Lithium battery degradation data intelligent generation method based on improved generative adversarial network

Abstract

The application relates to a lithium battery degradation data intelligent generation method based on an improved generative adversarial network, and belongs to the field of battery degradation data generation and equipment health protection, and comprises the following steps: obtaining lithium battery historical degradation data, including discharge voltage and capacity data; normalizing the capacity data, and one-to-one matching the discharge voltage and the capacity data to correspond to each other, constructing feature-label pairs, and forming a training data set; training LSTM and the improved GAN network respectively by using the training data set to obtain a battery capacity prediction model and a degradation data generation model; collecting a lithium battery current cycle discharge voltage curve, and predicting the current capacity of the battery by using the battery capacity prediction model; designing and executing a battery capacity generation mechanism program according to the physical characteristics of a target battery, and setting an expected personalized battery capacity; and inputting the preset target capacity into the degradation data generation model to generate synthesized degradation data.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of battery degradation data generation and device health protection, and relates to a lithium battery degradation data intelligent generation method based on an improved generative adversarial network. BACKGROUND

[0002] Lithium batteries are widely used in many important fields such as consumer electronics, electric vehicles, unmanned aerial vehicles, smart power station energy storage systems and many Internet of Things devices due to their high energy density, decreasing cost, low self-discharge rate and long service life. However, as the number of charge and discharge cycles increases, the performance of lithium batteries gradually degrades, resulting in capacity attenuation and power decline, which not only affects the reliability of the device, but also increases the risk of thermal runaway. Therefore, it is necessary to understand the health status of lithium batteries at all times to ensure stable operation of the system. Currently, the mainstream method for estimating the health status of lithium batteries is a data-driven prediction method. However, this method has high requirements for the quantity and quality of data. However, obtaining complete and representative battery degradation data is time-consuming and costly, and the process is also risky. The lithium battery degradation data intelligent generation method based on the improved generative adversarial network generates lithium battery degradation data with specified capacity results according to user requirements, and provides data support for lithium battery health status estimation models and lithium battery safety monitoring systems. SUMMARY

[0003] Therefore, the purpose of the present application is to provide a lithium battery degradation data intelligent generation method based on an improved generative adversarial network to solve the problem of lithium battery health status estimation data generation.

[0004] To achieve the above purpose, the present application provides the following technical solutions:

[0005] A lithium battery degradation data intelligent generation method based on an improved generative adversarial network comprises the following steps:

[0006] Obtain lithium battery historical degradation data including lithium battery discharge voltage data and lithium battery capacity degradation data using a battery management system;

[0007] Normalize the capacity degradation data and match the discharge voltage and capacity degradation data one by one to form a feature-label pair and form a training data set;

[0008] Train a long short-term memory network (LSTM) and an improved generative adversarial network (GAN) using the training data set to obtain a battery capacity prediction model based on the LSTM network and an intelligent degradation data generation model based on the improved GAN network.

[0009] Collecting the current cycle discharge voltage curve of the lithium battery, and predicting the current capacity of the battery by using the battery capacity prediction model;

[0010] According to the physical characteristics of the target battery, a battery capacity individual generation program is designed and executed to tamper with the predicted current capacity of the battery and generate a desired target capacity;

[0011] The preset target capacity is input into the degradation data intelligent generation model to generate synthesized degradation data.

[0012] Further, the capacity degradation data is normalized, and the formula is as follows:

[0013]

[0014] Wherein, c t represents the actual capacity measured at t cycles, c max represents the rated maximum capacity of the lithium battery in the initial state, represents the normalized capacity degradation data.

[0015] Further, the training data set formation step is as follows:

[0016] Collecting the discharge voltage data V of the lithium battery = {v1, v2, v3, …, v n}, and synchronously recording the normalized capacity degradation data Forming a training data set for health state estimation.

