Lithium ion battery RUL prediction method and device

Abstract

The invention provides a lithium ion battery RUL prediction method and device, and belongs to the technical field of lithium ion batteries, and the method comprises the steps: obtaining a voltage curve, a current curve, a temperature curve and capacity data of a sample lithium ion battery in each charging and discharging process, respectively extracting characteristic data related to the capacity data degradation from the voltage curve, the current curve and the temperature curve; determining the correlation between each feature data and the capacity data, and inputting the capacity data and the feature data corresponding to the correlation greater than a preset threshold value as health factors into a neural network for training; and determining the health factor of a target lithium ion battery, and inputting the health factor of the target lithium ion battery into the trained neural network to obtain an RUL predicted value of the target lithium ion battery. According to the method, the performance of the neural network is improved, so that the accuracy of lithium ion battery RUL prediction is improved, and higher robustness is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium-ion batteries, and in particular to a method and device for predicting RUL of lithium-ion batteries. Background Art

[0002] Lithium batteries have significant advantages such as high energy and power density, high efficiency, wide operating temperature range, low self-discharge rate, long service life, and environmental friendliness, so they have become the preferred energy source for electric vehicles. The demand for lithium-ion batteries is increasing day by day. However, during the long-term use of lithium batteries, with the increase in the number of charge and discharge cycles, some irreversible chemical reactions occur inside the battery, resulting in increased internal resistance, performance degradation, and even some accidents. Therefore, how to accurately predict the remaining service life of lithium batteries is particularly important for whether lithium batteries can be used safely, and has become a hot topic of research at home and abroad.

[0003] At present, the prediction methods for the RUL (Remaining Useful Life) of lithium batteries can be divided into direct and indirect methods. The direct method uses coulomb counting or impedance spectroscopy to predict the remaining life, which has limitations such as cumulative errors and long prediction time. The battery impedance needs to be measured using professional instruments, which is difficult to apply online. The indirect method is mainly model-driven and data-driven. The model-based methods mainly include electrochemical models and equivalent circuit models; the data-driven methods mainly include machine learning, deep learning, and the fusion of multiple methods for prediction, which requires a large amount of high-quality data, and the optimization of model parameters is still very challenging. Traditional methods extract fewer health factors, and most of them use a single capacity feature for estimation, which has problems such as insufficient features and poor robustness. Summary of the invention

[0004] In order to solve the shortcomings of the prior art, the present invention proposes a lithium-ion battery RUL prediction method and device, which considers the influence of multiple health factors on the remaining service life of the battery during the charging and discharging process of the lithium battery, and uses an optimization algorithm to optimize the neural network parameters, thereby improving the neural network performance and prediction accuracy, thereby improving the safety and reliability of power battery-related applications and improving energy utilization.

[0005] The present invention provides a lithium ion battery RUL prediction method, comprising:

[0006] Obtaining a voltage curve, a current curve, a temperature curve and capacity data of a sample lithium-ion battery during each charge and discharge process, and extracting characteristic data related to degradation of the capacity data from the voltage curve, the current curve and the temperature curve respectively;

[0007] Determine the correlation between each characteristic data and the capacity data, and input the characteristic data corresponding to the capacity data and the correlation greater than a preset threshold into a neural network as a health factor for training;

[0008] The health factor of the target lithium-ion battery is determined, and the health factor of the target lithium-ion battery is input into the trained neural network to obtain a predicted RUL value of the target lithium-ion battery.

[0009] According to a lithium-ion battery RUL prediction method provided by the present invention, the characteristic data includes one or more of constant voltage charging time, constant current charging time, constant current discharging time, temperature change corresponding to the constant current discharging time, time when the charging temperature reaches a peak value, and time when the discharging temperature reaches a peak value;

[0010] The constant voltage charging time is the charging time for charging the sample lithium ion battery at a constant voltage until the current of the sample lithium ion battery drops to a preset charging current;

[0011] The constant current charging time is the charging time for charging the sample lithium ion battery with a constant current until the voltage of the sample lithium ion battery reaches a preset charging voltage;

[0012] The constant current discharge time is the discharge time for discharging the sample lithium ion battery at a constant current until the voltage of the sample lithium ion battery drops to a preset discharge voltage;

[0013] The temperature change corresponding to the constant current discharge time is the difference between the temperature at the start time and the temperature at the end time of the constant current discharge time.

