Receding Horizon SOH Estimation of Lithium-ion Battery based on Novel Health Indicator

State of health (SOH) is a crucial part of the battery management system. Accurate SOH estimation is crucial for maintaining battery safety and stability. Extracting reliable and effective health indicators (HI) to describe the aging state of the battery and constructing accurate and stable estimation model are the main problems we are facing at present. In order to improve the accuracy of SOH estimation, a receding horizon estimation method for lithium-ion battery based on new HI is proposed. Firstly, the fuzzy entropy value of the convolution sequence is extracted as the HI of the battery by analyzing the discharge voltage data and performing the convolution operation on it. Secondly, a non-parametric state-space model is constructed to describe battery aging based on the metabolic grey model (MGM) with a sliding window of variable step size and long short-term memory (LSTM) network. Finally, the closed-loop estimation of battery SOH is realized by particle filter (PF). In addition, the proposed method for estimating SOH has been validated using the NASA battery dataset and is compared with various other approaches in the field. The results indicate that the estimation error remains consistently stable within 3%, and the estimation accuracy of the proposed approach is improved by up to about 80% compared with similar methods, which verifies the feasibility and superiority of the proposed approach.