[0017] Further, the training process of the battery capacity prediction model based on the LSTM network is as follows:

[0018] Taking the discharge voltage data V as input, and the corresponding capacity as label for supervised training; wherein the input data is organized as a three-dimensional tensor (batch_size, sequence_length, input_size) by batch, and the output of the LSTM layer is linearly transformed through a fully connected layer to finally output the capacity estimation value; by minimizing the error between the estimation value and the label, a regression relationship of the degradation behavior is established; and the loss function is:

[0019]

[0020] Wherein E ω represents the loss function value, Y t represents the label value of the tth capacity, Ψ LSTM represents the LSTM model, x t represents the discharge voltage data of the tth cycle, and Θw Θ represents a set of parameters of the LSTM model; Θ is updated by a back propagation algorithm w , when E ω stops training when E

[0021] Further, the degradation data intelligent generation model based on the improved GAN network comprises:

[0022] The generator is composed of a multi-layer perception mechanism, and is configured to receive a random noise vector z and condition information y, and generate synthesized discharge voltage data wherein θ g represents trainable parameters of the generator;

[0023] The discriminator is composed of an LSTM network, and is configured to receive input data x, and the formula is:

[0024] D(x; θ d )→[0, 1]

[0025] wherein θ d represents trainable parameters of the discriminator, and the closer the output value is to 1, the higher the probability that the input data is real discharge voltage data;

[0026] The value function is:

[0027]

[0028] wherein x represents a real discharge voltage data sample, and is subject to a real sample distribution P r , z represents random noise, and is subject to a uniform distribution z~U(-0.01, 0.01), y represents capacity degradation information as condition information of the GAN network, D(·) is the discriminator, D(x|y) represents the score of the discriminator on the input x being real data, G(·) is the generator, G(z|y) is a sample generated by the generator based on random noise z and condition y, λ is a weight parameter, and controls the weight of the gradient penalty term, is a sample obtained by linear interpolation of real data x and generated data G(z), i.e. wherein ∈~U(0, 1), represents the gradient module length of the discriminator on the input sample;

[0029] The generator loss function is:

[0030]

[0031] wherein, represents synthesized discharge voltage data generated by the generator, and satisfies z represents a random noise vector subject to a uniform distribution, y is capacity degradation condition information, P g is a generated data distribution;

[0032] The discriminator loss function is:

[0033]

[0034] where x is the real discharge voltage data, which is subject to distribution P r , is an interpolation sample, and λ is a gradient penalty coefficient.

[0035] Further, the training process of the degradation data intelligent generation model based on the improved GAN network is as follows:

[0036] Step (1): randomly initialize the weights of the generator G and the discriminator D;

[0037] Step (2): fix the generator G, and update the discriminator parameters θ D with the Adam optimizer at a learning rate η D to minimize the discriminator loss

[0038] Step (3): fix the discriminator D, and update the generator parameters θ g with the Adam optimizer at a learning rate η G to minimize the generator loss

[0039] Step (4): repeat steps (2)-(3) until the maximum number of iterations T max is reached.

[0040] The improved GAN network minimizes the discriminator loss to measure the Wasserstein distance between the generated data distribution P g and the real data distribution P r , wherein the ratio of the number of times k of each round of training of the discriminator to the number of times of training of the generator is 2:1, and the data distribution output by the generator is approximated to the real data distribution through adversarial training.

[0041] Further, the current cycle discharge voltage curve of the lithium battery is collected, and the battery current capacity is predicted by using the battery capacity prediction model, specifically including:

[0042] The current discharge voltage curve x t is input into the battery capacity prediction model based on the LSTM network to obtain the current predicted capacity The formula is:

[0043]

[0044] where Θ w is a set of solidification parameters of the LSTM model, and ΨLSTM represents the mapping relationship of the LSTM network containing the full connection layer.

[0045] Further, the design of the battery capacity individualization generation program meets the following physical constraint conditions:

[0046] (a) The generated target capacity value C adv satisfies:

[0047] 0 < C adv ≤ C rated

[0048] where C rated is the rated capacity of the battery;

[0049] (b) The capacity change amount △C between adjacent cycles satisfies:

[0050] |△C| < δ max

[0051] where δ max is the maximum allowed capacity fluctuation threshold preset according to the battery chemical characteristics.

