MDGN: Circuit design of memristor‐based denoising autoencoder and gated recurrent unit network for lithium‐ion battery state of charge estimation

Abstract

Abstract Due to the highly complex and non‐linear physical dynamics of lithium‐ion batteries, it is unfeasible to measure the state of charge (SOC) directly. Designing systems capable of accurate SOC estimation has become a key technology for battery management systems (BMS). Existing mainstream SOC estimation approaches still suffer from the limitations of low efficiency and high‐power consumption, owing to the great number of samples required for training. To address these gaps, this paper proposes a memristor‐based denoising autoencoder and gated recurrent unit network (MDGN) for fast and accurate SOC estimation of lithium‐ion batteries. Specifically, the DAE circuit module is designed to extract useful feature representation with strong generalization and noise immunity. Then, the gated recurrent unit (GRU) circuit module is designed to learn the long‐term dependencies between high‐dimensional input and output data. The overall performance is evaluated by root mean square error (RMSE) and mean absolute error (MAE) at 0, 25, and 45°C, respectively. Compared with the current state‐of‐the‐art methods, the entire scheme shows its superior performance in accuracy, robustness, and operation cost (referring to time cost).

Keywords

• circuit design
• denoising autoencoder
• gated recurrent unit
• memristor
• state of charge estimation