Lithium iron phosphate battery fault diagnosis algorithm

Abstract

The invention discloses a lithium iron phosphate battery fault diagnosis algorithm. The algorithm comprises the following steps: collecting an information source data set of lithium iron phosphate battery faults; a genetic algorithm is adopted to search the information source data set, a search result serves as an initial clustering center to be input into a fuzzy C-means clustering algorithm for means clustering, and the clustering number and distance measurement are determined; the expert weight is calculated by using the distance measurement of the trapezoidal fuzzy number, and the attribute weight is calculated by using a deviation maximization method; calculating positive and negative ideal solutions based on the expert weight; and based on the attribute weights and the positive and negative ideal solutions, calculating group benefits, personal regret values and compromise compromise solutions to obtain a fuzzy fault sequence. According to the method, the genetic algorithm is combined with the FCM, the FCM is prevented from falling into a local minimum value in the iteration process, and finally the search result of the genetic algorithm is used as the initial clustering center of the FCM, so that the fault recognition rate of the algorithm is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of new energy power batteries, and more particularly to a lithium iron phosphate battery fault diagnosis algorithm. BACKGROUND

[0002] Lithium iron phosphate batteries are widely used in the field of new energy vehicles due to their high safety performance, good high-temperature performance, large capacity, light weight, and environmental friendliness. However, due to product quality and the material limitations of lithium iron phosphate itself, the fire cause accounts for as high as 61%, directly causing double losses of personal and property. Through a large amount of data on battery online analysis for fault identification, the occurrence of lithium iron phosphate battery failure rate can be effectively reduced. However, during data uploading, problems such as uploading frequency and loss may occur, thereby increasing the difficulty of battery fault judgment.

[0003] Therefore, how to improve the identification of battery faults has attracted widespread attention. For example, Huang Zhaojun et al. (Huang Zhaojun, Su Jianhui, Xie Bao, et al. PEMFC fault diagnosis method based on fuzzy C-means clustering and probabilistic neural network [J]. Solar Energy, 2024, 45(01): 475-483. DOI:10.19912 / j.0254-0096.tynxb.2022-1480.) use the fuzzy C-means and probabilistic neural network model to improve the fault identification rate, and the research results prove that the algorithm model effectively eliminates singular data and improves the accuracy of model classification. Qin Huan et al. (Qin Huan, Huang Bixiong, Yan Xiao, et al. Research on micro-short circuit diagnosis method of lithium iron phosphate battery pack [J]. Energy Storage Science and Technology, 2021, 10(02): 664-670. DOI:10.19799 / j.cnki.2095-4239.2020.0349.) diagnose the micro-short circuit problem of the battery pack according to the relative charging time change of the single body. According to the box plot, the most frequently repeated one is the short-circuit single body. Wu Chao (Wu Chao. Research on typical fault mechanism and diagnosis method of lithium iron phosphate power battery [D]. Harbin Institute of Technology, 2019. DOI:10.27061 / d.cnki.ghgdu.2019.005001.) uses the relative attenuation rate of internal resistance capacity as a feature to evaluate the battery attenuation rate, and discusses the internal mechanism of the battery with fast attenuation rate but not obvious fault by means of differential analysis, and uses the incremental capacity curve to extract and diagnose short-term decay faults, and realizes online fault detection according to the relationship between the incremental capacity curve and the state of charge-open circuit voltage curve.

[0004] In summary, the current lithium iron phosphate battery fault detection has the following problems: (1) the selection of algorithm parameters is based on experience, the algorithm itself is not suitable for processing multi-classification problems, and there is great limitation; (2) there is a problem that a large number of temporary sample data and algorithm quantization short circuit cannot be obtained; (3) a single algorithm model cannot quantitatively evaluate the battery fault, and has no universality; (4) large-scale data collection is often difficult, a large number of prior experiments are required, and the requirement for hardware resources is high; therefore, under the condition of sparse data, the improvement of the failure rate is the challenge and opportunity faced by the future lithium iron phosphate battery. SUMMARY

[0005] Therefore, the present application provides a sparse data lithium iron phosphate battery method research, which is suitable for sparse data lithium iron phosphate battery fault diagnosis method, and the maximum feature is that the genetic algorithm is combined with fuzzy C means (FCM), the local minimum value in the iterative process of FCM is avoided, and finally the search result of the genetic algorithm is used as the initial clustering center of FCM, so that the identification rate of the algorithm to the fault is improved.