[0052] Further, the execution process of the battery capacity individualization generation program includes:

[0053] According to the predicted capacity generate the expected target capacity The formula is:

[0054]

[0055] where GEN(·) is the user-set expected capacity generation function;

[0056] The expected capacity generation function GEN(·) is divided into three categories: incremental modification, decremental modification, or peak value modification according to different modification forms, and the formulas are respectively:

[0057] Peak value modification:

[0058] Incremental modification:

[0059] Decremental modification:

[0060] where, represents the preset target capacity, represents the capacity obtained by the LSTM network, ε represents the set modification amplitude, and k(t) represents the continuous modification strategy.

[0061] Further, the generated synthetic degradation data specifically includes:

[0062] The preset target capacity The data, along with random noise z, is fed into a degraded data intelligent generation model based on an improved GAN network to obtain synthesized discharge voltage data, the formula of which is:

[0063]

[0064] Where Θ G This indicates the parameters for improving the GAN network.

[0065] The beneficial effects of this invention are as follows: Based on an improved generative adversarial network (GAN) intelligent generation method for lithium battery degradation data, this invention provides a novel intelligent method for generating degradation data tailored to the characteristics of lithium battery degradation data. By improving the GAN network, degradation data corresponding to the specified capacity can be accurately generated, guiding the lithium battery health state estimation results to the set expected results. This method does not require an internal physicochemical model of the lithium battery; it only requires some historical data. Compared with existing methods, this invention can generate samples with greater flexibility and generalization ability.

[0066] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0067] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0068] Figure 1 This is a flowchart of the intelligent generation method for lithium battery degradation data based on improved generative adversarial networks described in this invention;

[0069] Figure 2 This is a schematic diagram of the intelligent generation method for lithium battery degradation data based on improved generative adversarial networks as described in this invention.

[0070] Figure 3 This is a schematic diagram of the improved generative adversarial network structure of this invention;

[0071] Figure 4 These are the modification results of the case described in this invention, where (a) is the peak value modification, (b) is the incremental modification, and (c) is the reduction modification. Detailed Implementation

[0072] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0073] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0074] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0075] Example 1:

[0076] like Figure 1 As shown, this invention provides an intelligent generation method for lithium battery degradation data based on an improved generative adversarial network. The improved generative adversarial network used in this invention is as follows: Figure 2 As shown; the generator is a multilayer perceptron, including an input layer, hidden layers, and an output layer. The input layer has a dimension of 85, the hidden layers are two layers, each with a dimension of 128, and the output layer has a dimension of 85. The discriminator is an LSTM network with a time step length of 85 and a feature count of featuredim = 2 at each time step. This method includes the following steps:

[0077] Step 1: Collect historical degradation data of lithium batteries, resulting in 2547 discharge voltage and capacity data points. Each discharge voltage data point contains 85 sampling points. The capacity data for each battery was normalized using the following mathematical expression:

[0078]

[0079] Where c t c represents the capacity of the t-th loop. max This indicates the maximum capacity of the battery. This represents the capacity of the t-th cycle after normalization.

[0080] Step 2: Match the collected discharge voltage data with the capacity data one by one, and divide the dataset into training dataset and test dataset. The training dataset contains 2267 discharge voltage and capacity data, and the test dataset contains 280 discharge voltage and capacity data.

[0081] Step 3: Train the LSTM network using the training dataset to obtain a model that can accurately predict capacity data based on discharge voltage data.

[0082] Step 4: Train the improved GAN network using the training dataset to obtain a model that can generate discharge voltage data that matches the capacity condition from random noise and capacity condition.

