关键词: 孤岛检测, 小波包对数能量熵, BP神经网络, 特征量

Abstract:

The detecting time is long and non-detection zone (NDZ) is large for traditional passive islanding detection methods, while active methods have some negative effects on power quality. A novel islanding detection method is proposed based on wavelet packet log energy entropy (WPLEE) and BP neural network. Firstly, the voltage of point of common coupling (PCC) are gathered, and then WPLEE is adopted to analyze the voltage signal. Secondly, by log energy entropy to get adapted to the characteristics of islanding detection algorithm processing variable. Lastly, BP neural network used these characteristic variable to pattern recognition and determine whether there is an island phenomenon,especially in the case of the inverter output power and load power matching. The simulation and experiment results show that this method is faster than the traditional passive methods in islanding detection, and the non-detection zone is smaller and there isn’t a negative impact on power quality.

Key words: Islanding detection, wavelet packet log energy entropy, BP neural network, characteristic variable