performing fault detection on the power battery according to the fault vector and the fault threshold value to finish the fault detection of the power battery cascade utilization.
Further, the specific obtaining method of the plurality of enhanced feature vectors of each fault type is as follows:
calculating the enhancement degree of each feature chain according to the feature chains formed by the feature vectors of different echelon initial matrixes of each fault type; taking any fault type as a target fault type, taking a feature chain with the enhancement degree larger than an enhancement threshold value in the target fault type as an enhancement feature chain, and taking a feature vector corresponding to each element in each enhancement feature chain as an enhancement feature vector of the target fault type to obtain a plurality of enhancement feature vectors of the target fault type; several enhanced feature vectors for each fault type are obtained.
Further, the calculating the enhancement degree of each characteristic chain comprises the following specific methods:
acquiring a plurality of feature chains of each fault type according to the matching result of the feature vectors of the adjacent echelon initial matrix of each fault type; taking any one fault type as a target fault type, taking any one feature chain in the target fault type as a target feature chain, taking the sequence value of each element in the target feature chain as an abscissa, taking the element value as an ordinate, converting each element in the target feature chain into a coordinate point in a coordinate system, carrying out PCA analysis on the coordinate point to obtain the maximum projection vector of each coordinate point of the target feature chain, obtaining the direction of the maximum projection vector, and taking the ratio of the direction to 90 degrees as the enhancement degree of the target feature chain; the enhancement degree of each characteristic chain of each fault type is obtained.
Further, the specific acquisition method of the feature chains of each fault type is as follows:
performing KM matching on the feature vectors of adjacent echelon initial matrixes of the same fault type to obtain a plurality of feature vector pairs of adjacent echelons; taking any fault type as a target fault type, dividing the first echelon and the last echelon in the target fault type, wherein each feature vector of each echelon has a feature vector corresponding to a feature vector pair in an adjacent echelon, and obtaining a plurality of feature vector sequences of the target fault type according to the feature vector pairs, wherein two adjacent elements in each feature vector sequence are feature vector pairs;
each feature vector corresponds to a feature value, each element in the feature vector sequence is replaced by the corresponding feature value, and the obtained sequence is recorded as a plurality of feature chains of the target fault type; a number of feature chains for each fault type are acquired.
Further, the specific method for obtaining the plurality of feature vector pairs of the adjacent echelons includes the following steps:
taking any fault type as a target fault type, acquiring all echelon related data of a plurality of batteries corresponding to the target fault type, and arranging the related data of a plurality of batteries in each echelon according to the same battery sequence to obtain a related data sequence of each echelon;
constructing an initial matrix of each echelon according to the number of elements for the related data sequence of each echelon, and obtaining a plurality of feature vectors and corresponding feature values for each initial matrix through SVD decomposition;
performing KM matching on the feature vectors in the two initial matrixes of the adjacent echelons, and taking the feature vectors as bipartite graph nodes and cosine similarity of the feature vectors as side values to obtain a plurality of feature vector pairs of the adjacent echelons;
and acquiring a plurality of feature vector pairs of adjacent echelons of each fault type.
Further, the specific obtaining method of the fault vector and the fault threshold of each fault type includes:
taking any fault type as a reference fault type, and obtaining a matching feature value set of each enhanced feature vector in the reference fault type according to feature vector matching results of each enhanced feature vector in the reference fault type in other fault types;
for a matching characteristic value set of any enhancement characteristic vector in the reference fault type, obtaining a maximum value and a next-maximum value in the matching characteristic value set, and taking the ratio of a difference value obtained by subtracting the next-maximum value from the maximum value to the maximum value as the larger of the enhancement characteristic vector; acquiring the maximization of each enhanced feature vector in the reference fault type, and arranging all the maximization in a descending order to obtain a maximization sequence of the reference fault type, wherein the enhanced feature vector corresponding to the maximum element in the maximization sequence is used as the fault vector of the reference fault type, and the second element in the maximization sequence is used as the fault threshold of the reference fault type;
and obtaining fault vectors and fault thresholds of each fault type.
