基于DEPSO-RBF神经网络的锌银电池SOC估计

Estimation for Silver-zinc Battery State of Charge Based on DEPSO-RBF Neural Network

摘要:

电池容量是判断电池性能状态的重要指标。针对锌银电池的荷电状态估计问题,利用电池放电过程中放电时间、放电电流和电池电压3个参数作为径向基神经网络的输入,电池荷电状态为输出,建立电池放电的径向基神经网络模型;为克服径向基神经网络收敛精度不高、易陷入局部极小值的缺点,采用差分进化算法和粒子群算法结合的混合算法优化RBF神经网络;MATLAB仿真结果表明,经过混合优化算法优化的径向基神经网络与仅使用粒子群优化的径向基神经网络相比,估计精度得到大大提高。

Abstract:

The battery capacity is the key characteristic which determines the performance of battery. The estimation of zinc silver battery state of charge is the key research point. Using battery discharge time of whole discharge process, the discharge current and battery voltage as radial basis function (RBF) neural network input parameters, battery status of charge as the output parameter, this paper establishes RBF neural network model of battery discharge. In order to overcome the shortcoming of the RBF neural network with low convergence accuracy and easy to fall into local minimum, the hybrid algorithm, which is based on differential evolution algorithm and particle swarm optimization, optimizes the RBF neural network. After Simulating with MATLAB, results display that, RBF neural network optimized by hybrid optimization algorithm possesses higher estimation accuracy and less estimation error compared with RBF neural network optimized by particle swarm optimization algorithm.