Method, device and equipment for predicting residual life of battery and medium

Abstract

The invention discloses a battery remaining life prediction method, device, equipment and medium, and relates to the technical field of new energy, the method comprises the following steps: obtaining charge-discharge cycle data of a lithium battery to be detected, preprocessing the charge-discharge cycle data, and extracting capacity sequence characteristics in the charge-discharge cycle data; inputting the capacity sequence characteristics into a trained battery residual life prediction model to obtain the residual battery capacity of the lithium battery to be detected after each charge-discharge cycle output by the battery residual life prediction model; determining the number of charge-discharge cycles required when the residual battery capacity of the lithium battery to be detected reaches the preset capacity; and determining the residual battery cycle life of the current charge-discharge cycle according to the charge-discharge cycle times and the charge-discharge cycle times corresponding to the current charge-discharge cycle. According to the method, the key feature information of the data is effectively enhanced, so that the battery residual life prediction model has robustness and higher prediction precision on the prediction of the battery residual cycle life.

Description

Technical Field

[0001] The present invention relates to the field of new energy technologies, and in particular to a method, device, equipment and medium for predicting remaining battery life. Background Art

[0002] Lithium-ion batteries are widely used as energy sources in many fields due to their high energy and power density, low self-discharge rate, low price and long life. Similar to many other electrochemical systems, the internal resistance of lithium batteries will gradually increase after repeated charge and discharge cycles. The increase in internal resistance will cause the battery to generate significant heat during the charge and discharge process, which will cause irreversible aging and performance degradation of lithium batteries. That is, lithium-ion batteries will inevitably experience energy and power attenuation during use and storage. Therefore, in battery management systems, the remaining capacity at the beginning of the battery life and the ratio of the internal resistance to its initial value are often used as battery health status (SOH) indicators. This indicator can be used to indicate the remaining life of lithium batteries. As the number of charge and discharge cycles of lithium batteries increases, SOH will decrease, that is, the remaining battery life will decrease and the available capacity will gradually decrease. When SOH drops to a certain threshold, such as below 80%, it is necessary to stop using the lithium battery or limit its use.

[0003] Considering the application of lithium-ion batteries in more and more fields, such as battery design adjustment, life prediction and warranty cost estimation, the prediction of their degradation state has received increasing attention. How to accurately use the capacity data series at the beginning of the battery life to predict the end of the battery life is an important issue that needs to be urgently solved in the industry. Summary of the invention

[0004] In view of this, the embodiments of the present invention provide a method, device, equipment and medium for predicting the remaining life of a battery, so as to solve the problem of capacity attenuation and capacity regeneration during the use of lithium-ion batteries.

[0005] According to a first aspect, an embodiment of the present invention provides a method for predicting remaining battery life, the method comprising: Acquire the charge and discharge cycle data of the lithium battery to be tested, pre-process the charge and discharge cycle data, and extract the capacity sequence features in the charge and discharge cycle data; Input the capacity sequence features into the trained battery remaining life prediction model to obtain the remaining battery capacity of the lithium battery to be tested after each subsequent charge and discharge cycle output by the battery remaining life prediction model; The battery remaining life prediction model is obtained by training the sample charge and discharge cycle data of the sample lithium battery. The battery remaining life prediction model includes a double Gaussian model and a multi-layer perceptron after the attenuation parameters are adjusted by the Levenberg-Marquardt method algorithm and the particle swarm adjustment algorithm; the double Gaussian model is used to predict the degradation process of the long-term trend sequence to obtain the long-term degradation trend of the subsequent charge and discharge cycles of the lithium battery to be tested, and the multi-layer perceptron is used to predict the fluctuation of the capacity sequence characteristics to obtain the fluctuation sequence of the subsequent charge and discharge cycles of the lithium battery to be tested. The remaining battery capacity is obtained by combining the long-term degradation trend and the fluctuation sequence; The long-term trend series is obtained by performing variational mode decomposition on the characteristics of the capacity series, and the fluctuation of the characteristics of the capacity series is obtained by performing variational mode decomposition on the characteristics of the capacity series; Determine the number of charge and discharge cycles required for the remaining battery capacity of the lithium battery to be tested to reach a preset capacity according to the ratio of the remaining battery capacity of the lithium battery to be tested after each sample charge and discharge cycle to the battery capacity of the initial sample charge and discharge cycle; The remaining battery cycle life of the current charge and discharge cycle is determined based on the number of charge and discharge cycles required when the remaining battery capacity reaches the preset capacity and the number of charge and discharge cycles corresponding to the current charge and discharge cycle.

[0006] In combination with the first aspect, in a first implementation of the first aspect, the battery remaining life prediction model is trained by the following steps: Obtaining sample charge and discharge cycle data of a sample lithium battery, preprocessing the sample charge and discharge cycle data, and extracting sample capacity sequence features from the sample charge and discharge cycle data; Decomposing the sample volume sequence characteristics into a preset number of sample eigenmodes with different center frequencies and limited bandwidths; The sample long-term trend sequence in the sample intrinsic mode is used as the input data for training, and the attenuation parameters of the double Gaussian model are adjusted by the Levenberg-Marquardt algorithm and the particle swarm adjustment algorithm to train the double Gaussian model. The sample fluctuations in the sample intrinsic mode are used as input data for training. The features are extracted layer by layer through forward propagation and the multilayer perceptron is trained by iterating layer by layer through the gradient descent method.

[0007] In combination with the first implementation of the first aspect, in the second implementation of the first aspect, decomposing the sample capacity sequence characteristics into a preset number of sample eigenmodes with different center frequencies and limited bandwidths specifically includes: Decomposing the sample capacity sequence characteristics into a preset number of sample eigenmodes with different center frequencies and limited bandwidths by using a variational mode decomposition algorithm; the sample eigenmodes include a sample long-term trend sequence and two sample fluctuations; Hilbert transform is performed on each sample eigenmode to obtain the single-sided spectrum corresponding to the sample eigenmode; Each single-sided spectrum is modulated, and constraints are established on the modulated single-sided spectrum; The established constraints are transformed into unconstrained ones using the preset penalty terms and Lagrangian factors. The unconstrained optimal solution is determined by the alternating multiplier method, and the unconstrained subproblem is transformed into a minimum value problem to obtain the optimal solution of the sample eigenmode.

