A customized power battery matching method based on dual-coding genetic clustering

Abstract

The present invention discloses a method for customized power battery matching based on dual-coding genetic clustering, which belongs to the technical field of power battery manufacturing. Before formulating the objective function for power battery matching, the method innovatively introduces the Delphi expert scoring method to propose a method for establishing an objective function based on customized needs, and the improved genetic clustering algorithm that encodes the class center and class attributes at the same time avoids the problem that the existing power battery matching algorithm falls into the defect of local optimality and cannot obtain the consistent optimal matching result, and theoretically proves the global convergence of the genetic clustering algorithm with improved gene encoding method, that is, the method can finally obtain the optimal result of power battery matching, which is improved compared with the ordinary genetic algorithm in both matching speed and matching result.

Description

Technical Field

[0001] The invention relates to a customized power battery matching method based on dual-coding genetic clustering, and belongs to the technical field of power battery manufacturing. Background Art

[0002] With the global popularization of low-carbon consensus, the new energy industry has ushered in rapid development in recent years, especially in the lithium battery industry, with the vigorous development of electric vehicles and power storage. Power batteries have attracted widespread attention and use due to their excellent characteristics of high energy density. However, single power batteries are often unable to be used alone in life and production due to the limitations of their performance and parameters such as capacity, voltage, and power. Instead, they need to be grouped in a certain series-parallel manner, which is called the battery grouping process in the production process.

[0003] In the matching process of power batteries, in order to prevent the "short board effect" of grouped batteries, the ultimate goal of matching is to ensure the performance consistency between batteries. The common matching method is mainly the multi-parameter matching method, that is, on the basis of the single parameter matching indicators of capacity, voltage, and internal resistance, multiple external parameters such as pressure difference, self-discharge rate, and voltage platform are introduced to evaluate the comprehensive performance of power battery matching. At present, the matching method of battery index rating is often used in industry. According to the rating of battery indicators such as capacity, voltage, and internal resistance, the matching is carried out according to different levels. However, this matching method is too broad, resulting in poor consistency of battery packs. Therefore, an optimization algorithm is introduced in the matching process, and an objective function is constructed based on the index data difference of each single battery combined with the group attributes, and different optimization algorithms are used to optimize the objective function to obtain the matching result. However, the construction of the ordinary objective function lacks scientificity for the proportion of each indicator. On this basis, the idea of ​​customized battery matching is introduced. According to the actual application of different battery packs, the Delphi expert scoring method is used to evaluate each parameter index, so as to complete the customized matching.

[0004] The optimization algorithms proposed in the existing literature for the power battery assembly process usually use the Kmeans clustering algorithm or its improved version for assembly. Although simple mean clustering is simple, it is easy to embed the local optimal solution, resulting in the final successful battery pack performance is not good enough, especially when the number of batteries required to be assembled is large, the finished battery pack will have worse performance than the optimal one.

[0005] In view of the defects of simple clustering algorithms, some scholars have introduced genetic algorithms to improve the intelligent algorithm of clustering ideas. On the basis of clustering ideas, genetic algorithms are introduced to encode individual battery categories, and genetic methods such as crossover, mutation, and elimination are used to update the battery matching results according to the objective function. However, in clustering analysis problems, the quality of the initial class center plays a decisive role in the convergence speed of the algorithm. The genetic clustering algorithm with separate encoding randomly initializes the class center, and the iterative improvement of the class center is completely based on the Kmeans clustering thinking. The algorithm optimization speed is highly dependent on the quality of the initial class center. However, in the work of battery matching, it is difficult to define a group of highly excellent initial center points at the beginning. Therefore, the genetic improved clustering of the separate encoding battery category cannot obtain the best matching results; in addition, in the later stage of the algorithm iteration, the algorithm for encoding the battery category completely relies on the clustering idea to determine the cluster center. The cluster center is difficult to change in the later stage. It is easy to fail to jump out of the local optimal solution due to the lack of diversity of the cluster center, resulting in its slow optimization speed. At the same time, the traditional genetic clustering algorithm relies on the roulette elimination method of the classic genetic algorithm, which results in the failure to retain the genetic high-quality chromosomes, resulting in slow initial convergence. Summary of the invention