[0006] In the first aspect, the present application provides a lithium iron phosphate battery fault diagnosis algorithm based on fuzzy C means clustering, which comprises:

[0007] Collecting information source data set of lithium iron phosphate battery fault in a certain period;

[0008] Using genetic algorithm to search the information source data set, using the search result as the initial clustering center input into the fuzzy C means clustering algorithm for mean clustering, determining the clustering number and distance measure;

[0009] Using the distance measure of trapezoidal fuzzy number to calculate the expert weight, and using the maximum deviation method to calculate the attribute weight;

[0010] Calculating positive and negative ideal solutions based on the expert weight;

[0011] Based on the attribute weight and the positive and negative ideal solutions, the group benefit, the individual regret value and the compromise solution are calculated, and the fuzzy fault sorting is obtained.

[0012] Optionally, the information source data set is a set X=(x1, x2, …, xm) of m alternative schemes. m );

[0013] Each scheme corresponds to an attribute set C=(c1, c2, …, cn). n );

[0014] Wherein, m and n are positive integers greater than 1.

[0015] Optionally, the searching of the information source dataset by the genetic algorithm includes:

[0016] The information source dataset is input into the genetic algorithm GA for searching, and the searching result is input as an initial clustering center into a fuzzy C-means clustering FCM algorithm for mean clustering. c m max

[0017] Optionally, the searching of the information source dataset by the genetic algorithm includes:

[0018] The original data of the information source dataset is normalized, given parameters are initialized, initial chromosomes are randomly selected from the information source dataset and the best chromosome is updated, t chromosomes are selected and put into a selection pool by roulette selection, and the selection probability of chromosome i (1≤i≤t) is P i , wherein,

[0019] f i and f j are fitness values of chromosomes i and j, respectively.

[0020] Two chromosomes are randomly selected as parents and the crossover probability is checked to verify whether the crossover step is performed, and the probability of generating a new chromosome in the crossover process is defined as wherein and are offspring genes after the crossover, X g and Y g are parent genes, r is a random number between 0 and 1, and the crossover probability is set to be between 0.85 and 0.95.

[0021] The mutation rate P m is checked to determine whether to continue the mutation step, and the encoding genetic mutation probability is a g = 2 -uk , wherein s g and u are random numbers, s g ∈{-1,1} and u∈{0,1}, r g is a mutation range, r g ∈[10 -6 ,0.1], and k∈{4,5,…,20} is a mutation accuracy.

[0022] ​​​Optionally, the step of using the search results as initial cluster centers as input to the fuzzy C-means clustering algorithm for mean clustering includes:

[0023] The search results of the genetic algorithm (GA) are used as the initial cluster centers in the fuzzy C-means clustering (FCM) algorithm. The objective function J of the new fault diagnosis method is... FCM The definitions and constraints are as follows:

[0024] d ik =||u k -v i || 2 ,

[0025] Where m is the total number of data vectors in the given dataset, with the value of m corresponding to the fault type, C is the number of clusters, M is the fuzzy filter, M∈[1.25,2.5], u ik Let u be the membership degree of the k-th sample data belonging to the i-th cluster center, and satisfy the condition that the sum of the membership degrees of a single sample data belonging to all classes is 1. ik Construct a membership matrix U, u k This is the kth fault sample data, v i It is the i-th cluster center, d ik (u k ,v i () is a distance metric, which is a data object u k and cluster v i The distance between two points is measured using Euclidean metric, which is one of the commonly used methods.

