Lithium battery charging and discharging data generation method based on hierarchical convolution generative adversarial network

Abstract

The invention discloses a lithium battery charging and discharging data generation method based on a hierarchical convolution generative adversarial network, and the method comprises the steps: collecting the basic parameter information of the charging and discharging process data of a target lithium battery, and building a real lithium battery charging and discharging data set; features of an input data set are extracted by adopting a convolutional layer, the number of parameters and the calculation amount are reduced through a pooling layer, and finally classification is performed through a full connection layer; constructing a generator and a discriminator of a generative adversarial network (GAN) by using a convolutional neural network (CNN), and optimizing data through adversarial training; the generator learns to generate realistic data samples from random noise, and the discriminator learns to distinguish the realistic data samples from the samples generated by the generator. According to the method, the hierarchical convolution generative adversarial network is adopted to act on lithium battery charging and discharging data generation, so that the model can better capture a complex structure of the data, and real lithium battery charging and discharging data can be better collected.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium battery charge and discharge data generation, and in particular to a lithium battery charge and discharge data generation method based on a layered convolutional generative adversarial network. Background Art

[0002] The charging and discharging data of lithium batteries are important indicators for evaluating battery performance and life. They mainly include battery capacity, voltage, current and other parameters. These parameters are very important for evaluating battery performance and service life.

[0003] In recent years, deep learning has made breakthroughs in many fields, but everyone seems to have discovered such a reality, that is, the work that has made breakthroughs in deep learning is basically related to discriminant models. Generative adversarial networks contain a generative model and a discriminative model. Among them, the generative model is responsible for capturing the distribution of sample data, while the discriminative model is generally a binary classifier that determines whether the input is real data or a generated sample.

[0004] However, the training process of GANs is often unstable. The generator and the discriminator compete with each other like tug-of-war during the training process. When the discriminator is too powerful, the generator cannot get effective feedback, and vice versa, resulting in poor training results. In addition, GANs also have the problem of unstable "Nash equilibrium", that is, the generator and the discriminator may oscillate around each other during the game, making it difficult to reach a stable equilibrium state. Therefore, it is necessary to introduce convolutional neural networks to improve the efficiency and stability of model training. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] In view of the shortcomings of the prior art, the present invention adopts a layered convolutional generative adversarial network to generate data on lithium battery charging and discharging, which enables the model to better capture the complex structure of the data and thus better collect real data on lithium battery charging and discharging.

[0007] (II) Technical solution

[0008] To achieve the above object, the present invention provides the following technical solution: a method for generating lithium battery charge and discharge data based on a layered convolutional generative adversarial network, comprising the following steps:

[0009] (1) Collect basic parameter information of the target lithium battery charge and discharge process data to establish a real lithium battery charge and discharge data set;

[0010] (2) The convolutional layer is used to extract the features of the input data set, and the pooling layer is used to reduce the number of parameters and the amount of calculation, and finally the classification is performed through the fully connected layer;

[0011] (3) Use convolutional neural networks (CNNs) to build the generator and discriminator of generative adversarial networks (GANs) and optimize data through adversarial training;

[0012] (4) The generator learns to generate realistic data samples from random noise, and the discriminator learns to distinguish between real data samples and samples generated by the generator;

[0013] (5) After the GAN completes offline training, the lithium-ion battery is subjected to constant current-constant voltage charging and discharging operations and the voltage, current, temperature and other measurement values ​​of each sampling point are obtained and input into the trained GAN to obtain the lithium battery charging and discharging data.

[0014] As a preferred embodiment, the basic parameters of the lithium battery charging and discharging process data in step (1) are composed of charging parameters and discharging parameters;

[0015] The charging parameters include the charging limit voltage, charging temperature range, maximum charging current and recommended charging current; the discharging parameters include the discharge termination voltage, discharge cut-off voltage, maximum discharge current, recommended discharge current and discharge temperature range.

[0016] As a preferred solution, the step (2) specifically comprises:

[0017] (21) The input layer is responsible for inputting image information such as data sets;

[0018] (22) Extract the underlying features of the image through convolution operation;

[0019] (23) Perform nonlinear transformation on the output of the convolutional layer;

[0020] (24) Reduce the number of parameters by downsampling to prevent overfitting;

[0021] (25) Summarize the feature information of the convolution layer and the pooling layer;

[0022] (26) The final result is obtained based on the information of the fully connected layer and classified.

[0023] As a preferred solution, the step (3) specifically comprises:

[0024] (31) Generator

[0025] Receive random noise as input and generate new data samples, with the goal of generating data that is sufficient to "fool" the discriminator;

[0026] (32) Discriminator

[0027] Determine whether the input data is real data or data generated by the generator, and output a probability value indicating the possibility that the data is real;

[0028] (33) Confrontation

[0029] The generator and the discriminator are trained alternately. The generator tries to generate more realistic data to deceive the discriminator, while the discriminator strives to improve the accuracy of judgment. This process continues until the Nash equilibrium is reached, that is, the discriminator cannot distinguish between real data and generated data.

