Energy storage battery screening integrated optimization method

Abstract

The application relates to the technical field of electrochemical energy storage battery screening integrated optimization, and provides an energy storage battery screening integrated optimization method. First, according to the design requirements of a battery energy storage system, the number of batteries in a battery string and the number of parallel connection of the battery string of the battery energy storage system are determined based on the performance parameters of selected battery monomers. Then, the feature vectors representing the performance of the battery samples are defined according to the test data of the battery samples and are normalized, a target function for battery screening is constructed through binary coding individuals, and a tree-type screening structure for iteratively optimizing the genetic binary classification battery sample set is established. Finally, the battery set in the leaf node is defined to form a potential battery string and a parallel potential solution under different battery strings, and the battery energy storage system integration is completed. The application is favorable for improving the overall performance level of the batteries in each group of strings and promoting the more extensive application of the battery energy storage system in a new power system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrochemical energy storage battery screening and integration optimization, and relates to an energy storage battery screening and integration optimization method based on genetic optimization clustering under a tree architecture, and in particular to a screening and series-parallel integration optimization process under consideration of the difference of energy storage batteries. BACKGROUND

[0002] Energy storage systems are key elements in new power systems, and battery energy storage is widely concerned and applied due to its good flexibility and high adaptability. In practical applications, a battery energy storage system is often composed of several battery monomers, and the overall performance of the battery energy storage system will be determined by the performance of the battery monomers and the integration method. The battery type, grouping method, integration design, and source network load control strategy of the power station are of great significance to the application of large-scale energy storage system power stations on the grid side. The wood barrel effect caused by the inconsistent performance of each battery monomer poses a hidden danger to the reliability and safety of daily operation and maintenance of the battery energy storage power station, and the battery energy storage screening and integration technology puts forward more stringent requirements, which has become a key to the application of battery energy storage systems.

[0003] At present, in the battery energy storage power station, the integration of actual energy storage batteries is relatively rough, and generally, the performance of batteries in the same batch is assumed to be similar, and after measuring part of the batteries by sampling, the batteries are directly connected in series and parallel according to the requirements of voltage, power and capacity of the energy storage power station, to form a battery energy storage system, and then to build a battery energy storage power station. This method lacks performance system estimation and optimization for the integrated battery energy storage system, and in practical applications, the battery energy management system and the corresponding circuit topology are needed to control the performance of the battery, and a large number of battery monomer electrical quantities need to be monitored to ensure the safety and efficiency of the battery application. In the existing information about battery energy storage screening and integration, the relevant research mainly focuses on how to establish a characterization quantity that can accurately reflect the performance of the battery according to the battery test experimental data, and how to apply statistical theory and artificial intelligence such as machine learning and deep learning to cluster the battery groups with similar performance, that is, the two aspects of feature engineering and clustering analysis of battery energy storage. According to the formed clustering cluster family and the requirements of voltage, power and capacity of the direct current bus of the battery energy storage power station, the clustered batteries are screened and assembled to form an energy storage system that meets the actual requirements for energy storage power stations.

[0004] The simple and rough screening method has adverse effects on the operation maintenance and safety monitoring of the battery energy storage in practical application, so it is crucial to promote the existing research on the clustering screening and integration method to practical application. The research on the battery performance characterization is mostly based on subjective judgment combined with the battery electrochemical process and mathematical methods based on deep learning, and it is difficult to form a unified evaluation method. In the research on the battery clustering screening method, the open-loop and simple clustering algorithm is mostly directly applied to form the cluster division of all batteries, which has the disadvantage that the clustering process lacks the guidance of the optimization direction, resulting in general clustering effect, and the number of battery individuals in the same battery cluster formed by clustering is difficult to control, which is not conducive to the integration of the battery box / pack in a specific composition. SUMMARY

[0005] The present application establishes a multi-level optimization clustering battery screening and integration optimization method under a tree structure in order to alleviate the performance decline problem of the overall battery energy storage system after the integration of batteries with inconsistent performance, and improve the application level of the battery energy storage system.