Further, the specific method for obtaining the matching eigenvalue set of each enhancement eigenvector in the reference fault type includes:
performing KM matching on each enhanced feature vector in the reference fault type and all feature vectors in other fault types to obtain a plurality of matched feature vectors of each enhanced feature vector in the reference fault type;
for any one enhancement feature vector in the reference fault type, a plurality of matching feature vectors of the enhancement feature vector are obtained, and a set formed by feature values corresponding to all the matching feature vectors is recorded as a matching feature value set of the enhancement feature vector; and acquiring a matching characteristic value set of each enhanced characteristic vector in the reference fault type.
Further, the specific method for obtaining the plurality of matching feature vectors of each enhancement feature vector in the reference fault type includes:
taking each enhanced feature vector in the reference fault type as a node on the left side of the bipartite graph, taking all feature vectors in any other fault type as nodes on the right side of the bipartite graph, acquiring edge values between the nodes through the enhanced feature vectors corresponding to the nodes on the two sides and the DTW distance of the feature vectors, and acquiring feature vectors matched with each enhanced feature vector in the reference fault type in other fault types through KM matching;
and obtaining the feature vector matched with each enhanced feature vector in the reference fault type in each other fault type, and recording the feature vector as a plurality of matched feature vectors of each enhanced feature vector.
The beneficial effects of the invention are as follows: according to the invention, a plurality of echelons of related data are acquired for each fault type, the enhanced feature vector capable of reflecting the features of the related data is analyzed, the fault vector capable of singly reflecting the fault type is screened out, the fault threshold is obtained, the fault detection of the echelon utilization of the power battery is completed, and the problem that similar feature changes in different fault modes influence the fault detection result is avoided; the method comprises the steps of constructing an initial matrix for each echelon of each fault type, obtaining feature chains through matching of adjacent echelon feature vectors, analyzing trends of the feature chains, extracting feature chains with larger enhancement trends to obtain enhancement feature vectors, and preliminarily obtaining feature vectors capable of reflecting the fault types; and the degree of the fault type is quantized through single reflection of the enhanced feature vector, and finally, the fault vector of each fault type with the largest maximization is obtained through the maximization analysis, so that the one-to-one correspondence between the fault vector and the fault type is ensured, confusion is avoided, and the fault detection precision and efficiency of the gradient utilization of the power battery are improved.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only some embodiments of the invention, and that other drawings can be obtained according to these drawings without inventive faculty for a person skilled in the art.
Fig. 1 is a schematic flow chart of a method for detecting a power battery cascade utilization fault according to an embodiment of the invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, a flowchart of a method for detecting a power battery cascade utilization fault according to an embodiment of the invention is shown, the method includes the following steps:
and S001, collecting related data of different echelons of a plurality of power batteries, and obtaining the fault type of each power battery.
The purpose of this embodiment is to perform fault detection on the cascade utilization of the power battery, so related data of different cascades in the cascade utilization process of the power battery needs to be obtained first, where the related data are data generated in the use process of the power battery, such as current data, voltage data, temperature data, etc., and this embodiment is described with the current data as related data; meanwhile, in order to ensure that the related data of each fault type can be obtained, a large amount of related data of power batteries need to be acquired for analysis, and the embodiment acquires the related data of 200 power batteries in total; in the echelon utilization process of the power battery, the echelon represents one charge-discharge cycle process of the power battery, so that each echelon corresponds to current data in the use process of one power battery, wherein other parameters in each use process have small changes, and the embodiment ignores the changes; the power battery gradient utilization process has different stages, namely the initial stage is used for an electric vehicle, the next stage is used as an energy storage system, and the next stage in the lower use field is entered, so that current data of all gradients of each power battery in different stages are obtained, all charging and discharging circulation processes of all gradients, namely all power batteries in the whole life cycle, are determined, and the fault type of each power battery is determined.
