Lithium battery capacity prediction method based on XGBoost and particle swarm optimization

By using the combination method of XGBoost and multi-objective particle swarm algorithm in the production process of lithium-ion batteries, a lithium battery capacity prediction model is constructed based on the pre-capacity separation process data, which solves the problems of high energy consumption and low prediction accuracy of the capacity separation process, and realizes the Efficient and accurate lithium-ion battery capacity prediction and production efficiency improvement.

CN118917176APending Publication Date: 2024-11-08WUYI UNIV
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202410927026.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

In the production process of lithium-ion batteries, the capacity separation process consumes high energy and has many process parameters, resulting in low battery capacity prediction accuracy, which is difficult to effectively solve the problem in the existing technology.

Method used

Using the lithium battery capacity prediction method based on XGBoost and an improved multi-objective particle swarm algorithm, the lithium-ion battery capacity prediction model is constructed by obtaining production data before the capacity separation process and selecting excellent feature combinations to build a lithium-ion battery capacity prediction model instead of the traditional capacity separation process data to improve the Prediction accuracy.

Benefits of technology

It realizes high-precision prediction of lithium-ion battery capacity, reduces production energy consumption, improves production efficiency, can completely replace the capacity separation process, and reduces factory energy consumption and costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917176A_ABST
    Figure CN118917176A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a lithium battery capacity prediction method based on XGBoost and a particle swarm algorithm. The method comprises the following steps: acquiring target production data of a to-be-detected lithium ion battery before a capacity grading process; the target production data is input into a trained lithium ion battery capacity target prediction model for battery capacity prediction, a prediction result corresponding to the lithium ion battery to be detected is obtained, the lithium ion battery capacity target prediction model is constructed based on an XGBoost algorithm and a target feature combination, and the lithium ion battery capacity prediction result corresponding to the lithium ion battery to be detected is obtained. The target feature combination is selected from the production data by adopting an IMOPSO algorithm. The embodiment of the invention does not depend on the production data of the capacity grading process, but constructs the lithium ion battery capacity prediction model by acquiring the production data before the capacity grading process, so that the prediction model can completely replace the capacity grading process. And an IMOPSO algorithm is adopted to select an excellent feature combination, so that the prediction precision of the prediction model is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of lithium-ion batteries, and in particular to a lithium battery capacity prediction method based on XGBoost and particle swarm algorithm. Background Art

[0002] Lithium-ion batteries (LI Bs) are widely used in laptops, mobile phones, electric vehicles, forklifts, military communications and other fields due to their advantages such as high efficiency, high energy density, high power and low self-discharge. In the production process of lithium-ion batteries, charge and discharge tests are required to determine their capacity. This production process is called the capacity separation process, which is a very energy-consuming process. The battery capacity of lithium-ion batteries is closely related to the process parameters of the production process, which makes it possible to build a battery capacity prediction model. There are a large number of process parameters in the production process of lithium-ion batteries, and it is impossible to determine in advance which process parameters will affect the prediction model. If all process parameters are used as input features of the prediction model, it will not only reduce the operating efficiency of the prediction model, but also reduce the prediction accuracy of the model, resulting in low prediction accuracy of lithium-ion battery capacity. Summary of the invention

[0003] The embodiment of the present invention provides a lithium battery capacity prediction method based on XGBoost and particle swarm algorithm, which can improve the prediction accuracy of lithium-ion battery capacity.

[0004] In a first aspect, an embodiment of the present invention provides a lithium battery capacity prediction method based on XGBoost and particle swarm algorithm, comprising:

[0005] Obtain target production data of the lithium-ion battery to be tested before the capacity division process;

[0006] The target production data is input into a trained lithium-ion battery capacity target prediction model to perform battery capacity prediction, and a prediction result corresponding to the lithium-ion battery to be tested is obtained, wherein the lithium-ion battery capacity target prediction model is constructed based on an XGBoost algorithm and a target feature combination, and the target feature combination is selected from the production data of the lithium-ion battery before the capacity division process by using an improved multi-objective particle swarm algorithm.

[0007] In some embodiments, the training method of the lithium-ion battery capacity target prediction model includes:

[0008] Obtain production data of lithium-ion batteries before the capacity division process;

[0009] Extracting capacity data of the lithium-ion battery from the production data;

[0010] Dividing the capacity data into a training set, a validation set, and a test set according to a preset ratio;

[0011] Selecting feature combinations using an improved multi-objective particle swarm algorithm, building a lithium-ion battery capacity prediction model using an XGBoost algorithm, and training the lithium-ion battery capacity prediction model using each group of the feature combinations and the training set to obtain each group of the lithium-ion battery capacity prediction model;

[0012] Substituting each group of the feature combination and the test set into each group of the lithium-ion battery capacity prediction model, calculating the fitness value of each group of the feature combination, and looping in sequence until the improved multi-objective particle swarm algorithm completes iteration;

[0013] Determine a target feature combination according to the fitness value, substitute the target feature combination and the validation set into the lithium-ion battery capacity prediction model, calculate the fitness value of each group of the target feature combination, and select the best target feature combination;

[0014] The trained lithium-ion battery capacity target prediction model is generated according to the best target feature combination.

[0015] In some embodiments, after extracting the capacity data of the lithium-ion battery from the production data, the method further includes:

[0016] The capacity data is preprocessed, and the preprocessing includes removing the capacity data having abnormal values.