[0030] (34) Study

[0031] The generator and discriminator learn crosswise, and when one of them learns, the parameters of the other remain unchanged.

[0032] As a preferred solution, in the adversarial training process, in step (3), the generator and the discriminator are trained alternately; first, the parameters of the generator are fixed and the discriminator is trained so that it can accurately distinguish between real samples and generated samples; then, the parameters of the discriminator are fixed and the generator is trained so that the samples it generates can deceive the discriminator. This process is repeated until the performance of the generator and the discriminator reaches a balance.

[0033] As a preferred solution, the adversarial training process is specifically as follows:

[0034] First, fix the generator, use the generator to simulate G(z) as a negative sample based on the hidden random vector z, and sample the positive sample x from the real data;

[0035] Then the positive and negative samples are input to the discriminator for binary classification prediction, and finally the discriminator parameters are updated using its binary classification cross entropy loss;

[0036] Then the discriminator is fixed to optimize the generator. For the generator, in order to deceive the discriminator as much as possible, that is, to make the discriminator judge the generated "fake" samples as positive samples as much as possible, it is generally considered to maximize the discrimination probability of the generated samples for optimization.

[0037] As a preferred solution, let z be random noise and x be real data. The generative network and the discriminative network can be represented by G and D respectively, where D can be regarded as a binary classifier. Then, using cross entropy representation, it can be written as:

[0038] minGmaxDVD,G=Ex·pdata(x)logDx+Ez·pzz[log(1-D(G(z)))]

[0039] The first term logD(x) represents the discriminator's judgment on the real data, and the second term log(1-D(G(z))) represents the synthesis and judgment of the data. Through such a Max-min game, G and D are optimized alternately to train the required generative network and discriminative network until the Nash equilibrium point is reached.

[0040] As a preferred solution, the lithium battery charge and discharge data set in step (1) refers to detailed data of the lithium battery during the charge and discharge process obtained through actual testing and monitoring, and the data set usually includes multiple parameters such as voltage, current, temperature and capacity.

[0041] (III) Beneficial effects

[0042] Compared with the prior art, the present invention provides a method for generating lithium battery charge and discharge data based on a layered convolutional generative adversarial network, which has the following beneficial effects:

[0043] The present invention adopts a layered convolutional generative adversarial network to collect basic parameter information of the target lithium battery charging and discharging process data, and conducts adversarial exercises by establishing a generator model and a discriminator model. Through the layered training method, the risk of mode collapse is reduced, the training is made more stable, and the generation quality and authenticity of the lithium battery charging and discharging data are further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a flow chart of the method for generating charge and discharge data of the present invention;

[0045] Figure 2 Schematic diagram showing the functional mode of the charge and discharge data generation method of the present invention. DETAILED DESCRIPTION

[0046] In order to better understand the purpose, structure and function of the present invention, the following will further illustrate a method for generating lithium battery charging and discharging data based on a layered convolutional generative adversarial network in conjunction with the accompanying drawings.

[0047] refer to Figure 1-2 The present invention provides a method for generating lithium battery charge and discharge data based on a layered convolutional generative adversarial network, comprising the following steps:

[0048] (1) Collect basic parameter information of the target lithium battery charge and discharge process data to establish a real lithium battery charge and discharge data set;

[0049] (2) The convolutional layer is used to extract the features of the input data set, and the pooling layer is used to reduce the number of parameters and the amount of calculation, and finally the classification is performed through the fully connected layer;

[0050] (3) Use convolutional neural networks (CNNs) to build the generator and discriminator of generative adversarial networks (GANs) and optimize data through adversarial training;

[0051] (4) The generator learns to generate realistic data samples from random noise, and the discriminator learns to distinguish between real data samples and samples generated by the generator;

[0052] (5) After the GAN completes offline training, the lithium-ion battery is subjected to constant current-constant voltage charging and discharging operations and the voltage, current, temperature and other measurement values ​​of each sampling point are obtained and input into the trained GAN to obtain the lithium battery charging and discharging data.

[0053] Specifically, the lithium battery charge and discharge data set refers to the detailed data of the lithium battery during the charge and discharge process obtained through actual testing and monitoring. The data set usually contains multiple parameters such as voltage, current, temperature and capacity. The basic parameters of the lithium battery charge and discharge process data are composed of charging parameters and discharging parameters.

[0054] The charging parameters include the charging limit voltage, charging temperature range, maximum charging current and recommended charging current; the discharging parameters include the discharge termination voltage, discharge cut-off voltage, maximum discharge current, recommended discharge current and discharge temperature range.