[0006] The battery screening and integration optimization method comprises the following steps:

[0007] Step one, for the battery energy storage system composed of a plurality of battery strings to be constructed, according to the design requirements of the system and the performance parameters of the selected battery monomers, the number of batteries m in each battery string and the number of parallel connection of battery strings n are determined, and the number of batteries required for the integration of the battery energy storage system is obtained.

[0008] The design requirement parameters of the battery energy storage system include the DC bus voltage U dc (V), the rated power P eb (W) and the rated capacity Q eb (W·h) of the battery energy storage system; wherein the upper limit and the lower limit of the DC bus voltage are and

[0009] The performance parameters of the selected battery monomers include the rated voltage U b (V) of the battery monomer and the nominal capacity C b (A·h) of the battery;

[0010] The number of batteries in each battery string in the battery energy storage system is:

[0011] The number of parallel connection of battery strings in the battery energy storage system is:

[0012] The total number of batteries required for the integration of the battery energy storage system is: m×n

[0013] Step two, according to the battery performance test requirements, obtain the test data of N (N » m x n) battery samples, and calculate the characteristic variables representing the battery performance based on the test data of the battery samples, construct the characteristic vector X representing the performance of each battery and normalize the processing;

[0014] The battery performance test includes nominal capacity test, different temperature charge and discharge test, mixed power pulse characteristic test and reliability test, etc.

[0015] The test data of each battery sample includes voltage data, current data, temperature data and direct current resistance data, etc.

[0016] The characteristic variables representing the performance of the battery include the actual discharge capacity, the battery platform voltage, the maximum temperature rise and the direct current resistance, etc., and constitute the S-dimensional characteristic vector X, wherein the S-dimensional characteristic vector of the i-th battery is represented as X i =[x i,1 ,x i,2 ,…,x i,S ];

[0017] The S-dimensional characteristic vector of N batteries is normalized, that is, the characteristic values of N batteries under a certain characteristic are normalized, and the s-dimensional characteristic x i,s of the i-th battery is normalized and represented as y i,s , and the calculation formula is as follows:

[0018]

[0019] Where max(·) and min(·) represent the functions of finding the maximum and minimum values, respectively.

[0020] Step three, based on the normalized characteristic vector representing each battery sample, construct the binary code of the same length to form the genetic individual, compose the genetic sample set, and optimize the battery sample by using the genetic algorithm to complete the tree type screening structure of the battery and obtain the sub node set.

[0021] Specifically:

[0022] First, define the genetic sample set composed of W battery monomer S-dimensional characteristic vectors, and W = N initially; At the same time, define the binary code length of the individual in the genetic algorithm as the number of battery monomers W, and the genetic algorithm parameters include the number of individuals K in the population, the selection rate P s , the crossover rate P c , the mutation rate P m and the maximum number of iterations G, etc.; In the genetic sample set, the binary code of the k-th genetic individual is represented as The i-th code bit is a k,i ∈{0,1},i=1,2,...,W;

[0023] Then, the objective function J of the binary coded individual is calculated for the W battery cell screening, and the metric of the W battery cell grouping is calculated by the binary coded individual.

[0024] The objective function calculation formula is as follows:

[0025]

[0026] wherein D In and D Ot respectively represent the intra-class distance and the inter-class distance, and the calculation formula is

[0027]

[0028] wherein Y i (j) represents the feature vector of the i-th sample in the j-th class (j=0 or 1, determined by the coding bit a k,i of the genetic algorithm individual), Y i (j)=[y i,1 (j),y i,2 (j),…,y i,S (j)] S represents the number of feature variables, and n j represents the number of battery samples in the j-th class; represents the mean vector of all samples in the j-th class, represents the mean vector of all samples.

