Research for SOC Prediction of Lithium Battery Based on GA-ESN

摘要

For the problem of low prediction and difficulty to estimate the state of charge (SOC), the GA-ESN model, a combination of genetic algorithm (GA) with echo state network (ESN), is applied to predicting the SOC. The genetic algorithm is used to select, crossover and mutate the undefined parameters of the echo state network. According to the calculation of the fitness function value, the parameters that make the echo state network optimal are selected to complete the establishment of the GA-ESN model. The established GA-ESN model is used to predict SOC under NYCC, UDDS and US06 operating conditions, the simulation results show that the SOC error of the echo state network after optimization by the genetic algorithm can be controlled within 4%, which verifies the feasibility of the model applied to SOC estimation.