基于长短期记忆网络自编码器的台区线损预测方法  

Station Area Line Loss Prediction Method Based on Long Short Term Memory Network Autoencoder

作　　者：史晨豪 戴尉阳 王菁怡 张怡 SHI Chenhao;DAI Weiyang;WANG Jingyi;ZHANG Yi(Jiangsu Electric Power Company Extra High Voltage Branch,Nanjing 211102,China;Department of Electric Power Engineering,Shanghai University of Electric Power,Shanghai 200090,China)

机构地区：[1]国网江苏省电力有限公司超高压分公司,江苏南京211102 [2]上海电力大学电气工程学院,上海200090

出　　处：《供用电》2022年第1期52-57,共6页Distribution & Utilization

基　　金：国家自然科学基金项目(61872230)。

摘　　要：随着低压配电网接线方式和用电规模的不断变化,台区线损的理论计算愈发困难。提出一种基于长短期记忆网络改进自编码器的台区线损预测方法,首先考虑峰谷段负荷的不同变化特征,提取6种不同的台区电气指标,并利用改进的层次聚类算法进行数据聚类处理;然后通过长短期记忆网络改进的自编码器对不同负荷时间段的台区数据进行学习训练,得到线损预测值;最后在50个台区数据基础上进行仿真求解,证明了所提预测方法的有效性。With the continuous changes in the wiring mode and the scale of power consumption in the low-voltage distribution network,the theoretical calculation of the line loss in the station area becomes more and more difficult.This paper proposes a prediction method of station area line loss based on convolutional neural network autoencoder.Firstly taking into account the different change characteristics of the peak and valley load,six different electrical indicators of the station area are extracted.And use an improved hierarchical clustering algorithm for data clustering,and then use the convolutional neural network autoencoder to learn and train the station data in different load periods to obtain the line loss prediction value.Finally,the simulation solution based on the data of 50 stations has proved the effectiveness of the prediction method proposed in this paper.

关 键 词：低压配电网 台区线损 聚类 长短期记忆网络 自编码器 

分 类 号：TM71[电气工程—电力系统及自动化]