[0029] Finally, the battery samples are optimized by the genetic algorithm using the objective function, and the battery screening clustering is performed, and the specific process is as follows:

[0030] Step 301, arrange the objective function value sequence under the K coded individuals of the g-th iteration in descending order, select the first values and the corresponding coded individuals

[0031] represents the down rounding symbol;

[0032] Step 302, randomly combine the selected coded individuals two by two, and perform the crossover operation of the genetic algorithm to form new coded strings

[0033] Step 303, perform the mutation operation of the genetic algorithm on each coding bit of each coded individual in the new coded string , and update the coded string​

[0034] Step 304, selecting the coding individual with the minimum target function value from the remaining coding individuals in step 301, combining the coding strings and

[0035] Step 305, calculating the target function values of the new K coding individuals, and obtaining a sequence of target function values selecting the coding individual corresponding to the minimum target function value in the sequence as the optimal individual, denoted as minα (g) ;

[0036] Step 306, judging whether the iteration number g at this time reaches the maximum iteration number G, if yes, executing step 307, if not, returning to step 301 for the next iteration.

[0037] Step 307, using the optimal individual to divide the battery sample set A in the parent node into two complementary and incompatible subsets B and C as child nodes, completing the tree-type screening structure of the battery;

[0038] The 2G child nodes after G iterations are taken as the child node set.

[0039] Step four, judging whether the number of battery samples in each child node in the child node set in the tree-type screening structure of the battery is greater than the number of batteries m in the battery string, if yes, defining the child node as a branch node and returning to step three as a new parent node to perform the binary splitting process of the set and generate a new two child node set, if not, defining the child node as a leaf node and no longer performing the binary splitting process, entering step five;

[0040] Step five, judging whether all child nodes have become leaf nodes, if yes, stopping the binary splitting process and entering step six, if not, returning the internal battery sample subset of the child node that is not a leaf node to step three for the binary splitting process of the set;

[0041] Step six, pruning the tree-type screening structure according to the two child node types generated by each parent node in the tree-type screening structure until the tree-type structure does not change;

[0042] The two child node types under the same parent node specifically include:

[0043] ​​​​​(1) If both of the two child nodes under the same parent node are leaf nodes, then respectively judge whether the battery sample quantity in the two leaf nodes is greater than the battery quantity m of the battery string; if at least one of the two leaf nodes has battery sample quantity greater than the battery string quantity m, then do not perform any operation, and keep the two leaf nodes; if the battery sample quantity in the two leaf nodes is less than the battery string quantity m, then remove the two leaf nodes, and change the parent node into a new leaf node;

[0044] (2) If both of the two child nodes under the same parent node are branch nodes or are a branch node and a leaf node respectively, then do not perform any operation;

[0045] Step seven, define the battery sample set in each leaf node after pruning as a battery cluster family, judge whether the battery sample in each battery cluster family is greater than the battery quantity m of the battery string, if yes, then the battery cluster family constitutes a potential battery string which can be used for battery energy storage system integration, and step eight is executed; if not, then abandon the battery cluster family.

[0046] Step eight, judge whether the quantity of all potential battery strings is greater than the parallel quantity n of the required battery string of the battery energy storage system; if yes, then step nine is executed; otherwise, return to step two to obtain a new batch of batteries to re-perform battery screening.

[0047] Step nine, calculate the center of the battery sample in each potential battery string in the feature space, and select the battery string composed of the n closest battery cluster families to perform parallel integration;

[0048] The center of the battery sample in the feature space is represented by the mean value of each feature vector.

[0049] Step ten, select the m closest battery samples to the internal center in the n battery cluster families to perform series integration;

[0050] Step eleven, complete the series-parallel integration of the battery energy storage system according to the battery samples selected in step nine and step ten.

