基于小波变换与神经网络的孤岛检测技术

Islanding Detection Based on Wavelet Transform and Neural Network

为保障安全与电力用户供电质量,基于并网逆变器的分布式发电(distributed generation DG)系统要求具备孤岛检测功能。针对被动式孤岛检测法检测盲区(non-detection zone,NDZ)大、检测时间长以及主动式孤岛检测法影响分布式发电系统供电质量的缺点,提出了一种新的被动式孤岛检测方法。该方法利用小波变换从公共耦合点(point of common coupling,PCC)处的电压信号及逆变器输出电流信号中提取特征量,再通过BP神经网络进行模式识别来判断是否出现孤岛现象。仿真与实验结果表明,该方法比传统的被动式孤岛检测方法检测速度快,检测盲区小。同时,由于所提供的孤岛检测法没有向控制信号中加入扰动量,因而不会对电能质量产生不良影响,克服了主动式孤岛检测方法的不足,并具有很高的准确性与可靠性。

The function of islanding detection is required for the grid-connected inverter-based distributed generation system due to safety reasons and to maintain the quality of power supply. Passive methods have a large non detection zone and the detecting time is long, while active schemes have negative influence on power quality, so a novel passive islanding detection method was proposed. In this method, wavelet transform was adopted to extract feature vectors from the voltage of point of common coupling （PCC） point and the output current of inverter, and then pattern recognition was exerted by BP neural network to determine whether there was an island phenomenon. 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. At the same time, because no disturbance was added to the control signal in the method, there isn’t a negative impact on power quality. The method overcomes the shortcoming of active methods and has high accuracy and reliability.