Lithium battery capacity prediction method based on short-term charging data and enhanced whale optimization algorithm

Abstract

The embodiment of the invention discloses a lithium battery capacity prediction method based on short-term charging data and an enhanced whale optimization algorithm, which can accurately predict the battery capacities of different types of batteries, improve the capacity prediction accuracy and reduce the training data volume at the same time. The method comprises the steps that in the cyclic charging and discharging test process of different types of batteries, charging and discharging data of the batteries are collected and recorded in real time, short-term charging data are extracted, and the charging and discharging data comprise charging time, voltage, current and capacity data; the short-term charging data is obtained by performing spline interpolation processing on the charging voltage and current data within a preset short-term charging duration; based on an enhanced whale optimization algorithm, battery aging characteristics are extracted from short-term charging data through iterative optimization, and the enhanced whale optimization algorithm is obtained by introducing a pooling mechanism, a migration search strategy, a preferential selection search strategy and a surrounding prey search strategy into a classic whale optimization algorithm. The battery aging characteristic is the voltage increment of the voltage interval corresponding to the charging time interval with the strongest correlation with the battery capacity; and for different types of batteries or the same type of batteries, charging and discharging data and aging characteristics are input into a Gaussian process regression model to predict a battery capacity prediction value, and the Gaussian process regression model is constructed based on a Gaussian process regression algorithm.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of lithium battery technology, and in particular to a lithium battery capacity prediction method based on charging data and an enhanced whale algorithm. Background Art

[0002] As a clean and efficient energy storage solution, lithium batteries are widely used in electric vehicles and have a wide application market. As the use time and frequency of use increase, the capacity of lithium batteries will decay, affecting the reliability of the battery. Therefore, accurately estimating the battery capacity is of great significance to improving the safety of electric vehicles and fully utilizing their performance.

[0003] Machine learning algorithms are also gradually being used to predict battery capacity. However, the current machine learning method is to establish a corresponding prediction model based on the battery type, and use a large amount of data for data training to predict the battery capacity. This method not only requires a large amount of training data and high requirements for its characteristic parameters, but also when the battery type changes, the prediction model cannot be used and needs to be re-modeled and trained. Therefore, in the battery capacity prediction solution, how to improve the universal applicability of the prediction model and reduce the amount of training data while accurately predicting the battery capacity has become a key issue that needs to be solved urgently. Summary of the invention

[0004] In order to overcome the above technical defects, an embodiment of the present invention provides a lithium battery capacity prediction method based on short-term charging data and an enhanced whale optimization algorithm, which can accurately predict the battery capacity of different types of batteries, improve the accuracy of capacity prediction and reduce the amount of training data.

[0005] A lithium battery capacity prediction method based on short-term charging data and an enhanced whale optimization algorithm is provided in an embodiment of the present invention, comprising:

[0006] During the cyclic charge and discharge test of different types of batteries, the battery charge and discharge data are collected and recorded in real time, and short-term charging data are extracted. The charge and discharge data include charging time, voltage, current and capacity data. The short-term charging data is obtained by performing spline interpolation processing on the charging voltage and current data within the preset short-term charging time.

[0007] The battery aging characteristics are extracted from short-term charging data through iterative optimization based on the enhanced whale optimization algorithm. The enhanced whale optimization algorithm is improved by introducing pooling mechanism, migration search strategy, priority selection search strategy and encirclement prey search strategy into the classic whale optimization algorithm. The battery aging characteristics are the voltage increment of the voltage interval corresponding to the charging time interval with the strongest correlation with the battery capacity.

[0008] For different types of batteries or batteries of the same type, the charge and discharge data and aging characteristics are input into the Gaussian process regression model to predict the battery capacity prediction value, wherein the Gaussian process regression model is constructed based on the Gaussian process regression algorithm.

[0009] Optionally, in a possible implementation of the above-mentioned lithium battery capacity prediction method, the specific steps of setting the parameters of the enhanced whale optimization algorithm include: taking the Spearman correlation coefficient of the voltage increment and the battery capacity as the objective function, and defining the optimization target as minimizing the Spearman correlation coefficient of the voltage increment and the battery capacity; setting the optimization object as the upper and lower limits of the charging time interval; setting the enhanced whale optimization algorithm parameters: the number of search agents, the number of decision variables, the maximum number of iterations, the number of executions, and the upper and lower limits of the optimization object.

[0010] Optionally, in a possible implementation of the above-mentioned lithium battery capacity prediction method, the specific steps of iterative optimization include: based on the enhanced whale optimization algorithm, randomly initializing a group of whales in the hyperparameter space, that is, the candidate solutions of the algorithm; extracting the voltage increment corresponding to the upper and lower limits of the charging time interval represented by each whale, and calculating the Spearman correlation coefficient between the voltage increment and the battery capacity, and evaluating the algorithm performance with the fitness function; updating the position of the whale, that is, the upper and lower limits of the charging time interval according to the rules of the enhanced whale optimization algorithm; continuously updating the position of the whale until the maximum number of iterations is reached, obtaining the global optimal solution for the charging time interval with the strongest correlation with the battery capacity, finding the voltage increment of the corresponding voltage interval and using it as an aging feature.

[0011] Optionally, in a possible implementation of the above lithium battery capacity prediction method, the pooling mechanism is defined as: given a matrix of size k, Pool = (p1, p2, ..., p k ), at the end of each iteration, the element P in the matrix Pool is generated i =(p i,1 ,p i,2 ,...,p i,D ), element P i The resulting formula is as follows:

[0012]

[0013] Where P i t represents the element of the tth iteration, is the binary random vector in the tth iteration, is its inverse vector, is the worst solution in the tth iteration, It represents the optimal solution generated in the tth iteration A random location in the vicinity of the corresponding whale.