[0051] Compared with the closest prior art, the excellent effects of the present application are:

[0052] The present application combines tree structure and genetic optimization clustering, effectively forms a number of subsets of battery samples with similar performance for constructing battery series structure, completes the battery integration optimization process, is beneficial to improving the overall performance level of each group of battery string, improves the consistency of the operation of the battery energy storage system after integration, and can promote the more widespread application of the battery energy storage system in the new power system. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 is a flowchart of a battery screening and integration optimization method of the present application;

[0054] Figure 2 is a schematic diagram of battery energy storage system integration;

[0055] Figure 3 is a schematic diagram of battery sample distribution in the embodiment of the application, taking a 2-dimensional feature vector as an example;

[0056] Figure 4 is a schematic diagram of a binary battery sample set of an individual of a certain genetic algorithm, taking a 2-dimensional feature vector as an example in the embodiment of the application;

[0057] Figure 5 is a flowchart of the process of optimizing battery sample clustering by using a genetic algorithm in the application;

[0058] Figure 6 is a schematic diagram of splitting a parent node battery sample set into two complementary and non-contain subsets by using an optimal individual of a genetic algorithm in the application;

[0059] Figure 7 is a schematic diagram of a tree structure of a battery screening process before pruning and definition of branch nodes and leaf nodes in the application;

[0060] Figure 8 is a schematic diagram of the process of grouping and stringing battery samples in each leaf node after pruning, taking a 2-dimensional feature vector as an example in the embodiment of the application. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application.

[0062] The application is based on a feature space composed of a plurality of battery characteristic representation quantities, takes the distance of a plurality of battery samples in the feature space as a measure, constructs a comprehensive objective function with the minimum distance of same cluster batteries and the maximum distance of different cluster batteries, and uses binary coding of a genetic algorithm to represent the plurality of battery samples, iteratively optimizes the best two cluster families, realizes effective clustering of energy storage batteries, and then uses a tree structure to continuously divide the battery sample subsets until the stop condition is met, such as the number of battery samples in a subset being less than the number of integrated series-connected batteries required by a battery energy storage system, completes the screening of each series-connected battery group, and integrates the battery energy storage system according to the subset condition of the leaf nodes of the tree structure.

[0063] A battery screening and integration optimization method, as shown in Figure 1 includes the following steps:

[0064] Step 1: For a battery energy storage system to be constructed composed of a plurality of batteries, according to the design requirements of the system and the performance parameters of the selected battery monomers and other conditions, the number of batteries m in each battery string and the number of parallel-connected battery strings n are determined, the number of batteries required for battery energy storage system integration is obtained, and a schematic diagram of battery energy storage system integration is as shown in Figure 2shown;

[0065] Step 1.1, according to the design requirements of the battery energy storage system, define its DC bus voltage as U dc (V), the rated power P eb (W) and the rated capacity Q eb (W·h) of the battery energy storage system, the upper and lower limits of the DC bus voltage are and

[0066] Step 1.2, determine the performance parameters of the selected battery monomer, define the rated voltage of the battery monomer as U b (V), the battery nominal capacity C b (A·h);

[0067] Step 1.3, based on the requirements of the battery energy storage system determined in steps 1.1 and 1.2 and the parameters of the selected battery monomer, calculate the battery data m in each battery string in the battery energy storage system.

[0068] Taking the upper and lower limits of the DC bus voltage of the battery energy storage system and the rated voltage of the battery monomer U b , the number of battery strings m is calculated, and the calculation formula is as follows:

[0069]

[0070] where, and respectively represent the upward rounding and downward rounding functions, represent a set of positive real numbers.

[0071] If the cost of the battery monomer is considered to be reduced or the reliability of the system caused by the battery series is reduced, as few batteries as possible can be selected for stringing, that is,

[0072] Step 1.4, based on the number of batteries in each battery string in the battery energy storage system determined in step 1.3, calculate the number of parallel connections n of the battery strings in the battery energy storage system.

[0073] Taking the rated capacity Q eb (W·h) of the battery energy storage system as an example, according to the battery monomer nominal capacity C b (A·h) and the rated capacity Q eb (W·h) of the battery energy storage system, the number of parallel connections n of the battery strings in the battery energy storage system is calculated, and the calculation formula is as follows:

[0074]

[0075] Similarly, if the cost of the battery monomer is considered and the battery capacity redundancy is not considered, the least number of battery strings in parallel can be selected, that is

[0076] Step 1.5, based on the number of batteries m in the battery string and the number of parallel battery strings n, the total number of batteries required for battery energy storage system integration is obtained: m x n.