[0026] Optionally, the step of calculating expert weights using the distance metric of trapezoidal fuzzy numbers includes: determining the group expert decision matrix and the positive ideal decision matrix and the negative ideal decision matrix, calculating the distance between the individual expert decision matrix and the positive ideal decision matrix and the negative ideal decision matrix, and obtaining the expert weights.

[0027] Optionally, the calculation of expert weights using the distance metric of trapezoidal fuzzy numbers is used to construct an expert model λ using the distance metric. p , include:

[0028] Expert E P Fuzzy decision matrix X P =(x P ij ) m×n Transform into a standardized matrix R P =(r P ij ) m×n r (p) ij=([r (p) 1i (I j ),r (p) 2i (I j ),r (p) 3i (I j ),r (p) 4i (I j )];[u (p) ij ,v (p) ij ]),in,

[0029] Benefit-oriented attribute

[0030] Cost-type attribute indicators are

[0031] Optionally, determining the group expert decision matrix, the positive ideal decision matrix, and the negative ideal decision matrix includes:

[0032] The group expert decision matrix is ​​as follows: s = 1, 2, ..., m, j = 1, 2, ..., n, where s represents the number of experts;

[0033] The positive ideal decision matrix is:

[0034] The negative ideal decision matrix is:

[0035] Optionally, calculating the distance between the individual expert decision matrix and the positive ideal decision matrix and the negative ideal decision matrix includes:

[0036] The distance between the individual expert decision matrix and the average decision matrix is ​​calculated as follows:

[0037]

[0038] Calculate the distance D′ between the individual expert decision matrix and the positive ideal decision matrix and the negative ideal decision matrix. p ,D″ p ;in,

[0039] The distance between the expert individual decision matrix and the positive ideal decision matrix is:

[0040]

[0041] The distance between the expert individual decision matrix and the negative ideal decision matrix is:

[0042]

[0043] The expert weight is obtained, and the expert weight comprises:

[0044] The expert weight is obtained according to the distance between the decision matrices:

[0045] Wherein

[0046] Optionally, the positive and negative ideals are calculated based on the expert weight, and the positive and negative ideals comprise:

[0047] According to the expert weight, the s-th expert P s The attribute weight vector given is The attribute weight vector q = {q1, q2,..., q m} T is

[0048] The positive and negative ideal solutions are solved;

[0049] The positive ideal solution A' = {a'1, a'2,..., a' n} is calculated under the attribute weight vector q = {q1, q2,..., q m} T The membership interval of the maximum fuzzy number is [1, 1], and the non-membership interval is [0, 0];

[0050] The negative ideal solution A'' = {a''1, a''2,..., a'' n} is calculated under the attribute weight vector q = {q1, q2,..., q m} T The membership interval of the maximum fuzzy number is [0, 0], and the non-membership interval is [1, 1];

[0051] Wherein,

[0052]

[0053] In the application, a lithium iron phosphate battery fault detection algorithm in a sparse data state is provided. The algorithm collects feature data as an initial clustering center of fuzzy C-means through a genetic algorithm, avoids falling into a local optimal solution during fuzzy C-means clustering, considers the lack of experience of experts in an industry in a certain aspect and information uncertainty, converts a language variable into a trapezoidal fuzzy number, simultaneously calculates attribute weights, simultaneously brings the attribute weights into a TOPSIS model, converts the attribute weights into standardized data, calculates optimal and worst solutions, calculates the similarity of each decision item to the optimal and worst solutions, thereby obtains a comprehensive score of each decision option, and determines a final fault option.

[0054] The application is directed to a fault detection algorithm for sparse data in a lithium iron phosphate battery, which is obtained by using the above principle.

[0055] Compared with the prior art, the lithium iron phosphate battery fault diagnosis algorithm based on fuzzy C-means clustering provided by the application at least achieves the following beneficial effects:

[0056] Due to the strong universality of the algorithm, the language variables of industry experts can be converted into a fuzzy matrix, and a trapezoidal fuzzy number algorithm is introduced to improve the problems of data sparsity and fuzziness. Generally, the experience of industry experts is fuzzy, and there is no specific standard in general, so the trapezoidal fuzzy number is introduced to fuzz the experience of industry experts to calculate the similarity, so that the accuracy of predicting the failure rate is improved.