[0017] In some embodiments, substituting each group of the feature combination and the test set into each group of the lithium-ion battery capacity prediction model, calculating the fitness value of each group of the feature combination, and looping in sequence until the improved multi-objective particle swarm algorithm completes the iteration, includes:

[0018] Acquire characteristic data from the production data, use the characteristic data as the position vector of individual particles in the multi-objective particle swarm algorithm, and initialize the generated population with N individuals;

[0019] Calculate the fitness value corresponding to the initial population individuals through the fitness function calculation formula;

[0020] Initialize individual historical optimal files, global leader files, and abandoned global leader files;

[0021] The improved multi-objective particle swarm optimization algorithm is used to update the feature combination and calculate the fitness value of each particle;

[0022] Update the individual historical optimal file based on the individual historical optimal file of the previous iteration and the population of this iteration;

[0023] Based on the global leader profile of the previous iteration and the population of this iteration, the global leader profile is updated using the Pareto dominance principle;

[0024] The global leader profile is maintained according to the global leader profile maintenance method based on Pearson correlation coefficient and feature selection probability;

[0025] When the iterative termination condition of the improved multi-objective particle swarm algorithm is met, the final feature combination and the corresponding fitness value are output.

[0026] In some embodiments, the step of calculating the fitness value corresponding to the initial population individual by using the fitness function calculation formula includes:

[0027] Discretize the particle population and select feature combinations based on the discretization results;

[0028] The feature combination selected by each particle is substituted into the lithium-ion battery capacity prediction model to predict the lithium-ion battery capacity of the test set, and the fitness value of each particle is calculated.

[0029] In some embodiments, the use of an improved multi-objective particle swarm optimization algorithm to update the feature combination and calculate the fitness value of each particle includes:

[0030] Calculate the Pearson correlation coefficient between each feature and lithium-ion battery capacity;

[0031] Calculate the feature combination and the sum of Pearson correlation coefficients for each global leader;

[0032] Re-rank all global leader particles from small to large according to the sum of the Pearson correlation coefficients of each global leader to obtain a re-ranked global leader profile;

[0033] According to the iteration state of the improved multi-objective particle swarm algorithm and the sum of the Pearson coefficients of the global leader particles, a global leader particle is selected for each swarm particle;

[0034] The corresponding feature combination is obtained according to the position vector of the population particle, and the feature combination is substituted into the lithium-ion battery capacity prediction model to calculate the fitness value of each particle.

[0035] In some embodiments, maintaining the global leader profile according to the global leader profile maintenance method based on the Pearson correlation coefficient and the feature selection probability includes:

[0036] Calculate the number of particles Kn in the global leader file of this iteration, and merge the global leader file of this iteration and the abandoned global leader file into population P1;

[0037] Calculate the quality index of each particle in population P1;

[0038] Randomly select Kn particles from population P1 twice to form two parent populations, and obtain the quality indicators corresponding to the Kn particles at the same time;

[0039] The quality indicators of the two parent populations are compared with each other, and the corresponding simulated binary crossover formula is selected according to the quality indicators to perform crossover operations to obtain sub-populations;

[0040] Perform polynomial mutation operation on the sub-population to obtain the final sub-population P2;

[0041] Obtaining a corresponding feature combination according to the position vector of the final sub-population P2, substituting it into the lithium-ion battery capacity prediction model, and calculating the fitness value of the sub-population;

[0042] Based on the Pareto dominance principle, all target particles with non-dominated solutions are selected from population P1 and the final sub-population P2. If the number of target particles exceeds N, N particles are selected in descending order according to the Manhattan crowding distance to form the global leader file of this iteration; if the number of target particles does not exceed N, the global leader file of this iteration is composed of target particles.

[0043] In a second aspect, an embodiment of the present invention further provides a lithium battery capacity prediction device based on XGBoost and particle swarm algorithm, the device comprising:

[0044] An acquisition module, used to acquire target production data of the lithium-ion battery to be tested before the capacity division process;

[0045] A prediction module is used to input the target production data into a trained lithium-ion battery capacity target prediction model to perform battery capacity prediction, and obtain a prediction result corresponding to the lithium-ion battery to be tested, wherein the lithium-ion battery capacity target prediction model is constructed based on an XGBoost algorithm and a target feature combination, and the target feature combination is selected from the production data of the lithium-ion battery before the capacity division process by using an improved multi-objective particle swarm algorithm.

[0046] In a third aspect, an embodiment of the present invention further provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the lithium battery capacity prediction method based on XGBoost and particle swarm algorithm as described in the first aspect is implemented.

[0047] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the lithium battery capacity prediction method based on XGBoost and particle swarm algorithm as described in the first aspect.

[0048] According to the lithium battery capacity prediction method based on XGBoost and particle swarm algorithm provided by the embodiment of the present invention, the lithium ion battery capacity prediction method includes: obtaining the target production data of the lithium ion battery to be tested before the capacity division process; inputting the target production data into the trained lithium ion battery capacity target prediction model to predict the battery capacity, and obtaining the prediction result of the corresponding lithium ion battery to be tested, wherein the lithium ion battery capacity target prediction model is constructed based on the XGBoost algorithm and the target feature combination, and the target feature combination is obtained by selecting from the production data of the lithium ion battery before the capacity division process by using the improved multi-objective particle swarm algorithm. The embodiment of the present invention does not rely on the production data of the capacity division process, but constructs the lithium ion battery capacity prediction model by obtaining the production data before the capacity division process, so that the lithium ion battery capacity prediction model proposed by the present invention can completely replace the capacity division process. Then, by using the improved multi-objective particle swarm algorithm to select excellent feature combinations, the prediction accuracy of the lithium ion battery capacity prediction model is improved. Based on this, the embodiment of the present invention can improve the prediction accuracy of the lithium ion battery capacity. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a flow chart of a lithium battery capacity prediction method based on XGBoost and particle swarm algorithm provided by an embodiment of the present invention;