[0014] According to a lithium-ion battery RUL prediction method provided by the present invention, the determining of the correlation between each characteristic data and the capacity data includes:

[0015] The Pearson correlation coefficient is used to determine the correlation between each characteristic data and the capacity data during multiple charge and discharge processes.

[0016] According to a lithium-ion battery RUL prediction method provided by the present invention, before inputting the capacity data and the characteristic data corresponding to the correlation greater than a preset threshold as health factors into a neural network for training, the method further includes:

[0017] The hyperparameters of the neural network are optimized using the SandCat optimization algorithm.

[0018] According to a lithium-ion battery RUL prediction method provided by the present invention, the hyperparameters of the neural network are optimized using a sandcat optimization algorithm, including:

[0019] Determine multiple initial population positions of the sand cat according to the preset population number of the sand cat, the problem dimension, and the upper and lower limits of the search space, wherein the problem dimension is the number of the hyperparameters, the upper and lower limits of the search space are the upper and lower limits of each hyperparameter, and the values ​​of the hyperparameters contained in different population positions are not completely the same;

[0020] Determine the fitness of each current population position of the sand cat, and update the global optimal position of the sand cat population according to the current population position with the highest fitness;

[0021] Determine the guidance parameters of the sand cat according to its sensitivity range;

[0022] When the guiding parameter is greater than 1, updating each current population position according to the best position of the sand cat at the current moment, each current population position and the sensitivity range;

[0023] When the guide parameter is less than or equal to 1, a random position is generated according to the global best position of the sand cat population and each current population position, and each current population position is updated according to the global best position of the sand cat population, the random position and the sensitivity range;

[0024] When the number of times the current population position of the sand cats is updated reaches a preset number, the global best position of the sand cat population is taken as the optimal value of the hyperparameter.

[0025] According to a lithium-ion battery RUL prediction method provided by the present invention, the determination of the fitness of each current population position of sand cats includes:

[0026] Using each current population position as a hyperparameter of the neural network, and inputting the health factor into the neural network to obtain a predicted RUL value of the sample lithium-ion battery;

[0027] The fitness of each current population position is determined according to the RUL predicted value and the RUL actual value of the sample lithium-ion battery.

[0028] According to a lithium-ion battery RUL prediction method provided by the present invention, the capacity data and the characteristic data corresponding to the correlation greater than a preset threshold are input as health factors into a neural network for training, including:

[0029] Inputting the health factor corresponding to the sample lithium-ion battery into the CNN in the neural network to obtain feature data extracted by the CNN;

[0030] Inputting the feature data extracted by the CNN into the BiGRU in the neural network to obtain the RUL prediction value of the sample lithium-ion battery output by the BiGRU;

[0031] The weight parameters in the neural network are adjusted according to the RUL predicted value and the RUL actual value of the sample lithium-ion battery.

[0032] According to a lithium-ion battery RUL prediction method provided by the present invention, the hyperparameters of the neural network include the number of layers of the CNN, the number of layers of the BiGRU, and the learning rate and number of training times of the neural network.

[0033] The present invention also provides a lithium-ion battery RUL prediction device, comprising:

[0034] An extraction module, used to obtain a voltage curve, a current curve, a temperature curve and capacity data of a sample lithium-ion battery during each charge and discharge process, and to extract characteristic data related to the degradation of the capacity data from the voltage curve, the current curve and the temperature curve respectively;

[0035] A training module, used to determine the correlation between each characteristic data and the capacity data, and input the characteristic data corresponding to the capacity data and the correlation greater than a preset threshold as health factors into a neural network for training;

[0036] A prediction module is used to determine the health factor of the target lithium-ion battery, input the health factor of the target lithium-ion battery into the trained neural network, and obtain the RUL prediction value of the target lithium-ion battery.

[0037] The present invention also 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 program, any of the above-mentioned lithium-ion battery RUL prediction methods is implemented.

[0038] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for predicting the RUL of a lithium-ion battery as described above is implemented.

[0039] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method for predicting the RUL of a lithium-ion battery as described above is implemented.

