Capacity Estimation of Lithium Ion Battery Considering Hybrid Charging Data

Abstract

Accurate and effective battery capacity estimation is very important for the electric vehicles' safety. Currently, the battery capacity estimation method combined with health factors extraction has received widespread attention, however, most studies fail to take into account that the charging data of each cycle in the actual application of batteries will have different charging data structures according to the different charging and discharging situations, which will lead to the inability to continuously and effectively extract health factors, invalid or missing health factor sequences would fail to effectively estimate battery capacity, so a capacity estimation method for lithium-ion batteries considering hybrid charging data is carried out. First, three most common charging data structures are considered to form the hybrid charging data. Effective health factors are extracted according to different data structures, and then particle swarm optimization algorithm is used to obtain the best health factors. Second, the complete health factor sequences are obtained by using the method of health factor estimation based on the relevance vector regression(RVM). Third, the complete health factor sequences are used to train the LSTM to estimate the future battery capacity. The simulation result shows that the relative errors of RVM estimation of health factors are kept within 1%, and the relative errors of future battery capacity are basically kept within 2%, which achieves high accuracy and may meets certain practical application requirements.