[0050] Figure 2 It is a flow chart of a training method of a lithium-ion battery capacity target prediction model provided by one embodiment of the present invention;

[0051] Figure 3 This is a general flow chart of a lithium battery capacity prediction method based on XGBoost and particle swarm algorithm provided by an embodiment of the present invention;

[0052] Figure 4 It is a detailed flow chart of a lithium battery capacity prediction method based on XGBoost and particle swarm algorithm provided by an embodiment of the present invention;

[0053] Figure 5 This is a feature selection engineering flow chart of a lithium-ion battery capacity prediction model provided by one embodiment of the present invention;

[0054] Figure 6 is a schematic diagram of a lithium battery capacity prediction device based on XGBoost and particle swarm algorithm provided by an embodiment of the present invention;

[0055] Figure 7 is a schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0057] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification and claims and the following drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0058] In the embodiments of the present invention, the words "further", "exemplarily" or "optionally" are used to indicate examples, illustrations or descriptions, and should not be interpreted as being more preferred or more advantageous than other embodiments or designs. The use of the words "further", "exemplarily" or "optionally" is intended to present related concepts in a specific way.

[0059] First, some terms involved in the present invention are analyzed:

[0060] LI Bs: (Lithium-ion batteries) lithium-ion batteries;

[0061] XGBoost: distributed gradient boosting library;

[0062] IMOPSO: Improved multi-objective particle swarm optimization algorithm.

[0063] In order to more conveniently describe the working principle of the embodiment of the present invention later, an introduction to relevant technical scenarios is first given below.

[0064] Lithium-ion batteries (LI Bs) are widely used in laptops, mobile phones, electric vehicles, forklifts, military communications and other fields due to their advantages such as high efficiency, high energy density, high power and low self-discharge. In the production process of lithium-ion batteries, charge and discharge tests are required to determine their capacity. This production process is called the capacity separation process, which is a very energy-consuming process. The battery capacity of lithium-ion batteries is closely related to the process parameters of the production process, which makes it possible to build a battery capacity prediction model. There are a large number of process parameters in the production process of lithium-ion batteries, and it is impossible to determine in advance which process parameters will affect the prediction model. If all process parameters are used as input features of the prediction model, it will not only reduce the operating efficiency of the prediction model, but also reduce the prediction accuracy of the model, resulting in low prediction accuracy of lithium-ion battery capacity.

[0065] Based on this, the present invention provides a lithium battery capacity prediction method based on XGBoost and particle swarm algorithm. Wherein, the lithium-ion battery capacity prediction method includes obtaining the target production data of the lithium-ion battery to be tested before the capacity division process; inputting the target production data into the trained lithium-ion battery capacity target prediction model to predict the battery capacity, and obtaining the prediction result of the corresponding lithium-ion battery to be tested, wherein the lithium-ion battery capacity target prediction model is constructed based on the XGBoost algorithm and the target feature combination, and the target feature combination is obtained by selecting from the production data of the lithium-ion battery before the capacity division process by using an improved multi-objective particle swarm algorithm. The embodiment of the present invention does not rely on the production data of the capacity division process, but constructs the lithium-ion battery capacity prediction model by obtaining the production data before the capacity division process, so that the lithium-ion battery capacity prediction model proposed by the present invention can completely replace the capacity division process. Then, by using the improved multi-objective particle swarm algorithm to select excellent feature combinations, the prediction accuracy of the lithium-ion battery capacity prediction model is improved. Based on this, the embodiment of the present invention can improve the prediction accuracy of the lithium-ion battery capacity.

[0066] The embodiments of the present invention are further described below in conjunction with the accompanying drawings.

[0067] like Figure 1 As shown, Figure 1 It is a flowchart of a lithium battery capacity prediction method based on XGBoost and particle swarm algorithm provided by an embodiment of the present invention. The lithium ion battery capacity prediction method may include but is not limited to steps S101 to S102.

[0068] Step S101: obtaining target production data of the lithium-ion battery to be tested before the capacity division process;

[0069] Step S102: input the target production data into the trained lithium-ion battery capacity target prediction model to perform battery capacity prediction, and obtain the prediction result of the corresponding lithium-ion battery to be tested, wherein the lithium-ion battery capacity target prediction model is constructed based on the XGBoost algorithm and the target feature combination, and the target feature combination is selected from the production data of the lithium-ion battery before the capacity division process by adopting the improved multi-objective particle swarm algorithm.

[0070] The embodiment of the present invention does not rely on the production data of the capacity division process, but constructs a lithium-ion battery capacity prediction model by acquiring the production data before the capacity division process, so that the lithium-ion battery capacity prediction model proposed by the present invention can completely replace the capacity division process. Then, by adopting an improved multi-objective particle swarm algorithm to select excellent feature combinations, the prediction accuracy of the lithium-ion battery capacity prediction model is improved. Based on this, the embodiment of the present invention can improve the prediction accuracy of the lithium-ion battery capacity.

[0071] It should be pointed out that the lithium-ion battery capacity prediction method based on improved multi-objective particle swarm optimization and XGBoost provided by the present invention can completely replace the capacity division process, and fully consider the impact of each production step on the capacity of the lithium-ion battery, thereby reducing the production energy consumption of lithium batteries, improving production efficiency, optimizing the production process, and realizing the direct prediction of the capacity of lithium-ion batteries.

[0072] like Figure 2 As shown, the training method of the lithium-ion battery capacity target prediction model of the present invention includes but is not limited to steps S201 to S207.