[0057] Compared with the prior art, the application can well predict the occurrence of the failure rate of the lithium iron phosphate battery even under sparse data, and simultaneously perform fuzzy sorting on the failure rate reasons. In addition, the method has the advantages of good universality, simplicity, reliability and the like, and is expected to realize industrial production.

[0058] Of course, any product implementing the application does not necessarily need to achieve all the technical effects described above.

[0059] Other features of the application and its advantages will become apparent from the following detailed description of exemplary embodiments thereof, taken in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0060] The accompanying drawings incorporated in and forming a part of the specification, illustrate embodiments of the application and, together with the description, serve to explain the principles of the application.

[0061] Figure 1 is a flowchart of the lithium iron phosphate battery fault diagnosis algorithm based on fuzzy C-means clustering of the application;

[0062] Figure 2 is a fuzzy number graph in step S300 in the lithium iron phosphate battery fault diagnosis algorithm based on fuzzy C-means clustering of the application;

[0063] Figure 3 is a trapezoidal fuzzy number graph in step S300 in the lithium iron phosphate battery fault diagnosis algorithm based on fuzzy C-means clustering of the application;

[0064] Figure 4 is a genetic algorithm program block diagram in the lithium iron phosphate battery fault diagnosis algorithm based on fuzzy C-means clustering of the application;

[0065] Figure 5 is a program block diagram of the lithium iron phosphate battery fault diagnosis algorithm based on fuzzy C-means clustering of the application. Detailed Implementation

[0066] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention.

[0067] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use.

[0068] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0069] In all the examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0070] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0071] This invention provides a fault diagnosis algorithm for lithium iron phosphate batteries based on fuzzy C-means clustering, referring to... Figure 1 , Figure 5 , Figure 1 This is a flowchart illustrating a fault diagnosis algorithm for lithium iron phosphate batteries based on fuzzy C-means clustering, as described in this invention. Figure 5 This is a flowchart of the fault diagnosis algorithm for lithium iron phosphate batteries based on fuzzy C-means clustering, as described in this invention. Figure 1 As shown, the present invention provides a fault diagnosis algorithm for lithium iron phosphate batteries based on fuzzy C-means clustering, comprising the following steps (steps S100 to S500):

[0072] S100, Collect data source information on lithium iron phosphate battery faults;

[0073] S200. Use a genetic algorithm to search the information source dataset, and use the search results as initial cluster centers to input the fuzzy C-means clustering algorithm for mean clustering to determine the number of clusters and distance measure.

[0074] S300. Calculate expert weights using the distance metric of trapezoidal fuzzy numbers, and calculate attribute weights using the deviation maximization method.

[0075] S400. Calculate the positive and negative ideal solutions based on the expert weights;

[0076] S500, based on the attribute weight and the positive and negative ideal solution, calculating the group benefit, the individual regret compromise solution, obtaining the fuzzy fault ranking.

[0077] In step S100, specifically refers to collecting fault information in a certain period, forming an information source data set, the information source data set considers X=(x1, x2, …, x m ) is a set of m alternative schemes, C=(c1, c2, …, c n ) is a set of n attributes corresponding to each scheme, wherein m and n are positive integers greater than 1.

[0078] In step S200, the genetic algorithm is used to search the information source data set, and the result of searching the feature data set by the genetic algorithm (Genetic Algorithms, GA) is input into the fuzzy C-means clustering FCM algorithm as the initial clustering center, and the mean clustering is performed to determine the number of clusters and eliminate singular data with insufficient membership.

