光伏阵列故障类型的改进型RBF神经网络识别算法

光伏阵列是光伏系统中非常重要的组成部分。传统的BP神经网络诊断算法有着精度低、收敛速度慢等缺点,为了精确地诊断出光伏阵列内部的故障位置及其类型,通过分析阵列开路、短路、老化、阴影和电池板裂片5种故障,提出了一种改进型RBF神经网络的故障诊断识别算法。首先,建立RBF神经网络的光伏阵列故障诊断模型,确定基于遗传算法的故障模型隐层中心的确定方法,然后针对基于粒子群优化算法的网络模型进行自适应权重寻优的仿真实验。最后,将优化的算法与传统RBF神经网络算法进行对比。结果表明:该优化算法不仅可以有效地诊断光伏阵列的故障类型,还可以提高故障诊断的准确率。

Photovoltaic Array Fault Identification Algorithm Based on Improved RBF Neural Network

Photovoltaic(PV) array is an important part of the PV system. The traditional BP neural network diagno-sis algorithm has some disadvantages, such as low accuracy and slow convergence speed. To diagnose the location and types of fault in the PV array accurately, a fault diagnosis and identification algorithm based on the improved RBF neural network is put forward through analyzing five types of fault, i.e., open circuit, short circuit, aging, shadow, and panel fragmentation. Firstly, a PV array fault diagnosis model based on radial basis function(RBF) neural network is esta-blished. The method of determining the center of hidden layer of the fault model is formulated based on genetic algorithm, and then simulation experiments are conducted using the adaptive network weight optimization method based on particle swarm optimization(PSO) algorithm. Finally, the optimized algorithm and the traditional RBF neural network algorithm are com-pared. Results show that the proposed algorithm can not only diagnose the fault types of PV array effectively, but also improve the accuracy of fault diagnosis.