Retired Battery Screening Based on Rebooted Auxiliary Classifier Generative Adversarial Network and Improved Gramian Angular Field

Lithium-ion batteries (LIBs) are widely used in electronic gadgets, electric cars, and energy storage applications due to their high energy density and long cycle lifespan. The precise evaluation of retired batteries significantly hinges on utilizing optimal health features that are both highly informative and easily obtainable. In particular, for time-series data, there are current challenges related to insufficient feature capture and the difficulty of capturing effective features. This article introduces an innovative classification approach for retired batteries by integrating an improved Gramian angular field (IGAF) with a rebooted auxiliary classifier generative adversarial network (REACGAN). The IGAF method transforms subtle variations in battery charging voltage curves into 2-D images, utilizing the fast Fourier transform (FFT) to extract amplitude and phase features, thereby preserving both temporal and spatial characteristics. The REACGAN model enhances classification stability and generated image quality by optimizing input vector projection and incorporating a novel loss function. To assess the effectiveness of the proposed method, comparative experiments were conducted using 284 retired batteries. The experimental findings indicate that the suggested approach attains an average classification accuracy of 95. 41%, surpassing other models in both classification performance and processing efficiency.