[0014] Optionally, in a possible implementation of the above lithium battery capacity prediction method, the working principle of the migration search strategy can be summarized as follows:

[0015]

[0016] Where: represents the position of the i-th whale in the t+1th iteration; represents the random position of the whale in the search space in the tth iteration; rand is a random number between [0,1) and it follows a uniform distribution; δ max and δ min Corresponding to the upper and lower bounds of the problem respectively; δ best-max and δ best-min Then they correspond to the optimal solutions The upper and lower bounds of .

[0017] Optionally, in a possible implementation of the above lithium battery capacity prediction method, the working principle of the priority selection search strategy can be summarized as follows:

[0018]

[0019] Where: represents the position of the i-th whale in the t-th iteration; and Both are positions 1 and 2 randomly selected from the matrix Pool in the tth iteration; MaxIt represents the maximum number of iterations set in the enhanced whale optimization algorithm.

[0020] Optionally, in a possible implementation of the above lithium battery capacity prediction method, the working principle of the encirclement prey search strategy can be summarized as the following equation:

[0021]

[0022] Where: is the position 3 randomly selected from the matrix Pool in the tth iteration.

[0023] Optionally, in a possible implementation of the above-mentioned lithium battery capacity prediction method, the specific steps of training the Gaussian process regression model include: using the extracted first 15% aging characteristics of a battery and the battery capacity data as a model training set, using the training set to run the Gaussian process regression model to find correlations, and saving the resulting model.

[0024] Optionally, in a possible implementation of the above-mentioned lithium battery capacity prediction method, it is characterized in that, for different batteries of different types, the specific steps of battery capacity prediction include: inputting the first 15% battery capacity data and aging characteristics collected from a certain battery into a Gaussian process regression model for model training, and inputting the last 85% battery capacity data and aging characteristics collected from the battery into the trained Gaussian process regression model for model verification, and all characteristics of other batteries of the same type are input into the trained Gaussian process regression model to obtain the battery capacity prediction value of the battery.

[0025] Optionally, in a possible implementation of the above-mentioned lithium battery capacity prediction method, for batteries of the same type, the specific steps of battery capacity prediction include: for different batteries of the same type, the last 85% of the battery capacity data obtained during the battery cycle charge and discharge test and the aging characteristics of the battery are directly input into the trained Gaussian process regression model to obtain the battery capacity prediction value of the corresponding battery.

[0026] The above technical solution can achieve the following technical effects in the embodiments of the present invention:

[0027] In the lithium battery capacity prediction method based on short-term charging data and enhanced whale optimization algorithm in the embodiment of the present invention, short-term charging data is first obtained by interpolating the charging voltage and current data within the short-term charging time. Secondly, the classic whale optimization algorithm is improved by introducing a pooling mechanism, a migration search strategy, a priority search strategy and a surrounding prey search strategy, so that the algorithm's search ability and the ability to jump out of the local optimum are improved, and the local search ability and the global search ability of the algorithm are balanced. Therefore, when the enhanced whale optimization algorithm is used to extract battery aging features from short-term charging data, the extracted battery aging features can be made more correlated with the battery capacity. The battery aging features extracted using this method can improve the prediction accuracy of battery capacity prediction. At the same time, since the features are extracted from the charging current and voltage data within the short-term charging time of different similar batteries, the prediction method can also predict the battery capacity of different types of batteries, and at the same time can reduce the amount of training data required for prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings exemplarily illustrate the embodiments and constitute a part of the specification, and together with the text description of the specification, are used to explain the exemplary implementation of the embodiments. The embodiments shown are for illustrative purposes only and do not limit the scope of the claims. In all drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0029] Figure 1 A schematic diagram of a flow chart of a lithium battery capacity prediction method based on short-term charging data and an enhanced whale optimization algorithm in Example 1;

[0030] Figure 2 A flow chart of parameter setting for enhancing the whale optimization algorithm in Example 2;

[0031] Figure 3 A schematic diagram of a process of iterative optimization in Example 3;

[0032] Figure 4 A graph showing the relationship between the selected characteristics and capacity aging at different cycle times in Example 5;

[0033] Figure 5 A diagram showing the prediction results of the test set after training with the first 15% of the data of cell 1 in Example 5;

[0034] Figure 6 A diagram showing the prediction results of similar batteries after training using the first 15% data of cells 2 and 3 in Example 5;

[0035] Figure 7 A diagram showing prediction results of applying the method in Embodiment 5 to different types of batteries;

[0036] Figure 8 A schematic diagram of the structure of the lithium battery capacity prediction system in Example 6;

[0037] Fig. 9 The figure is a hardware structure diagram of a computer device suitable for implementing the lithium battery capacity prediction method in the seventh embodiment. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical scheme 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 intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0039] It should be noted that the descriptions involving "first", "second", etc. in the embodiments of the present invention are only for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0040] In order to facilitate understanding of the technical solutions of the embodiments of the present invention, the technical solutions of the embodiments of the present invention will be described below with reference to the accompanying drawings, as follows:

[0041] Embodiment 1

[0042] See also Figure 1 The lithium battery capacity prediction method based on short-term charging data and enhanced whale optimization algorithm in the embodiment of the present invention is described as follows:

[0043] S1. During the cyclic charge and discharge test of different types of batteries, the battery charge and discharge data is collected and recorded in real time and short-term charging data is extracted.