[0055] Furthermore, the present invention is based on a convolutional neural network to generate adversarial methods. The convolutional neural network is a type of neural network specifically used to process network structure data. Its core is to use convolutional layers and pooling layers to extract features from images, and finally perform classification or regression tasks through a fully connected layer.

[0056] Its convolutional neural network model is shown in step (2), which specifically includes:

[0057] (21) The input layer is responsible for inputting image information such as data sets;

[0058] (22) Extract the underlying features of the image through convolution operation;

[0059] (23) Perform nonlinear transformation on the output of the convolutional layer;

[0060] (24) Reduce the number of parameters by downsampling to prevent overfitting;

[0061] (25) Summarize the feature information of the convolution layer and the pooling layer;

[0062] (26) The final result is obtained based on the information of the fully connected layer and classified.

[0063] Among them, the basic structure of a convolutional neural network usually consists of the following parts:

[0064] (1) Input layer: receives raw image data or other types of grid structure data.

[0065] (2) Convolutional layer: extracts local features of the input data through convolution operations. Each convolution kernel in the convolutional layer can extract a specific feature, and multiple convolution kernels can work in parallel to extract different types of features.

[0066] (3) Pooling layer: Downsample the output of the convolutional layer (or reduce dimensionality) to reduce the number of parameters and improve computational efficiency. Common pooling operations include maximum pooling and average pooling.

[0067] (4) Fully connected layer: It combines the features extracted by the previous layers for tasks such as classification or regression. Each neuron in the fully connected layer is connected to all neurons in the previous layer.

[0068] The generative adversarial network of the present invention is composed of two parts: a generator and a discriminator. They coordinate with each other in an adversarial way to improve the performance of the network. The generator attempts to generate samples similar to real data, and the discriminator attempts to distinguish between generated samples and real samples.

[0069] Furthermore, the generative adversarial network is performed as shown in step (3), which specifically includes:

[0070] (31) Generator

[0071] Receive random noise as input and generate new data samples, with the goal of generating data that is sufficient to "fool" the discriminator;

[0072] (32) Discriminator

[0073] Determine whether the input data is real data or data generated by the generator, and output a probability value indicating the possibility that the data is real;

[0074] (33) Confrontation

[0075] The generator and the discriminator are trained alternately. The generator tries to generate more realistic data to deceive the discriminator, while the discriminator strives to improve the accuracy of judgment. This process continues until the Nash equilibrium is reached, that is, the discriminator cannot distinguish between real data and generated data.

[0076] (34) Study

[0077] The generator and discriminator learn crosswise, and when one of them learns, the parameters of the other remain unchanged.

[0078] Furthermore, let z be random noise and x be real data. The generative network and the discriminative network can be represented by G and D respectively, where D can be regarded as a binary classifier. Then, using cross entropy representation, it can be written as:

[0079] minGmaxDVD,G=Ex·pdata(x)logDx+Ez·pzz[log(1-D(G(z)))]

[0080] The first term logD(x) represents the judgment of the discriminator on the real data, and the second term log(1-D(G(z))) represents the synthesis and judgment of the data. Through such a Max-min game, G and D are optimized alternately to train the required generative network and discriminative network until the Nash equilibrium point is reached.

[0081] In the above adversarial training process, the adversarial training process is specifically as follows:

[0082] First, fix the generator, use the generator to simulate G(z) as a negative sample based on the hidden random vector z, and sample the positive sample x from the real data;

[0083] Then the positive and negative samples are input to the discriminator for binary classification prediction, and finally the discriminator parameters are updated using its binary classification cross entropy loss;

[0084] Then the discriminator is fixed to optimize the generator. For the generator, in order to deceive the discriminator as much as possible, that is, to make the discriminator judge the generated "fake" samples as positive samples as much as possible, it is generally considered to maximize the discrimination probability of the generated samples for optimization.

[0085] Specifically, the generator and discriminator are trained alternately. First, the parameters of the generator are fixed and the discriminator is trained so that it can accurately distinguish between real samples and generated samples. Then, the parameters of the discriminator are fixed and the generator is trained so that the samples it generates can deceive the discriminator. This process is repeated until the performance of the generator and the discriminator is balanced.

[0086] The present invention adopts a layered convolutional generative adversarial network to act on the data generation of lithium battery charging and discharging, and conducts adversarial exercises by establishing a generator model and a discriminator model, thereby further improving the generation quality and authenticity of lithium battery charging and discharging data.

[0087] It is to be understood that the present invention is described by some embodiments, and it is known to those skilled in the art that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. In addition, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the scope of protection of the present invention.

