基于改进的VAE-GAN模型在电池EIS数据增强中的应用

Application of Improved VAE-GAN Model in Battery EIS Data Enhanced Applications

电化学阻抗谱(electrochemical impedance spectroscopy,EIS)是一种用于表征电池内部电化学过程的测试方法。电化学阻抗谱数据可以用于分析、评估和优化电池性能。测试EIS数据需要使用专业的仪器设备,成本较高,测试数据的数量往往不多,可以使用数据增强方法来增加EIS数据的数量。变分自编码器(variational autoencoder,VAE)是一种生成模型,可以通过对潜在分布中的采样来生成新的样本。生成对抗网络(generative adversarial networks,GAN)也是一种生成模型,其原理是通过两个相互对抗的网络模型来实现生成数据和判别数据的任务。VAE模型和GAN模型都可以单独用于数据增强,但是VAE和GAN模型都存在一些缺点,通过组合VAE和GAN的方法,构建VAE-GAN模型,一定程度上弥补各自的缺点,达到更好的生成效果和性能。对VAE-GAN模型的网络结构进行优化,将Transformer(转换器)模型用于VAE模型的编码器和解码器以及GAN模型的判别器中,提升了模型效果。使用改进的VAE-GAN模型,将EIS数据作为输入数据,构建EIS的预测模型,由生成器来生成EIS增强数据,由判别器来判断新生成的EIS数据是否是有效的增强数据。实验表明,提出的方法能够生成质量较好的EIS数据。

Electrochemical Impedance Spectroscopy(EIS)is a testing method used to characterize the internal electrochemical processes of batteries.Electrochemical impedance spectroscopy data can be used to analyze,evaluate,and optimize battery performance.Testing EIS data requires the use of professional instruments and equipment,which is costly.The amount of test data is often limited,and data augmentation methods can be used to increase the amount of EIS data.Variational Autoencoder(VAE)is a generative model that can generate new samples by sampling from potential distributions.Generative Adversarial Networks(GANs)are also a type of generative model,whose principle is to achieve the task of generating and discriminating data through two opposing network models.Both VAE and GAN models can be used separately for data augmentation,but both have some drawbacks.By combining VAE and GAN methods to construct VAE-GAN models,some shortcomings can be compensated and achieve better generation results and performance.In traditional VAE-GAN models,the encoder and decoder in the VAE model,as well as the discriminator in the GAN model,generally use fully connected neural networks or convolutional neural network models.The network structure of the VAE-GAN model was optimized by using the Transformer model in the encoder and decoder of the VAE model,as well as in the discriminator of the GAN model,which improved the model performance.Improved VAE-GAN model was used to construct a prediction model for EIS using EIS data as input.The generator generates EIS enhancement data,and the discriminator determines whether the newly generated EIS data is effective enhancement data.Experiments have shown that the method proposed in this article can generate high-quality EIS data.