基于自适应权重粒子群优化BP神经网络的光伏阵列故障诊断与定位

Fault Diagnosis&Location of Photovoltaic Array Based on BP Neural Network Trained by Adaptive Inertia Weight Particle Swarm Optimization Algorithm

BP神经网络算法易陷入局部极小点,而自适应权重粒子群算法全局搜索能力强,将两者结合起来,提出了一种基于自适应权重粒子群优化BP神经网络的光伏阵列故障诊断方法,仿真验证该算法可避免网络陷入局部极小点,提高网络预测精度。提取光照强度S、组件温度T、光伏阵列开路电压U_(oc)、最大功率P_m和电压表U_2、U_1与电流表I_2、I_1作为光伏阵列的特征量,经仿真测试比较,表明了所选特征量合理,降低了诊断复杂度,提高了预测精度。

BP neural network algorithm can easily fall into local minimum point, while adaptive inertia weight particle swarm optimization （AWPSO） has great performance in global searching. Combined with the above two mentioned algorilhms, a new algorithm based on BP neural network trained by AWPSO is presented for fault diagnosis and its location of photovohaie（PV） array. Simulation test validate that the algorithm refrains the network from falling into local minimum point and improves the network prediction accuracy. The two ambient parameters including light intensity S ＆ environmental temperature T, and the two PV array parameters including open circuit voltage U∝ and maximum power Pm, and the metering data ineluding U2,U1 of vohmeter and I2, I1 of ammeter are extracted as the characteristics of PV array. Simulation test shows that the selected characteristics are reasonable, and the complexity of the diagnosis process is reduced,while the prediction accuracy is improved.