[0040] The lithium-ion battery RUL prediction method and device provided by the present invention extract a variety of feature data from the charging and discharging process of the sample lithium-ion battery, use the Pearson correlation coefficient to analyze the correlation between the feature data and capacity degradation, extract health factors with higher correlation as a data set to train the neural network, improve the performance of the neural network, thereby improving the accuracy of the lithium-ion battery RUL prediction, and having higher robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 This is one of the flow charts of the lithium-ion battery RUL prediction method provided by the present invention;

[0043] Figure 2 It is a schematic diagram of a sliding window in the lithium-ion battery RUL prediction method provided by the present invention;

[0044] Figure 3 It is a curve diagram showing the change of charging voltage with charging time at different cycle times in the battery data set B0005 used in the experiment in the lithium-ion battery RUL prediction method provided by the present invention;

[0045] Figure 4 It is a curve diagram of the change of charging current with charging time at different cycle times in the battery data set B0005 used in the experiment in the lithium-ion battery RUL prediction method provided by the present invention;

[0046] Figure 5 It is a schematic diagram of the capacity degradation curve of the battery data set used in the experiment of the lithium-ion battery RUL prediction method provided by the present invention;

[0047] Figure 6 It is a schematic diagram of a normalized curve of all health factors of the battery data set B0005 used in the experiment in the lithium-ion battery RUL prediction method provided by the present invention;

[0048] Figure 7 This is the second flow chart of the lithium-ion battery RUL prediction method provided by the present invention;

[0049] Figure 8 It is a structural schematic diagram of the GRU in the lithium-ion battery RUL prediction method provided by the present invention;

[0050] Fig. 9 It is a structural schematic diagram of BiGRU in the lithium-ion battery RUL prediction method provided by the present invention;

[0051] Fig.10 It is a structural schematic diagram of a lithium-ion battery RUL prediction device provided by the present invention;

[0052] Fig.11 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Combine the following Figure 1 A lithium-ion battery RUL prediction method of the present invention is described, comprising:

[0055] Step 101, obtaining a voltage curve, a current curve, a temperature curve and capacity data of a sample lithium-ion battery during each charge and discharge process, and extracting characteristic data related to the degradation of the capacity data from the voltage curve, the current curve and the temperature curve respectively;

[0056] In the charge and discharge cycle of the sample lithium-ion battery, characteristic data related to capacity degradation is extracted from the voltage curve, current curve and temperature curve as an indirect health factor of capacity degradation of the sample lithium-ion battery. This embodiment does not limit the type of indirect health factor. The capacity data is used as a direct health factor of capacity degradation.

[0057] Step 102, determining the correlation between each characteristic data and the capacity data, and inputting the capacity data and the characteristic data corresponding to the correlation greater than a preset threshold into a neural network as a health factor for training;

[0058] The extracted indirect health factors and capacity data may be normalized, and then the correlation between each indirect health factor and the capacity data may be determined. The correlation between each indirect health factor and the capacity data may be measured by using cosine distance, Euclidean distance, etc., which is not specifically limited in this embodiment.

[0059] Feature data and capacity data with high correlation are selected as the health factor dataset to train the neural network.

[0060] like Figure 2 As shown, a sliding window can be used to select the health factor in the previous n charge and discharge processes, and input into the neural network to predict the capacity data in the n+1th charge and discharge process, that is, the RUL predicted value, where n is a positive integer. The weight parameter of the neural network is adjusted according to the difference between the predicted value and the actual value of the capacity data in the n+1th charge and discharge process to minimize the difference.

[0061] For example, a sliding window is used to process the health factor, and the health factors of the past 6 cycles are used to predict the capacity data of the next cycle. The data set processed by the sliding window is divided into a training set and a test set, and the ratio of the training set to the test set can be 7:3. The training set and the test set are used to train and test the neural network respectively.

[0062] Step 103, determining the health factor of the target lithium-ion battery, inputting the health factor of the target lithium-ion battery into the trained neural network, and obtaining the RUL predicted value of the target lithium-ion battery.

[0063] The target lithium-ion battery is a lithium-ion battery for which RUL prediction is required. The health factor of the target lithium-ion battery is obtained in the same manner, and the health factor of the target lithium-ion battery is of the same type as the health factor of the sample lithium-ion battery.

[0064] The trained neural network is used to predict the RUL predicted value of the target lithium-ion battery according to the health factor of the target lithium-ion battery. This embodiment does not limit the type of neural network.

[0065] The health factor of the target lithium-ion battery during the most recent n charge and discharge processes can be input into the neural network to predict the capacity data of the target lithium-ion battery during the next charge and discharge process.

[0066] This embodiment extracts a variety of feature data from the charging and discharging process of the sample lithium-ion battery, uses the Pearson correlation coefficient to analyze the correlation between the feature data and capacity degradation, extracts health factors with higher correlation as a data set to train the neural network, improves the performance of the neural network, and thus improves the accuracy of the lithium-ion battery RUL prediction, with higher robustness.

