基于长短期记忆网络和模糊C均值聚类的配电网故障预警方法研究  

Research on distribution network fault warning method based on long short-term memory network and fuzzy C-means clustering

作　　者：吴佳庆 顾洁[1] 金之俭[1] 游铭豪 王耀健 WU Jiaqing;GU Jie;JIN Zhijian;YOU Minghao;WANG Yaojian(School of Electronic Information and Electrical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]上海交通大学电子信息与电气工程学院,上海200240

出　　处：《供用电》2023年第10期63-72,共10页Distribution & Utilization

基　　金：上海市科学技术委员会重点项目(18DZ1100303)。

摘　　要：为辅助电网运行部门能够及时采取有效的风险防控措施,提高配电网供电可靠性,提出一种基于长短期记忆网络(long short-term memory network,LSTM)和模糊C均值(fuzzy C-means,FCM)聚类算法的配电网故障预警模型。首先,通过对电网运行数据进行归类,筛选出配电网25个特征变量,并采用主成分分析法进行特征变量融合,形成可表征配电网运行状态的低维特征变量;然后,基于已融合的特征变量,选用FCM聚类算法对配电网故障状态和非故障状态区分的边界条件进行辨识,并采用LSTM对短期内配电网特征量进行预测,进一步结合特征量预测结果和所得出的状态区分边界条件,预测未来短期内的故障发生时刻,实现故障预警;最后,以某地级市39条馈线配电网运行数据为例,验证了所提方法的有效性。In order to assist the power grid operation department in taking effective risk prevention and control measures in a timely manner,and improve the reliability of power supply in the distribution network,this paper proposes a distribution network fault warning model based on long short-term memory network and fuzzy C-means clustering algorithm.Firstly,by screening the operation data of the power grid,25 characteristic variables of the distribution network are selected,and the principal component analysis algorithm is used to fuse the characteristic variables to form a low dimensional feature quantity that can characterize the operation status of the distribution network.Secondly,based on the fused feature variables,the fuzzy C-means clustering algorithm is selected to identify the boundary conditions for distinguishing fault and non fault states in the distribution network,and using long short-term memory network algorithms to predict the characteristics of the distribution network in the short term.Further combining the predicted results of feature quantities with the obtained state differentiation boundary conditions,predict the time of future short-term faults and achieve fault warning.Finally,the effectiveness of the proposed method is verified by the operation data of 39 feeder distribution networks in a prefecture-level city.

关 键 词：故障预警 特征变量 长短期记忆网络 模糊C均值聚类 边界条件 

分 类 号：TM74[电气工程—电力系统及自动化]