[0083] Step 5: Based on the target battery's physical information, the capacity personalization generation program is designed with three modification modes: peak value modification, incremental modification, and decremental modification. The formula for peak value modification is: in Here, c represents the desired target capacity, ε represents the capacity captured by the LSTM network, and ε represents the set modification increment. The incremental modification formula is: Where k(t) represents the strategy of continuous modification; the formula for decreasing modification is:

[0084] Step 6: Collect the current discharge voltage data v from the lithium battery. t The current capacity c is obtained by feeding it into the LSTM network. t The mathematical expression is c t =Ψ LSTM (v t ;Θ w ).

[0085] Step 7: Based on the current capacity, utilize the different generation modes of the capacity personalization generation program: In peak modification, when t = 100~102, ε = 0.02; when t = 200~202, ε = 0.04; in incremental modification, ε = 0.0003, k(t) = 0.25t; in decrement modification, ε = 0.0007, k(t) = 0.25t, to obtain the desired target capacity.

[0086] Step 8: Set the target capacity The random noise z is fed into the improved GAN network to obtain the synthesized discharge voltage data. in z~μ(-0.01,0.01); The results of the guided test using the generated discharge voltage data are as follows: Figure 4As shown, (a) represents peak value modification, (b) represents incremental modification, and (c) represents decremental modification.

[0087] Example 2:

[0088] This embodiment provides a training process for a battery capacity prediction model based on an LSTM network, as shown below:

[0089] Discharge voltage data V is used as input, and the corresponding capacity C is used as the label for supervised training. The input data is organized into a three-dimensional tensor (batch_size, sequence_length, input_size) in batches and fed into an LSTM network. The output of the LSTM layer undergoes a linear transformation through a fully connected layer, ultimately outputting a capacity estimate. By minimizing the error between the estimate and the label, a regression relationship for degradation behavior is established. The loss function is:

[0090]

[0091] Where E ω Y represents the loss function value. t Ψ represents the tag value of the t-th capacity. LSTM Describing the LSTM model, x t Θ represents the discharge voltage data for the t-th cycle. w This represents the parameter set of the LSTM model; Θ is updated using the backpropagation algorithm. w , when E ω Training stops when the value is less than the threshold ε.

[0092] Example 3:

[0093] In this embodiment, a specific structure for an improved GAN network is provided, such as... Figure 3 As shown, it includes:

[0094] The generator consists of a multilayer sensing mechanism, used to receive a random noise vector z and conditional information y, and generate synthesized discharge voltage data. Where, θ g Represents the trainable parameters of the generator;

[0095] The discriminator, constructed from an LSTM network, receives input data x, and its formula is:

[0096] D(x;θ d → [0,1]

[0097] Where, θ d This represents the trainable parameters of the discriminator. The closer the output value is to 1, the higher the probability that the input data is the real discharge voltage data.

[0098] The value function is:

[0099]

[0100] Where x represents a sample of actual discharge voltage data, which follows the actual sample distribution P. r Let z represent random noise, following a uniform distribution z ~ U(-0.01, 0.01), y represent capacity degradation information used as conditional information in the GAN network, D(·) be the discriminator, D(x|y) represent the discriminator's score for input x being real data, G(·) be the generator, G(z|y) are samples generated by the generator based on random noise z and conditional y, and λ be the weight parameter, controlling the weight of the gradient penalty term. It is a sample obtained by linear interpolation of real data x and generated data G(z), that is... Where ∈~U(0,1), This represents the gradient magnitude of the discriminator with respect to the input sample;

[0101] The generator loss function is:

[0102]

[0103] in, This indicates that the synthesized discharge voltage data generated by the generator satisfies... z represents a random noise vector that follows a uniform distribution, y represents capacity degradation condition information, and P g To generate a data distribution;

[0104] The discriminator loss function is:

[0105]

[0106] Where x represents the actual discharge voltage data, following the distribution P r , λ represents the interpolated sample, and λ is the gradient penalty coefficient.

