基于神经网络的黑启动操作过电压的快速预测

在电力系统黑启动方案制定的过程中,需要对不同的黑启动方案进行反复的空充输电线路过电压的仿真计算和校验。文中提出了利用神经网络快速预测黑启动合闸过电压的方法,通过选择一组有效的输入特征形成总的样本空间,并采用电磁暂态仿真程序(EMTP)建立训练样本集和测试样本集,误差反向传播神经网络用于构造输入特征与过电压峰值之间的映射关系,从而实现黑启动合闸过电压的快速预测。算例分析表明了该方法的有效性。

Fast Determination of Switching Overvoltage in Black Start Process Based on Artificial Neural Networks

During the process to work out black start scheme of power system, it is necessary to repeatedly calculate and verify the switching overvoltage of the energized transmission line for various black start schemes. The authors propose a method to quickly determine black start caused switching overvoltage by artificial neural network. By means of choosing a group of effective input features a total sample space is composed and using EMTP software package the training sample set and test sample set are built up. The back-propagation neural network is used to construct the mapping relation between input feature and peak values of overvoltages, thereby the fast determination of black start caused switching overvoltage is realized. The results of numerical example show that the proposed method is effective.