So far, the related data of different echelons of a plurality of power batteries are obtained, and meanwhile, the fault type of each power battery is obtained.
Step S002, respectively constructing initial matrixes for related data of different echelons of the same fault type, acquiring feature vectors for each initial matrix, acquiring a plurality of feature chains of each fault type through matching among feature vectors acquired by adjacent echelons of the same fault type, and acquiring a plurality of enhanced feature vectors of each fault type according to trend changes of the feature chains.
It should be noted that, for each fault type, features of each fault type need to be acquired first, an initial matrix is constructed by related data of the same echelon under each fault type, and a plurality of feature vectors and corresponding feature values are acquired for the initial matrix through singular value decomposition; after the feature vectors are obtained, matching the feature vectors obtained by adjacent echelons to obtain a feature chain, and extracting the enhanced feature vectors, namely the feature vectors with larger trend changes, by analyzing the trend changes of the feature chain; and the larger trend change indicates that the feature vector can reflect the fault feature more, so as to provide a basis for the subsequent analysis of the fault vector.
Specifically, taking any fault type as an example, acquiring all echelon related data of a plurality of batteries corresponding to the fault type, and arranging the related data of a plurality of batteries in each echelon according to the same battery sequence to obtain a related data sequence of each echelon; taking any one echelon related data sequence as an example, obtaining the element number of the related data sequence, taking a result obtained by rounding an element number evolution value as the number of rows of the echelon initial matrix, filling the related data sequence into the initial matrix row by adopting a rounding method, wherein the number of columns of the initial matrix is the minimum number of columns which can meet the requirement of completely filling the initial matrix, and if the last row can not be completely filled, completing the construction of the initial matrix by supplementing 0, and obtaining the initial matrix of the echelon of the fault type; according to the method, each echelon initial matrix of each fault type is obtained, a plurality of feature vectors are obtained for each initial matrix through SVD decomposition, and meanwhile, feature values corresponding to each feature vector are obtained, wherein SVD decomposition, namely singular value decomposition, is used for decomposing the matrix through singular value decomposition to obtain a plurality of feature vectors and corresponding feature values, which are known technologies, and the embodiment is not repeated.
Further, taking two initial matrixes of adjacent echelons in any fault type as an example, performing KM matching on feature vectors of the two initial matrixes, taking a plurality of feature vectors of one initial matrix as a node on the left side of a bipartite graph, taking a plurality of feature vectors of the other initial matrix as a node on the right side of the bipartite graph, taking cosine similarity of feature vectors corresponding to the nodes on the left side and the right side as a boundary value between the nodes, and performing matching on the nodes on the left side and the right side of the bipartite graph to obtain a plurality of feature vector pairs of the adjacent echelons, wherein the fact that the sizes of the initial matrixes of different echelons in the same fault type are equal, and the number of the feature vectors of different echelons is equal; according to the method, a plurality of feature vector pairs of any adjacent echelon in the fault type are obtained, the first echelon and the last echelon are removed, each feature vector of each echelon has a feature vector corresponding to the feature vector pair in the adjacent echelon, a plurality of feature vector sequences of the fault type can be obtained according to the feature vector pairs, two adjacent elements in each feature vector sequence are feature vector pairs obtained through KM matching, each feature vector corresponds to a feature value, each element in the feature vector sequence is replaced by a corresponding feature value, and the obtained sequence is recorded as a plurality of feature chains of the fault type; and acquiring a plurality of characteristic chains of each fault type according to the method.
