【摘要】 BP神经网络在进行光伏阵列故障检测时,故障诊断性能易受局部极值影响,且收敛速度过慢。针对这一问题,利用遗传算法(GA)的全局搜索能力,找到BP神经网络初始权值阈值的最优解,可以有效克服BP神经网络上述缺陷,并高效完成对光伏阵列的故障诊断。对比光伏阵列正常状态与不同故障状态下的输出特性,选定电压电流作为输入,故障状态为输出,建立GA-BP故障诊断模型,将经过遗传算法优化后的BP神经网络与经典的BP神经网络相比较,通过仿真验证,GA-BP神经网络不仅可以更快速地进行故障诊断,还提高了故障诊断的准确率。更多还原

【Abstract】 The fault diagnosis performance of BP neural network is easily affected by local extreme values, and the convergence speed is too slow. In order to solve these problems, using the global search ability of genetic algorithm(GA) to find the optimal solution of initial weight and threshold of BP neural network. This way not only can be effectively overcome the above defects of BP neural network, but also can complete the fault diagnosis efficiently.Compared the output characteristics of the normal state of the photovoltaic array with different fault states, a fault diagnosis model was established. Compared the BP neural network optimized by genetic algorithm with the basic BP neural network, the simulation results show that, GA-BP neural network can not only perform fault diagnosis more quickly, but also improve the accuracy of fault diagnosis.更多还原