딥러닝 기반 모빌리티 배터리의 건전성 상태 추정 방법

초록

Accurate real‑time estimation and monitoring of the state of health (SOH) in lithium-ion batteries is critical for mobility applications, as battery failures can result in significant personal injury. This study introduces a method for SOH estimation employing a deep learning-based hybrid neural network that integrates a temporal convolutional network (TCN) and gated recurrent unit (GRU). To facilitate applicability in mobility onboard systems, the proposed model utilizes input data obtained by resampling raw current, voltage, and time signals, which can be directly measured by the battery management system (BMS). The TCN model, trained on both past and current input data, enhances reliability by effectively extracting non-linear features observed during battery degradation, such as local capacity recovery. These extracted features are then processed by the GRU model, which learns temporal patterns and produces precise SOH estimation results. The proposed method was validated using CALCE battery degradation data, achieving consistent and superior performance across all battery cells, with mean absolute error and root mean square error values not exceeding 0.55 and 0.7, respectively.

키워드