Further, taking any one of the feature chains in any one of the fault types as an example, taking the sequence value of each element in the feature chain as an abscissa and the element value as an ordinate, converting each element in the feature chain into a coordinate point in a coordinate system, performing PCA analysis on the coordinate point to obtain a maximum projection vector of each coordinate point of the feature chain, wherein the maximum projection vector obtained by PCA analysis is in the prior art, namely, principal component direction analysis is not repeated in the embodiment; after the maximum projection vector is obtained, the direction of the maximum projection vector is obtained, the ratio of the direction to 90 degrees is used as the enhancement degree of the characteristic chain, and the larger the enhancement degree is, the more obvious the increase trend of the characteristic chain is; according to the method, the enhancement degree of each feature chain of the fault type is obtained, an enhancement threshold is preset, the enhancement threshold is described by adopting 0.6, the feature chains with the enhancement degree larger than the enhancement threshold are taken as enhancement feature chains, the feature chains smaller than the enhancement threshold are not processed, and the feature vector corresponding to each element in each enhancement feature chain is taken as the enhancement feature vector of the fault type, so that a plurality of enhancement feature vectors of the fault type can be obtained; and obtaining a plurality of enhanced feature vectors of each fault type according to the method.
So far, an initial matrix is built for different echelons of each fault type, a feature chain is obtained, and a plurality of enhanced feature vectors of each fault type are obtained by analyzing trend changes of the feature chain.
Step S003, according to the feature vector matching result of each enhanced feature vector of each fault type in other fault types, obtaining the matching feature value combination of each enhanced feature vector, calculating the maximization of each matching feature value set, obtaining the maximization sequence of each fault type, and obtaining the fault vector and the fault threshold of each fault type according to the maximization sequence.
It should be noted that, several enhancement feature vectors of each fault type correspond to features with larger increasing trend, however, these features may also have larger increasing trend in other fault types, so that these enhancement feature vectors cannot reflect the fault type, so that feature vectors corresponding to other fault types need to be obtained for the enhancement feature vectors through matching, and the feature values of these feature vectors are analyzed to obtain the bigger nature of each enhancement feature vector, where the bigger nature reflects the degree that it can reflect the fault type independently, and the bigger nature can reflect the bigger nature, so as to obtain the fault vector finally.
In particularTaking any fault type as a reference fault type, taking each enhanced feature vector in the reference fault type as a node on the left side of a bipartite graph, taking all feature vectors in any other fault type as nodes on the right side of the bipartite graph, and describing that because the feature vectors matched by the enhanced feature vectors in other fault types do not necessarily correspond to large change trends, the enhanced feature vectors are possibly not enhanced feature vectors, the matching results need to be obtained from all feature vectors, and regarding the edge values between the nodes on the left side and the right side of the bipartite graph, the dimension numbers of the feature vectors on the two sides can be differentThe edge between the nodes on the left and right sides is represented as the similarity between feature vectors, wherein +.>For the DTW distance between the enhanced feature vector and the feature vector corresponding to the nodes at two sides, < + >>Representing an exponential function based on natural constants, the present embodiment employs +.>The function presents an inverse proportion relation, and an implementer can set an inverse proportion function according to actual conditions; performing KM matching on the bipartite graph to obtain feature vectors matched with each enhanced feature vector in the reference fault type in the other fault types, obtaining feature vectors matched with each enhanced feature vector in the reference fault type in the other fault types according to the method, and recording the feature vectors as a plurality of matched feature vectors of each enhanced feature vector.
Further, taking any one enhancement feature vector in the reference fault type as an example, acquiring a plurality of matching feature vectors of the enhancement feature vector, and recording a set formed by feature values corresponding to all the matching feature vectors as a matching feature value set of the enhancement feature vector; obtaining the maximum value and the next maximum value in the matched characteristic value set, and subtracting the next maximum value from the maximum value to obtain the ratio of the difference value to the maximum value, wherein the ratio is used as the larger of the enhanced characteristic vector; and obtaining the maximization of each enhanced feature vector in the reference fault type according to the method, and arranging all the maximization in a descending order to obtain a maximization sequence of the reference fault type, wherein the enhanced feature vector corresponding to the largest element in the maximization sequence is used as the fault vector of the reference fault type, and the second element in the maximization sequence is used as the fault threshold of the reference fault type.
Further, according to the method, a matching characteristic value set of each enhancement characteristic vector in each fault type is obtained, the maximization is calculated, a maximization sequence of each fault type is obtained, and then a fault vector and a fault threshold of each fault type are obtained; the fault vector corresponds to the maximum value in the larger sequence, and then the fault vector can be indicated to be an enhanced feature vector which singly reflects the fault type.