[0006] In order to improve the final matching performance of the battery pack, the present invention provides a customized matching method for power batteries based on dual-coding genetic clustering, which combines genetic algorithm chromosomes to simultaneously encode class centers and categories for optimization; introduces elite retention to replace the roulette update population of traditional genetic ideas; and introduces the idea of ​​local Kmeans clustering to weaken the natural randomness of the genetic algorithm in the early stage of the algorithm, and improves the global convergence under the premise that local clustering does not change the convergence of the final iteration; thereby improving the final matching performance of the battery pack, and compared with the slow optimization speed of genetic improved clustering that only encodes battery categories, the method of the present application also has a significant improvement in matching speed.

[0007] A method for customized power battery matching based on dual-coding genetic clustering, the method comprising:

[0008] Step 1: Determine the weight of each matching index according to the demand; the matching index includes voltage index, capacity index and resistance index;

[0009] Step 2, formulating a matching objective function according to the weights of the matching indicators determined in step 1;

[0010] Step 3: Get the group to be matched Power battery The characteristic variable data of the battery is obtained, the characteristic variables correspond to the configuration indicators one by one, and the characteristic variable data is normalized; Groups, randomly initialized Genetic chromosome coding sequence, chromosomes encode both battery category attributes and class feature centers, length is ;

[0011] Step 4: Update the first k The chromosome encoding population under iterations is obtained to obtain the new generation of chromosome encoding population ;

[0012] Step 5, determine whether the current number of iterations is within the preset clustering applicable range. If it is within the preset clustering applicable range, jump to step 6; if the current number of iterations is not within the preset clustering applicable range, jump to step 7;

[0013] Step 6: Calculate the chromosome encoding population The objective function value of each individual in the update retains the optimal individual corresponding to the minimum value of the objective function; Mean clustering, based on the objective function, calculates new cluster centers, redistributes class attributes, and updates chromosome encoding populations ;

[0014] Step 7, battery grouping according to the final optimal individual result;

[0015] Optionally, in step 1, determining the weight of each matching index according to requirements includes:

[0016] According to the customer's customized needs for battery packs, the expert scoring method is used to determine the weight of each matching index. , Indicates the battery d The weight of the feature variables:

[0017] (1)

[0018] in Indicates the expert authority value, Indicates e Experts on batteries d The weight value given to each feature variable.

[0019] Optionally, step 2 includes:

[0020] Get the group to be matched Power battery feature variables to form a data set ,in , Indicates Battery characteristic variables;

[0021] According to the class attribute, the sum of the Euclidean distances between the characteristic variables of each battery in different classes and the average point of the class is traversed, and the weights of each characteristic variable calculated in step 1 and the characteristic variable information of the battery are combined to determine the group matching objective function as follows:

[0022] (3)

[0023] in, is the first d The average value of the characteristic variables; Battery category attributes, battery information b Expressed by various characteristic variables of the battery.

[0024] Optionally, in step 3, the battery category attribute represents the group to which the battery belongs, and the class feature center represents the feature variable center of all batteries in the group to which the battery belongs; the genetic chromosome coding sequence is , among which The code is the battery type, , ,back The variables are coded as class characteristics;

[0025] The chromosome is initialized randomly. The code is A random positive integer, then A random number encoded as (0, 1].

[0026] Optionally, step 4 includes:

[0027] According to The chromosome information in the population under the iteration is used to determine the classification; In the iteration The chromosome codes A battery collection expression:

[0028] (5)

[0029] Calculate the objective function value corresponding to each chromosome individual in the population ;

[0030] According to the calculated objective function value, the chromosome population is cross-operated in combination with the adaptive improvement idea. After mutation operation, it becomes ;by As the basis for judgment, two sets of chromosome coding sequences were selected. is the upper limit of the crossover probability, and a random number is used to judge each point to perform a crossover operation;

[0031] After the crossover, the chromosome Perform mutation transformation to get the next stage Chromosome individuals, mutation operators combined with adaptive ideas, As the basis for judgment, a group of individuals are selected to enter the mutation stage, mut is the crossover mutation threshold, and the subsequent mutation is judged by random numbers for point mutations, and the mutation range is consistent with the coding range;

[0032] The mutated chromosomes are subjected to natural selection, micro-mutation and elite retention operations to form a new generation Chromosome populations.