[0107] The training process of the above-mentioned intelligent generation model for degraded data based on the improved GAN network is as follows:

[0108] Step (1): Randomly initialize the weights of the generator G and the discriminator D;

[0109] Step (2): Fix the generator G, and use the Adam optimizer with a learning rate η D Update discriminator parameters θ D Minimize the discriminator loss

[0110] Step (3): Fix the discriminator D, and use the Adam optimizer with a learning rate η g Update generator parameter θ G Minimize generator loss

[0111] Step (4): Repeat steps (2)-(3) until the maximum number of iterations T is reached. max ;

[0112] The improved GAN network minimizes the discriminator loss. Measuring the distribution of generated data P g Compared with the actual data distribution P r The Wasserstein distance is calculated, where the ratio of the number of training iterations (k) of the discriminator to the number of training iterations of the generator is 2:1, and adversarial training is used to make the data distribution output by the generator approximate the real data distribution.

[0113] In the above embodiments, the reference to "this embodiment" in the specification indicates that a specific feature, structure, or characteristic described in connection with the embodiment is included in at least some embodiments, but not necessarily all embodiments. Multiple appearances of "this embodiment" do not necessarily refer to the same embodiment.

[0114] In the above embodiments, although the invention has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory structures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed. The embodiments of the invention are intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims.

[0115] This embodiment also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the methods in this embodiment.

[0116] This embodiment also provides an electronic terminal, including: a processor and a memory;

[0117] The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory to cause the terminal to perform any of the methods in this embodiment.

[0118] As will be understood by those skilled in the art, the computer-readable storage medium described in this embodiment allows for the implementation of all or part of the steps in the above method embodiments by computer program-related hardware. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0119] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver, and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication between them. The memory is used to store computer programs, the communication interface is used to perform communication, and the processor and the transceiver are used to run the computer programs, so that the electronic terminal performs the steps of the above method.

[0120] In this embodiment, the memory may include random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0121] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0122] This invention can be used in a wide range of general-purpose or special-purpose computing system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices, etc.

[0123] This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligently generating lithium battery degradation data based on an improved generative adversarial network, characterized in that: Includes the following steps: Use the battery management system to obtain historical degradation data of lithium batteries, including lithium battery discharge voltage data and lithium battery capacity degradation data; The capacity degradation data is normalized, and the discharge voltage is matched one by one with the capacity degradation data to construct feature-label pairs and form a training dataset. The Long Short-Term Memory (LSTM) network and the improved Generative Adversarial Network (GAN) were trained using the training dataset to obtain a battery capacity prediction model based on the LSTM network and a degradation data generation model based on the improved GAN network. Collect the current cycle discharge voltage curve of the lithium battery, and use the battery capacity prediction model to predict the current battery capacity. Based on the physical characteristics of the target battery, a personalized battery capacity generation program is designed and executed to predict the current battery capacity and generate the desired target capacity. The target capacity is input into the degradation data generation model to generate synthetic degradation data; The intelligent generation model for degraded data based on the improved GAN network includes: The generator consists of a multilayer sensing mechanism used to receive random noise vectors. With conditional information Generate synthesized discharge voltage data ,in, Represents the trainable parameters of the generator; The discriminator is composed of an LSTM network and is used to receive input data. Its formula is: in, This represents the trainable parameters of the discriminator. The closer the output value is to 1, the higher the probability that the input data is the real discharge voltage data. The value function is: in This represents a sample of actual discharge voltage data, which follows the actual sample distribution. , Represents random noise, following a uniform distribution. , This indicates that capacity degradation information is used as conditional information in the GAN network. For discriminator, This indicates that the discriminator reacts to the input. Scoring based on real data, For generator, The generator is based on random noise. and conditions The generated samples, The weight parameter controls the weights of the gradient penalty term. Based on real data and generating data The sample obtained by linear interpolation, i.e. ,in , This represents the gradient magnitude of the discriminator with respect to the input sample; The generator loss function is: in, This indicates that the synthesized discharge voltage data generated by the generator satisfies... , This represents a random noise vector that follows a uniform distribution. Information on capacity degradation conditions, To generate a data distribution; The discriminator loss function is: in, The data represents actual discharge voltages and follows a distribution. , For interpolation samples, This is the gradient penalty coefficient; The training process of the intelligent generation model for degraded data based on the improved GAN network is as follows: Step (1): Randomly initialize the weights of the generator G and the discriminator D; Step (2): Fix the generator Through the Adam optimizer with learning rate Update discriminator parameters Minimize the discriminator loss ; Step (3): Fix the discriminator Through the Adam optimizer with learning rate Update generator parameters Minimize generator loss ; Step (4): Repeat steps (2)-(3) until the maximum number of iterations is reached. ; The improved GAN network minimizes the discriminator loss. Measuring the distribution of generated data Compared with the actual data distribution The Wasserstein distance, where the discriminator is trained 100 times per round. The ratio of the number of training iterations to the number of generator iterations is 2:1, and adversarial training is used to make the data distribution output by the generator approximate the real data distribution. The design of the personalized battery capacity generation program satisfies the following physical constraints: (a) Target capacity value generated satisfy: in This refers to the battery's rated capacity. (b) Capacity change between adjacent cycle periods satisfy: in This is the maximum allowable capacity fluctuation threshold preset based on the battery's chemical characteristics; The execution process of the personalized battery capacity generation program includes: Based on predicted capacity Generate desired target capacity Its formula is: in Define the desired capacity generation function for the user; The expected capacity generation function Based on different modification methods, modifications are categorized into three types: incremental modifications, decremental modifications, and peak value modifications. The formulas are as follows: Peak value modified: ; Incremental modification: ; Reduced quantity modification: ; in, This indicates the preset target capacity. This indicates the capacity acquired by the LSTM network. express The actual capacity measured during cycling. This indicates the set modification range. Indicates a strategy of continuous modification; Generate synthetic degradation data, specifically including: The preset target capacity With random noise The data are fed into a degraded data intelligent generation model based on an improved GAN network to obtain synthesized discharge voltage data, the formula of which is: in This indicates the parameters for improving the GAN network.