So far, the fault vector and the fault threshold value of each fault type are obtained by quantifying the degree that the enhanced feature vector singly reflects the fault type.
And S004, performing fault detection on the power battery according to the fault vector and the fault threshold value, and completing the fault detection of the power battery cascade utilization.
After obtaining fault vectors and fault thresholds of each fault type, obtaining all echelon related data of the power battery to be detected practically, namely, from the beginning of use to the current of all echelons, constructing a matrix for the obtained current data according to an initial matrix construction method, marking the obtained matrix as a history matrix of the power battery, and decomposing the history matrix through SVD to obtain a plurality of feature vectors and corresponding feature values; taking the characteristic vector of the power battery as a node on the left side of the bipartite graph, taking the fault vector of each fault type as a node on the right side of the bipartite graph, and acquiring the edge value between the nodes by adopting the DTW distance between the characteristic vector and the fault vector, namelyAs a boundary value, wherein->Represents the DTW distance between the corresponding feature vector of the left and right nodes and the fault vector,/and the like>Representing an exponential function based on natural constants, the present embodiment employs +.>The function presents an inverse proportion relation, and an implementer can set an inverse proportion function according to actual conditions; obtaining a plurality of matching pairs through KM matching, taking any matching pair as an example, if the characteristic value of the characteristic vector corresponding to the left node in the matching pair is larger than the fault threshold value of the fault vector corresponding to the right node, indicating that the power battery has the fault type of the fault vector corresponding to the right node, and if the characteristic value is smaller than or equal to the fault threshold value, not having the corresponding fault type; and the fault detection of the power battery is finished according to the method.
Thus, the fault detection of the power battery cascade utilization process is completed.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, alternatives, and improvements that fall within the spirit and scope of the invention.
Claims (8)
1. A method for detecting a power battery cascade utilization fault, the method comprising the steps of:
collecting related data of different echelons of a plurality of power batteries, and obtaining fault types of each power battery;
acquiring a plurality of enhanced feature vectors of each fault type according to a feature chain formed by feature vectors of different echelon initial matrixes of each fault type;
according to the feature vector matching result of each enhanced feature vector of each fault type in other fault types, obtaining the fault vector and the fault threshold of each fault type;
and performing fault detection on the power battery according to the fault vector and the fault threshold value to finish the fault detection of the power battery cascade utilization.
2. The method for detecting power battery cascade utilization faults as claimed in claim 1, wherein the specific obtaining method of the enhanced feature vectors of each fault type is as follows:
calculating the enhancement degree of each feature chain according to the feature chains formed by the feature vectors of different echelon initial matrixes of each fault type; taking any fault type as a target fault type, taking a feature chain with the enhancement degree larger than an enhancement threshold value in the target fault type as an enhancement feature chain, and taking a feature vector corresponding to each element in each enhancement feature chain as an enhancement feature vector of the target fault type to obtain a plurality of enhancement feature vectors of the target fault type; several enhanced feature vectors for each fault type are obtained.
3. The method for detecting the cascade utilization fault of the power battery according to claim 2, wherein the step of calculating the enhancement degree of each characteristic chain comprises the following specific steps:
acquiring a plurality of feature chains of each fault type according to the matching result of the feature vectors of the adjacent echelon initial matrix of each fault type; taking any one fault type as a target fault type, taking any one feature chain in the target fault type as a target feature chain, taking the sequence value of each element in the target feature chain as an abscissa, taking the element value as an ordinate, converting each element in the target feature chain into a coordinate point in a coordinate system, carrying out PCA analysis on the coordinate point to obtain the maximum projection vector of each coordinate point of the target feature chain, obtaining the direction of the maximum projection vector, and taking the ratio of the direction to 90 degrees as the enhancement degree of the target feature chain; the enhancement degree of each characteristic chain of each fault type is obtained.