[0033] Optionally, the crossover operation in step 4 is only performed in the front Position carried out, is the crossover operator:

[0034] (7)

[0035] in, , used to limit the size of the intersection operator; min( ) represents the objective function value of the chromosome with the smallest objective function, iter represents the total number of iterations, and rand represents a random decimal between 0 and 1.

[0036] Optionally, the mutation operator in step 4 for:

[0037] (8)

[0038] in , used to limit the size of mutation operators.

[0039] Optionally, in the elite retention operation in step 4, With the previous generation The objective functions are sorted from small to large, and the top ten percent of elite chromosomes are selected to compare the objective functions to determine whether to retain the replacement.

[0040] Optionally, the pairing indicators also include expected battery life, discharge temperature, discharge voltage curve change rate, and self-discharge rate.

[0041] The present application also provides application of the above method in a battery production process.

[0042] The beneficial effects of the present invention are:

[0043] The method of the present application combines genetic algorithm chromosomes to simultaneously encode class centers and categories, update elite retention methods, and introduce local clustering. On the one hand, dual encoding can increase class center diversity in the early stage, quickly find the optimal class center range, to ensure the best matching results in the later stage, elite retention increases the probability of retaining high-quality chromosomes, and local clustering also plays a role in early rapid convergence for the genetic clustering algorithm, improving the optimization speed. At the same time, in the later stage of the algorithm's iteration, the dual encoding encoding method can also ensure the diversity of cluster centers in the later stage, giving the algorithm the possibility of jumping out of the local optimum, so the optimization speed of the genetic improved clustering of dual encoding battery categories and class centers is fast and the optimization results are more excellent. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0045] Figure 1 An embodiment of the present invention provides a flow chart of a method for customized power battery matching based on dual-coding genetic clustering.

[0046] Figure 2 It is a normalized battery grouping index characteristic spatial distribution diagram in one embodiment of the present invention.

[0047] Figure 3 is a schematic diagram of chromosome individual crossover in one embodiment of the present invention, where crossover occurs only in the front The first In the iteration Chromosome code and Chromosome No. Take single-point crossover as an example.

[0048] Figure 4 is a schematic diagram of chromosome individual variation in one embodiment of the present invention, with the first In the iteration Chromosome No. Take single-point mutation as an example.

[0049] Figure 5 It is a diagram showing the effect of applying the genetic clustering algorithm disclosed in one embodiment of the present invention to battery grouping. The smaller the vertical coordinate objective function is, the better the grouping result is. ADC-GC is the genetic clustering algorithm in the present invention.

[0050] Fig. 6A yes Figure 5A zoomed-in view of the battery pack simulation results at the initial moment of the middle iteration;

[0051] Figure 6B yes Figure 5 A zoomed-in picture of the battery pack simulation results at 300 iterations;

[0052] Figure 6C yes Figure 5 A zoomed-in view of the battery pack simulation results at the end of the middle iteration.

[0053] Fig. 7A This is a comparison chart of the application effect of the present application method and two other commonly used battery grouping clustering algorithms and their corresponding improved algorithms;

[0054] Figure 7B It is an enlarged view corresponding to the vertical coordinate values ​​between 35-45 in 7A. DETAILED DESCRIPTION

[0055] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0056] The "short board effect" in the power battery assembly process refers to the accelerated attenuation of battery performance after assembly due to inconsistent parameters of single cells. Usually, after the single cells are manufactured, there are certain differences in their initial performance. After the single cells of the same specification and model are assembled into a battery pack, there are certain differences in parameters such as voltage, capacity, internal resistance, life, temperature influence, and self-discharge rate. As the battery is used, the performance differences between single cells continue to accumulate. At the same time, since the use environment of each single cell in the battery pack is not exactly the same, the inconsistency of the single cells is gradually amplified, thereby accelerating the attenuation of battery performance and ultimately causing premature failure of the battery pack.

[0057] The inconsistency of single cell parameters is mainly caused during the manufacturing process and use. Every link in the manufacturing process, such as the uniformity of the slurry during batching, the control of surface density and surface tension during coating, will cause differences in the performance of single cells.