[0077] Step 2: According to the battery performance test requirements, obtain the test data of N (N>>m x n) battery samples, calculate the characteristic variables representing the performance of each battery sample using the test data of each battery sample, construct the characteristic vector X describing the performance of each battery and normalize Y. Take 2-dimensional feature vector as an example, N battery samples are distributed in the feature space as shown in Figure 3 ;

[0078] Step 2.1, test the performance of N (N>>m x n) battery samples, such as nominal capacity test, different temperature charge and discharge test, mixed power pulse characteristic test and reliability test, etc. Obtain the test data of each battery sample, mainly including voltage data, current data, temperature data and direct current resistance data, etc.

[0079] Step 2.2, calculate the characteristics representing the performance of each battery according to the test data obtained in step 2.1, such as actual discharge capacity, battery platform voltage, maximum temperature rise and direct current resistance, etc., and form S-dimensional feature vector X, where the S-dimensional feature vector of the i-th battery is represented as X i =[x i,1 ,x i,2 ,…,x i,S ];

[0080] Step 2.3, normalize the S-dimensional feature vector of N batteries calculated in step 2.2, that is, normalize the feature values of N batteries under a certain feature, where the s-dimensional feature x i,s of the i-th battery after normalization is represented as y i,s , and the calculation formula is as follows:

[0081]

[0082] Where max(·) and min(·) represent the functions of finding the maximum and minimum values, respectively.

[0083] Step 3: Constructing equal-length binary codes from each battery sample characterized by normalized feature vectors to form a genetic individual, composing a genetic sample set, establishing an optimization objective function (such as taking the intra-class distance and the inter-class distance as the objective function), optimizing the battery sample through genetic algorithm, dividing the original battery sample set into two complementary and incompatible subsets, and continuously optimizing the split subsets until the stopping condition is met, forming several indivisible leaf node sets, and completing the tree-type screening structure of the battery;

[0084] As shown in Figure 5 , the clustering process of the genetic algorithm to optimize the battery sample is as follows:

[0085] Step 3.1, define a sample set composed of W battery monomer S-dimensional feature vectors, W == N in the first division, and define the binary code length of the individual in the genetic algorithm as the number of battery monomers W. Genetic algorithm parameters include the number of individuals K in the population, selection rate P s , crossover rate P c , mutation rate P m , and maximum iteration number G, etc. Let the iteration number g = 0, and design a genetic optimization algorithm objective function J for evaluating the clustering effect of the battery sample. The specific process is as follows:

[0086] Step 3.1.1, define K binary code individuals equal in length to the number of W battery monomer samples, where the kth code individual can be represented as α k = [a k,1 a k,2 … a k,W ], the ith code bit a k,i ∈{0,1}, i = 1, 2,..., W;

[0087] Step 3.1.2, calculate the objective function J of the binary code individual for the W battery monomer screening, and form the W battery monomer grouping metric calculation with this binary code individual.

[0088] Taking the minimum intra-class distance and the maximum inter-class clustering as an example, the objective function calculation formula is as follows:

[0089]

[0090] Where D In and D Ot represent the intra-class distance and the inter-class distance, respectively, and the calculation formula is

[0091]

[0092] Where Y i (j) represents the jth class (j = 0 or 1, represented by the code bit a k,i= j e {0,1} determines the i-th sample in the j-th class, Y i (j) = [y i,1 (j), y i,2 (j), …, y i,S (j)], S denotes the number of feature variables, n j denotes the number of battery samples in the j-th class. denotes the mean vector of all samples in the j-th class, denotes the mean vector of all samples.

[0093] As shown in Figure 4 , taking a 2-dimensional feature vector as an example, a binary coded individual is used to divide the battery sample set by means of a genetic algorithm.

