A Hybrid State of Health Estimation Method for Lithium Ion Battery

Abstract

In order to improve the accuracy and practicability of state of health estimation for lithium-ion batteries, the Improved Arithmetic Particle Swarm Optimization Algorithm (IAPSOA) is proposed in this work. Combined with Deterministically Constructed Cycle Reservoirs with Regular Jumps(CRJ), IAPSOA-CRJ estimation method is proposed. The constant current charging time of the battery is extracted as a health indicator to predict the real capacity series. Then, by improving the Arithmetic Optimization Algorithm(AOA),the search ability and stability of AOA algorithm are improved. This paper also studied the influence of different training set length on the model. Finally, the generalization performance is tested with the trained model on another set of battery data of the same type. IAPSOA algorithm is used to optimize the input matrix parameter, reservoir parameters and regularization coefficient of CRJ network, and compared with Radial Basis Function Neural Network, Elman Neural Network and Optimized Kernel Extreme Learning Machine. The results show that the proposed IAPSOA-CRJ estimation model performs best in all aspects, and has strong robustness and generalization ability.