트랜스포머 모델을 활용한 전기차 배터리 SOH 예측 연구

Abstract

This paper presents a Transformer-based State of Health(SoH) estimation algorithm that is applied to publicly available NASA data. Data preprocessing involves the application of a moving average to reduce noise and utilize wavelet transformers to extract features. The preprocessed data serves as input to the Transformer model in estimating SoH. Next, a comparative analysis with an LSTM-based model is conducted by using the Root Mean Square Error(RMSE) metric. The proposed model demonstrates a superior SoH estimation accuracy, surpassing the LSTM-based model by up to 39.58 %.