State of health estimation of the lithium-ion power battery based on the principal component analysis-particle swarm optimization-back propagation neural network

Abstract

State of Health (SOH) estimation of the lithium-ion power battery has become the focus of the research and it has important scientific significance for optimizing the battery energy management strategy as well as prolonging the battery life. However, the reaction mechanism of lithium-ion power battery is complex with strong nonlinear and time-varying. Meanwhile, the complex and varied external operating environment and operating conditions increase the uncertainty of the lithium-ion power battery performance decline and further increase the difficulty of SOH estimation. The SOH estimation method of the lithium-ion power battery based on the Principal Component Analysis-Particle Swarm Optimization-Back Propagation Neural Network (PCA-PSO-BPNN) is proposed in this paper. The PCA is adopted to reduce the system input dimension, the PSO is used to optimize the weights of BPNN and the optimized BPNN is applied to estimate SOH accurately. Experimental results on the lithium-ion power battery of the NASA battery aging test data demonstrate the effectiveness of the proposed method and it can reach more excellent SOH estimation results. The Mean Absolute Error is no more than 0.51%, the Root Mean Square Error is no more than 0.65% and the Maximum Absolute Error is no more than 1.86%, respectively.