[0008] In combination with the first embodiment of the first aspect, in the third embodiment of the first aspect, the process of training the attenuation parameters of the double Gaussian model by the Levenberg-Marquardt algorithm is: Establish an initial set of attenuation parameters for the double Gaussian model; Determine the residual between the predicted and actual values ​​of the double Gaussian model; Construct the Jacobian matrix based on the residual and the initial decay parameters; Use the Jacobian matrix to construct an approximate Hessian matrix; Update the attenuation parameters of the double Gaussian model according to the approximate Hessian matrix and the residual; Determine whether the update amount of the attenuation parameter is less than a preset value or whether the number of iterations of the parameter update reaches a first preset number; when the update amount of the attenuation parameter is less than the preset value, it indicates that the convergence condition is met; When it is determined that the convergence condition is met or the first preset number is reached, the attenuation parameter is stopped from being updated; when it is determined that the convergence condition is not met or the first preset number is not reached, the attenuation parameter is continuously updated until the convergence condition is met or the first preset number is reached.

[0009] In combination with the third implementation of the first aspect, in the fourth implementation of the first aspect, the process of the particle swarm adjustment algorithm training the attenuation parameters of the double Gaussian model is: Adding a preset disturbance to the attenuation parameter adjusted by the Levenberg-Marquardt algorithm, initializing it to the initial position and velocity of the particle, and obtaining the initial position and initial velocity of each particle; Determine algorithm parameters; the algorithm parameters include inertia parameter, first learning factor, second learning factor and second iteration number; The fitness value of each particle is determined according to a preset fitness function, and the individual optimal position and the group optimal position of each particle are updated according to the fitness value until the second iteration number is reached or the fitness value of the particle is less than the preset residual sum.

[0010] In combination with the first implementation of the first aspect, in the fifth implementation of the first aspect, the sample fluctuations in the sample intrinsic mode are used as input data for training, features are extracted layer by layer by forward propagation, and layer by layer iteration is performed by the gradient descent method to train the multilayer perceptron, which specifically includes: The sample fluctuations in the sample intrinsic mode are input into the hidden layer of the multi-layer perceptron for forward propagation, and features are extracted layer by layer; According to the loss function of the multi-layer perceptron, the weights and biases of each hidden layer are reversely obtained using the gradient descent method, and the update is continuously iterated until the preset training cycle is reached.

[0011] In combination with the first aspect, in the first implementation of the first aspect, the hidden layer of the multilayer perceptron is set to 2 layers, the number of neurons in the first hidden layer is 128, the number of neurons in the second hidden layer is 32, the activation function of the hidden layer is the Relu function, and the number of neurons in the output layer of the multilayer perceptron is 1.

[0012] According to a second aspect, an embodiment of the present invention further provides a battery remaining life prediction device, the device comprising: A sequence extraction module is used to obtain the charge and discharge cycle data of the lithium battery to be tested, and pre-process the charge and discharge cycle data to extract the capacity sequence features in the charge and discharge cycle data; A capacity prediction module is used to input the capacity sequence features into the trained battery remaining life prediction model to obtain the remaining battery capacity of the lithium battery to be tested after each subsequent charge and discharge cycle output by the battery remaining life prediction model; The battery remaining life prediction model is obtained by training the sample charge and discharge cycle data of the sample lithium battery. The battery remaining life prediction model includes a double Gaussian model and a multi-layer perceptron after the attenuation parameters are adjusted by the Levenberg-Marquardt method algorithm and the particle swarm adjustment algorithm; the double Gaussian model is used to predict the degradation process of the long-term trend sequence to obtain the long-term degradation trend of the subsequent charge and discharge cycles of the lithium battery to be tested, and the multi-layer perceptron is used to predict the fluctuation of the capacity sequence characteristics to obtain the fluctuation sequence of the subsequent charge and discharge cycles of the lithium battery to be tested. The remaining battery capacity is obtained by combining the long-term degradation trend and the fluctuation sequence; The long-term trend series is obtained by performing variational mode decomposition on the characteristics of the capacity series, and the fluctuation of the characteristics of the capacity series is obtained by performing variational mode decomposition on the characteristics of the capacity series; A cycle determination module, used to determine the number of charge and discharge cycles required for the remaining battery capacity of the lithium battery to be detected to reach a preset capacity based on the ratio of the remaining battery capacity of the lithium battery to be detected after each sample charge and discharge cycle to the battery capacity of the initial sample charge and discharge cycle; The life prediction module is used to determine the remaining battery cycle life of the current charge and discharge cycle based on the number of charge and discharge cycles required when the remaining battery capacity reaches the preset capacity and the number of charge and discharge cycles corresponding to the current charge and discharge cycle.

[0013] According to the third aspect, an embodiment of 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, the steps of any of the above-mentioned methods for predicting remaining battery life are implemented.

[0014] According to a fourth aspect, an embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of any of the above-mentioned methods for predicting remaining battery life.

[0015] The battery remaining life prediction method, device, equipment and medium of the present invention adopt a multi-scale modal decomposition method, and use variational modal decomposition to decompose the capacity sequence characteristics extracted from the charge and discharge cycle data of the lithium battery to be detected into a long-term trend sequence and a series of fluctuations, so as to avoid modal aliasing and effectively enhance the key feature information of the data; by considering the capacity sequence characteristics as the weighting of the double Gaussian function, the LM algorithm is combined with the PSO particle swarm algorithm to accurately optimize the attenuation parameters of the double Gaussian model, effectively fit the long-term trend sequence, and then use the characteristics of the MLP structure, which is simple, easy to implement and fast in calculation speed, to effectively perceive the fluctuation of the capacity sequence characteristics; considering the poor generalization performance of a single model and the problem that it is difficult to accurately fit the degradation trend and fluctuation of the original capacity sequence, the VMD, LM-PSO algorithm, the double Gaussian model and the MLP are pre-integrated, so that the battery remaining life prediction model has robustness and higher prediction accuracy for the prediction of the battery remaining cycle life. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings: Figure 1 A schematic diagram showing the flow of a method for predicting remaining battery life provided by the present invention is shown; Figure 2 The battery remaining life prediction method provided by the present invention is decomposed by the VMD algorithm. Schematic diagram of the mode; Figure 3 The battery remaining life prediction method provided by the present invention is decomposed by the VMD algorithm. Schematic diagram of the mode; Figure 4 The battery remaining life prediction method provided by the present invention is decomposed by the VMD algorithm. Schematic diagram of the mode; Figure 5 A schematic diagram showing the fitting results in the battery remaining life prediction method provided by the present invention; Figure 6 A schematic diagram showing the prediction results of the battery remaining life prediction method provided by the present invention; Figure 7 A schematic diagram showing the verification results of the battery remaining life prediction method provided by the present invention; Figure 8 The structure diagram of the battery remaining life prediction device provided by the present invention is shown; Fig. 9 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0018] Lithium-ion batteries are widely used as energy sources in many fields due to their high energy and power density, low self-discharge rate, low price and long life. Similar to many other electrochemical systems, the internal resistance of lithium batteries will gradually increase after repeated charge and discharge cycles. The increase in internal resistance will cause the battery to generate significant heat during the charge and discharge process, which will cause irreversible aging and performance degradation of lithium batteries. That is, lithium-ion batteries will inevitably experience energy and power attenuation during use and storage. Therefore, in battery management systems, the remaining capacity at the beginning of the battery life and the ratio of the internal resistance to its initial value are often used as SOH indicators. This indicator can be used to indicate the remaining life of lithium batteries. As the number of charge and discharge cycles of lithium batteries increases, SOH will decrease, that is, the remaining life of the battery will decrease and the available capacity will gradually decrease. When SOH drops to a certain threshold, such as below 80%, it is necessary to stop using the lithium battery or limit its use.

