基于GA优化卷积神经网络的含分布式光伏低压台区电能质量扰动特征识别研究

Feature recognition research of power quality disturbance in low⁃voltage region with distributed PV based on GA optimized convolutional neural network

针对扰动特征识别效果较差的问题,采用基于GA优化卷积神经网络的含分布式光伏低压台区电能质量扰动特征识别方法。获取各尺度扰动信号的分解系数,求解电扰动信号各尺度的能量值与能量熵,将能量值与能量熵当成扰动特征提取的依据,完成扰动特征提取;利用GA算法优化卷积神经网络结构参数,在优化后的卷积神经网络内输入提取的扰动特征,输出扰动特征识别结果。通过仿真实验得出,该方法能有效提取扰动特征,提取到的特征类型数量与实际类型数量一致;在噪声环境下,该方法依旧能够有效识别扰动特征,最低识别精度高达98.7%。

In order to solve the problem of poor effect of disturbance feature recognition,a method of power quality disturbance feature recognition for distributed photovoltaic low⁃voltage region based on GA optimization convolutional neural network is proposed.The decomposition coefficients of disturbance signals at various scales were obtained to solve the energy value and energy entropy of electric disturbance signals at various scales.The energy value and energy entropy were taken as the basis for the extraction of disturbance features to complete the extraction of disturbance features.GA algorithm is used to optimize the structural parameters of convolutional neural network,and the extracted disturbance features are input into the optimized convolutional neural network,and the recognition results of disturbance features are output.Simulation results show that this method can extract disturbance features effectively,and the number of extracted feature types is consistent with the actual number of types.In the noise environment,the method can still identify the disturbance features effectively,with the lowest recognition accuracy up to 98.7%.