[0073] Step S201: Acquire production data of lithium-ion batteries before the capacity division process;

[0074] Step S202: extracting capacity data of the lithium-ion battery from the production data;

[0075] Step S203: Divide the capacity data into a training set, a validation set, and a test set according to a preset ratio;

[0076] Step S204: using an improved multi-objective particle swarm algorithm to select feature combinations, constructing a lithium-ion battery capacity prediction model through an XGBoost algorithm, and using each set of feature combinations and a training set to train the lithium-ion battery capacity prediction model to obtain each set of lithium-ion battery capacity prediction models;

[0077] Step S205: Substitute each feature combination and test set into each lithium-ion battery capacity prediction model, calculate the fitness value of each feature combination, and repeat until the improved multi-objective particle swarm algorithm completes the iteration;

[0078] Step S206: determining a target feature combination according to the fitness value, substituting the target feature combination and the validation set into the lithium-ion battery capacity prediction model, calculating the fitness value of each target feature combination, and selecting the best target feature combination;

[0079] Step S207: Generate a trained lithium-ion battery capacity target prediction model based on the best target feature combination.

[0080] It can be understood that the lithium-ion battery capacity prediction method of the present invention includes feature selection, model training and parameter optimization.

[0081] like Figures 3 to 5 As shown, the training method of the lithium-ion battery capacity target prediction model of the present invention mainly includes the following steps:

[0082] Step 1: Obtain the production data of lithium-ion batteries before the capacity division process, and extract the capacity data of lithium-ion batteries from them, and divide these data into training set, verification set and test set. At the same time, pre-process these data and remove some batteries with abnormal data;

[0083] A total of 20,000 battery data and 124 feature data (including lithium-ion battery capacity data) were collected. After data preprocessing, 19,324 battery data remained. After extracting the lithium-ion battery capacity data, 123 feature data remained.

[0084] Step 2: Build a lithium-ion battery capacity prediction model framework based on the XGBoost algorithm, and divide the training set, test set, and validation set into a ratio of 8:1:1;

[0085] Step 3: Use the improved multi-objective particle swarm algorithm (IMOPSO) to select a feature combination, substitute it into the lithium-ion battery capacity prediction model, and calculate the fitness value of each particle; cycle in sequence until the improved multi-objective particle swarm algorithm completes the iteration;

[0086] The improved multi-objective particle swarm algorithm has the following two improvements: 1. A global leader selection strategy based on Pearson correlation coefficient is proposed. 2. A global leader profile maintenance method based on Pearson correlation coefficient and feature selection probability is proposed;

[0087] Step 3.1: Initialize feature combination: Use the above 123 features as the position vector p of the individual particle in the multi-objective particle swarm algorithm i , initialize the generated population with N individuals, and calculate the fitness value corresponding to the individuals in the initial population through the fitness function calculation formula. i represents the position vector of the i-th particle in the population, p i ={pi,1 ,p i,2 ,...p i,M}, where M is the number of features, the value of M is 123, p i,M The value range of is [0,1];

[0088] The construction method of the fitness function is:

[0089] (1) Discretize the particle population and select the feature combination based on the discretization result. The discretization method is as follows:

[0090] p i,j >0.5, the jth feature of the ith particle is selected, otherwise the feature is not selected.

[0091] p1 = [0.3, 0.4, 0.5, 0.6, 0.7]. After discretization, p1 = [0, 0, 0, 1, 1], where 0 means that the feature is not selected and 1 means that the feature is selected.

[0092] (2) Substituting the feature combination selected by each particle into the lithium-ion battery capacity prediction model, predicting the lithium-ion battery capacity of the validation set, and calculating the fitness value of each particle;

[0093] Assume that the lithium-ion battery capacity value of the validation set is Tc = {Tc1, Tc2, ...Tc i}, the predicted lithium-ion battery capacity of the validation set is Pc = {Pc1, Pc2, ... Pc i}, where Tc i Represents the i-th battery capacity value in the validation set, Pc i Represents the predicted capacity value of the ith lithium-ion battery in the validation set. The fitness values ​​that need to be calculated are the mean error (MAPE) and the goodness of fit (IRS).

[0094] and the percentage of samples exceeding the preset error (PESE). The smaller the three optimization goals, the better. Their calculation formulas are

[0095] The formula is as follows:

[0096]

[0097] Where l represents the number of lithium-ion batteries in the validation set.

[0098] Step 3.2: Initialize the individual historical optimal archive, the global leader archive, and the abandoned global leader archive;

[0099] The initialized individual optimal history archive is the same as the initialized population. The initialized global leader archive is to find all non-dominated particles in the initialized population using the Pareto dominance principle. The initialized abandoned global leader archive is an empty set.

[0100] Step 3.3: Use the improved multi-objective particle swarm optimization algorithm to update the feature combination and calculate the fitness value of each particle;

[0101] In the particle swarm algorithm, in order to simulate the social behavior of birds looking for food, the swarm particles will learn from the global leader particle and the individual historical best particle. Since there are often multiple global leader particles in the multi-objective optimization algorithm, how to better select the global leader particle is an urgent problem to be solved. The speed and position update formula of the particle in the particle swarm algorithm is as follows:

[0102]

[0103] in represents the velocity of the ith particle in the t+1th iteration, represents the position of the ith particle in the t+1th iteration, ω represents the velocity inertia coefficient, c1 represents the global learning factor, c2 represents the individual learning factor, represents the best individual historical position of the ith particle in the t+1th iteration, represents the position of the global leader of the ith particle in the t+1th iteration, and r1 and r2 represent a random number between 0 and 1.