[0044] Specifically, the charge and discharge data include charging time, voltage, current and capacity data. For example, different types of lithium-ion batteries are subjected to cyclic charge and discharge tests until the discharge capacity of each type of battery is less than 80% of the nominal capacity, and the battery charge and discharge data are recorded in real time. Short-term charging data is obtained by performing spline interpolation processing on the charging voltage and current data within a preset short-term charging time. For example, the number of spline interpolations can be three times, such as: performing three spline interpolations on the recorded voltage and current data with a sampling interval of 30 seconds, so that the interval between every two data points is 1 second.

[0045] S2. Based on the enhanced whale optimization algorithm, battery aging characteristics are extracted from short-term charging data through iterative optimization.

[0046] Specifically, the enhanced whale optimization algorithm is obtained by introducing a pooling mechanism, a migration search strategy, a priority search strategy and a prey encirclement search strategy into the classic whale optimization algorithm. The battery aging characteristic is the voltage increment of the voltage interval corresponding to the charging time interval with the strongest correlation with the battery capacity.

[0047] The above battery aging characteristics can be extracted from the charge and discharge data interpolated 30 minutes before the start of charge and discharge, where the battery capacity can be recorded as Q, and the charging time interval with the strongest correlation with the battery capacity, the corresponding voltage interval and the voltage increment are recorded as [X best1 X best2 ]s,[Y best1 Y best2 ]V and ΔY best .

[0048] S3. For different types of batteries or batteries of the same type, the charge and discharge data and aging characteristics are input into the Gaussian process regression model to predict the battery capacity prediction value.

[0049] Specifically, the Gaussian process regression model is constructed based on the Gaussian process regression algorithm.

[0050] The training process of the Gaussian process regression model in embodiment S3 of the present invention specifically includes the following steps: using the extracted first 15% aging characteristics of a battery and the battery capacity data as a model training set, using the training set to run the Gaussian process regression model to find correlations, and saving the resulting model.

[0051] Further optionally, the specific technical solution for battery capacity prediction in embodiment S3 of the present invention is as follows:

[0052] On the one hand, for different types of different batteries, the specific steps of battery capacity prediction include: inputting the first 15% battery capacity data and aging characteristics collected from a certain battery into the Gaussian process regression model for model training, and inputting the last 85% battery capacity data and aging characteristics collected from the battery into the trained Gaussian process regression model for model verification, and inputting all characteristics of other batteries of the same type into the trained Gaussian process regression model to obtain the battery capacity prediction value of the battery.

[0053] On the other hand, for batteries of the same type, the specific steps for predicting the battery capacity include: for different batteries of the same type, the last 85% of the battery capacity data obtained during the battery cycle charge and discharge test and the aging characteristics of the battery are directly input into the trained Gaussian process regression model to obtain the battery capacity prediction value of the corresponding battery.

[0054] In the lithium battery capacity prediction method based on short-term charging data and enhanced whale optimization algorithm in the embodiment of the present invention, short-term charging data is first obtained by interpolating the charging voltage and current data within the short-term charging time. Secondly, the classic whale optimization algorithm is improved by introducing a pooling mechanism, a migration search strategy, a priority search strategy and a surrounding prey search strategy, so that the algorithm's search ability and the ability to jump out of the local optimum are improved, and the local search ability and the global search ability of the algorithm are balanced. Therefore, when the enhanced whale optimization algorithm is used to extract battery aging features from short-term charging data, the extracted battery aging features can be made more correlated with the battery capacity. The battery aging features extracted using this method can improve the prediction accuracy of battery capacity prediction. At the same time, since the features are extracted from the charging current and voltage data within the short-term charging time of different similar batteries, the prediction method can also predict the battery capacity of different types of batteries, and at the same time can reduce the amount of training data required for prediction.

[0055] Embodiment 2

[0056] See also Figure 2 The parameter setting of the enhanced whale optimization algorithm in S2 in the embodiment of the present invention is described, which specifically includes the following steps:

[0057] S2.1. Taking the Spearman correlation coefficient between the voltage increment and the battery capacity as the objective function, the optimization objective is defined as minimizing the Spearman correlation coefficient between the voltage increment and the battery capacity.

[0058] That is, the Spearman correlation coefficient between ΔY and Q is used as the objective function, and the optimization goal is defined as minimizing the Spearman correlation coefficient between ΔY and Q.

[0059] S2.2, setting the optimization object as the upper limit and lower limit of the charging time interval, wherein the upper limit and the lower limit can be recorded as X1 and X2 respectively.

[0060] S2.3. Set the enhanced whale optimization algorithm parameters: the number of search agents, the number of decision variables, the maximum number of iterations, the number of executions, and the upper and lower limits of the optimization object.

[0061] First of all, the key point involved in this step is how to achieve effective prediction of lithium battery capacity through specific numerical indicators, which is based on an objective function and corresponding parameter settings. The core of taking the Spearman correlation coefficient of voltage increment and battery capacity as the objective function is to define the best result to be achieved in the optimization process. In this case, it is to minimize this relationship as much as possible to reduce the correlation between them, which means seeking the prediction independence of voltage increment for battery capacity under given conditions. To achieve this, we set the optimization goal to minimize the Spearman correlation coefficient of voltage increment and battery capacity.

[0062] Next, it is very important to consider the various possible interval distribution characteristics of the battery charging process and choose a suitable range for constraint; this constitutes the second important link, namely, determining the optimization object. In this link, we focus on the definition of the charging time interval of the lithium battery, and regard the setting of the upper and lower limits as a means of adjusting the efficiency of the entire system to avoid ineffective or overcharging.

[0063] In the specific optimization parameter design process (such as the parameter setting of the enhanced whale optimization algorithm), these values ​​directly affect the behavior of the algorithm and whether it can successfully converge to the ideal solution set: including the number of search agents that determine the size of the search space; the number of decision variables that specify the dimensions of the space; and the number of cycles to ensure that the search process has enough time for effective iterations. In addition, operations may be repeated repeatedly during the execution of the algorithm to explore the global optimal solution in more detail. In addition, boundary conditions for optimization variables are set so that the problem can be solved within a physically feasible and effective search range.