[0094] Step 3.2, according to the metric evaluation function of the binary coded individual in step 3.1, complete the battery screening clustering under genetic optimization;

[0095] Step 3.2.1, sort the target function value sequence of the K coded individuals in step 3.1 in the g-th generation , select the largest value and the corresponding coded individual denotes the floor symbol;

[0096] Step 3.2.2, randomly combine the selected coded individuals in step 3.2.1 in pairs to perform the conventional crossover operation in the genetic algorithm, forming new coded strings

[0097] Step 3.2.3, perform the conventional mutation operation in the genetic algorithm on each coded bit of the coded individuals obtained after the crossover in step 3.2.2, update the coded strings

[0098] Step 3.2.4, select the coded individuals with the smallest target function value from the coded individuals reserved in step 3.2.1, combine each of the coded individuals obtained by the crossover in step 3.2.2 and the mutation in step 3.2.3 to form the new K coded individuals

[0099] Step 3.2.5, according to the new K coded individuals obtained in step 3.2.4 ​​​The encoding bit value is calculated based on step 3.1 to calculate the objective function value of each encoding individual;

[0100] Step 3.2.6, the minimum value of the objective function value sequence of the K encoding individuals obtained in step 3.2.5 is calculated The minimum value of the objective function value is recorded, and the encoding individual corresponding to the minimum value is recorded as the optimal individual, denoted as minα (g) ; whether the iteration number g at this time is less than the maximum iteration number G is judged, if yes, the next iteration is returned to step 3.2.1; if not, step 3.3 is entered;

[0101] Step 3.3, the optimal individual is used to divide the battery sample set A in the parent node into two complementary and incompatible subsets B and C, and the parent node set A and the two child node sets B and C are formed as child nodes, as shown in Figure 6 2G child nodes after G iterations are taken as the child node set;

[0102] Step 3.4, whether the number of battery samples in each child node in the child node set is greater than the number of batteries m in the battery string is judged; if yes, the child node is defined as a branch node, which needs to return to steps 3.1-3.3 for the set splitting process and generate new two child node sets; if not, the child node is defined as a leaf node, and the splitting process is not performed, and step 3.5 is entered, as shown in Figure 7 ;

[0103] Step 3.5, whether all child nodes have become leaf nodes is judged, if yes, the splitting process is stopped, and step 3.6 is entered; if not, the internal battery sample subset of the child node which is not a leaf node is returned to steps 3.1-3.3 for the set splitting process;

[0104] Step 3.6, according to the type of two child nodes generated by each parent node in the tree structure, the tree screening structure is pruned until the tree structure does not change;

[0105] Specifically:

[0106] (1) if both child nodes under the same parent node are leaf nodes, whether the number of battery samples in the two leaf nodes is greater than the number of batteries m in the battery string is judged respectively; if the number of battery samples in the two leaf nodes is greater than the number of batteries m in the battery string, no operation is performed; if the number of battery samples in the two leaf nodes is less than the number of batteries m in the battery string, the two leaf nodes are removed, and the parent node is changed to a new leaf node; if the number of battery samples in one leaf node under the same parent node is greater than the number of batteries m in the battery string, and the number of battery samples in the other leaf node is less than the number of batteries m in the battery string, no operation is performed, and the two leaf nodes are retained;

[0107] (2) If both child nodes under the same parent node are branch nodes or are a branch node and a leaf node respectively, no operation is performed;

[0108] Step 4: Define the battery set in each pruned leaf node as a battery cluster family, and determine whether the battery samples in each battery cluster family are greater than the number of batteries m in the battery string. If yes, the battery cluster family is determined to be a potential battery string that can be used for battery energy storage system integration. If not, the battery samples in the battery cluster family are determined to be a battery string that cannot be used for battery energy storage system integration and are not used for battery energy storage system integration.

[0109] Step 5: Determine whether the number of potential battery strings that can be used for battery energy storage system integration is greater than the parallel number n of battery strings required by the battery energy storage system. If yes, go to Step 6. If not, a new batch of batteries needs to be obtained (purchased or produced) and returned to Step 2 to re-screen the batteries.