[0104] The selection strategy of the global leader particle of each particle is as follows:

[0105] (1) Calculate the Pearson correlation coefficient between each feature and the lithium-ion battery capacity. The calculation formula is as follows:

[0106]

[0107] Among them, Pe j represents the Pearson correlation coefficient between the jth feature and the lithium-ion battery capacity, y represents the lithium-ion battery capacity, and f j Represents the relevant data of the jth feature, cov(y,f j ) represents the covariance between the capacity of the lithium-ion battery and the jth feature, σ(y) represents the standard deviation of the lithium-ion battery, σ(f j ) represents the standard deviation of the jth feature;

[0108] (2) Calculate the sum of the feature combination and Pearson correlation coefficient of each global leader. The calculation formula is as follows:

[0109]

[0110] GBP i,j represents the jth position component of the i-th global leader particle; GBD i,j represents the selection of the jth feature of the i-th global leader particle, 0 means that this feature is not selected, and 1 means that this feature is selected; GBPe i represents the sum of the Pearson coefficients of the i-th global leader particle;

[0111] (3) According to the result of formula (8), all global leader particles are re-ranked from small to large to obtain the re-ranked global leader file (GS). GS = {GS1, GS2, ..., GS K}, where K represents the number of global leader particles in the global leader file of this iteration;

[0112] (4) According to the iteration state of the multi-objective particle swarm algorithm and the sum of the Pearson coefficients of the global leader particles,

[0113] The population particles select the global leader particle, and the selection scheme is as follows:

[0114]

[0115] Where FE represents the current function evaluation times of the multi-objective particle swarm algorithm, maxFE represents the maximum function evaluation times of the multi-objective particle swarm algorithm; GB ra Represents a random selection of a global leader particle from the global leader archive; Represents FE / maxFE*K rounded up; Represents the global leader particle of the i-th swarm particle at the t-th iteration;

[0116] (5) Obtain the corresponding feature combination based on the position vector of the population particles and substitute it into the lithium-ion battery capacity prediction model.

[0117] Calculate its fitness value.

[0118] Step 3.4: Update the individual historical optimal file based on the individual historical optimal file of the previous iteration and the population of this iteration;

[0119] The optimal individual history archive update scheme is as follows: Can control for if Can control for if and It is a relationship of non-dominance. for

[0120] Step 3.5: Based on the global leader file of the previous iteration and the population of this iteration, the global leader file is updated using the Pareto dominance principle;

[0121] The global leader file update scheme is as follows: Assuming that the maximum number of particles stored in the global leader file is N, all particles with non-dominated solutions are selected from the global leader file of the previous iteration and the population of this iteration using the Pareto dominance principle. If the number of these particles exceeds N, N particles are selected in descending order according to the Manhattan congestion distance to form the global leader file of this iteration. If the number of these particles does not exceed N, the global leader file of this iteration consists of these particles.

[0122] Step 3.6: Update the abandoned global leader profile according to the global leader profile of the previous iteration and the global leader profile of this iteration;

[0123] The abandoned global leader file update scheme is as follows: first, compare the global leader file of the previous iteration with the global leader file of this iteration to find the global leader particle (P1) abandoned in this iteration; use the Pareto dominance principle to select all particles with non-dominated solutions from P1 and the abandoned global leader file of the previous iteration; assume that the maximum number of stored particles in the abandoned global leader file is N, if the number of this part of particles does not exceed N, then this part of particles is the abandoned global leader file of this iteration; if the number of this part of particles exceeds N, then calculate this part of particles based on the Manhattan crowding distance, and select N particles from large to small in order as the abandoned global leader particles of this iteration.

[0124] Step 3.7: Maintain the global leader profile according to the global leader profile maintenance method based on the Pearson correlation coefficient and feature selection probability;

[0125] The global leader profile maintenance method based on Pearson correlation coefficient and feature selection probability is as follows:

[0126] (1) Calculate the number of particles in the global leader file of this iteration (Kn), and merge the global leader file of this iteration and the abandoned global leader file into a population (P1);

[0127] (2) Calculate the quality index of each particle in the population. The quality index formula is as follows;

[0128]

[0129] GBD i,j and GBPe i It can be calculated by formula (8) and formula (9)

[0130] (3) Kn particles are randomly selected from this population twice to form two parent populations (pa1 t With pa2 t ),in At the same time, quality indicators corresponding to these particles are obtained;

[0131] (4) pa1 t With pa2 t The quality indicators of the two are compared with each other, and the corresponding simulated binary exchange is selected according to the quality indicator.

[0132] The crossover formula is used to perform a crossover operation to obtain the sub-population, as shown in the following formula:

[0133]

[0134] The size of i and j is the same, but i represents pa1 t Particles, j represents pa2 t of particles.

[0135] (5) The final subpopulation (P2) is obtained through the polynomial mutation operation. The polynomial mutation operation is shown in the following formula:

[0136]

[0137] MCP k represents the kth position component of the offspring, u k represents the upper limit of the kth position component, l k represents the lower limit of the kth position component; r and γ both represent random numbers in [0,1]; η m An index representing the distribution of variation.

[0138] (6) Obtain the corresponding feature combination according to the position vector of P2, substitute it into the lithium-ion battery capacity prediction model, and calculate the fitness value of the subpopulation;

[0139] (7) Using the Pareto dominance principle, select all particles with non-dominated solutions from P1 and P2. If the number of these particles exceeds N, select N particles in descending order according to the Manhattan crowding distance to form the global leader profile of this iteration. If the number of these particles does not exceed N, the global leader profile of this iteration consists of these particles.

[0140] Step 3.8: Determine whether the iterative termination condition of the multi-objective particle swarm algorithm is met. If so, output the final feature combination and the corresponding fitness value; if the iterative termination condition of the multi-objective particle swarm algorithm is not met, repeat steps 3.3 to 3.7.

[0141] The IMOPSO is a multi-objective optimization algorithm, so IMOPSO will eventually output multiple sets of feature combinations.