[0064] Embodiment 3

[0065] See also Figure 3The specific steps of iterative optimization in S2 in the embodiment of the present invention are described as follows, which specifically include the following steps:

[0066] S2.4. Based on the enhanced whale optimization algorithm, a group of whales, i.e., candidate solutions of the algorithm, are randomly initialized in the hyperparameter space.

[0067] S2.5. Extract the voltage increments corresponding to the upper and lower limits of the charging time interval represented by each whale, calculate the Spearman correlation coefficient between the voltage increment and the battery capacity, and use the fitness function to evaluate the algorithm performance.

[0068] S2.6, updating the position of the whale according to the rules of the enhanced whale optimization algorithm, that is, the upper and lower limits of the charging time interval;

[0069] S2.7. Continuously update the position of the whale until the maximum number of iterations is reached, obtain the global optimal solution for the charging time interval that is most correlated with the battery capacity, find the voltage increment of the corresponding voltage interval and use it as the aging feature.

[0070] First, before starting the iterative optimization, it is necessary to randomly generate a group of initial candidate solutions in a predetermined lithium battery hyperparameter space, that is, randomly initialize a group of whales. Specifically, each whale is a position point in the coordinate representation within a specific search space; these coordinates may represent the charging voltage or other related optimization hyperparameters in a certain time interval or certain time intervals during the battery charging process. In one embodiment, in a prediction scheme for an electric vehicle battery management system, these hyperparameters may refer to different stages of the charging process, and their value range can be limited according to the maximum charging power and voltage limit of the lithium battery.

[0071] Next, the voltage increments of these initialized whales are calculated during the charging time intervals, and used to solve the Spearman correlation coefficient between the battery voltage change and the capacity estimation. The Spearman rank correlation coefficient is a non-parametric method for measuring the statistical correlation between two variables. It is suitable for situations where the data is of sequential type or the data has been transformed into its ranking form. Here, it is to determine whether there is a linear relationship between the voltage change sequence and the estimated battery capacity value sequence. .

[0072] After completing the fitness value evaluation, the next step is to perform a position or direction update for each current virtual whale's position or state, which refers to the hyperparameter value, according to the rules of the enhanced whale optimization algorithm. This means that positions that do not have good performance or low fitness values ​​(such as the Spearman coefficient) will be gradually discarded or moved to areas that are more likely to contain good solutions to seek improved solutions, thereby approaching the algorithm convergence conditions. A specific example is that when a set of experiments shows that the actual available energy level of the battery obtained under a certain time partition charging strategy is lower than expected, the corresponding parameters will be corrected according to a certain mechanism, and subsequent searches will be more focused on those control schemes that can theoretically produce larger battery capacity or at least will not reduce the original energy storage.

[0073] The above calculations need to be repeated and parameters adjusted throughout the process until the pre-defined termination conditions are met. For example, when the fitness value converges to a sufficiently stable value or the number of iterations reaches the upper limit, this means that the entire search optimization process is basically completed and the final recommended optimal charging solution can be obtained, which is used as the basis for improving the accuracy of the lithium battery capacity prediction method.

[0074] Embodiment 4

[0075] The following further describes the enhanced whale optimization algorithm after introducing the pooling mechanism, migration search strategy, optimization selection search strategy and surrounding prey search strategy in Example S2 of the present invention, as follows:

[0076] First: Improvement of the whale optimization algorithm based on the pooling mechanism

[0077] The pooling mechanism uses the worst solution in the current iteration and a randomly generated promising solution to crossover at the end of each iteration to increase population diversity. This mechanism improves the algorithm's search ability by mixing the worst solution with promising solutions, avoiding premature convergence and local optimal traps.

[0078] In the embodiment of the present invention, the pooling mechanism is defined as:

[0079] Given a matrix of size k, Pool = (p1, p2, ..., p k ), at the end of each iteration, the element P in the matrix Pool is generated i =(p i,1 ,p i,2 ,...,p i,D ), element P i The resulting formula is as follows:

[0080]

[0081] Where Pi t represents the element of the tth iteration, is the binary random vector in the tth iteration, is its inverse vector, is the worst solution in the tth iteration, It represents the optimal solution generated in the tth iteration A random location in the vicinity of the corresponding whale.

[0082] Second: Improvement of the whale optimization algorithm based on migration search strategy

[0083] The migration search strategy increases exploration ability and population diversity by randomly separating a portion of whales to cover unvisited areas. This strategy improves the algorithm's ability to escape local optimality by introducing randomness.

[0084] In the embodiment of the present invention, the working principle of the migration search strategy can be summarized as follows:

[0085]

[0086] Where: represents the position of the i-th whale in the t+1th iteration; represents the random position of the whale in the search space in the tth iteration; rand is a random number between [0,1) and it follows a uniform distribution; δ max and δ min Corresponding to the upper and lower bounds of the problem respectively; δ best-max and δ best-min Then they correspond to the optimal solutions The upper and lower bounds of .

[0087] This step ensures that the search scope is not limited to positive direction exploration at each update, increasing search flexibility.

[0088] Next, the worst solution, that is, the inferior individual in the current search space, is determined. The purpose is to avoid wasting resources in subsequent iterations to repeatedly search directions that have been proven to be poor. This helps the algorithm focus on the improvement space more intelligently. Finally, a random position adjacent to the optimal solution is generated, which means that the search individual makes further random exploration based on the existing successful cases.