[0067] On the basis of the above embodiment, the characteristic data in this embodiment includes one or more of constant voltage charging time, constant current charging time, constant current discharging time, temperature change corresponding to the constant current discharging time, time when the charging temperature reaches a peak value, and time when the discharging temperature reaches a peak value;

[0068] The constant voltage charging time is the charging time for charging the sample lithium ion battery at a constant voltage until the current of the sample lithium ion battery drops to a preset charging current;

[0069] The constant current charging time is the charging time for charging the sample lithium ion battery with a constant current until the voltage of the sample lithium ion battery reaches a preset charging voltage;

[0070] The constant current discharge time is the discharge time for discharging the sample lithium ion battery at a constant current until the voltage of the sample lithium ion battery drops to a preset discharge voltage;

[0071] The temperature change corresponding to the constant current discharge time is the difference between the temperature at the start time and the temperature at the end time of the constant current discharge time.

[0072] The capacity data is taken as the direct health factor, denoted as HI1, and the constant voltage charging time, constant current charging time, constant current discharging time, the temperature change corresponding to the constant current discharging time, the time when the charging temperature reaches the peak value, and the time when the discharge temperature reaches the peak value are taken as indirect health factors, denoted as HI2, HI3, HI4, HI5, HI6 and HI7 respectively.

[0073] For example, at room temperature of 24 degrees, charge in constant current (CC) mode of 1.5A and obtain the charging voltage curve until the battery voltage reaches 4.2V. Figure 3 Then charge in constant voltage (CV) mode and obtain the charging current curve until the charging current drops to 20mA, as shown in Figure 4 As shown, the constant voltage charging time is obtained. Discharging is performed in a constant current (CC) mode of 2A until the voltages of batteries 5, 6, 7, and 18 drop to 2.7V, 2.5V, 2.2V, and 2.5V, thereby obtaining the constant current discharge time. The degradation curve of the battery capacity during discharge is shown in Figure 5 As shown, the capacity data of each battery discharge is obtained.

[0074] When the battery reaches the end of life (EOF) standard, the rated capacity drops to 30% of it, that is, the rated capacity of batteries 5, 6, 7 and 18 decreases from 2Ahr to 1.4Ahr. Batteries 5, 6 and 7 are selected as the data set, and the health factor is extracted from the data set.

[0075] Based on the above embodiment, the determination of the correlation between each characteristic data and the capacity data in this embodiment includes:

[0076] The Pearson correlation coefficient is used to determine the correlation between each characteristic data and the capacity data during multiple charge and discharge processes.

[0077] A value of each characteristic data is obtained during each charge and discharge process, and a sequence of each characteristic data is obtained during multiple charge and discharge processes.

[0078] Normalize each feature data to between [0,1], the formula is as follows:

[0079]

[0080] Among them, x min is the minimum value in the sequence of each feature data, x maxis the maximum value in the sequence of each feature data, x is each value in the sequence, x new is the result of normalizing each value in the sequence. The normalized curve for all health factors is as follows Figure 6 shown.

[0081] The Pearson correlation coefficient is a statistical indicator for measuring the linear correlation between two continuous variables. It is simple and easy to understand and widely used in various fields. This embodiment uses the Pearson correlation coefficient to calculate the correlation between each normalized feature data and the normalized capacity data. The formula is as follows:

[0082]

[0083] Among them, X is the sequence of each feature data after normalization, Y is the sequence of capacity data after normalization, and E is the mean calculation. The value range of Pearson is between -1 and 1. The closer the result is to 1, the stronger the positive correlation is; the closer it is to -1, the stronger the negative correlation is; and the closer it is to 0, the weaker the correlation is.

[0084] The Pearson correlation coefficient was used to analyze the correlation between the indirect health factor and the direct health factor capacity data after normalization. The results of the Pearson correlation coefficient are shown in Table 1.

[0085] Table 1 Correlation between indirect health factors HI2 to HI7 and direct health factor capacity HI 1

[0086] Pearson HI2 HI3 HI4 HI5 HI6 HI7 B0005 -0.9132 0.9975 0.999948 -0.9817 0.04478 0.999815 B0006 -0.9399 0.9934 0.999914 -0.9254 0.011169 0.999868 B0007 -0.8846 0.9949 0.999728 -0.8829 -0.1039 0.999213

[0087] It can be seen from the table that HI3, HI4, HI7 and HI1 have a strong positive correlation, and HI2, HI5 and HI1 have a strong negative correlation. HI3, HI4, HI5, HI7 and HI1 have a high correlation, so HI1, HI3, HI4, HI5 and HI7 are selected as health factors to train the neural network.