[0142] Step 4: Substitute the feature combination of the final result of IMOPSO and the validation set data into the lithium-ion battery capacity prediction model in turn, calculate the fitness value, select the optimal feature combination according to the fitness value, and form the final lithium-ion battery capacity prediction model.

[0143] The beneficial effect of adopting the above technical solution is that the present invention provides a lithium-ion battery capacity prediction method based on improved multi-objective particle swarm optimization and XGBoost, and determines the optimal lithium-ion battery capacity prediction model by combining the fast convergence ability of the multi-objective particle swarm algorithm and the prediction ability of the XGBoost algorithm. Using the lithium-ion battery capacity prediction model of the present invention, the capacity of the lithium-ion battery can be predicted by partial lithium-ion battery production data before the capacity division process. The use of this model can replace the capacity division process, thereby reducing factory energy consumption and improving factory production efficiency. At the same time, its feature combination can provide a certain reference value for optimizing the factory industrial structure.

[0144] It should be pointed out that the present invention proposes an improved multi-objective particle swarm algorithm, which can find better feature combinations compared with other multi-objective optimization algorithms.

[0145] It should be pointed out that the present invention combines the multi-objective particle swarm algorithm with the machine learning algorithm, improves the feature selection capability of the multi-objective particle swarm algorithm, and enhances the accuracy of the lithium-ion battery capacity prediction model.

[0146] It should be pointed out that the present invention can completely replace the capacity division process. Compared with the general lithium-ion battery capacity prediction model, the present invention does not require the production data of the capacity division process. It only needs the lithium-ion battery production data before the capacity division process to predict the capacity of the lithium-ion battery, thereby reducing factory energy consumption, saving production costs, and achieving green development. It has great application significance.

[0147] In this embodiment, the main parameters of IMOPSO are shown in Table 1.

[0148] Table 1 The main parameters of IMOPSO are as follows:

[0149] Population size 100 Number of function evaluations 5000 Speed ​​Inertia Weight [0.1,0.5] Individual / Global Learning Factors 1 Crossover probability 1.0 Cross-distribution index 20 Mutation probability 1 / 123 Mutation Distribution Index 20

[0150] The IMOPSO listed in Table 1 is applied to the feature selection project of the lithium-ion battery capacity prediction model, and the feature combination that can improve the performance of the lithium-ion battery capacity prediction model is obtained, as shown in Table 2. It can be seen that the feature combination selected by IMOPSO can effectively improve the prediction performance of the lithium-ion battery capacity prediction model.

[0151] Table 2 Comparison of the results of the lithium-ion battery capacity prediction model after feature selection and without feature selection (substituting the test set data) is as follows:

[0152]

[0153] After IMOPSO selects features, several sets of excellent feature combinations can be obtained. The average value of the lithium-ion battery capacity prediction results after feature selection engineering in Table 2 is the result of substituting these sets of feature combinations into the lithium-ion battery capacity prediction model, calculating the results of MAPE, PESE, and I RS, and finally finding the average value. It can be seen from Table 2 that the performance of the lithium-ion battery capacity prediction model after feature selection engineering is better than that without feature selection engineering.

[0154] In order to verify the effectiveness of the feature selection engineering of IMOPSO in the lithium-ion battery capacity prediction model based on the XGBoost algorithm, this embodiment also provides LMOCSO, MMOPSO, MOEADDCWV, NMPSO, and SMPSO instead of IMOPSO, and performs feature selection engineering in the lithium-ion battery capacity prediction model based on the XGBoost algorithm to find excellent feature combinations that can improve the lithium-ion battery capacity prediction model. The average values ​​of the optimal fitness values ​​of the above five multi-objective optimization algorithms at the maximum number of function evaluations are compared. In order to ensure the fairness of the algorithm, the population size and the maximum number of function evaluations of the above five multi-objective optimization algorithms are the same as those of IMOPSO, as shown in Table 1. The results of the above five multi-objective optimization algorithms are statistically analyzed, as shown in Table 3. It can be seen from Table 3 that compared with the above five multi-objective optimization algorithms, the improved multi-objective particle swarm algorithm can find a better feature combination and improve the performance of the lithium-ion battery capacity prediction model. Thus, the effectiveness of the feature selection engineering of IMOPSO proposed in the present invention in the lithium-ion battery capacity prediction model is proved.

[0155] Table 3 The optimization results of other multi-objective optimization algorithms are as follows:

[0156]

[0157] This embodiment also provides an ablation experiment. In the ablation experiment, only one improvement point is used respectively, and then the algorithm is applied to the lithium-ion battery capacity prediction model based on the XGBoost algorithm for feature selection engineering, as shown in Table 4. As can be seen from Table 4, compared with using only one improvement point, using two improvement points at the same time can find a better feature combination, where IMOPSO-1 represents the global leader selection strategy based only on the Pearson correlation coefficient, and IMOPSO-2 represents the global leader profile maintenance method based only on the Pearson correlation coefficient and feature selection probability.

[0158] The ablation experiment results in Table 4 are as follows:

[0159]

[0160] Finally, this embodiment provides that after the feature selection process is performed by IMOPSO, the selected feature combinations are sequentially substituted into the lithium-ion battery capacity prediction model together with the verification data to verify the effectiveness of these feature combinations, as shown in Table 5. As can be seen from Table 5, when the verification set data is substituted into the lithium-ion battery capacity prediction model, the result after the feature combination selected by IMOPSO is better than the result without the feature selection process.