[0110] Step 6: Calculate the center of the battery samples in the battery cluster family that becomes a potential battery string in the feature space (which can be represented by the mean of each feature vector), and select the battery string composed of the n closest battery cluster families to the center for parallel integration.

[0111] Step 7: Within the n battery cluster families, select the battery string composed of the m closest battery samples to the center within the n battery cluster families for series integration.

[0112] Step 8: Complete the series-parallel integration of the battery energy storage system according to the battery samples selected in Steps 6 and 7.

[0113] Taking a 2-dimensional feature vector as an example, when m = 5 and n = 3, the battery string parallel integration selection process of the battery energy storage system is as shown in Figure 8 .

[0114] Finally, it should be noted that the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

Claims

1. An energy storage battery screening integration optimization method, characterized in that, Comprising the following steps: Step one, for the battery energy storage system to be constructed consisting of a plurality of battery strings, according to the design requirements of the system and the performance parameters of the selected battery monomers, determine the number of batteries m in each battery string and the number of parallel battery strings n, and obtain the number of batteries required for the integration of the battery energy storage system; The total number of batteries required for the integration of the battery energy storage system is: m x n; Step two, according to the battery performance test requirements, obtain the test data of N battery samples, and calculate the characteristic variables representing the performance of the battery based on the test data of the battery samples, construct the characteristic vector X describing the performance of each battery and normalize the processing; N >>m x n, the characteristic variables representing the performance of the battery constitute an S-dimensional feature vector X, where the S-dimensional feature vector of the i-th battery is expressed as X i = [x i,1 , x i,2 , …, x i,S ] ; Step three, based on the normalized characteristic vector representing each battery sample, construct an equal-length binary code to form a genetic individual, form a genetic sample set, and optimize the battery sample using a genetic algorithm to complete the tree-type screening structure of the battery and obtain a sub-node set; Specifically: Firstly, define a genetic sample set consisting of W battery monomer S eigenvectors, and W=N at the beginning; meanwhile, define the binary coding length of an individual in the genetic algorithm as the battery monomer number W, and the genetic algorithm parameters include the number of individuals K in the population, the selection rate P s , the crossover rate P c , the mutation rate P m , and the maximum iteration number G; In this set of genetic samples, the binary encoding of the kth genetic individual is denoted as , The i-th encoding bit is ; Then, calculate the objective function J of the binary code individual for the selection of W battery monomers; Finally, use the objective function to optimize the battery sample using a genetic algorithm, and perform battery selection clustering, the specific process being: Step 301, arrange the objective function value sequence under K coded individuals of the gth iteration in descending order, , select the first values and the corresponding coded individuals ; ​ si = 1, 2,..., , denotes the floor symbol; Step 302, for the selected individuals , randomly pair them up and perform the crossover operation of the genetic algorithm to form new coded strings ; Step 303, performing mutation operation on each coding bit of each coding individual in the new coding string Step 304, performing crossover operation on each coding bit of each coding individual in the new coding string Step 305, performing selection operation on each coding bit of each coding individual in the new coding string Step 306, updating the coding string Step 304, selecting the coded individual with the minimum target function value from the remaining coded individuals in step 301 ;​​​​ Step 305, calculate the objective function values of the new K coded individuals , and obtain the objective function value sequence , select the coded individual corresponding to the minimum objective function value in the sequence as the optimal individual, denoted as ; Step 306, determine whether the iteration number g at this time reaches the maximum iteration number G, if yes, execute step 307; if not, return to step 301 for the next iteration; Step 307, use the optimal individual to divide the battery sample set A in the parent node into two complementary and incompatible subsets B and C as child nodes, and complete the tree-type screening structure of the battery; The 2G sub-nodes after G iterations are taken as a sub-node set; Step four, determine whether the number of battery samples in each sub-node in the sub-node set of the tree-type screening structure of the battery is greater than the number of batteries m in the battery string, if yes, define the sub-node as a branch node and return to step three as a new parent node for the binary splitting process of the set and generate two new sub-node sets; if not, define the sub-node as a leaf node and no longer perform the binary splitting process, and proceed to step five; Step five, determine whether all sub-nodes have become leaf nodes, if yes, stop the binary splitting process and proceed to step six; if not, return the internal battery sample subset of the sub-node that is not a leaf node to step three for the binary splitting process of the set; Step six, according to the type of the two sub-nodes generated by each parent node in the tree-type screening structure, prune the tree-type screening structure until the tree-type structure does not change; Step seven, define the battery sample set in each leaf node after pruning as a battery cluster family, and determine whether the battery samples in each battery cluster family are greater than the number of batteries m in the battery string, if yes, the battery cluster family constitutes a potential battery string that can be used for the integration of the battery energy storage system, and step eight is executed; if not, discard the battery cluster family; Step eight, determine whether the number of all potential battery strings is greater than the parallel number n of battery strings required for the battery energy storage system; if yes, execute step nine; otherwise, return to step two to obtain a new batch of batteries for battery selection; Step nine, calculate the center of the battery samples in each potential battery string in the feature space, and select the battery string composed of the n closest battery cluster families to the center for parallel integration; Step ten, in the n battery cluster families, select the m battery samples closest to the center of the internal to series integration; Step eleven, according to the battery samples selected in step nine and step ten, complete the series-parallel integration of the battery energy storage system.