[0079] In some optional embodiments, the information source data set in step 100 is input into the genetic algorithm GA for searching, and the search result of the GA is input into the fuzzy C-means clustering FCM algorithm as the initial clustering center, and the mean clustering is performed, and the population size M, the membership factor m, the number of clustering clusters c, the crossover rate P c , the mutation rate P m , the convergence precision epsilon and the maximum iteration number t max are given.

[0080] Referring to Figure 4 , Figure 4 is the genetic algorithm program block diagram in the lithium iron phosphate battery fault diagnosis algorithm based on fuzzy C-means clustering of the present application, as shown in Figure 4 , the genetic algorithm searches the information source data set, randomly selects an initial population, defines a fitness function, performs fitness calculation, selects individuals from the population according to fitness, selects individuals with high fitness by roulette, generates new solutions by gene recombination, randomly changes individual genes slightly to avoid the algorithm falling into local optimum, increases population diversity, repeatedly selects, crosses and mutates until the optimal population is found and the termination condition is met.

[0081] In step S300, the industry expert weight and the attribute weight are considered. Due to lack of experience and information uncertainty, industry experts cannot accurately calculate the alternative schemes without considering various attributes. The distance measure of trapezoidal fuzzy numbers is used to calculate the expert weight, and the language variable is converted into trapezoidal fuzzy numbers. At the same time, the attribute weight is calculated by the maximum deviation method to obtain the attribute characteristics.

[0082] Specifically, referring to Figure 5 ,Figure 5 is a program block diagram of the lithium iron phosphate battery fault diagnosis algorithm based on fuzzy C-means clustering of the present application, as shown in Figure 5 The information source data set of the lithium iron phosphate battery fault in a period of time is collected; the collected information source data set is processed and input into the genetic algorithm for searching; the search result is input into the fuzzy C-means clustering algorithm as the initial clustering center for mean clustering, the similarity is calculated, the number of clusters and the distance measure are determined; the distance measure of the trapezoidal fuzzy number is used to calculate the expert weight to obtain the expert feature; the attribute weight is calculated by the maximum deviation method to obtain the attribute feature; the group expert decision matrix, the positive ideal solution matrix and the negative ideal solution matrix are calculated based on the expert weight, and then the positive and negative ideal solutions are calculated; the group benefit, the personal regret compromise solution and the fuzzy fault ranking are obtained based on the attribute weight and the positive and negative ideal solutions.

[0083] In actual application, the above embodiments and preferred embodiments can be fused according to actual application requirements. In order to facilitate those skilled in the art to understand how to realize the actual application and how to combine the preferred embodiments, the specific algorithm of the specific steps is described in detail below.

[0084] In step S200, the genetic algorithm is used to search the information source data set, the search result is input into the fuzzy C-means clustering algorithm as the initial clustering center for mean clustering, the number of clusters and the distance measure are determined, and the specific steps and algorithms include the following steps and algorithms:

[0085] Figure 4 The genetic algorithm program block diagram in the lithium iron phosphate battery fault diagnosis algorithm based on fuzzy C-means clustering of the present application is shown in Figure 4 The original data is preprocessed, normalized, and initialized for the given parameters, and the data of 0 and the like is discarded, the initial chromosome is randomly selected from the information source data set and the best chromosome is updated, t chromosomes are selected, and the roulette selection is used to put them into the selection pool P i is the selection probability of chromosome i (1≤i≤t), f i and f j are the fitness values of i and j, respectively.

[0086] In order to eliminate contingency and improve the accuracy of the algorithm, two chromosomes are randomly selected as parents and the crossover probability is checked to verify whether the crossover step is performed, and the probability of generating new chromosomes in the crossover process is defined as wherein and are the offspring genes after the crossover, X g and Y g are the parent genes, and r is a random number between 0 and 1, and the crossover probability is set to [0.85, 0.95].

[0087] Verify mutation P m Probability to determine whether to continue the mutation step, the implementation of the genetic mutation probability is a g = 2 -uk , where s g and u are random numbers, and s g ∈{-1,1} and u∈{0,1}, r g is a mutation range, r g ∈[10 -6 ,0.1], k∈{4,5,…,20} is the precision of mutation.

