WGAN-GP with Residual Network Model for Lithium Battery Thermal Image Data Expansion with Quantitative Metrics

Energy storage systems, especially lithium battery energy storage, play a significant role in microgrids. Thermal images of lithium batteries are crucial for the study of lithium batteries since lithium batteries are strongly impacted by temperature. In order to increase the sample of thermal images from lithium batteries, this paper designs the WGAN-GP with ResNet model (Wasserstein generative adversarial network with gradient penalty and residual networks). A complete WGAN-GP with ResNet model is built with the embedding of a generative adversarial network (GAN) structure, introducing gradient penalties, and combining residual networks. The experimental results demonstrate that WGAN-GP with ResNet have a greater improvement in training stability and generated image quality than GAN and Wasserstein GAN by using four evaluation indicators to examine the generated image quality.