[0058] In order to minimize the "short board effect", battery manufacturers strictly control the production process on the one hand, and sort the batteries through the matching process on the other hand. The battery pack uses batteries of uniform specifications and models, and the battery voltage, capacity, internal resistance, etc. must be measured to ensure the consistency of the initial performance of the battery. This application proposes a new matching method to improve the final matching performance of the battery pack.

[0059] Embodiment 1:

[0060] This embodiment provides a customized power battery matching method based on dual-coding genetic clustering, see Figure 1, the method comprising:

[0061] Step 1: Determine the weight of each matching index according to the demand; the matching index includes voltage index, capacity index and resistance index;

[0062] Due to the different application scenarios of battery packs, the requirements for battery pack indexes are also different. For example, some application scenarios have higher requirements for battery capacity index, and slightly lower requirements for battery voltage index or resistance index. There are also some scenarios that have higher requirements for battery voltage index or resistance index, but the requirements for battery capacity index can be slightly reduced. According to different needs, this application first determines the weight of each packing index in the packing process.

[0063] In addition to voltage, capacity and resistance, the commonly used matching indicators include expected life, discharge temperature, discharge voltage curve, self-discharge rate and other parameters. Different matching indicators are selected, and the corresponding objective functions determined later are also different.

[0064] The weight of each matching indicator in the matching process can be determined by experience based on customer needs, or by other methods, such as multi-expert scoring.

[0065] Step 2, formulating a matching objective function according to the weights of the matching indicators determined in step 1;

[0066] Step 3: Get the group to be matched Power battery The characteristic variable data of the battery is obtained, the characteristic variables correspond to the configuration indicators one by one, and the characteristic variable data is normalized; Groups, randomly initialized The chromosome codes the battery category attributes and the class feature center at the same time. The length of the chromosome coding sequence is ; Among them, the battery category attribute represents the group to which the battery belongs, and the class feature center represents the feature variable center of all batteries in the group to which the battery belongs;

[0067] Battery Extraction After the features are formed, a data set , which is in the form of ,in , Indicates Battery characteristic variables;

[0068] In this embodiment, m=3, and the three characteristic variables of the battery correspond to three matching indicators, namely, voltage, capacity and resistance.

[0069] Step 4: Update the first k The chromosome encoding population under iterations is obtained to obtain the new generation of chromosome encoding population ;

[0070] Step 5, determine whether the current number of iterations is within the preset clustering applicable range. If it is within the preset clustering applicable range, jump to step 6; if the current number of iterations is not within the preset clustering applicable range, jump to step 7;

[0071] Step 6: Calculate the chromosome encoding population The objective function value of each individual in the update retains the optimal individual corresponding to the minimum value of the objective function; Mean clustering, based on the objective function, calculates new cluster centers, redistributes class attributes, and updates chromosome encoding populations ;

[0072] Step 7, battery grouping based on the final optimal individual results.

[0073] Embodiment 2

[0074] This embodiment provides a method for customized power battery assembly based on dual-coding genetic clustering, including:

[0075] Step 1: Select the power battery matching indicators and evaluate the matching indicator weights based on customized requirements;

[0076] Select the multi-parameter matching method, determine the matching indicators, seek multiple experts to assign weights to the matching battery indicators in terms of battery performance evaluation and screening, and request experts to score according to the customized needs of the battery group. Determine the expert authority value based on each expert's qualifications in the battery industry and their familiarity with each influencing indicator in their job position , the sum of all experts’ authority is 1.

[0077] Taking into account the different levels of authority of experts, the weights given by them are averaged. The relative importance of each battery indicator under customized requirements is scored, and the total score of each expert is 100. Calculate the authoritative value set , formula (1) is the authoritative value set:

[0078] (1)

[0079] in, Indicates e Experts on batteries d The weight value given to each feature variable.

[0080] In this example, eight experts were invited to meet the temperature customization requirements. e The battery temperature matching index is scored, and the scoring table is shown in Table 1:

[0081] Table 1: Expert scoring table for battery operating surface temperature indicators

[0082]

[0083] According to formula (1), the authority value set is That is, in response to temperature customization requirements, the weights of the three matching indicators of voltage, capacity and resistance are 0.5825, 0.2397 and 0.1778 respectively.