[0089] Specifically, in one embodiment, suppose we are trying to use the enhanced whale optimization algorithm to improve the capacity prediction accuracy of lithium batteries; suppose we have accumulated a large amount of battery short-time charging data as reference samples through previous generation searches. Then, before we are ready to start a new iteration, we select the most failed solution and the best solution according to the solution after the previous round of evaluation. The most failed position represents the least credible solution among all the tested lithium battery capacity estimates; then we will select a random position around the most successful prediction value as the new starting point for the next update solution.

[0090] Third: Improvement of the whale optimization algorithm based on optimized selection search strategy

[0091] In order to improve the exploration ability of the classic whale optimization algorithm's method of searching for prey, the enhanced whale optimization algorithm adopts a preferential selection search strategy, which discovers diverse solutions by dispersing whales in different areas of the search space, so a larger step size is required. This strategy uses a heavy-tailed Cauchy distribution, which has a very high probability of producing large values.

[0092] In the embodiment of the present invention, the working principle of the priority selection search strategy can be summarized as follows:

[0093]

[0094] Where: represents the position of the i-th whale in the t-th iteration; and Both are positions 1 and 2 randomly selected from the matrix Pool in the tth iteration; MaxIt represents the maximum number of iterations set in the enhanced whale optimization algorithm.

[0095] First, determine the current iteration number i and the position vector Xi of all whales, which need to be updated at each iteration during the algorithm. Select two random position vectors Xr1 and Xr2 (i.e., position 1 and position 2), which are randomly selected from the elements of the matrix M generated in the search space containing all possible solutions. In addition, set the maximum number of iterations of the algorithm, MaxIt.

[0096] In this process, each individual (i.e. a virtual whale) corresponds to a potential solution and tries to gradually approach the global optimal point through iteration. After each cycle update, the individual will evaluate its own adaptability and compare it with the current best position in order to make updates.

[0097] In the specific implementation, for example, in order to optimize the parameters in the lithium battery capacity estimation model, we can assume that in the initial state, there are 100 virtual whales representing 100 sets of candidate parameters. The combination of these parameters determines the accuracy of the battery charging mode prediction. Each iteration means simulating a prediction operation on a batch of newly arrived charging and discharging experimental data sets in a real environment and adjusting the corresponding whale group layout according to the model output until the predetermined maximum iteration limit (such as MaxIt = 50) is met. At this time, if Xi is taken as the position vector of an individual selected for analysis at a certain moment, it will change during the entire solution stage due to the influence of two completely random samples Xr1 and Xr2; this selection strategy allows some of the better but not absolutely top options to have the opportunity to be inherited and carried forward instead of just blindly following the strongest.

[0098] The motivation for adopting this approach is to avoid the search process falling into local optimality early and missing more efficient configuration points in the global scope. The reason for setting the formula is mainly to be able to better balance the conflict between using historical accumulated information and exploring unfamiliar areas, and to further improve the ability to adapt to uncertain scenarios on the basis of inheriting traditional experience while also enhancing the overall learning efficiency of the system. Therefore, the rules given here are essentially a mathematical reproduction of the predation habits in the animal kingdom in real life (such as the collaborative hunting method of killer whale groups). Under this setting, the optimal value changes dynamically depending on factors such as the difference in the nature of the task and its corresponding accuracy target requirements.

[0099] Fourth: Improvement of whale optimization algorithm based on encirclement prey search strategy

[0100] The rich surrounding prey search strategy improves the surrounding prey method in the classic whale optimization algorithm by adding a randomly selected pool matrix member to balance the local search capability and the global search capability.

[0101] In the embodiment of the present invention, the working principle of the encirclement prey search strategy can be summarized as the following equation:

[0102]

[0103] Where: is the position 3 randomly selected from the matrix Pool in the tth iteration.

[0104] This method is to improve the exploration ability of individuals in the whale optimization algorithm by increasing the diversity of the algorithm and improving the convergence of the algorithm. The core equation in this step represents a model of how the encirclement strategy randomly selects new positions in the search space. Under this step, each algorithm iteration will evaluate the position of the existing predator or solution, and then select a new position in the search space according to the principles of algorithm design. For example, in one embodiment, consider a lithium battery capacity prediction method based on short-term charging data and an enhanced whale optimization algorithm. Our matrix may contain the measured values ​​of the battery capacity at different periods and the related charging state, and the randomly selected position represents a feasible battery capacity prediction value or a parameter combination that the algorithm considers as a new approximation to the solution to the prediction problem at a given iteration.

[0105] Embodiment 5

[0106] See next Figure 4-7 The specific implementation methods and technical effects corresponding to the technical solutions of the present invention are described as follows:

[0107] In the embodiment of the present invention, three nickel-cobalt-manganese batteries with a rated capacity of 4Ah and four soft-pack batteries with a rated capacity of 0.74Ah are used as the test objects, and the above seven lithium batteries are named cell 1-7 respectively, which specifically includes the following steps:

[0108] 1) Data collection:

[0109] Conduct cyclic charge and discharge tests on different types of lithium-ion batteries until the discharge capacity of each type of battery is less than 80% of the nominal capacity, and record the charge and discharge data in real time.

[0110] 2) Data processing:

[0111] The charging voltage and current data obtained in 1) are processed, and the voltage and current data recorded with a sampling interval of 30 seconds are interpolated by cubic spline, so that the interval between every two data points is 1 second.

[0112] 3) Aging feature extraction based on enhanced whale optimization algorithm:

[0113] The relationship between charging voltage and battery aging is analyzed, and the voltage increment corresponding to a certain charging time interval is used as the aging feature. The feature with the strongest correlation with capacity is extracted from the charging data after interpolation of the first 30 minutes.

