Abnormal Degradation Detection for Lithium-Ion Batteries With Denoising Diffusion Implicit Model

Abstract

This paper proposes a method for the immediate detection of abnormal degradation in lithium-ion batteries by applying the denoising diffusion implicit model (DDIM). The proposed approach utilizes a diffusion-based generative model to emphasize and reconstruct battery operation data. In this framework, the DDIM compresses and reconstructs lithium-ion battery operational data, and the resulting reconstruction error is used as an indicator of anomalies. Experimental results demonstrate that the proposed method can detect abnormal degradation at least 10 cycles earlier than the benchmark models.