[0084] Step 2: According to the matching weight index obtained in step 1, formulate a customized matching objective function:

[0085] For the unnormalized f Battery d feature variables, For unnormalized battery The minimum value of the characteristic variable, is the unnormalized original battery The maximum value of the characteristic variable. Before using the battery data, the data must be normalized to obtain the normalized battery information. , formula (2) gives the data normalization calculation method:

[0086] (2)

[0087] Figure 2 The middle dots represent the layout of normalized battery data in the feature space.

[0088] According to the class attributes, the sum of the Euclidean distances between the characteristic variables of each battery in different classes and the average point of the class is traversed, combined with the authoritative value set calculated in step 1 , and normalized battery information With characteristic variables , the category attribute is , is the first d The average value of the characteristic variable.

[0089] Based on this, the customized matching objective function is obtained. Formula (3) gives the expression of the customized matching objective function:

[0090] (3)

[0091] Step 3: Get Power battery The characteristic variables of the battery are corresponding to the configuration indicators one by one. The characteristic data are normalized to determine Groups, randomly initialized Genetic chromosome coding sequence, chromosomes encode both battery category attributes and class feature centers, length is :

[0092] The parameters selected in this case are shown in Table 2, where A random integer between 1 and K:

[0093] Table 2: Case parameter setting table

[0094]

[0095] Battery Extraction After the features are formed, a data set , which is in the form of ,in . Determined based on actual situation There are 3 matching groups, which define the population chromosomes, and the content of the chromosome increases with the number of iterations Update, but the number of chromosomes is fixed, the coding length of the chromosome , formula (4) gives the encoding length Calculate the expression:

[0096] (4)

[0097] The chromosome design method is: The first iteration The coding sequence of chromosome , among which The code is the battery type, , ,back The code is the center of the class feature variable. After normalization, .

[0098] The chromosome is initialized randomly. The code is A random positive integer, then A random number coded as (0, 1]. After randomly initializing the chromosome, it enters the iteration.

[0099] Step 4: Update the chromosome population , updated by crossover, mutation, natural selection micro-mutation, and elite retention time Chromosome-encoded population:

[0100] According to The chromosome information in the population under the iteration is used to determine the classification. Formula (5) gives In the iteration The chromosome codes A battery collection expression:

[0101] (5)

[0102] Calculate the objective function value corresponding to each chromosome individual in the population , formula (6) gives the calculation method of the objective function value:

[0103] (6)

[0104] Figure 3 According to the calculated objective function value, combined with the adaptive improvement idea, the chromosome population is cross-operated. After mutation operation, it becomes .by As the basis for judgment, two sets of chromosome coding sequences were selected. is the upper limit of the crossover probability. For each point, a random number is used to determine the crossover operation. Considering the practical significance, the crossover is only performed in the front Position carried out, is the crossover operator, formula (7) is Calculation method:

[0105] (7)

[0106] in , limit the size of the intersection operator. min( ) represents the objective function value of the chromosome with the smallest objective function, iter represents the total number of iterations, and rand represents a random decimal between 0 and 1.

[0107] Figure 4 In the process of mutation, the chromosome individuals after crossover are Perform mutation transformation to get the next stage Chromosome individuals, mutation operators combined with adaptive ideas, As the basis for judgment, a group of individuals are selected to enter the mutation stage. Mut is the crossover mutation threshold. For point mutations, subsequent mutations are judged by random numbers. The mutation range is consistent with the coding range. is the mutation operator, formula (8) is Calculation method:

[0108] (8)

[0109] in , limit the size of mutation operators.

[0110] The mutated chromosomes are subjected to natural selection, micro-mutation and elite retention operations. Natural selection is full of randomness, and the probability of each individual entering the next generation is exactly the same. After the population is updated, micro-mutation is performed to increase diversity, and elite retention operations are performed. Select chromosomes to enter the next generation and times, copy and update the chromosome. Formula (9) is Calculation method:

[0111] (9)

[0112] is an upward rounding function. After natural selection, the updated chromosomes are subjected to micro-mutation, which can reduce premature maturity caused by selection operations to a certain extent. The micro-mutation method is similar to the previous section, but the point mutation rate Reduce, only improve chromosome diversity. Finally, the elite is retained, With the previous generation The objective functions are sorted from small to large, and the top ten percent of elite chromosomes are selected to compare the objective functions to determine whether to retain the replacement.