[0114] Figure 4The relationship between the selected features and capacity aging under different cycle numbers, where X1 and X2 correspond to the upper and lower limits of the optimal charging time interval, respectively, and the voltage increment corresponding to the charging time interval, that is, the feature is ΔY. Among them, the voltage increment corresponding to the charging time interval [X1X2] in the first cycle is ΔY1, with a value of 174.4mV; the voltage increment corresponding to the charging time interval [X1X2] in the 301st cycle is ΔY1, with a value of 216.8mV; the voltage increment corresponding to the charging time interval [X1X2] in the 6011th cycle is ΔY1, with a value of 278.2mV. It can be observed that as the number of battery cycles increases, the voltage increment corresponding to the charging time interval [X1X2] continues to increase.

[0115] 4) Enhanced whale optimization algorithm parameter settings:

[0116] The enhanced whale optimization algorithm is applied to the optimization of aging characteristics. The Spearman correlation coefficient between the voltage increment and the battery capacity corresponding to a certain charging time interval is used as the objective function, and the population size and initialization parameters are set. The starting point and end point of the charging time interval are specified as the optimization target. The parameters of the enhanced whale optimization algorithm are shown in Table 1:

[0117] Table 1 Grey correlation coefficient of cell 1-7

[0118] Parameter name Search Agent Quantity Number of decision variables Maximum number of iterations Execution times Upper Bound The Nether Settings 100 2 200 5 [0,1] [1799,1800]

[0119] 5) Iterative optimization and optimal feature output:

[0120] Calculate the corresponding fitness value, and the minimum fitness value is the optimal solution for this iteration. Compare the calculation result with the global optimal solution. If the calculation result is better than the global optimal solution, replace the global optimal solution with the calculation result of this calculation. When the maximum number of iterations is reached, the optimization stops. Calculate the starting point and end point of the optimal charging time interval, and use the voltage increment corresponding to this charging time interval as the aging feature of the model. The calculation results of 5 executions are: [X best1 X best2 ]s=[1425,1789]s, and the corresponding fitness is -0.995097. It can be seen that the correlation between the solved features and the capacity is very high.

[0121] 6) Single cell battery capacity prediction:

[0122] The capacity data obtained in 2) and the optimal aging features obtained in 5) are used for model training. The training set is the first 15% feature data of cell 1, that is, the feature data of the 1st to 103rd cycles. The trained model is saved, and the last 85% feature data, that is, the 104th to 689th cycles, are input into the model to obtain the capacity prediction value of the corresponding battery. The results are as follows Figure 5As shown, the solid line in the figure represents the capacity measurement value, the dotted line represents the capacity prediction value, the training data is the first 15% of the solution features and battery capacity values, the prediction data is the solution features of the last 85% of the cycles, and the solution result is the battery capacity. From the results, it can be seen that the maximum error of the prediction value of the present invention is less than 2.5%, and the maximum error of more than 93.5% of the prediction values ​​is controlled within 1%.

[0123] 7) Model generalization verification:

[0124] For different batteries of the same type, the optimal features of cell2-3 obtained in 5) are directly input into the model saved in 6) to obtain the capacity prediction value of the corresponding battery; the result is as follows Figure 6 As shown in Table 2, Figure 6 The solid line in the middle represents the measured capacity value, and the dotted line represents the predicted capacity value. From the results, we can see that the R-squared (R 2 ), maximum absolute error (MAE), mean absolute error (MSE), root mean square error (RMSE) and maximum error (Max) are within 0.992, 1.513%, 0.462%, 1.904% and 2.49% respectively, which verifies that the proposed method only needs short-term low-precision charging data of the power battery and only needs to train 15% of the data of a certain battery to accurately and stably predict the capacity of this type of battery, indicating the high efficiency and strong prediction performance of the method, which makes it show broad application prospects in the field of battery health status prediction.

[0125] For different types of batteries, similarly, the capacity data obtained in 2) and the optimal aging features obtained in 5) are used for model training. The training set is the first 15% feature data of cell 4. The trained model is saved, and the last 85% feature data of cell 4 and all feature data of cells 5-7 are input into the model to obtain the capacity prediction value of the corresponding battery. The results are as follows: Figure 7 As shown in Table 2, Figure 7 The solid line in the middle represents the capacity measurement value, and the dotted line represents the capacity prediction value. From the results, it can be seen that the maximum error of the present invention is less than 1% and the R-squared is 0.999, which shows that the proposed method has good prediction accuracy and generalization.

[0126] Table 2 Prediction error of cell 2-5

[0127] cell <![CDATA[R 2 ]]> MAE(%) AAE (%) RMSE(%) Max(%) 2 0.993 1.342 0.416 1.766 2.49 3 0.992 1.513 0.462 1.904 2.21 4 0.999 0.126 0.216 0.141 -0.46 5 0.999 0.145 0.230 0.187 -0.99 6 0.999 0.119 0.191 0.144 -0.63 7 0.999 0.128 0.209 0.166 -0.89

[0128] In the embodiment of the present invention, the voltage increment of a specific charging time interval is extracted from the charging voltage curve of the first 30 minutes as a feature, and the enhanced whale optimization algorithm is used to achieve the strongest correlation feature extraction, and the battery capacity prediction model is built using the Gaussian process regression algorithm. Only the first 15% of the data of a certain battery is used for training, and the relative error of the capacity prediction result of this type of battery can be controlled within 2.5%, which greatly reduces the number of training sets while ensuring the prediction accuracy. On this basis, the capacity prediction values ​​of different types of batteries are obtained using this method, and the relative error of the capacity prediction result of this type of battery is controlled within 1%, which verifies that the constructed model has good prediction accuracy and generalization ability.