[0088] The search result of GA, that is, the output optimal population, is used as the initial clustering center of FCM, and the objective function J FCM of the new fault diagnosis method is defined and its constraint condition is:

[0089] d ik =||u k -v i || 2 ,

[0090] Wherein, m is the total number of data vectors in the given data set, which corresponds to the fault category, C is the number of clusters, M is the fuzzy device, generally M∈[1.25,2.5], u ik is the membership degree of the kth sample data to the i th clustering center, and satisfies that the sum of the membership degrees of a single sample data to all classes is 1, and u ik of each sample data constitutes a membership matrix U, u k is the kth fault sample data, v i is the i th clustering center, d ik (u k ,v i ) is the distance value between data objects u k and cluster v i , that is, the distance measure, which adopts the Euclidean measure.

[0091] Step S300 utilizes the distance measure of trapezoidal fuzzy numbers to calculate the expert weight, which needs to convert the language variables such as fuzzy evaluation (such as "high", "medium", "low") of experts into trapezoidal fuzzy numbers, and then measure the consistency of expert opinions through the distance measure, and further determine the weight.

[0092] Specifically, it refers to calculating the group expert decision matrix and positive and negative ideal decision matrix, determining the distance between the individual expert decision matrix and the positive and negative ideal decision matrix, and thus giving the expert weight;

[0093] Specifically comprising: constructing an expert weight model λ ik (u k ,v i ) by using the distance measure d p in step S200

[0094]

[0095] Referring to Figure 2 、 Figure 3 , Figure 2 is a fuzzy number graph in step S300 in the lithium iron phosphate battery fault diagnosis algorithm based on fuzzy C-means clustering of the application; Figure 3 is a trapezoidal fuzzy number graph in step S300 in the lithium iron phosphate battery fault diagnosis algorithm based on fuzzy C-means clustering of the application, as shown in Figure 2 、 Figure 3 , the fuzzy decision matrix X P of the expert E P = (x P ij ) m×n is converted into a standardized matrix R P = (r P ij ) m×n , r (p) ij = ([r (p) 1i (I j ), r (p) 2i (I j ), r (p) 3i (I j ), r (p) 4i (I j )];[u (p) ij ,v (p) ij ])

[0096] The benefit type attribute is:

[0097]

[0098] The cost type attribute index is:

[0099]

[0100] Determine the group expert decision matrix, positive ideal decision matrix, and negative ideal decision matrix, wherein

[0101] The group expert decision matrix is:

[0102] s represents the number of experts;

[0103] The positive ideal decision matrix is:

[0104]

[0105] The negative ideal decision matrix is:

[0106]

[0107] The distance between the individual decision matrix of an expert and the average decision matrix is calculated:

[0108]

[0109] wherein, and r′ ij , r″ ij have a subordination relationship;

[0110] The distance D′ p between the individual decision matrix of an expert and the positive and negative ideal decision matrices is calculated: p ′,

[0111] The distance between the individual decision matrix of an expert and the positive ideal decision matrix is:

[0112]

[0113] The distance between the individual decision matrix of an expert and the negative ideal decision matrix is:

[0114]

[0115] The weight of an expert is given according to the distance between the decision matrices:

[0116] wherein

[0117] The positive and negative ideal solutions are calculated in step S400, which is a core step in the TOPSIS analysis method adopted by the present application. The weight of an expert is solved according to step S300, and the s-th expert P s is given an attribute weight vector The attribute weight vector q = {q1, q2, …, q m} T is calculated. The positive and negative ideal solutions are solved at the same time.

[0118] The positive ideal solution A′ = {a′1, a′2, …, a′ n} is calculated according to the attribute weight vector q = {q1, q2, …, q m} TThe membership interval of the lower part relative to the maximum fuzzy number is [1, 1], and the non-membership interval is [0, 0].