[0019] Considering the application of lithium-ion batteries in more and more fields, such as battery design adjustment, life prediction and warranty cost estimation, the prediction of their degradation state has received increasing attention. How to accurately use the capacity data series at the beginning of the battery life to predict the end of the battery life is an important issue that needs to be urgently solved in the industry.

[0020] In order to solve the above problems, a battery remaining life prediction method is provided in this specification, which aims to solve the above-mentioned capacity attenuation and capacity regeneration problems during the use of lithium-ion batteries, so as to effectively complete the task of battery cycle life prediction according to the battery capacity sequence, and accurately predict the remaining cycle life of lithium batteries. The battery remaining life prediction method provided in this specification can be applied to electronic devices capable of predicting the SOH of lithium batteries. The electronic devices may include notebooks, desktop computers, smart phones, smart wearable devices (virtual reality glasses, smart watches, etc.), tablet computers, etc. Of course, the battery remaining life prediction method provided in this specification can also be applied to applications running in the above-mentioned electronic devices. For example, the battery remaining life prediction method can be applied to a browser capable of predicting the SOH of lithium batteries, and can also be applied to software capable of predicting the SOH of lithium batteries. Figure 1 FIG. 1 is a flow chart of a method for predicting remaining battery life according to an embodiment of the present invention. Figure 1 As shown, the method may include the following steps: S10, acquiring the charge and discharge cycle data of the lithium battery to be tested, and preprocessing the charge and discharge cycle data to extract the capacity sequence features in the charge and discharge cycle data.

[0021] In this embodiment, during preprocessing, capacity sequence features are first screened out from the sample charge and discharge cycle data. The capacity sequence features include: temperature, current, voltage, power, time and battery capacity.

[0022] In this embodiment, the charge and discharge cycle data of the lithium battery to be tested may be stored in advance in the electronic device, or may be acquired by the electronic device from the outside. For example, the electronic device acquires the charge and discharge cycle data from an external acquisition device, and the specific acquisition form of the charge and discharge cycle data is not limited, and it is only necessary to ensure that the electronic device can acquire the charge and discharge cycle data.

[0023] S20, inputting the capacity sequence features into the trained battery remaining life prediction model to obtain the remaining battery capacity of the lithium battery to be tested after each subsequent charge and discharge cycle output by the battery remaining life prediction model.

[0024] In this embodiment, the battery remaining life prediction model has two major modules, namely a double Gaussian model after attenuation parameters are adjusted by the Levenberg-Marquardt (LM) algorithm and the particle swarm optimization (PSO) algorithm, and a multilayer perceptron (MLP).

[0025] The double Gaussian model is used to predict the degradation process of the long-term trend sequence and obtain the long-term degradation trend of the subsequent charge and discharge cycles of the lithium battery to be tested. The application of the double Gaussian model on the long-term trend sequence can eliminate the impact of short-term fluctuations caused by capacity regeneration on the long-term trend of the predicted sequence. At the same time, the LM-PSO algorithm is used to adjust the parameters of the double Gaussian model to effectively fit the long-term trend sequence, thereby reducing the prediction error; MLP is used to predict the fluctuations of the capacity sequence characteristics and obtain the fluctuation sequence of the subsequent charge and discharge cycles of the lithium battery to be tested. MLP will fit the fluctuations of the original capacity sequence characteristics, which are displayed through a series of intrinsic mode functions and represent the short-term fluctuations in the original capacity sequence. The remaining battery capacity is obtained by combining the long-term degradation trend and the fluctuation sequence.

[0026] Among them, the long-term trend series is obtained by performing variational mode decomposition (VMD) on the characteristics of the capacity series, and the fluctuation of the characteristics of the capacity series is obtained by performing VMD on the characteristics of the capacity series.

[0027] The VMD algorithm can decompose the sample capacity sequence characteristics into a preset number of eigenmodes with different center frequencies and limited bandwidths. In this embodiment, the number of modes decomposed by the VMD algorithm is set to 3. After the capacity sequence characteristics are extracted from the charge and discharge cycle data of the lithium battery to be tested, the VMD algorithm is used to perform modal decomposition to obtain three modes: , such as Figures 2 to 4 As shown. Among them, A long-term trend sequence that represents the characteristics of the capacity sequence, and represents the fluctuation of the capacity series characteristics, that is, The first fluctuation that characterizes the capacity series, The second fluctuation representing the characteristics of the capacity sequence. VMD can effectively decouple different information components in the original capacity sequence and reduce the complexity of the signal.

[0028] The integration of the above two modules can effectively reduce the prediction error of the battery remaining life prediction model for subsequent cycles.

[0029] S30, determining the number of charge and discharge cycles required for the remaining battery capacity of the lithium battery to be tested to reach a preset capacity based on the ratio of the remaining battery capacity of the lithium battery to be tested after each sample charge and discharge cycle to the battery capacity of the initial sample charge and discharge cycle.

[0030] S40, determining the remaining battery cycle life of the current charge and discharge cycle according to the number of charge and discharge cycles required for the remaining battery capacity to reach the preset capacity and the number of charge and discharge cycles corresponding to the current charge and discharge cycle. The remaining battery cycle life indicates how many charge and discharge cycles the sample lithium battery can still experience.