[0088] Based on the above embodiment, before the capacity data and the characteristic data corresponding to the correlation greater than the preset threshold are input as health factors into the neural network for training, the present embodiment further includes:

[0089] The hyperparameters of the neural network are optimized using the SandCat optimization algorithm.

[0090] The sand cat optimization algorithm simulates the two main behaviors of the sand cat, searching for prey and attacking prey. The algorithm is simple to implement, can well avoid local optimum, and has strong robustness. Therefore, the present embodiment uses the sand cat optimization algorithm to optimize the hyperparameters of the neural network.

[0091] Set the network optimizer to Adam, input the training set and the validation set for training, until the number of iterations of the set Sand Cat optimization algorithm is reached. The neural network hyperparameters at this time are the optimal network hyperparameters. Use the optimal network model to predict the RUL of lithium batteries and obtain the predicted results.

[0092] This embodiment improves the performance of the network by extracting multiple health factors and using the Sand Cat optimization algorithm to optimize the hyperparameters of the neural network, so that the neural network has higher robustness and learning ability, thereby improving the accuracy of lithium battery RUL prediction.

[0093] Based on the above embodiment, the hyperparameters of the neural network are optimized using the Sand Cat optimization algorithm in this embodiment, including:

[0094] Determine multiple initial population positions of the sand cat according to the preset population number of the sand cat, the problem dimension, and the upper and lower limits of the search space, wherein the problem dimension is the number of the hyperparameters, the upper and lower limits of the search space are the upper and lower limits of each hyperparameter, and the values ​​of the hyperparameters contained in different population positions are not completely the same;

[0095] Determine the fitness of each current population position of the sand cat, and update the global optimal position of the sand cat population according to the current population position with the highest fitness;

[0096] Determine the guidance parameters of the sand cat according to its sensitivity range;

[0097] When the guiding parameter is greater than 1, updating each current population position according to the best position of the sand cat at the current moment, each current population position and the sensitivity range;

[0098] When the guide parameter is less than or equal to 1, a random position is generated according to the global best position of the sand cat population and each current population position, and each current population position is updated according to the global best position of the sand cat population, the random position and the sensitivity range;

[0099] When the number of times the current population position of the sand cats is updated reaches a preset number, the global best position of the sand cat population is taken as the optimal value of the hyperparameter.

[0100] The specific steps of the Sand Cat optimization algorithm are as follows:

[0101] The first step is to initialize the population. Assume that the population size of sand cats is N, the problem dimension is D, and the upper and lower limits of the search space are ub = [ub1, ub2, ..., ub d ] and lb = [lb1, lb2, ..., lb d]. Where D is the number of hyperparameters optimized by the neural network, and the data at the same position in ub and lb represent the upper and lower limits of a hyperparameter respectively. The initialization formula is:

[0102]

[0103] Where X is the position matrix of the sand cats, and rand(N, D) is a random vector with N rows and D columns in the range of (0, 1).

[0104] The second step is to search for prey. Each sand cat will find its best position at the current moment according to the randomly generated and its current location and its sensitive range Update its own position. The best position is calculated by calculating the fitness of each current population position in each iteration. If the highest fitness is greater than the fitness of the best position, the current population position corresponding to the highest fitness is used as the best position, otherwise the best position remains unchanged.

[0105]

[0106] Among them, iter c is the current iteration number, iter max is the maximum number of iterations, is the boot parameter, is the sensitivity range, S M The maximum sensitivity is preset.

[0107] The third step is to attack the prey, modeling the attack phase and using the global optimal position of the population (The population position with the highest fitness in all iterations) and the current position Generate a random position. Assuming that the sensitivity range of the sand cat is a circle, use the roulette wheel selection algorithm to select a random angle θ for each sand cat. Using a random angle can also avoid falling into the local optimum.

[0108]

[0109] in, A random location for a sand cat.

[0110] The fourth step is exploration and development. The exploration and development of the sand cat optimization algorithm is through parameters The two parameters, determined by R and R, realize the two-stage transformation. When R≤1, the sand cat is guided to attack the prey, otherwise, the task of the sand cat is to find new possible solutions in the global area.