[0129] Embodiment 6

[0130] See also Figure 8 The lithium battery capacity prediction system in the embodiment of the present invention is described as follows:

[0131] Figure 8 The block diagram of the lithium battery capacity prediction system according to the embodiment of the present invention is schematically shown. The device can be divided into one or more program modules, and one or more program modules are stored in a storage medium and executed by one or more processors to complete the embodiment of the present invention. The program module referred to in the embodiment of the present invention refers to a series of computer program instruction segments that can complete specific functions. The following description will specifically introduce the functions of each program module in this embodiment.

[0132] like Figure 8 As shown, the lithium battery capacity prediction system 80 may include: a data acquisition module 801, a feature extraction module 802 and a battery capacity prediction module 803, wherein:

[0133] The data acquisition module 801 is used to collect and record the battery charge and discharge data in real time and extract short-term charging data during the cyclic charge and discharge test of different types of batteries. The charge and discharge data include charging time, voltage, current and capacity data. The short-term charging data is obtained by performing spline interpolation processing on the charging voltage and current data within a preset short-term charging time.

[0134] The feature extraction module 802 is used to extract the battery aging feature from the short-term charging data through iterative optimization based on the enhanced whale optimization algorithm, wherein the enhanced whale optimization algorithm is obtained by introducing a pooling mechanism, a migration search strategy, a priority selection search strategy and a surrounding prey search strategy into the classic whale optimization algorithm, and the battery aging feature is the voltage increment of the voltage interval corresponding to the charging time interval with the strongest correlation with the battery capacity;

[0135] The battery capacity prediction module 803 is used to: for different types of batteries or the same type of batteries, input the charge and discharge data and aging characteristics into the Gaussian process regression model to predict the battery capacity prediction value, wherein the Gaussian process regression model is constructed based on the Gaussian process regression algorithm.

[0136] As an optional embodiment, the data acquisition module 801, feature extraction module 802 and battery capacity prediction module 803 may also be used to execute the steps described in the method embodiments of the second to sixth embodiments.

[0137] Embodiment 7

[0138] Fig. 9 The hardware architecture diagram of the computer device 10000 suitable for implementing the lithium battery capacity prediction method according to the first to fifth embodiments of the present invention is schematically shown. In some embodiments, the computer device 10000 can be a terminal device such as a smart phone, a wearable device, a tablet computer, a personal computer, a vehicle terminal, a game console, a virtual device, a workbench, a digital assistant, a set-top box, a robot, etc. In other embodiments, the computer device 10000 can be a rack server, a blade server, a tower server or a cabinet server (including an independent server, or a server cluster composed of multiple servers), etc. Fig. 9 As shown, the computer device 10000 includes but is not limited to: a memory 10010, a processor 10020, and a network interface 10030 that can communicate with each other through a system bus. Among them:

[0139] The memory 10010 includes at least one type of computer-readable storage medium, and the readable storage medium includes a flash memory, a hard disk, a multimedia card, a card-type memory (such as an SD or DX memory), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 10010 may be an internal storage module of the computer device 10000, such as a hard disk or a memory of the computer device 10000. In other embodiments, the memory 10010 may also be an external storage device of the computer device 10000, such as a plug-in hard disk equipped on the computer device 10000, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the memory 10010 may also include both the internal storage module of the computer device 10000 and its external storage device. In this embodiment, the memory 10010 is generally used to store the operating system and various application software installed in the computer device 10000, such as the program code of the lithium battery capacity prediction method, etc. In addition, the memory 10010 can also be used to temporarily store various data that have been output or will be output.

[0140] In some embodiments, the processor 10020 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other chips. The processor 10020 is generally used to control the overall operation of the computer device 10000, such as performing control and processing related to data interaction or communication with the computer device 10000. In this embodiment, the processor 10020 is used to run the program code stored in the memory 10010 or process data.

[0141] The network interface 10030 may include a wireless network interface or a wired network interface, and the network interface 10030 is generally used to establish a communication link between the computer device 10000 and other computer devices. For example, the network interface 10030 is used to connect the computer device 10000 to an external terminal through a network, and to establish a data transmission channel and a communication link between the computer device 10000 and the external terminal. The network may be a wireless or wired network such as an intranet, the Internet, the Global System of Mobile communication (GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G network, Bluetooth, Wi-Fi, etc.

[0142] It should be pointed out that Fig. 9 Only a computer device having components 10010 - 10030 is shown, but it should be understood that implementation of all of the components shown is not a requirement, and more or fewer components may alternatively be implemented.

[0143] In this embodiment, the lithium battery capacity prediction method stored in the memory 10010 can also be divided into one or more program modules and executed by one or more processors (such as processor 10020) to complete the embodiment of the present invention.

[0144] Embodiment 8

[0145] A computer-readable storage medium is also provided in an embodiment of the present invention. A computer-readable storage medium stores a computer program thereon. When the computer program is executed by a processor, the steps of the lithium battery capacity prediction method in the embodiment are implemented.

[0146] In this embodiment, the computer-readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the computer-readable storage medium can be an internal storage unit of a computer device, such as a hard disk or memory of the computer device. In other embodiments, the computer-readable storage medium can also be an external storage device of a computer device, such as a plug-in hard disk equipped on the computer device, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the computer-readable storage medium can also include both the internal storage unit of the computer device and its external storage device. In this embodiment, the computer-readable storage medium is generally used to store the operating system and various application software installed on the computer device, such as the program code of the lithium battery capacity prediction method in the embodiment. In addition, the computer-readable storage medium can also be used to temporarily store various types of data that have been output or are to be output.