[0031] Specific:

[0032] in, Indicates the remaining battery cycle life of the current charge and discharge cycle; Indicates the number of charge and discharge cycles required for the remaining battery capacity to reach the preset capacity, that is, using Represents the number of charge and discharge cycles required when the remaining battery capacity drops below a certain cycle capacity; Indicates the number of charge and discharge cycles corresponding to the current charge and discharge cycle, that is, the current charge and discharge cycle is the number of charge and discharge cycles.

[0033] Preferably, the preset capacity is set to 80% of the initial battery capacity, that is, when the predicted capacity value drops to 80% of the initial capacity value of the initial capacity sequence, it is considered that the life of the lithium battery to be tested has reached the end point EOL.

[0034] In this embodiment, the battery remaining life prediction model is trained by the following steps: A10. Obtain sample charge and discharge cycle data of the sample lithium battery, preprocess the sample charge and discharge cycle data, and extract sample capacity sequence features from the sample charge and discharge cycle data.

[0035] Similarly, in this embodiment, the sample charge and discharge cycle data of the sample lithium battery may be stored in advance in the electronic device, or may be obtained by the electronic device from the outside. For example, the electronic device obtains the sample charge and discharge cycle data from an external acquisition device, and there is no restriction on the specific acquisition form of the sample charge and discharge cycle data. It is only necessary to ensure that the electronic device can obtain the sample charge and discharge cycle data. By collecting all the sample charge and discharge cycle data of the offline sample lithium battery as sample data, a large amount of data support is provided for subsequent training.

[0036] A20. Decompose the sample volume sequence features into a preset number of sample eigenmodes with different center frequencies and finite bandwidths. VMD decomposes the signal into a finite number of modes, each of which represents a component with a specific frequency and amplitude in the signal.

[0037]

[0038] in, Indicates time; express The sample eigenmodes at time t; express The instantaneous amplitude at a moment; express The instantaneous phase of a moment; Indicates a preset quantity, such as 3.

[0039] Step A20 specifically includes: A21. Decompose the sample capacity sequence characteristics into a preset number of sample eigenmodes with different center frequencies and finite bandwidths through the VMD algorithm.

[0040] A22. Perform Hilbert transform on each sample eigenmode to obtain the single-sided spectrum corresponding to the sample eigenmode.

[0041]

[0042] in, express The one-sided spectrum corresponding to the sample eigenmode at time; represents the impulse function; Is an imaginary unit.

[0043] A23. Modulate each single-sided spectrum and establish constraints on the modulated single-sided spectrum.

[0044]

[0045] in, express The broadband range corresponding to the unilateral spectrum at the moment; Indicates the center frequency.

[0046] Establishing the constraint is to calculate the L2 norm of the above broadband range, and the corresponding calculation formula is:

[0047] in, express The partial derivative of .

[0048] A24. Using preset penalty items and the Lagrangian factor Convert the established constraints into unconstrained ones, and the corresponding calculation formula is:

[0049] In this embodiment, the optimal solution of the unconstrained problem is derived by the Alternating Direction Method of Multipliers (ADMM), and in order to update The subproblem of the unconstrained problem is converted into a minimum problem to solve the following equation, and then the sample eigenmode is obtained The optimal solution of .

[0050]

[0051] Therefore, the sample eigenmodes The optimal solution is converted to:

[0052] in, express Fourier transform of express Fourier transform of express Fourier transform of express The Fourier transform of .

[0053] A30. The long-term trend sequence of samples in the sample intrinsic mode is used as the input data for training, and the attenuation parameters of the double Gaussian model are adjusted through the LM-PSO algorithm to train the double Gaussian model.

[0054] A40. The sample fluctuations in the sample intrinsic mode are used as input data for training. The features are extracted layer by layer by forward propagation and the MLP is trained by iterating layer by layer through the gradient descent method.

[0055] The model setting of the multilayer perceptron is: the hidden layer is set to 2 layers, the number of neurons in the first hidden layer is 128, the number of neurons in the second hidden layer is 32, the number of neurons in the output layer is 1, and the activation function of the hidden layer is the Relu function.

[0056] In this embodiment, during the training of the battery remaining life prediction model, the adaptive moment estimation (Adaptive Moment Estimation, Adam) optimizer in the gradient descent method with an adaptive learning rate is used to train the MLP, and the learning rate is set to 0.001. This method can effectively handle the adjustment problem of large-scale data. In order to fully train the model and achieve a good convergence effect, the training cycle (epoch) is set to 500 to obtain better model performance.

[0057] After the dual Gaussian model and MLP are trained, the battery remaining life prediction model is trained.

[0058] In this embodiment, the attenuation of battery capacity is approximately regarded as the superposition of Gaussian distributions of the number of cycles, and two Gaussian models are used to weightedly describe the attenuation characteristics of the capacity, that is, a double Gaussian model is used to describe the degradation process of the battery capacity sequence. The specific model is:

[0059] in, Indicates the number of sample charge and discharge cycles that the sample lithium battery has undergone; Indicates the remaining battery capacity or current battery capacity of the sample lithium battery; , , , , and are all attenuation parameters of the double Gaussian model, corresponding to the first, second, third, fourth, fifth and sixth attenuation parameters respectively.

[0060] Considering that the attenuation parameters of the double Gaussian model will increase the difficulty of parameter identification, in this embodiment, the LM algorithm is used to adjust the attenuation parameters of the double Gaussian model and solve the unconstrained nonlinear least squares regression problem.

[0061] The training process of the LM algorithm for the double Gaussian model specifically includes: B10. Establish an initial set of attenuation parameters for the double Gaussian model , The values ​​in correspond to the initial values ​​of the first, second, third, fourth, fifth and sixth attenuation parameters respectively.

[0062] B20. Determine the residual between the predicted value and the actual value of the double Gaussian model , using the residual It can be used to represent the difference between the predicted value and the actual value. In this embodiment, the residual between the predicted value and the actual value of the double Gaussian model is calculated and the attenuation parameter of the double Gaussian model is preliminarily adjusted. The residual It can be expressed as:

[0063] in, Represents data points The true value of Represents the predicted value of the double Gaussian model calculated based on the current parameter values.