[0111]

[0112] The fifth step is to iterate until the maximum number of iterations is reached, i.e. iter c =iter max , obtain the global best position, that is, get the optimal solution of the hyperparameters. The complete schematic diagram of the lithium-ion battery RUL prediction method is as follows Figure 7 shown.

[0113] On the basis of the above embodiment, the determination of the fitness of each current population position of the sand cat in this embodiment includes:

[0114] Using each current population position as a hyperparameter of the neural network, and inputting the health factor into the neural network to obtain a predicted RUL value of the sample lithium-ion battery;

[0115] The fitness of each current population position is determined according to the RUL predicted value and the RUL actual value of the sample lithium-ion battery.

[0116] Each current population position of the sand cat is taken as a hyperparameter of the neural network, and the mean square error between the RUL predicted value of the sample lithium-ion battery by the neural network and the actual RUL value can be taken as the fitness of each current population position.

[0117] On the basis of the above embodiment, in this embodiment, the capacity data and the characteristic data corresponding to the correlation greater than the preset threshold are input as health factors into the neural network for training, including:

[0118] Inputting the health factor corresponding to the sample lithium-ion battery into the CNN (Convolutional Neural Network) in the neural network to obtain feature data extracted by the CNN;

[0119] Inputting the feature data extracted by the CNN into the Bidirectional Gate Recurrent Unit (Bidirectional Gate Recurrent Unit) in the neural network to obtain the RUL prediction value of the sample lithium-ion battery output by the BiGRU;

[0120] The weight parameters in the neural network are adjusted according to the RUL predicted value and the RUL actual value of the sample lithium-ion battery.

[0121] CNN usually includes convolutional layers, pooling layers, and fully connected layers. The one-dimensional convolution operation can be seen as a process in which a sliding window scans the input sequence and extracts local features. At each window position, the convolution kernel is dot-multiplied with the input sequence and summed to obtain a new value, which constitutes an element in the output sequence. By sliding the window on the input sequence, features at different positions of the input sequence can be extracted.

[0122] BiGRU is a form of applying the gated recurrent unit GRU to a bidirectional recurrent neural network. The network structure of GRU is as follows Figure 8 As shown in Figure 1. The bidirectional recurrent neural network includes two recurrent paths, forward and reverse, which process the sequence from two directions respectively. The network structure of BiGRU is shown in Figure 1. Fig. 9 As shown. The calculation process of the GRU network is as follows:

[0123] r t =σ(X t W xr +h t-1 U hr +b r ) (10)

[0124] z t =σ(X t W xz +h t-1 U hz +b z ) (11)

[0125]

[0126] Among them, r t is the reset gate, z t is the update gate, h t is the hidden layer, is a candidate hidden layer. xr , W xz , W xh , U hr , U hz and U hh Represent the weights of each gate and hidden layer, b r 、b z and b h Represents the corresponding bias, σ is the sigmoid function, and tanh is the hyperbolic tangent activation function.

[0127] This embodiment extracts multiple health factor sequences from the lithium battery data set, uses the Pearson correlation coefficient to analyze the correlation between health factors and capacity degradation, extracts health factors with higher correlation as data sets and divides them into training sets and test sets, and constructs a SCSO-CNN-BiGRU model to estimate the remaining service life of lithium batteries. This method extracts multiple health factors and uses an intelligent optimization algorithm to optimize network hyperparameters to improve network performance. The SCSO-CNN-BiGRU model has higher robustness and learning ability, thereby improving the accuracy of lithium battery RUL prediction.

[0128] Based on the above embodiment, the hyperparameters of the neural network in this embodiment include the number of layers of the CNN, the number of layers of the BiGRU, and the learning rate and number of training times of the neural network.

[0129] The initialization parameters of the sand cat optimization algorithm are set to include population size, problem dimension, upper and lower limits of the search space, fitness function, and number of iterations. The upper and lower limits of the search space are the upper and lower limits of the neural network hyperparameters to be optimized in this embodiment. The number of layers of CNN, the number of layers of BiGRU, the learning rate, and the number of training times are optimized, and the parameters set are shown in Table 2.

[0130] Table 2 Initialization parameters of the Sand Cat optimization algorithm

[0131] Population size Problem Dimensions Iterations Upper limit of search space Search space lower bound 20 4 50 60,60,0.01,600 5,5,0.001,100

[0132] The lithium-ion battery RUL prediction device provided by the present invention is described below. The lithium-ion battery RUL prediction device described below and the lithium-ion battery RUL prediction method described above can be referenced to each other.