Claims

1. A method for generating lithium battery charge and discharge data based on a layered convolutional generative adversarial network, comprising the following steps: (1) Collect basic parameter information of the target lithium battery charge and discharge process data to establish a real lithium battery charge and discharge data set; (2) The convolutional layer is used to extract the features of the input data set, and the pooling layer is used to reduce the number of parameters and the amount of calculation, and finally the classification is performed through the fully connected layer; (3) Use convolutional neural networks (CNNs) to build the generator and discriminator of generative adversarial networks (GANs) and optimize data through adversarial training; (4) The generator learns to generate realistic data samples from random noise, and the discriminator learns to distinguish between real data samples and samples generated by the generator; (5) After the GAN completes offline training, the lithium-ion battery is subjected to constant current-constant voltage charging and discharging operations and the voltage, current, temperature and other measurement values ​​of each sampling point are obtained and input into the trained GAN to obtain the lithium battery charging and discharging data.

2. According to claim 1, a method for generating lithium battery charge and discharge data based on a layered convolutional generative adversarial network is characterized in that: In the step (1), the basic parameters of the lithium battery charge and discharge process data consist of charging parameters and discharging parameters; The charging parameters include the charging limit voltage, charging temperature range, maximum charging current and recommended charging current; the discharging parameters include the discharge termination voltage, discharge cut-off voltage, maximum discharge current, recommended discharge current and discharge temperature range.

3. The method for generating lithium battery charge and discharge data based on a layered convolutional generative adversarial network according to claim 2, characterized in that: The step (2) specifically comprises: (21) The input layer is responsible for inputting image information such as data sets; (22) Extract the underlying features of the image through convolution operation; (23) Perform nonlinear transformation on the output of the convolutional layer; (24) Reduce the number of parameters by downsampling to prevent overfitting; (25) Summarize the feature information of the convolution layer and the pooling layer; (26) The final result is obtained based on the information of the fully connected layer and classified.

4. The method for generating lithium battery charge and discharge data based on a layered convolutional generative adversarial network according to claim 3, characterized in that: The step (3) specifically comprises: (31) Generator Receive random noise as input and generate new data samples, with the goal of generating data that is sufficient to "fool" the discriminator; (32) Discriminator Determine whether the input data is real data or data generated by the generator, and output a probability value indicating the possibility that the data is real; (33) Confrontation The generator and the discriminator are trained alternately. The generator tries to generate more realistic data to deceive the discriminator, while the discriminator strives to improve the accuracy of judgment. This process continues until the Nash equilibrium is reached, that is, the discriminator cannot distinguish between real data and generated data. (34) Study The generator and discriminator learn crosswise, and when one of them learns, the parameters of the other remain unchanged.

5. The method for generating lithium battery charge and discharge data based on a layered convolutional generative adversarial network according to claim 1, characterized in that: In the adversarial training process of step (3), the generator and the discriminator are trained alternately; first, the parameters of the generator are fixed and the discriminator is trained so that it can accurately distinguish between real samples and generated samples; then, the parameters of the discriminator are fixed and the generator is trained so that the samples it generates can deceive the discriminator. This process is repeated until the performance of the generator and the discriminator reaches a balance.

6. The method for generating lithium battery charge and discharge data based on a layered convolutional generative adversarial network according to claim 5, characterized in that: The adversarial training process is specifically as follows: First, fix the generator, use the generator to simulate G(z) as a negative sample based on the hidden random vector z, and sample the positive sample x from the real data; Then the positive and negative samples are input to the discriminator for binary classification prediction, and finally the discriminator parameters are updated using its binary classification cross entropy loss; Then the discriminator is fixed to optimize the generator. For the generator, in order to deceive the discriminator as much as possible, that is, to make the discriminator judge the generated "fake" samples as positive samples as much as possible, it is generally considered to maximize the discrimination probability of the generated samples for optimization.

7. The method for generating lithium battery charge and discharge data based on a layered convolutional generative adversarial network according to claim 6, characterized in that: Let z be random noise and x be real data. The generative network and discriminative network can be represented by G and D respectively, where D can be regarded as a binary classifier. Then, using cross entropy representation, it can be written as: minGmaxDVD,G=Ex·pdata(x)logDx+Ez·pzz[log(1-D(G(z)))] The first term logD(x) represents the discriminator's judgment on the real data, and the second term log(1-D(G(z))) represents the synthesis and judgment of the data. Through such a Max-min game, G and D are optimized alternately to train the required generative network and discriminative network until the Nash equilibrium point is reached.

8. The method for generating lithium battery charge and discharge data based on a layered convolutional generative adversarial network according to claim 1, characterized in that: The lithium battery charge and discharge data set in step (1) refers to the detailed data of the lithium battery during the charge and discharge process obtained through actual testing and monitoring. The data set usually includes multiple parameters such as voltage, current, temperature and capacity.

Images