[0161] Table 5 Comparison of the results of the lithium-ion battery capacity prediction model after feature selection and without feature selection (substituting the validation set data) is as follows:

[0162]

[0163] The existing lithium-ion battery capacity prediction model still requires some production data of the capacity division process. Currently, in the lithium-ion battery production process, no lithium-ion battery capacity prediction model can replace the capacity division process. By constructing a lithium-ion battery capacity prediction model and a multi-objective particle optimization feature combination through machine learning algorithms, the prediction accuracy of the lithium-ion battery capacity prediction model is improved, so that the lithium-ion battery capacity prediction model can completely replace the capacity division process, which can greatly improve the production efficiency of lithium-ion batteries, reduce factory production costs, and achieve energy conservation and emission reduction.

[0164] Therefore, compared with the prior art, the present invention has the following advantages:

[0165] 1. It does not rely on the production data of the capacity division process. The lithium-ion battery capacity prediction model is constructed through the production data before the capacity division process, so that the lithium-ion battery capacity prediction model proposed in the present invention can completely replace the capacity division process. Secondly, IMOPSO selects excellent feature combinations to improve the prediction accuracy of the lithium-ion battery capacity prediction model, and these model combinations can provide certain reference value for industrial optimization.

[0166] 2. The lithium-ion battery capacity prediction model of the present invention is aimed at predicting the battery capacity of the lithium-ion battery at the factory, rather than the battery pole piece. Therefore, the lithium-ion battery capacity prediction model proposed by the present invention can replace the capacity division process, accurately predict the battery capacity of the lithium-ion battery at the factory, and improve production efficiency.

[0167] The present invention has strong expansibility. Since the basic properties of lithium-ion batteries such as internal resistance, battery capacity, median voltage, etc. are fixed before the capacity division process, and the electrolyte activity has been activated, the internal resistance and median voltage of lithium-ion batteries can be predicted through the production data before the capacity division process. And by adjusting the parameters or algorithm structure of the lithium-ion battery capacity prediction model, the internal resistance and median voltage can also be predicted. Therefore, the embodiment of the present invention lays the foundation for building an intelligent platform that can predict all parameters of lithium-ion batteries in the future.

[0168] In addition, if Figure 6 As shown, one embodiment of the present invention further discloses a lithium battery capacity prediction device based on XGBoost and particle swarm algorithm, the device comprising:

[0169] An acquisition module 110 is used to acquire target production data of the lithium-ion battery to be tested before the capacity division process;

[0170] The prediction module 120 is used to input the target production data into the trained lithium-ion battery capacity target prediction model to perform battery capacity prediction, and obtain the prediction result of the corresponding lithium-ion battery to be tested, wherein the lithium-ion battery capacity target prediction model is constructed based on the XGBoost algorithm and the target feature combination, and the target feature combination is selected from the production data of the lithium-ion battery before the capacity division process by adopting the improved multi-objective particle swarm algorithm.

[0171] The lithium-ion battery capacity prediction device of the embodiment of the present invention is used to execute the lithium-ion battery capacity prediction method in the above embodiment. Its specific processing process is the same as the lithium-ion battery capacity prediction method in the above embodiment, and will not be repeated here.

[0172] In addition, if Figure 7 As shown, an embodiment of the present invention also discloses an electronic device, including: at least one processor 210; at least one memory 220, for storing at least one program; when the at least one program is executed by the at least one processor 210, a lithium battery capacity prediction method based on XGBoost and particle swarm algorithm as in any of the previous embodiments is implemented.

[0173] In addition, an embodiment of the present invention further discloses a computer-readable storage medium storing computer-executable instructions for executing a lithium battery capacity prediction method based on XGBoost and particle swarm algorithm as in any of the previous embodiments.

[0174] The system architecture and application scenarios described in the embodiments of the present invention are intended to more clearly illustrate the technical solutions of the embodiments of the present invention, and do not constitute a limitation on the technical solutions provided by the embodiments of the present invention. Those skilled in the art will appreciate that with the evolution of the system architecture and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present invention are equally applicable to similar technical problems.

[0175] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0176] In hardware implementation, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0177] The terms "component", "module", "system", etc. used in this specification are used to represent computer-related entities, hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an execution thread, a program, or a computer. By way of illustration, both applications running on a computing device and a computing device can be components. One or more components may reside in a process or an execution thread, and a component may be located on a computer or distributed between two or more computers. In addition, these components may be executed from various computer-readable media having various data structures stored thereon. Components may communicate, for example, through local or remote processes based on signals having one or more data packets (e.g., data from two components interacting with another component between a local system, a distributed system, or a network, such as the Internet interacting with other systems through signals).

Claims

1. A lithium battery capacity prediction method based on XGBoost and particle swarm algorithm, characterized in that: include: Obtain target production data of the lithium-ion battery to be tested before the capacity division process; The target production data is input into a trained lithium-ion battery capacity target prediction model to perform battery capacity prediction, and a prediction result corresponding to the lithium-ion battery to be tested is obtained, wherein the lithium-ion battery capacity target prediction model is constructed based on an XGBoost algorithm and a target feature combination, and the target feature combination is selected from the production data of the lithium-ion battery before the capacity division process by using an improved multi-objective particle swarm algorithm.

2. The lithium battery capacity prediction method based on XGBoost and particle swarm algorithm according to claim 1, characterized in that: The training method of the lithium-ion battery capacity target prediction model includes: Obtain production data of lithium-ion batteries before the capacity division process; Extracting capacity data of the lithium-ion battery from the production data; Dividing the capacity data into a training set, a validation set, and a test set according to a preset ratio; Selecting feature combinations using an improved multi-objective particle swarm algorithm, building a lithium-ion battery capacity prediction model using an XGBoost algorithm, and training the lithium-ion battery capacity prediction model using each group of the feature combinations and the training set to obtain each group of the lithium-ion battery capacity prediction model; Substituting each group of the feature combination and the test set into each group of the lithium-ion battery capacity prediction model, calculating the fitness value of each group of the feature combination, and looping in sequence until the improved multi-objective particle swarm algorithm completes iteration; Determine a target feature combination according to the fitness value, substitute the target feature combination and the validation set into the lithium-ion battery capacity prediction model, calculate the fitness value of each group of the target feature combination, and select the best target feature combination; The trained lithium-ion battery capacity target prediction model is generated according to the best target feature combination.