[0064] B30, construct Jacobian matrix, Jacobian matrix is the residual The matrix of partial derivatives with respect to the model parameters. For the double Gaussian model, the Jacobian matrix Each element of It can be expressed as:

[0065] in, represents the first An attenuation parameter, which is one of the attenuation parameters of the double Gaussian model; express The partial derivative of express The partial derivative of .

[0066] B40, using the Jacobian matrix Constructing the approximate Hessian matrix . Approximate Hessian matrix It can be expressed as:

[0067] in, represents the Jacobian matrix The transpose of represents the identity matrix; represents the damping factor, Used to strike a balance between the rapid descent method and the Gauss–Newton method.

[0068] B50, according to the approximate Hessian matrix and residual Update the attenuation parameters of the double Gaussian model, specifically:

[0069] in, Represents the attenuation parameter Relative to the attenuation parameter The update amount.

[0070] From this we can solve the linear system and obtain:

[0071] B60, Determine attenuation parameters Update volume Is it less than the preset value or the number of iterations of parameter update reaches the first preset number. Update volume When it is less than a preset value, the convergence condition is met, and the first preset number can be understood as the maximum number of iterations of the LM algorithm.

[0072] B70. If the convergence condition is met or the first preset number of times is reached, stop updating the attenuation parameter; otherwise, repeat steps B20 to B60 until the convergence condition is met or the first preset number of times is reached.

[0073] After the LM algorithm preliminarily adjusts the attenuation parameters of the dual Gaussian model, in this embodiment, the PSO algorithm is used to further adjust the attenuation parameters of the preliminarily adjusted Gaussian model.

[0074] The training process of the PSO algorithm for the double Gaussian model specifically includes: C10, by adding a preset disturbance to the attenuation parameter adjusted by the LM algorithm, the initial position and velocity of the particles in the PSO algorithm are initialized to obtain the initial position and velocity of each particle. Initial position and initial velocity .

[0075] C20, determine the algorithm parameters in the PSO algorithm, the algorithm parameters include inertia parameter, first learning factor, second learning factor and second number of iterations. The second preset number can be understood as the maximum number of iterations of the PSO algorithm. For example, the specific setting is inertia weight w=0.5, first learning factor =1, the second learning factor = 2 and the second iteration number =100.

[0076] C30, determine the fitness value of each particle according to the preset fitness function, and update the individual optimal position of each particle according to the fitness value and the optimal position of the group , until the second iteration is reached =100 or the fitness value of the particle is less than the preset residual sum.

[0077] In each iteration, For each particle and each dimension , dimension Corresponding to a decay parameter, the particle The speed is updated as follows:

[0078] in, represents the inertia weight, It can be used to indicate the extent to which a particle retains its previous velocity; Indicates Iteration Particles in the Speed ​​in dimension; Indicates Iteration Particles in the Speed ​​in dimension; Shidi Iteration Particles in the Position in dimension; Indicates Particles in the The optimal position of individuals on the dimension; Indicates Particles in the The optimal position of the group in the dimension; represents the first exploration coefficient, represents the second exploration coefficient, and Both are random numbers between 0 and 1, used to increase the randomness of the search.

[0079] particle The location is updated as follows:

[0080] in, Indicates Iteration Particles in the Position in dimension.

[0081] Fitness function To evaluate each particle That is, the quality of each set of attenuation parameters. For the double Gaussian model, the residual can be calculated The sum of squares is used to measure the goodness of fit, and the parameters are adjusted by minimizing the residual sum of squares, trying to find the minimum residual sum to obtain the corresponding best fitting parameter set. Specifically:

[0082] in, , , , , and They are , , , , and The corresponding attenuation parameters after adjustment by LM algorithm; It is the independent variable data set of sample charge and discharge cycle data; It is also the true value of the sample capacity sequence characteristic of the sample lithium battery; represents the double Gaussian model, That is, the long-term degradation trend of the sample predicted by the double Gaussian model, and the residual of the double Gaussian model That is and The difference between, that is:

[0083] In the iterative process of the PSO particle swarm algorithm, the fitness value of each particle is calculated according to the fitness function, and the individual optimal position of the particle is updated. and the optimal position of the group When the second iteration is reached =100 or the fitness value of the particle is less than the preset residual sum, the PSO particle swarm algorithm iteration stops.

[0084] By integrating the LM algorithm and the PSO particle swarm algorithm to adjust the attenuation parameters of the double Gaussian model, the long-term degradation trend can be effectively fitted. The fitting results are shown in Figure 5 shown.

[0085] The training process of MLP specifically includes: D10. The sample fluctuation in the sample eigenmode is the first fluctuation and the second wave The input is integrated into the hidden layer of MLP for forward propagation to extract features layer by layer.

[0086] The forward propagation formula is:

[0087] in, represents the weighted sum of the inputs to the neurons in the output layer of the MLP; Indicates The result after all inputs of the node are weighted; represents the activation function; represents the first Tier The weights of the neurons; Indicates Tier The output of a neuron; Indicates Tier The bias of a neuron.

[0088] D20, according to the loss function of MLP and the gradient descent method, the weight of each hidden layer is reversed and bias , and continuously iterate and update until the preset training cycle is reached.

[0089] In order to evaluate the performance of the battery remaining life prediction model and make adjustments, in this embodiment, MLP uses mean square error (MSE) as the loss function to intuitively reflect the degree of difference between the predicted value and the actual value. Specifically:

[0090] in, The total number of sample size features representing the true and predicted values; Represents the true value of the sample capacity sequence characteristic of the sample lithium battery; It represents the predicted sample fluctuation sequence after MLP processing.

[0091] The MLP parameters that meet the accuracy requirements are obtained layer by layer through the gradient descent method to obtain the weights and biases of the neurons.

[0092] When training the model, the first 25% of the historical charge and discharge cycle data collected from the selected lithium battery is allocated as a training set to train and adjust the battery remaining life prediction model, while the remaining data is used as an independent test set to evaluate the prediction and generalization performance of the battery remaining life prediction model, thereby verifying the effectiveness of the model in complex battery charge and discharge cycles. For example, a total of 1,600 historical charge and discharge cycle data of a sample lithium battery of model A are obtained, and the first 400 historical charge and discharge cycle data are used as sample charge and discharge cycle data. The battery remaining life prediction model is used to perform modal decomposition and fit the capacity sequence curve to predict the battery capacity sequence of subsequent cycles, and the data is compared with the remaining 1,200 historical charge and discharge cycle data.