4. A method for detecting a power battery cascade utilization fault according to claim 3, wherein the specific acquisition method is as follows:
performing KM matching on the feature vectors of adjacent echelon initial matrixes of the same fault type to obtain a plurality of feature vector pairs of adjacent echelons; taking any fault type as a target fault type, dividing the first echelon and the last echelon in the target fault type, wherein each feature vector of each echelon has a feature vector corresponding to a feature vector pair in an adjacent echelon, and obtaining a plurality of feature vector sequences of the target fault type according to the feature vector pairs, wherein two adjacent elements in each feature vector sequence are feature vector pairs;
each feature vector corresponds to a feature value, each element in the feature vector sequence is replaced by the corresponding feature value, and the obtained sequence is recorded as a plurality of feature chains of the target fault type; a number of feature chains for each fault type are acquired.
5. The method for detecting the power battery cascade utilization fault according to claim 4, wherein the step of obtaining a plurality of feature vector pairs of adjacent cascades comprises the following specific steps:
taking any fault type as a target fault type, acquiring all echelon related data of a plurality of batteries corresponding to the target fault type, and arranging the related data of a plurality of batteries in each echelon according to the same battery sequence to obtain a related data sequence of each echelon;
constructing an initial matrix of each echelon according to the number of elements for the related data sequence of each echelon, and obtaining a plurality of feature vectors and corresponding feature values for each initial matrix through SVD decomposition;
performing KM matching on the feature vectors in the two initial matrixes of the adjacent echelons, and taking the feature vectors as bipartite graph nodes and cosine similarity of the feature vectors as side values to obtain a plurality of feature vector pairs of the adjacent echelons;
and acquiring a plurality of feature vector pairs of adjacent echelons of each fault type.
6. The method for detecting the power battery cascade utilization faults according to claim 1, wherein the specific obtaining method is as follows:
taking any fault type as a reference fault type, and obtaining a matching feature value set of each enhanced feature vector in the reference fault type according to feature vector matching results of each enhanced feature vector in the reference fault type in other fault types;
for a matching characteristic value set of any enhancement characteristic vector in the reference fault type, obtaining a maximum value and a next-maximum value in the matching characteristic value set, and taking the ratio of a difference value obtained by subtracting the next-maximum value from the maximum value to the maximum value as the larger of the enhancement characteristic vector; acquiring the maximization of each enhanced feature vector in the reference fault type, and arranging all the maximization in a descending order to obtain a maximization sequence of the reference fault type, wherein the enhanced feature vector corresponding to the maximum element in the maximization sequence is used as the fault vector of the reference fault type, and the second element in the maximization sequence is used as the fault threshold of the reference fault type;
and obtaining fault vectors and fault thresholds of each fault type.
7. The method for detecting a power battery cascade utilization fault according to claim 6, wherein the obtaining the matched feature value set of each enhanced feature vector in the reference fault type comprises the following specific steps:
performing KM matching on each enhanced feature vector in the reference fault type and all feature vectors in other fault types to obtain a plurality of matched feature vectors of each enhanced feature vector in the reference fault type;
for any one enhancement feature vector in the reference fault type, a plurality of matching feature vectors of the enhancement feature vector are obtained, and a set formed by feature values corresponding to all the matching feature vectors is recorded as a matching feature value set of the enhancement feature vector; and acquiring a matching characteristic value set of each enhanced characteristic vector in the reference fault type.
