Online Estimation Algorithm of SOC and SOH Using Neural Network for Lithium Battery

Abstract

Lithium batteries are being employed as primary power sources in various applications, including cell phones, electric vehicles, unmanned submarines, and energy storage systems. Therefore, for stable and safe use of a system, it is important to quickly detect defects in the battery and effectively diagnose faults. In this work, we proposed an algorithm that evaluates the state of charge (SOC) and state of health (SOH) online using long short-term memory (LSTM). The SOC is estimated using an LSTM model bank with three LSTM models in which a battery data group has learned normal, caution, and fault. The SOH is estimated by receiving SOC and battery parameters from the LSTM model bank to output SOH as one of the three states: normal, caution, and fault. Experimental results show that the proposed battery SOC and SOH estimation algorithm have high accuracy.