[0093] During the verification, the prediction cycle error was used as the core evaluation indicator, and the constructed joint prediction model was compared with the single Holt-winters prediction research model. In the remaining cycle life prediction task of battery data set A, the battery remaining life end predicted by the joint prediction model differed by 3 cycles from the actual life end, and the predicted battery life end was close to the actual situation. The intuitive display of the prediction results is as follows Figure 6 The presented.

[0094] VMD variational mode decomposition decomposes the original capacity sequence into a long-term trend sequence and a series of intrinsic mode functions. The application of the double Gaussian model on the long-term trend sequence can eliminate the impact of short-term fluctuations caused by capacity regeneration on the long-term trend of the predicted sequence. At the same time, the LM-PSO algorithm is used to adjust the parameters of the double Gaussian model to effectively fit the long-term trend sequence, thereby reducing the prediction error. For the intrinsic mode function representing the short-term fluctuations in the original capacity sequence, a multi-layer perceptron network is used for fitting. Directly applying the Holt-winters algorithm to fit the long-term trend on the original capacity sequence has a large experimental error, and the prediction error of the end of battery life reaches 115 cycles. The verification results are as follows: Figure 7 shown.

[0095] The battery remaining life prediction method of the present invention adopts a multi-scale mode decomposition method, and uses variational mode decomposition to decompose the capacity sequence characteristics extracted from the charge and discharge cycle data of the lithium battery to be detected into a long-term trend sequence and a series of fluctuations, so as to avoid mode aliasing and effectively enhance the key feature information of the data; by considering the capacity sequence characteristics as the weighting of the double Gaussian function, the LM algorithm is combined with the PSO particle swarm algorithm to accurately optimize the attenuation parameters of the double Gaussian model, effectively fit the long-term trend sequence, and then use the characteristics of the MLP structure, which is simple, easy to implement and fast in calculation speed, to effectively perceive the fluctuation of the capacity sequence characteristics; considering the problem that a single model has poor generalization performance and is difficult to accurately fit the degradation trend and fluctuation of the original capacity sequence, the VMD, LM-PSO algorithm, double Gaussian model and MLP are pre-integrated, so that the battery remaining life prediction model has robustness and higher prediction accuracy for the prediction of the battery remaining cycle life.

[0096] The following is a description of a battery remaining life prediction device provided by an embodiment of the present invention. The battery remaining life prediction device described below and the battery remaining life prediction method described above can be referenced to each other.

[0097] In order to solve the above problems, a battery remaining life prediction device is provided in this specification, aiming to. Figure 8 is a schematic diagram of the structure of a battery remaining life prediction device according to an embodiment of the present invention. Figure 8 As shown, the device may include: The sequence extraction module 10 is used to obtain the charge and discharge cycle data of the lithium battery to be tested, and pre-process the charge and discharge cycle data to extract the capacity sequence features in the charge and discharge cycle data.

[0098] In this embodiment, during preprocessing, capacity sequence features are first screened out from the sample charge and discharge cycle data. The capacity sequence features include: temperature, current, voltage, power, time and battery capacity.

[0099] In this embodiment, the charge and discharge cycle data of the lithium battery to be tested may be stored in advance in the electronic device, or may be acquired by the electronic device from the outside. For example, the electronic device acquires the charge and discharge cycle data from an external acquisition device, and the specific acquisition form of the charge and discharge cycle data is not limited, and it is only necessary to ensure that the electronic device can acquire the charge and discharge cycle data.

[0100] The capacity prediction module 20 is used to input the capacity sequence features into the trained battery remaining life prediction model to obtain the remaining battery capacity of the lithium battery to be tested after each subsequent charge and discharge cycle output by the battery remaining life prediction model.

[0101] In this embodiment, the battery remaining life prediction model has two major modules, namely, a double Gaussian model after attenuation parameters are adjusted by the LM algorithm and the PSO algorithm, and an MLP.

[0102] The double Gaussian model is used to predict the degradation process of the long-term trend sequence and obtain the long-term degradation trend of the subsequent charge and discharge cycles of the lithium battery to be tested. The application of the double Gaussian model on the long-term trend sequence can eliminate the impact of short-term fluctuations caused by capacity regeneration on the long-term trend of the predicted sequence. At the same time, the LM-PSO algorithm is used to adjust the parameters of the double Gaussian model to effectively fit the long-term trend sequence, thereby reducing the prediction error; MLP is used to predict the fluctuations of the capacity sequence characteristics and obtain the fluctuation sequence of the subsequent charge and discharge cycles of the lithium battery to be tested. MLP will fit the fluctuations of the original capacity sequence characteristics, which are displayed through a series of intrinsic mode functions and represent the short-term fluctuations in the original capacity sequence. The remaining battery capacity is obtained by combining the long-term degradation trend and the fluctuation sequence.

[0103] Among them, the long-term trend sequence is obtained by performing VMD on the capacity sequence characteristics, and the fluctuation of the capacity sequence characteristics is obtained by performing VMD on the capacity sequence characteristics.

[0104] The VMD algorithm can decompose the sample capacity sequence characteristics into a preset number of eigenmodes with different center frequencies and limited bandwidths. In this embodiment, the number of modes decomposed by the VMD algorithm is set to 3. After the capacity sequence characteristics are extracted from the charge and discharge cycle data of the lithium battery to be tested, the VMD algorithm is used to perform modal decomposition to obtain three modes: , such as Figures 2 to 4 As shown. Among them, A long-term trend sequence that represents the characteristics of the capacity sequence, and represents the fluctuation of the capacity series characteristics, that is, The first fluctuation that characterizes the capacity series, The second fluctuation representing the characteristics of the capacity sequence. VMD can effectively decouple different information components in the original capacity sequence and reduce the complexity of the signal.

[0105] The integration of the above two modules can effectively reduce the prediction error of the battery remaining life prediction model for subsequent cycles.

[0106] The cycle determination module 30 is used to determine the number of charge and discharge cycles required when the remaining battery capacity of the lithium battery to be detected reaches a preset capacity based on the ratio of the remaining battery capacity of the lithium battery to be detected after each sample charge and discharge cycle to the battery capacity of the initial sample charge and discharge cycle.

[0107] The life prediction module 40 is used to determine the remaining battery cycle life of the current charge and discharge cycle according to the number of charge and discharge cycles required when the remaining battery capacity reaches the preset capacity and the number of charge and discharge cycles corresponding to the current charge and discharge cycle. The remaining battery cycle life indicates how many charge and discharge cycles the sample lithium battery can still experience.

