State of Health Prediction of Lithium-ion Battery Based on Feature Fusion

Aiming at the problems of low accuracy and poor stability of existing methods, a feature fusion and algorithm fusion based estimation method for the state of health (SOH) of lithiumion batteries is proposed. Firstly, six health features with high correlation with SOH are extracted from the charge and discharge curves of the lithium-ion batteries. Then, the extracted health features are classified and fused using principal component analysis to realize the reduction of the data dimensionality. Finally, the parameter of the bi-directional long and short-term memory network (BILSTM) is optimized by using Sparrow's algorithm (SSA) and Firefly's algorithm (FA) to obtain the SOH estimation model. To validate the usability of the method, experiments are done on a variety of battery aging data and compared with the BILSTM model, SSA-BILSTM model, and FA-BILSTM model. The experimental results show that the FA-SSA-BILSTM method can effectively improve the accuracy of SOH estimation.