[0119] The negative ideal solution A''={a''1, a''2,..., a''n} is obtained. n} in the weight vector q={q1, q2,..., qn} of the attribute. m} T The membership interval of the lower part relative to the maximum fuzzy number is [0, 0], and the non-membership interval is [1, 1].

[0120] Wherein,

[0121]

[0122] It can be seen from the above examples that the lithium iron phosphate battery fault diagnosis algorithm based on fuzzy C-means clustering provided by the present application at least achieves the following beneficial effects:

[0123] The algorithm used in the present embodiment has strong universality, can convert language variables of industry experts into fuzzy matrices, and introduces a trapezoidal fuzzy number algorithm to improve the problems of data sparseness and fuzziness. The experience of industry experts is generally fuzzy, and generally there is no specific standard, so the introduction of trapezoidal fuzzy numbers can fuzz the experience of industry experts to calculate the similarity, so as to improve the accuracy of predicting the failure rate by using fuzzy scores.

[0124] Compared with the prior art, the present embodiment can well predict the occurrence of the failure rate of the lithium iron phosphate battery even under sparse data, and at the same time, can perform fuzzy sorting on the failure rate reasons.

[0125] Although some specific embodiments of the present application have been described in detail through examples, those skilled in the art should understand that the above examples are only for illustration, and are not intended to limit the scope of the present application. Those skilled in the art should understand that the above embodiments can be modified without departing from the scope and spirit of the present application. The scope of the present application is defined by the appended claims.

Claims

1. A fault diagnosis algorithm for lithium iron phosphate batteries, characterized in that, include: Collect a dataset of information sources related to lithium iron phosphate battery failures; A genetic algorithm is used to search the information source dataset. The search results are used as the initial cluster centers to perform mean clustering in a fuzzy C-means clustering algorithm to determine the number of clusters and the distance measure. Expert weights are calculated using the distance metric of trapezoidal fuzzy numbers, and attribute weights are calculated using the deviation maximization method. Calculate positive and negative ideal solutions based on the aforementioned expert weights; Based on the attribute weights and the positive and negative ideal solutions, the group benefits, individual regret values, and compromise solutions are calculated to obtain the fuzzy fault ranking.

2. The lithium iron phosphate battery fault diagnosis algorithm according to claim 1, characterized in that, The information source dataset is a set of m alternative solutions X = (x1, x2, ..., xm) m ); Each scheme corresponds to a set of n attributes C = (c1, c2, ..., c...). n ); Where m and n are positive integers greater than 1.

3. The lithium iron phosphate battery fault diagnosis algorithm according to claim 1, characterized in that, The process involves using a genetic algorithm to search the information source dataset, and then using the search results as initial cluster centers in a fuzzy C-means clustering algorithm for mean clustering. This includes: The information source dataset is input into the genetic algorithm for searching. The search results are used as initial cluster centers and then input into the fuzzy C-means clustering algorithm for mean clustering, given the population size M, membership factor m, number of clusters c, and crossover rate P. c Variation rate P m The parameters t for convergence accuracy ε and maximum number of iterations max .

4. The lithium iron phosphate battery fault diagnosis algorithm according to claim 3, characterized in that, The process of using a genetic algorithm to search the information source dataset includes: The raw data of the information source dataset is normalized, the given parameters are initialized, an initial chromosome is randomly selected from the information source dataset and the optimal chromosome is updated, chromosome t is selected, and it is placed into the selection pool using roulette wheel selection. The selection probability of chromosome i (1≤i≤t) is P. i ,in, f i and f j These are the fitness values ​​of chromosomes i and j; Two chromosomes are randomly selected as parents, and the crossover probability is checked to verify whether the crossover step should be performed. The probability of generating a new chromosome during the crossover process is defined as follows: in and It is the gene of the offspring after crossing over, X g and Y g The genes are parental genes, r is a random number between 0 and 1, and the crossover probability is set to between [0.85, 0.95]. Verify the stated mutation rate P m To determine whether to continue the mutation step, the probability of encoding genetic mutation is... a g =2 -uk , where s g And u are random numbers, and s g ∈{-1,1}, u∈{0,1}, r g For the range of variation, r g ∈[10 -6 [,0.1], k∈{4,5,…,20} represents the precision of the mutation.