2. The intelligent generation method for lithium battery degradation data based on improved generative adversarial networks according to claim 1, characterized in that: The normalization formula for the capacity degradation data is as follows: in, express The actual capacity measured during cycling. This indicates the rated maximum capacity of the lithium battery in its initial state. This represents the normalized capacity degradation data.

3. The intelligent generation method for lithium battery degradation data based on improved generative adversarial networks according to claim 2, characterized in that: The training dataset is formed in the following steps: Collect lithium battery discharge voltage data Simultaneously record the normalized capacity degradation data within the corresponding cycle period. This forms a training dataset for health status estimation.

4. The intelligent generation method for lithium battery degradation data based on improved generative adversarial networks according to claim 1, characterized in that: The training process of the battery capacity prediction model based on the LSTM network is as follows: Discharge voltage data As input, the corresponding capacity Supervised training is performed using labels; the input data is organized into three-dimensional tensors in batches and fed into the LSTM network. These three-dimensional tensors include batch_size, sequence_length, and input_size. The output of the LSTM layer undergoes a linear transformation through a fully connected layer, ultimately outputting a capacity estimate. By minimizing the error between the estimate and the label, a regression relationship for the degradation behavior is established; the loss function is: in This represents the value of the loss function. Indicates the first The tag value of the capacity, Represents the LSTM model. Indicates the first Discharge voltage data for each cycle, This represents the parameter set of the LSTM model; it is updated using the backpropagation algorithm. ,when Less than the threshold Training should be stopped at this time.

5. The intelligent generation method for lithium battery degradation data based on improved generative adversarial networks according to claim 4, characterized in that: The process of collecting the current cycle discharge voltage curve of the lithium battery and predicting the current battery capacity using the battery capacity prediction model specifically includes: Current discharge voltage curve The data is fed into an LSTM-based battery capacity prediction model to obtain the current predicted capacity. Its formula is: in, For the LSTM model, a set of parameters is fixed. This represents the mapping relationship of an LSTM network containing fully connected layers.

Images