[0108] The battery remaining life prediction device of the present invention adopts a multi-scale modal decomposition method, and uses variational modal decomposition to decompose the capacity sequence characteristics extracted from the charge and discharge cycle data of the lithium battery to be detected into a long-term trend sequence and a series of fluctuations, so as to avoid modal aliasing and effectively enhance the key feature information of the data; by considering the capacity sequence characteristics as the weighting of the double Gaussian function, the LM algorithm is combined with the PSO particle swarm algorithm to accurately optimize the attenuation parameters of the double Gaussian model, effectively fit the long-term trend sequence, and then use the characteristics of the MLP structure, which is simple, easy to implement and fast in calculation speed, to effectively perceive the fluctuation of the capacity sequence characteristics; considering the poor generalization performance of a single model and the problem that it is difficult to accurately fit the degradation trend and fluctuation of the original capacity sequence, the VMD, LM-PSO algorithm, double Gaussian model and MLP are pre-integrated, so that the battery remaining life prediction model has robustness and higher prediction accuracy for the prediction of the battery remaining cycle life.

[0109] Fig. 9 An example of a physical structure diagram of an electronic device is shown in FIG. Fig. 9 As shown, the electronic device may include: a processor 910 (processor), a communication interface 920 (Communications Interface), a memory 930 (memory) and a communication bus 940, wherein the processor 910, the communication interface 920, and the memory 930 communicate with each other through the communication bus 940. The processor 910 may call the logic command in the memory 930 to execute the battery remaining life prediction method, which includes: Acquire the charge and discharge cycle data of the lithium battery to be tested, pre-process the charge and discharge cycle data, and extract the capacity sequence features in the charge and discharge cycle data; Input the capacity sequence features into the trained battery remaining life prediction model to obtain the remaining battery capacity of the lithium battery to be tested after each subsequent charge and discharge cycle output by the battery remaining life prediction model; Determine the number of charge and discharge cycles required for the remaining battery capacity of the lithium battery to be tested to reach a preset capacity according to the ratio of the remaining battery capacity of the lithium battery to be tested after each sample charge and discharge cycle to the battery capacity of the initial sample charge and discharge cycle; The remaining battery cycle life of the current charge and discharge cycle is determined based on the number of charge and discharge cycles required when the remaining battery capacity reaches the preset capacity and the number of charge and discharge cycles corresponding to the current charge and discharge cycle.

[0110] In addition, the logic instructions in the above-mentioned memory 930 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several 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 Memor), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0111] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, when the program instructions are executed by a computer, the computer can execute the battery remaining life prediction method provided by the above methods, the method comprising: Acquire the charge and discharge cycle data of the lithium battery to be tested, pre-process the charge and discharge cycle data, and extract the capacity sequence features in the charge and discharge cycle data; Input the capacity sequence features into the trained battery remaining life prediction model to obtain the remaining battery capacity of the lithium battery to be tested after each subsequent charge and discharge cycle output by the battery remaining life prediction model; Determine the number of charge and discharge cycles required for the remaining battery capacity of the lithium battery to be tested to reach a preset capacity according to the ratio of the remaining battery capacity of the lithium battery to be tested after each sample charge and discharge cycle to the battery capacity of the initial sample charge and discharge cycle; The remaining battery cycle life of the current charge and discharge cycle is determined based on the number of charge and discharge cycles required when the remaining battery capacity reaches the preset capacity and the number of charge and discharge cycles corresponding to the current charge and discharge cycle.

[0112] In another aspect, the present invention further 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 remaining battery life provided above is implemented, and the method includes: Acquire the charge and discharge cycle data of the lithium battery to be tested, pre-process the charge and discharge cycle data, and extract the capacity sequence features in the charge and discharge cycle data; Input the capacity sequence features into the trained battery remaining life prediction model to obtain the remaining battery capacity of the lithium battery to be tested after each subsequent charge and discharge cycle output by the battery remaining life prediction model; Determine the number of charge and discharge cycles required for the remaining battery capacity of the lithium battery to be tested to reach a preset capacity according to the ratio of the remaining battery capacity of the lithium battery to be tested after each sample charge and discharge cycle to the battery capacity of the initial sample charge and discharge cycle; The remaining battery cycle life of the current charge and discharge cycle is determined based on the number of charge and discharge cycles required when the remaining battery capacity reaches the preset capacity and the number of charge and discharge cycles corresponding to the current charge and discharge cycle.

[0113] 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.

[0114] 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, 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.

[0115] 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 method for predicting remaining battery life, characterized in that: The method comprises: Acquire the charge and discharge cycle data of the lithium battery to be tested, pre-process the charge and discharge cycle data, and extract the capacity sequence features in the charge and discharge cycle data; Input the capacity sequence features into the trained battery remaining life prediction model to obtain the remaining battery capacity of the lithium battery to be tested after each subsequent charge and discharge cycle output by the battery remaining life prediction model; The battery remaining life prediction model is obtained by training the sample charge and discharge cycle data of the sample lithium battery. The battery remaining life prediction model includes a double Gaussian model and a multi-layer perceptron after the attenuation parameters are adjusted by the Levenberg-Marquardt method algorithm and the particle swarm adjustment algorithm; the double Gaussian model is used to predict the degradation process of the long-term trend sequence to obtain the long-term degradation trend of the subsequent charge and discharge cycles of the lithium battery to be tested, and the multi-layer perceptron is used to predict the fluctuation of the capacity sequence characteristics to obtain the fluctuation sequence of the subsequent charge and discharge cycles of the lithium battery to be tested. The remaining battery capacity is obtained by combining the long-term degradation trend and the fluctuation sequence; The long-term trend series is obtained by performing variational mode decomposition on the characteristics of the capacity series, and the fluctuation of the characteristics of the capacity series is obtained by performing variational mode decomposition on the characteristics of the capacity series; Determine the number of charge and discharge cycles required for the remaining battery capacity of the lithium battery to be tested to reach a preset capacity according to the ratio of the remaining battery capacity of the lithium battery to be tested after each sample charge and discharge cycle to the battery capacity of the initial sample charge and discharge cycle; The remaining battery cycle life of the current charge and discharge cycle is determined based on the number of charge and discharge cycles required when the remaining battery capacity reaches the preset capacity and the number of charge and discharge cycles corresponding to the current charge and discharge cycle.