5. The lithium iron phosphate battery fault diagnosis algorithm according to claim 3, characterized in that, The step of using the search results as initial cluster centers in the fuzzy C-means clustering algorithm for mean clustering includes: The search results of the genetic algorithm are used as the initial cluster centers in the fuzzy C-means clustering algorithm. The objective function J of the new fault diagnosis method is... FCM The definition and its constraints are as follows: Where m is the total number of data vectors in the given dataset, with the value of m corresponding to the fault type, C is the number of clusters, M is the fuzzy filter, M∈[1.25,2.5], u ik Let u be the membership degree of the k-th sample data belonging to the i-th cluster center, and satisfy the condition that the sum of the membership degrees of a single sample data belonging to all classes is 1. ik Construct a membership matrix U, u k This is the kth fault sample data, v i It is the i-th cluster center, d ik (u k ,v i () is a distance metric, which is a data object u k and cluster v i The distance between them.

6. The lithium iron phosphate battery fault diagnosis algorithm according to claim 1, characterized in that, The method of calculating expert weights using the distance metric of trapezoidal fuzzy numbers includes: determining the group expert decision matrix and the positive ideal decision matrix and the negative ideal decision matrix; calculating the distance between the individual expert decision matrix and the positive ideal decision matrix and the negative ideal decision matrix; and obtaining the expert weights.

7. The lithium iron phosphate battery fault diagnosis algorithm according to claim 6, characterized in that, The expert weights are calculated using the distance metric of trapezoidal fuzzy numbers, and the expert model λ is constructed using the distance metric. p , include: Expert E P Fuzzy decision matrix X P =(x P ij ) m×n Transform into a standardized matrix R P =(r P ij ) m×n r (p) ij =([r (p) 1i (I j ), r (p) 2i (I j ), r (p) 3i (I j ), r (p) 4i (I j )];[u (p) ij v (p) ij ]),in, Benefit-oriented attribute Cost-type attribute indicators are 8. The lithium iron phosphate battery fault diagnosis algorithm according to claim 6, characterized in that, The determination of the group expert decision matrix, the positive ideal decision matrix, and the negative ideal decision matrix includes: The group expert decision matrix is ​​as follows: s represents the number of experts; The positive ideal decision matrix is: The negative ideal decision matrix is:

9. The lithium iron phosphate battery fault diagnosis algorithm according to claim 8, characterized in that, The calculation of the distance between the individual expert decision matrix and the positive ideal decision matrix and the negative ideal decision matrix includes: The distance between the individual expert decision matrix and the average decision matrix is ​​calculated as follows: Calculate the distance D′ between the individual expert decision matrix and the positive ideal decision matrix and the negative ideal decision matrix. p , D′ p ';in, The distance between the individual expert decision matrix and the positive ideal decision matrix is: The distance between the individual expert decision matrix and the negative ideal decision matrix is: The expert weights are derived as follows: Expert weights are derived from the distances between decision matrices. in 10. The lithium iron phosphate battery fault diagnosis algorithm according to claim 1, characterized in that, The calculation of positive and negative ideal solutions based on the expert weights includes: Based on the expert weights, let the s-th expert P... s The given attribute weight vector is Then the attribute weight vector q = {q1, q2, ..., q m } T for Solve for the positive and negative ideal solutions; The ideal solution A′={a1′,a′2,…,a′ n In the attribute weight vector q = {q1, q2, ..., q...} m } T The membership interval relative to the maximum fuzzy number is [1,1], and the non-membership interval is [0,0]. Negative ideal solution A″={a1″,a′2′,…,a′ n In the attribute weight vector q = {q1, q2, ..., q} m } T The membership interval relative to the maximum fuzzy number is [0,0], and the non-membership interval is [1,1]. in,

Images