[0133] like Fig.10 As shown, the device includes an extraction module 1001, a training module 1002 and a prediction module 1003, wherein:

[0134] The extraction module 1001 is used to obtain the voltage curve, current curve, temperature curve and capacity data of the sample lithium-ion battery during each charge and discharge process, and extract characteristic data related to the degradation of the capacity data from the voltage curve, current curve and temperature curve respectively;

[0135] The training module 1002 is used to determine the correlation between each characteristic data and the capacity data, and input the characteristic data corresponding to the capacity data and the correlation greater than a preset threshold into the neural network as a health factor for training;

[0136] The prediction module 1003 is used to determine the health factor of the target lithium-ion battery, input the health factor of the target lithium-ion battery into the trained neural network, and obtain the RUL prediction value of the target lithium-ion battery.

[0137] This embodiment extracts a variety of feature data from the charging and discharging process of the sample lithium-ion battery, uses the Pearson correlation coefficient to analyze the correlation between the feature data and capacity degradation, extracts health factors with higher correlation as a data set to train the neural network, improves the performance of the neural network, and thus improves the accuracy of the lithium-ion battery RUL prediction, with higher robustness.

[0138] Fig.11 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.11As shown, the electronic device may include: a processor 1110, a communication interface 1120, a memory 1130 and a communication bus 1140, wherein the processor 1110, the communication interface 1120 and the memory 1130 communicate with each other through the communication bus 1140. The processor 1110 may call the logic instructions in the memory 1130 to execute the RUL prediction method of the lithium-ion battery, the method comprising: obtaining the voltage curve, current curve, temperature curve and capacity data of the sample lithium-ion battery during each charge and discharge process, extracting the characteristic data related to the degradation of the capacity data from the voltage curve, current curve and temperature curve respectively; determining the correlation between each characteristic data and the capacity data, and inputting the characteristic data corresponding to the capacity data and the correlation greater than the preset threshold as health factors into the neural network for training; determining the health factor of the target lithium-ion battery, and inputting the health factor of the target lithium-ion battery into the trained neural network to obtain the RUL prediction value of the target lithium-ion battery.

[0139] In addition, the logic instructions in the above-mentioned memory 1130 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0140] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the lithium-ion battery RUL prediction method provided by the above methods, the method comprising: obtaining the voltage curve, current curve, temperature curve and capacity data of the sample lithium-ion battery during each charging and discharging process, and extracting characteristic data related to the degradation of the capacity data from the voltage curve, current curve and temperature curve respectively; determining the correlation between each characteristic data and the capacity data, and inputting the characteristic data corresponding to the capacity data and the correlation greater than a preset threshold as health factors into a neural network for training; determining the health factor of the target lithium-ion battery, and inputting the health factor of the target lithium-ion battery into the trained neural network to obtain the RUL prediction value of the target lithium-ion battery.

[0141] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the lithium-ion battery RUL prediction method provided by the above methods, the method comprising: obtaining the voltage curve, current curve, temperature curve and capacity data of the sample lithium-ion battery during each charging and discharging process, and extracting characteristic data related to the degradation of the capacity data from the voltage curve, current curve and temperature curve respectively; determining the correlation between each characteristic data and the capacity data, and inputting the characteristic data corresponding to the capacity data and the correlation greater than a preset threshold as health factors into a neural network for training; determining the health factor of the target lithium-ion battery, and inputting the health factor of the target lithium-ion battery into the trained neural network to obtain the RUL prediction value of the target lithium-ion battery.

[0142] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0143] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A lithium-ion battery RUL prediction method, characterized in that: include: Obtaining a voltage curve, a current curve, a temperature curve and capacity data of a sample lithium-ion battery during each charge and discharge process, and extracting characteristic data related to degradation of the capacity data from the voltage curve, the current curve and the temperature curve respectively; Determine the correlation between each characteristic data and the capacity data, and input the characteristic data corresponding to the capacity data and the correlation greater than a preset threshold into a neural network as a health factor for training; The health factor of the target lithium-ion battery is determined, and the health factor of the target lithium-ion battery is input into the trained neural network to obtain a predicted RUL value of the target lithium-ion battery.