[0113] The chromosome population forms a new generation after crossover, mutation, natural selection, micro-variation and elite retention. Chromosome populations.

[0114] Step 5: Determine if the number of iterations is within the applicable range of clustering: If yes, go to step 6. Optimize and update according to clustering conditions The chromosome encoding population at the moment; No, the number of iterations is not within the applicable range of clustering, jump to step seven.

[0115] Step 6: Calculation The objective function value of each individual in the chromosome population at the moment is updated, and the optimal individual corresponding to the minimum value of the objective function is retained:

[0116] Determine the current number of iterations Is it smaller than the maximum applicable range of local clustering? , that is, whether the number of iterations is within the applicable range of local clustering. If not, jump to step 6. If applicable to local clustering, follow The basic principle of mean clustering is based on formula (3) as the objective function, calculating the new cluster center, redistributing class attributes, and updating the population. .

[0117] Step 7: Perform battery matching based on the final optimal individual result obtained in step 6:

[0118] Using the genetic clustering algorithm proposed in this invention, the battery clustering effect at each moment obtained by simulation is as follows: Figure 5 As shown in Figure 6. Figure 5 The horizontal axis is the number of iterations, and the vertical axis is the value of the objective function. Fig. 6A , Figure 6B and Figure 6C They represent the initial iteration moment, 300 iterations, and the final battery pack distribution after the iteration ends.

[0119] The initial objective function in the initial iteration is random grouping and random selection of class centers. Since the initial objective function value is the optimal chromosome target value under the random population, the target value is large, lacks reference and affects the viewing of the image content. Figure 5 The number of iterations starts from 1. It can be seen that the objective function value drops from 18.2090 in the first iteration to 13.7025 in the final. Figure 6 highlights the grouping results under the iterations of 1, 300, and 3000. From the grouping results under the three iterations in the figure, it can be seen that as the number of iterations increases, the various samples screened also change from the initial chaotic and disordered state to a regionally significant aggregation form, and finally form three battery groups with significant differences, verifying that the proposed adaptive genetic clustering algorithm can effectively complete the battery screening and grouping work. The grouping result of 3000 iterations is the final battery grouping result corresponding to the final optimal individual output in step 6. The optimal chromosome is decompiled to obtain the actual battery grouping result.

[0120] This embodiment also compares the application effect of the method of the present application, two other commonly used battery grouping clustering algorithms, and the genetic clustering algorithm improved by introducing local Kmeansb. Fig. 7A and 7B As shown in the figure, the other two commonly used battery group clustering algorithms are the SC-GC algorithm and the WFE-CA algorithm. SC-GC is a genetic algorithm based on the clustering idea with a common encoding method. For details, please refer to (Ma Suliang, Li Jianlin, Li Yaxin, et al. Customized clustering optimization method for battery cascade utilization screening needs. Proceedings of the CSEE, 2022, 42(17): 6208-6220). WFE-CA is a weighted feature integration clustering algorithm. For details, please refer to (Yao Siyang. Improved research and application of K-means algorithm [D]. Beijing Jiaotong University, 2019. DOI: 10.26944 / d.cnki.gbfju.2019.001677.).

[0121] in addition Fig. 7AAmong them, EP-GC is a comparative algorithm that changes the roulette elimination method to elite retention based on SC-GC; LC-GC is a comparative algorithm that introduces local Kmean clustering based on EP-GC.

[0122] Fig. 7A and 7B In the figure, the ordinate represents the objective function value, and the smaller the objective function value, the better the matching result.

[0123] from Fig. 7A It can be seen that after the introduction of elite retention to SC-GC, that is, EP-GC, the grouping results are better, but due to the lack of local clustering and the unchanged encoding method, the optimization ability of these two forms of algorithms for the grouping problem is still far behind the other three. Fig. 7A After LC-GC further introduced dual coding on the basis of EP-GC, EP-GC with ordinary coding method has obvious advantages in terms of matching problem. Figure 7B The enlarged image shows that the method of the present application, i.e., the introduction of local Kmeans clustering based on LC-GC, has significantly better grouping results than the existing SC-GC and WFE-CA commonly used battery grouping clustering methods.