8. The method for detecting power battery cascade utilization faults as claimed in claim 7, wherein the obtaining of the plurality of matching feature vectors of each enhancement feature vector in the reference fault type comprises the following specific steps:
taking each enhanced feature vector in the reference fault type as a node on the left side of the bipartite graph, taking all feature vectors in any other fault type as nodes on the right side of the bipartite graph, acquiring edge values between the nodes through the enhanced feature vectors corresponding to the nodes on the two sides and the DTW distance of the feature vectors, and acquiring feature vectors matched with each enhanced feature vector in the reference fault type in other fault types through KM matching;
and obtaining the feature vector matched with each enhanced feature vector in the reference fault type in each other fault type, and recording the feature vector as a plurality of matched feature vectors of each enhanced feature vector.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311442766.0A CN117171588A (en) | 2023-11-02 | 2023-11-02 | Method for detecting gradient utilization faults of power battery |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311442766.0A CN117171588A (en) | 2023-11-02 | 2023-11-02 | Method for detecting gradient utilization faults of power battery |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117171588A true CN117171588A (en) | 2023-12-05 |
Family
ID=88932028
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311442766.0A Pending CN117171588A (en) | 2023-11-02 | 2023-11-02 | Method for detecting gradient utilization faults of power battery |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117171588A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117825973A (en) * | 2024-03-01 | 2024-04-05 | 深圳市健网科技有限公司 | Lithium battery state estimation method and system for distributed energy storage system |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111679158A (en) * | 2020-08-04 | 2020-09-18 | 国网内蒙古东部电力有限公司呼伦贝尔供电公司 | Power distribution network fault identification method based on synchronous measurement data similarity |
CN112287980A (en) * | 2020-10-14 | 2021-01-29 | 北方工业大学 | Power battery screening method based on typical feature vector |
CN113311344A (en) * | 2021-04-06 | 2021-08-27 | 国网江苏省电力有限公司南京供电分公司 | Thermal fault diagnosis method and system suitable for lithium ion battery with gradient utilization |
CN113537525A (en) * | 2021-07-23 | 2021-10-22 | 四川大学 | Self-adaptive early warning method for fault state of battery energy storage system |
CN113922438A (en) * | 2021-08-23 | 2022-01-11 | 国网辽宁省电力有限公司营口供电公司 | High-power active equalization battery management system |
CN114629195A (en) * | 2022-01-17 | 2022-06-14 | 国网吉林省电力有限公司经济技术研究院 | Partial power flexible grouping system for gradient battery utilization |
CN116089777A (en) * | 2023-04-10 | 2023-05-09 | 北京智蚁杨帆科技有限公司 | Intelligent new energy settlement method and system based on intelligent information matching |
CN116148656A (en) * | 2023-04-20 | 2023-05-23 | 国网山东省电力公司莱州市供电公司 | Portable analog breaker fault detection method |
-
2023
- 2023-11-02 CN CN202311442766.0A patent/CN117171588A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111679158A (en) * | 2020-08-04 | 2020-09-18 | 国网内蒙古东部电力有限公司呼伦贝尔供电公司 | Power distribution network fault identification method based on synchronous measurement data similarity |
CN112287980A (en) * | 2020-10-14 | 2021-01-29 | 北方工业大学 | Power battery screening method based on typical feature vector |
CN113311344A (en) * | 2021-04-06 | 2021-08-27 | 国网江苏省电力有限公司南京供电分公司 | Thermal fault diagnosis method and system suitable for lithium ion battery with gradient utilization |
CN113537525A (en) * | 2021-07-23 | 2021-10-22 | 四川大学 | Self-adaptive early warning method for fault state of battery energy storage system |
CN113922438A (en) * | 2021-08-23 | 2022-01-11 | 国网辽宁省电力有限公司营口供电公司 | High-power active equalization battery management system |
CN114629195A (en) * | 2022-01-17 | 2022-06-14 | 国网吉林省电力有限公司经济技术研究院 | Partial power flexible grouping system for gradient battery utilization |