2. The lithium-ion battery RUL prediction method according to claim 1, characterized in that: The characteristic data includes one or more of constant voltage charging time, constant current charging time, constant current discharging time, temperature change corresponding to the constant current discharging time, time when the charging temperature reaches a peak value, and time when the discharging temperature reaches a peak value; The constant voltage charging time is the charging time for charging the sample lithium ion battery at a constant voltage until the current of the sample lithium ion battery drops to a preset charging current; The constant current charging time is the charging time for charging the sample lithium ion battery with a constant current until the voltage of the sample lithium ion battery reaches a preset charging voltage; The constant current discharge time is the discharge time for discharging the sample lithium ion battery at a constant current until the voltage of the sample lithium ion battery drops to a preset discharge voltage; The temperature change corresponding to the constant current discharge time is the difference between the temperature at the start time and the temperature at the end time of the constant current discharge time.

3. The lithium-ion battery RUL prediction method according to claim 1, characterized in that: The determining of the correlation between each characteristic data and the capacity data comprises: The Pearson correlation coefficient is used to determine the correlation between each characteristic data and the capacity data during multiple charge and discharge processes.

4. The lithium-ion battery RUL prediction method according to claim 1, characterized in that: Before inputting the capacity data and the characteristic data corresponding to the correlation greater than a preset threshold as health factors into a neural network for training, the method further includes: The hyperparameters of the neural network are optimized using the SandCat optimization algorithm.

5. The lithium-ion battery RUL prediction method according to claim 4, characterized in that: The method of using the SandCat optimization algorithm to optimize the hyperparameters of the neural network includes: Determine multiple initial population positions of the sand cat according to the preset population number of the sand cat, the problem dimension, and the upper and lower limits of the search space, wherein the problem dimension is the number of the hyperparameters, the upper and lower limits of the search space are the upper and lower limits of each hyperparameter, and the values ​​of the hyperparameters contained in different population positions are not completely the same; Determine the fitness of each current population position of the sand cat, and update the global optimal position of the sand cat population according to the current population position with the highest fitness; Determine the guidance parameters of the sand cat according to its sensitivity range; When the guiding parameter is greater than 1, updating each current population position according to the best position of the sand cat at the current moment, each current population position and the sensitivity range; When the guide parameter is less than or equal to 1, a random position is generated according to the global best position of the sand cat population and each current population position, and each current population position is updated according to the global best position of the sand cat population, the random position and the sensitivity range; When the number of times the current population position of the sand cats is updated reaches a preset number, the global best position of the sand cat population is taken as the optimal value of the hyperparameter.

6. The lithium-ion battery RUL prediction method according to claim 5, characterized in that: Determining the fitness of each current population position of the sand cat comprises: Using each current population position as a hyperparameter of the neural network, and inputting the health factor into the neural network to obtain a predicted RUL value of the sample lithium-ion battery; The fitness of each current population position is determined according to the RUL predicted value and the RUL actual value of the sample lithium-ion battery.

7. The lithium-ion battery RUL prediction method according to claim 4, characterized in that: Inputting the capacity data and the characteristic data corresponding to the correlation greater than a preset threshold as health factors into a neural network for training, including: Inputting the health factor corresponding to the sample lithium-ion battery into the CNN in the neural network to obtain feature data extracted by the CNN; Inputting the feature data extracted by the CNN into the BiGRU in the neural network to obtain the RUL prediction value of the sample lithium-ion battery output by the BiGRU; The weight parameters in the neural network are adjusted according to the RUL predicted value and the RUL actual value of the sample lithium-ion battery.

8. The lithium-ion battery RUL prediction method according to claim 7, characterized in that: The hyperparameters of the neural network include the number of layers of the CNN, the number of layers of the BiGRU, and the learning rate and number of training times of the neural network.

9. A lithium-ion battery RUL prediction device, characterized in that: include: An extraction module, used to obtain a voltage curve, a current curve, a temperature curve and capacity data of a sample lithium-ion battery during each charge and discharge process, and to extract characteristic data related to the degradation of the capacity data from the voltage curve, the current curve and the temperature curve respectively; A training module, used to determine the correlation between each characteristic data and the capacity data, and input the characteristic data corresponding to the capacity data and the correlation greater than a preset threshold as health factors into a neural network for training; A prediction module is used to determine the health factor of the target lithium-ion battery, input the health factor of the target lithium-ion battery into the trained neural network, and obtain the RUL prediction value of the target lithium-ion battery.

10. 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 program, the lithium-ion battery RUL prediction method according to any one of claims 1 to 8 is implemented.

Images