State of Health Estimation for Lithium-Ion Batteries Using Long-Term Recurrent Convolutional Network

Abstract

Lithium-ion batteries are widely used as main energy storage devices in real life and industrial fields. Battery management system (BMS) observes battery state to ensure stability and reliability in the application using lithium-ion batteries. State-of-health (SOH) estimation technique helps the BMS to operate the appropriate profile by providing maximal releasable capacity to the BMS. This paper compares the estimation performance of SOH for data-driven methods which are convolutional neural network (CNN), long short-term memory (LSTM), and long-term recurrent convolutional network (LRCN). To compare three networks, the experimental data randomized battery usage dataset that consists of 4 cells provided by NASA Ames Research Center is used. The preprocessing is applied to increase for learning efficiency and the accuracy of the estimation network is compared based on evaluation criteria. The comparison of estimation results demonstrates LRCN achieves 4.268% and 3.773% higher MAPE than CNN and LSTM, due to LRCN can modeling the long-scale sequence data through feature extraction and sequence learning.