[0124] Some steps in the embodiments of the present invention may be implemented using software, and the corresponding software program may be stored in a readable storage medium, such as a CD or a hard disk.

[0125] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for customized power battery matching based on dual-coding genetic clustering, characterized in that: The method comprises: Step 1: Determine the weight of each matching index according to the demand; the matching index includes voltage index, capacity index and resistance index; Step 2, formulating a matching objective function according to the weights of the matching indicators determined in step 1; Step 3: Get the group to be matched Power battery feature variables, the feature variables correspond to the configuration indicators one by one, and the feature variable data is normalized; determine Groups, randomly initialized Genetic chromosome coding sequence, chromosomes encode both battery category attributes and class feature centers, length is ; Step 4: Update the first k The chromosome encoding population under iterations is obtained to obtain the new generation of chromosome encoding population ; Step 5, determine whether the current number of iterations is within the preset clustering applicable range. If it is within the preset clustering applicable range, jump to step 6; if the current number of iterations is not within the preset clustering applicable range, jump to step 7; Step 6: Calculate the chromosome encoding population The objective function value of each individual in the update retains the optimal individual corresponding to the minimum value of the objective function; Mean clustering, based on the objective function, calculates new cluster centers, redistributes class attributes, and updates chromosome encoding populations ; Step 7, battery grouping according to the final optimal individual result; Determining the weight of each matching index according to the requirements in step 1 includes: According to the customer's customized needs for battery packs, the expert scoring method is used to determine the weight of each matching index. , Indicates the battery d The weight of the feature variables: (1) in Indicates the expert authority value, Indicates e Experts on batteries d The weight value given by the feature variable; The step 2 comprises: Get the group to be matched Power battery feature variables to form a data set ,in , Indicates The battery characteristic variables; According to the class attributes, the sum of the Euclidean distances between the characteristic variables of each battery in different classes and the class average point is traversed, and the weights of each characteristic variable calculated in step 1 and the characteristic variable information of the battery are combined to determine the group matching objective function as follows: (3) in, is the first d The average value of the characteristic variables; Battery category attributes, battery information b Expressed by various characteristic variables of the battery; In step 3, the battery category attribute indicates the group to which the battery belongs, and the class feature center indicates the feature variable center of all batteries in the group to which the battery belongs; the genetic chromosome coding sequence is , among which the former The code is the battery type, , ,back The variables are coded as class characteristics; The chromosome is initialized randomly. The code is A random positive integer, then A random number encoded as (0, 1]; The step 4 comprises: According to The chromosome information in the population under the iteration is used to determine the classification; In the iteration The chromosome codes A battery collection expression: (5) Calculate the objective function value corresponding to each chromosome individual in the population ; According to the calculated objective function value, the chromosome population is cross-operated in combination with the adaptive improvement idea. After mutation operation, it becomes ;by As the basis for judgment, two sets of chromosome coding sequences were selected. is the upper limit of the crossover probability, and a random number is used to judge each point to perform a crossover operation; After the crossover, the chromosome Perform mutation transformation to get the next stage Chromosome individuals, mutation operators combined with adaptive ideas, As the basis for judgment, a group of individuals are selected to enter the mutation stage. Mut is the judgment threshold of crossover mutation. For point mutation, the subsequent mutation is judged by random numbers. The mutation range is consistent with the coding range. The mutated chromosomes are subjected to natural selection, micro-mutation and elite retention operations to form a new generation of chromosome populations ; The crossover operation in step 4 is only performed before Position carried out, is the crossover operator: (7) in, , used to limit the size of the intersection operator; min( ) represents the objective function value of the chromosome with the minimum objective function, iter represents the total number of iterations, and rand represents a random decimal between 0 and 1; The mutation operator in step 4 for: (8) in , used to limit the size of mutation operators.

2. The method according to claim 1, characterized in that In the elite retention operation in step 4, With the previous generation The objective functions are sorted from small to large, and the top ten percent of elite chromosomes are selected to compare the objective functions to determine whether to retain the replacement.

3. The method according to claim 1, characterized in that The matching indicators also include expected battery life, discharge temperature, discharge voltage curve change rate, and self-discharge rate.

Images