[0147] Embodiment 9

[0148] A computer program product is also provided in an embodiment of the present invention, including a computer program, and when the computer program is executed by a processor, the method in the above embodiment is implemented.

[0149] Obviously, those skilled in the art should understand that the modules or steps of the above-mentioned embodiments of the present invention can be implemented by general-purpose computer devices, they can be concentrated on a single computer device, or distributed on a network composed of multiple computer devices, and optionally, they can be implemented by executable program codes of computer devices, so that they can be stored in a storage device and executed by the computer device, and in some cases, the steps shown or described can be executed in a different order from that here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the embodiments of the present invention are not limited to any specific combination of hardware and software.

[0150] It should be noted that the above are only preferred embodiments of the present invention, and the patent protection scope of the present invention is not limited thereto. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A lithium battery capacity prediction method based on short-term charging data and enhanced whale optimization algorithm, It is characterized in that include: During the cyclic charge and discharge test of different types of batteries, the battery charge and discharge data are collected and recorded in real time and short-term charging data are extracted. The charge and discharge data include charging time, voltage, current and capacity data. The short-term charging data is obtained by performing spline interpolation processing on the charging voltage and current data within a preset short-term charging time. Extracting battery aging characteristics from short-term charging data through iterative optimization based on an enhanced whale optimization algorithm, wherein the enhanced whale optimization algorithm is obtained by introducing a pooling mechanism, a migration search strategy, a priority selection search strategy, and a surround prey search strategy into the classic whale optimization algorithm, and the battery aging characteristics are the voltage increment of the voltage interval corresponding to the charging time interval with the strongest correlation with the battery capacity; For batteries of different types or batteries of the same type, the charge and discharge data and the aging characteristics are input into a Gaussian process regression model to predict the battery capacity prediction value, wherein the Gaussian process regression model is constructed based on a Gaussian process regression algorithm.

2. The method according to claim 1, It is characterized in that The specific steps of parameter setting of the enhanced whale optimization algorithm include: The Spearman correlation coefficient between the voltage increment and the battery capacity is used as the objective function, and the optimization target is defined as the minimum Spearman correlation coefficient between the voltage increment and the battery capacity; The optimization object is set as the upper and lower limits of the charging time interval; Set the enhanced whale optimization algorithm parameters: the number of search agents, the number of decision variables, the maximum number of iterations, the number of executions, and the upper and lower limits of the optimization object.

3. The method according to claim 2, It is characterized in that The specific steps of the iterative optimization include: Based on the enhanced whale optimization algorithm, a group of whales are randomly initialized in the hyperparameter space, i.e., candidate solutions of the algorithm; The voltage increments corresponding to the upper and lower limits of the charging time interval represented by each whale are extracted, and the Spearman correlation coefficient between the voltage increment and the battery capacity is calculated. The algorithm performance is evaluated using the fitness function. Update the whale's position according to the rules of the enhanced whale optimization algorithm, that is, the upper and lower limits of the charging time interval; The position of the whale is continuously updated until the maximum number of iterations is reached, and the global optimal solution for the charging time interval with the strongest correlation with the battery capacity is obtained. The voltage increment of the corresponding voltage interval is found and used as the aging feature.

4. The method according to claim 3, It is characterized in that The pooling mechanism is defined as: given a matrix of size k, Pool = (p 1 ,p 2 ,...,p k ), at the end of each iteration, the element P in the matrix Pool is generated i =(p i,1 ,p i,2 ,...,p i,D ), element P i The resulting formula is as follows: In the formula, represents the element of the tth iteration, is the binary random vector in the tth iteration, is its inverse vector, is the worst solution in the tth iteration, It represents the optimal solution generated in the tth iteration A random location in the vicinity of the corresponding whale.

5. The method according to claim 4, It is characterized in that The working principle of the migration search strategy can be summarized as follows: Where: represents the position of the i-th whale in the t+1th iteration; represents the random position of the whale in the search space in the tth iteration; rand is a random number between [0,1) and it follows a uniform distribution; δ max and δ min Corresponding to the upper and lower bounds of the problem respectively; δ best-max and δ best-min Then they correspond to the optimal solutions The upper and lower bounds of .

6. The method according to claim 5, It is characterized in that The working principle of the priority search strategy can be summarized as follows: Where: represents the position of the i-th whale in the t-th iteration; and Both are positions 1 and 2 randomly selected from the matrix Pool in the tth iteration; MaxIt represents the maximum number of iterations set in the enhanced whale optimization algorithm.

7. The method according to claim 6, It is characterized in that The working principle of the encircling prey search strategy can be summarized in the following equation: Where: is the position 3 randomly selected from the matrix Pool in the tth iteration.

8. The method according to claim 7, It is characterized in that The specific steps of training the Gaussian process regression model include: The extracted first 15% aging characteristics of a battery and the battery capacity data are used as a model training set, the training set is used to run a Gaussian process regression model to find correlations, and the resulting model is saved.

9. The method according to claim 8, It is characterized in that For different types of batteries, the specific steps for battery capacity prediction include: The first 15% of the battery capacity data and aging characteristics collected from a certain battery are input into the Gaussian process regression model for model training, and the last 85% of the battery capacity data and aging characteristics collected from the battery are input into the trained Gaussian process regression model for model verification. All characteristics of other batteries of the same type are input into the trained Gaussian process regression model to obtain the battery capacity prediction value of the battery.

10. The method according to claim 8, It is characterized in that For the same type of battery, the specific steps of battery capacity prediction include: For different batteries of the same type, the last 85% of the battery capacity data obtained during the battery cycle charge and discharge test and the aging characteristics of the battery are directly input into the trained Gaussian process regression model to obtain the battery capacity prediction value of the corresponding battery.

Images