CN116089777A (en) * | 2023-04-10 | 2023-05-09 | 北京智蚁杨帆科技有限公司 | Intelligent new energy settlement method and system based on intelligent information matching |
CN116148656A (en) * | 2023-04-20 | 2023-05-23 | 国网山东省电力公司莱州市供电公司 | Portable analog breaker fault detection method |
Non-Patent Citations (10)
Title |
---|
刘璐;李相俊;贾学翠;惠东;: "梯次利用储能电池管理技术与试验应用", 电力建设, no. 01 * |
孙冬: "锂离子电池梯次利用关键技术研究", 《中国博士学位论文全文数据库工程科技Ⅱ辑》, pages 042 - 20 * |
张扬等: "锂离子电池故障诊断算法研究综述", 《重庆理工大学学报(自然科学)》, pages 1 - 13 * |
徐佳宁: "基于特征提取的退役电池健康状态快速评价方法研究", 《中国博士学位论文全文数据库工程科技Ⅱ辑》, pages 042 - 188 * |
徐寿臣;王春玲;赵泽昆;韩晓娟;: "基于GA-BP神经网络的电池储能***软故障模糊综合评价", 电器与能效管理技术, no. 13 * |
徐敏: "储能电池检测诊断及能量管理策略的研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》, pages 035 - 315 * |
徐艳民: "电动汽车退役锂离子动力电池故障诊断及梯次利用关键技术研究", 《中国博士学位论文全文数据库工程科技Ⅰ辑》, pages 027 - 87 * |
王瑶;赵泽昆;: "基于概率神经网络的UPS储能电池软故障诊断", 电器与能效管理技术, no. 06 * |
薛撬: "动力锂离子电池剩余寿命预测与故障诊断研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》, pages 035 - 1011 * |
车云弘: "基于数据驱动的锂离子电池健康状态估计及寿命预测研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》, pages 035 - 160 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117825973A (en) * | 2024-03-01 | 2024-04-05 | 深圳市健网科技有限公司 | Lithium battery state estimation method and system for distributed energy storage system |
CN117825973B (en) * | 2024-03-01 | 2024-05-17 | 深圳市健网科技有限公司 | Lithium battery state estimation method and system for distributed energy storage system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113887601B (en) | Retired power battery recombination method based on cluster sorting | |
CN112630662B (en) | Power battery SOH estimation method based on data driving and multi-parameter fusion | |
CN113109715B (en) | Battery health condition prediction method based on feature selection and support vector regression | |
CN113009349A (en) | Lithium ion battery health state diagnosis method based on deep learning model | |
CN109800446B (en) | Method and device for estimating voltage inconsistency in discharging process of lithium ion battery | |
CN117171588A (en) | Method for detecting gradient utilization faults of power battery | |
CN112834927A (en) | Lithium battery residual life prediction method, system, device and medium | |
CN116298934B (en) | Modeling method of prediction network for lithium battery health state estimation | |
CN112287980B (en) | Power battery screening method based on typical feature vector | |
CN111983474A (en) | Lithium ion battery life prediction method and system based on capacity decline model | |
CN114726045B (en) | Lithium battery SOH estimation method based on IPEA-LSTM model | |
CN115166563A (en) | Power battery aging state evaluation and decommissioning screening method and system | |
CN115980588A (en) | Lithium ion battery health state estimation method based on self-encoder extraction features | |
CN112686380A (en) | Neural network-based echelon power cell consistency evaluation method and system | |
CN115389946A (en) | Lithium battery health state estimation method based on isobaric rise energy and improved GRU | |
CN114280490B (en) | Lithium ion battery state of charge estimation method and system | |
CN115186579A (en) | Lithium battery health state estimation method based on capacity increment variable quantity curve | |
CN113805060B (en) | Lithium battery residual life detection method based on relevance vector regression | |
CN113791351A (en) | Lithium battery life prediction method based on transfer learning and difference probability distribution | |
CN117110884A (en) | Lithium battery remaining service life prediction method based on multi-core correlation vector machine | |
CN116736172A (en) | Lithium battery pack health state evaluation method based on generation type countermeasure network | |
CN116774045A (en) | Lithium battery health state prediction method based on HHO-SVR | |
CN114895206B (en) | Lithium ion battery SOH estimation method based on RBF neural network of improved gray wolf optimization algorithm | |
CN115481796A (en) | Method for predicting remaining service life of battery based on Bayesian hybrid neural network | |
CN115018133A (en) | Lithium battery capacity prediction method based on chaos theory and nonlinear system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |