关键词: SOC估计, 无迹卡尔曼滤波算法, 锂电池, 粒子群算法, BP神经网络

Abstract:

The state of charge (SOC) of lithium batteries is one of the core of quality management of lithium batteries. Based on effective SOC estimation is also necessary to ensure the safe and efficient operation of lithium batteries, A method for estimating the SOC of lithium batteries is proposed, which uses particle swarm algorithm (PSO) to optimize the backpropagation(BP) neural network as the observed value of the unscented Kalman filter(UKF). The proposed PSO-BP-UKF algorithm is compared with the GA-BP-UKF algorithm and the BP algorithm using FUDS operating condition battery test data from the University of Maryland. Taking the test results in 25 ℃ environment, the maximum deviation of PSO-BP-UKF algorithm is within 3.17%, the average error is within 6.44%, and the root-mean-square deviation is within 0.002 5, which is significantly improved than both GA-BP-UKF algorithm and BP method, and shows that the proposed algorithm is the effective and practical.

Key words: SOC estimation, unscented Kalman filter(UKF) algorithm, lithium battery, particle swarm algorithm(PSO), BP neural network