3. The lithium battery capacity prediction method based on XGBoost and particle swarm algorithm according to claim 2, characterized in that: After extracting the capacity data of the lithium-ion battery from the production data, the method further includes: The capacity data is preprocessed, and the preprocessing includes removing the capacity data having abnormal values.

4. The lithium battery capacity prediction method based on XGBoost and particle swarm algorithm according to claim 2, characterized in that: Substituting each group of the feature combination and the test set into each group of the lithium-ion battery capacity prediction model, calculating the fitness value of each group of the feature combination, and looping in sequence until the improved multi-objective particle swarm algorithm completes the iteration, includes: Acquire characteristic data from the production data, use the characteristic data as the position vector of individual particles in the multi-objective particle swarm algorithm, and initialize the generated population with N individuals; Calculate the fitness value corresponding to the initial population individuals through the fitness function calculation formula; Initialize individual historical optimal files, global leader files, and abandoned global leader files; The improved multi-objective particle swarm optimization algorithm is used to update the feature combination and calculate the fitness value of each particle; Update the individual historical optimal file based on the individual historical optimal file of the previous iteration and the population of this iteration; Based on the global leader profile of the previous iteration and the population of this iteration, the global leader profile is updated using the Pareto dominance principle; The global leader profile is maintained according to the global leader profile maintenance method based on Pearson correlation coefficient and feature selection probability; When the iterative termination condition of the improved multi-objective particle swarm algorithm is met, the final feature combination and the corresponding fitness value are output.

5. The lithium battery capacity prediction method based on XGBoost and particle swarm algorithm according to claim 4, characterized in that: The fitness value corresponding to the initial population individual is calculated by the fitness function calculation formula, including: Discretize the particle population and select feature combinations based on the discretization results; The feature combination selected by each particle is substituted into the lithium-ion battery capacity prediction model to predict the lithium-ion battery capacity of the test set, and the fitness value of each particle is calculated.

6. The lithium battery capacity prediction method based on XGBoost and particle swarm algorithm according to claim 4, characterized in that: The improved multi-objective particle swarm optimization algorithm is used to update the feature combination and calculate the fitness value of each particle, including: Calculate the Pearson correlation coefficient between each feature and lithium-ion battery capacity; Calculate the feature combination and the sum of Pearson correlation coefficients for each global leader; Re-rank all global leader particles from small to large according to the sum of the Pearson correlation coefficients of each global leader to obtain a re-ranked global leader profile; According to the iteration state of the improved multi-objective particle swarm algorithm and the sum of the Pearson coefficients of the global leader particles, a global leader particle is selected for each population particle; The corresponding feature combination is obtained according to the position vector of the population particle, and the feature combination and the test set are substituted into the lithium-ion battery capacity prediction model to calculate the fitness value of each particle.

7. The lithium battery capacity prediction method based on XGBoost and particle swarm algorithm according to claim 4, characterized in that: The method for maintaining a global leader profile based on a global leader profile maintenance method based on a Pearson correlation coefficient and a feature selection probability includes: Calculate the number of particles Kn in the global leader file of this iteration, and merge the global leader file of this iteration and the abandoned global leader file into population P1; Calculate the quality index of each particle in population P1; Randomly select Kn particles from population P1 twice to form two parent populations, and obtain the quality indicators corresponding to the Kn particles at the same time; The quality indicators of the two parent populations are compared with each other, and the corresponding simulated binary crossover formula is selected according to the quality indicators to perform crossover operations to obtain sub-populations; Perform polynomial mutation operation on the sub-population to obtain the final sub-population P2; Obtaining a corresponding feature combination according to the position vector of the final sub-population P2, substituting it into the lithium-ion battery capacity prediction model, and calculating the fitness value of the sub-population; Based on the Pareto dominance principle, all target particles with non-dominated solutions are selected from population P1 and the final sub-population P2. If the number of target particles exceeds N, N particles are selected in descending order according to the Manhattan crowding distance to form the global leader file of this iteration; if the number of target particles does not exceed N, the global leader file of this iteration is composed of target particles.

8. A lithium battery capacity prediction device based on XGBoost and particle swarm algorithm, characterized in that: The device comprises: An acquisition module, used to acquire target production data of the lithium-ion battery to be tested before the capacity division process; A prediction module is used to input the target production data into a trained lithium-ion battery capacity target prediction model to perform battery capacity prediction, and obtain a prediction result corresponding to the lithium-ion battery to be tested, wherein the lithium-ion battery capacity target prediction model is constructed based on an XGBoost algorithm and a target feature combination, and the target feature combination is selected from the production data of the lithium-ion battery before the capacity division process by using an improved multi-objective particle swarm algorithm.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the lithium battery capacity prediction method based on XGBoost and particle swarm algorithm as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the lithium battery capacity prediction method based on XGBoost and particle swarm algorithm as described in any one of claims 1 to 7.

Citation Information

Cited By

  • Lithium battery capacity grading capacity prediction method

    CN119916213A

  • Lithium battery capacity prediction method and system based on GTSFormer model

    CN121167661A

  • A lithium battery capacity prediction method and system based on the GTSFormer model

    CN121167661B