2. The method for predicting remaining battery life according to claim 1, characterized in that: The battery remaining life prediction model is trained by the following steps: Obtaining sample charge and discharge cycle data of a sample lithium battery, preprocessing the sample charge and discharge cycle data, and extracting sample capacity sequence features from the sample charge and discharge cycle data; Decomposing the sample volume sequence characteristics into a preset number of sample eigenmodes with different center frequencies and limited bandwidths; The sample long-term trend sequence in the sample intrinsic mode is used as the input data for training, and the attenuation parameters of the double Gaussian model are adjusted by the Levenberg-Marquardt algorithm and the particle swarm adjustment algorithm to train the double Gaussian model. The sample fluctuations in the sample intrinsic mode are used as input data for training. The features are extracted layer by layer through forward propagation and the multilayer perceptron is trained by iterating layer by layer through the gradient descent method.

3. The method for predicting remaining battery life according to claim 2, characterized in that: Decomposing the sample capacity sequence characteristics into a preset number of sample eigenmodes with different center frequencies and limited bandwidths specifically includes: Decomposing the sample capacity sequence characteristics into a preset number of sample eigenmodes with different center frequencies and limited bandwidths by using a variational mode decomposition algorithm; the sample eigenmodes include a sample long-term trend sequence and two sample fluctuations; Hilbert transform is performed on each sample eigenmode to obtain the single-sided spectrum corresponding to the sample eigenmode; Each single-sided spectrum is modulated, and constraints are established on the modulated single-sided spectrum; The established constraints are transformed into unconstrained ones using the preset penalty terms and Lagrangian factors. The unconstrained optimal solution is determined by the alternating multiplier method, and the unconstrained subproblem is transformed into a minimum value problem to obtain the optimal solution of the sample eigenmode.

4. The method for predicting remaining battery life according to claim 2, characterized in that: The process of the Levenberg-Marquardt algorithm to train the attenuation parameters of the double Gaussian model is: Establish an initial set of attenuation parameters for the double Gaussian model; Determine the residual between the predicted and actual values ​​of the double Gaussian model; Construct the Jacobian matrix based on the residual and the initial decay parameters; Use the Jacobian matrix to construct an approximate Hessian matrix; Update the attenuation parameters of the double Gaussian model according to the approximate Hessian matrix and the residual; Determine whether the update amount of the attenuation parameter is less than a preset value or whether the number of iterations of the parameter update reaches a first preset number; When the update amount of the attenuation parameter is less than the preset value, it means that the convergence condition is met; When it is determined that the convergence condition is met or the first preset number is reached, the attenuation parameter is stopped from being updated; when it is determined that the convergence condition is not met or the first preset number is not reached, the attenuation parameter is continuously updated until the convergence condition is met or the first preset number is reached.

5. The method for predicting remaining battery life according to claim 4, characterized in that: The process of the particle swarm adjustment algorithm training the attenuation parameters of the double Gaussian model is: Adding a preset disturbance to the attenuation parameter adjusted by the Levenberg-Marquardt algorithm, initializing it to the initial position and velocity of the particle, and obtaining the initial position and initial velocity of each particle; Determine algorithm parameters; the algorithm parameters include inertia parameter, first learning factor, second learning factor and second iteration number; The fitness value of each particle is determined according to a preset fitness function, and the individual optimal position and the group optimal position of each particle are updated according to the fitness value until the second iteration number is reached or the fitness value of the particle is less than the preset residual sum.

6. The method for predicting remaining battery life according to claim 2, characterized in that: The sample fluctuations in the sample intrinsic mode are used as input data for training, features are extracted layer by layer by forward propagation, and layer by layer iteration is performed by gradient descent method to train the multilayer perceptron, specifically including: The sample fluctuations in the sample intrinsic mode are input into the hidden layer of the multilayer perceptron for forward propagation, and features are extracted layer by layer; According to the loss function of the multi-layer perceptron, the weights and biases of each hidden layer are reversely obtained using the gradient descent method, and the update is continuously iterated until the preset training cycle is reached.

7. The method for predicting remaining battery life according to claim 1, characterized in that: The hidden layer of the multilayer perceptron is set to 2 layers, the number of neurons in the first hidden layer is 128, the number of neurons in the second hidden layer is 32, the activation function of the hidden layer is selected as the Relu function, and the number of neurons in the output layer of the multilayer perceptron is 1.

8. A battery remaining life prediction device, characterized in that: The device comprises: A sequence extraction module is used to obtain the charge and discharge cycle data of the lithium battery to be tested, and pre-process the charge and discharge cycle data to extract the capacity sequence features in the charge and discharge cycle data; A capacity prediction module is used to input the capacity sequence features into the trained battery remaining life prediction model to obtain the remaining battery capacity of the lithium battery to be tested after each subsequent charge and discharge cycle output by the battery remaining life prediction model; The battery remaining life prediction model is obtained by training the sample charge and discharge cycle data of the sample lithium battery. The battery remaining life prediction model includes a double Gaussian model and a multi-layer perceptron after the attenuation parameters are adjusted by the Levenberg-Marquardt method algorithm and the particle swarm adjustment algorithm; the double Gaussian model is used to predict the degradation process of the long-term trend sequence to obtain the long-term degradation trend of the subsequent charge and discharge cycles of the lithium battery to be tested, and the multi-layer perceptron is used to predict the fluctuation of the capacity sequence characteristics to obtain the fluctuation sequence of the subsequent charge and discharge cycles of the lithium battery to be tested. The remaining battery capacity is obtained by combining the long-term degradation trend and the fluctuation sequence; The long-term trend series is obtained by performing variational mode decomposition on the characteristics of the capacity series, and the fluctuation of the characteristics of the capacity series is obtained by performing variational mode decomposition on the characteristics of the capacity series; A cycle determination module, used to determine the number of charge and discharge cycles required for the remaining battery capacity of the lithium battery to be detected to reach a preset capacity based on the ratio of the remaining battery capacity of the lithium battery to be detected after each sample charge and discharge cycle to the battery capacity of the initial sample charge and discharge cycle; The life prediction module is used to determine the remaining battery cycle life of the current charge and discharge cycle based on the number of charge and discharge cycles required when the remaining battery capacity reaches the preset capacity and the number of charge and discharge cycles corresponding to the current charge and discharge cycle.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the battery remaining life prediction method according to any one of claims 1 to 7 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the battery remaining life prediction method as claimed in any one of claims 1 to 7 are implemented.

Images