2. The method of claim 1, wherein the method further comprises: In step one, the design requirement parameters of the battery energy storage system include a DC bus voltage U dc (V), a rated power P eb (W) and a rated capacity Q eb (W·h) of the battery energy storage system; wherein the upper and lower limits of the DC bus voltage are Uu dc(V) and Ud dc(V), respectively. The performance parameters of the selected battery cell include a rated voltage of the battery cell of U b (V) and a nominal capacity of the battery C b (A-h); The number of cells in each cell string in a battery energy storage system is: ; The number of parallel connections of the battery string in the battery energy storage system is: .

3. The method of claim 1, wherein the method further comprises: The battery performance test in step two includes nominal capacity test, different temperature charge-discharge test, hybrid power pulse characteristic test and reliability test; the test data of each battery sample obtained includes voltage data, current data, temperature data and DC resistance data.

4. The method of claim 1, wherein the method further comprises: The S-dimensional eigenvectors of the N batteries in step two are normalized, that is, the eigenvalues of the N batteries under a certain feature are normalized. The S-dimensional feature x of the i-th battery is normalized as follows: i,s After normalization, it is represented as y i,s The calculation formula is as follows: ； wherein and denote the functions to find the maximum and minimum, respectively.

5. The method of claim 1, wherein: The objective function calculation formula in step three is as follows: ； wherein, and denote the intra-class distance and the inter-class distance, respectively, and are calculated as follows ； Among them, Y i (j) represents the feature vector of the i-th sample in the j-th class, Y i (j)=[y i,1 (j), y i,2 (j), …, y i,S (j)], j = 0 or 1, the coding bit of the genetic algorithm individual Determine, S represents the number of characteristic variables, n j represents the number of battery samples in the jth category; represents the mean vector of all samples in the jth class, Represents the mean vector of all samples.

6. The method of claim 1, wherein: The process of judging the types of two child nodes under the same parent node and pruning the tree screening structure in step six specifically includes: (1) if both of the two child nodes under the same parent node are leaf nodes, then respectively judge whether the number of battery samples in the two leaf nodes is greater than the number of batteries m in the battery string; if at least one of the two leaf nodes has a number of battery samples greater than the number of battery strings m, then no operation is performed and the two leaf nodes are retained; if the number of battery samples in the two leaf nodes is less than the number of battery strings m, then the two leaf nodes are removed and the parent node is changed to a new leaf node; (2) if both of the two child nodes under the same parent node are branch nodes or are respectively a branch node and a leaf node, then no operation is performed.

7. The method of claim 1, wherein the method further comprises: The center of the battery sample in the feature space is represented by the mean value of each feature vector.

Images