基于神经网络的分布式光伏故障外特性聚合等值建模  

A Neural Network-based Aggregated Modelling Approach for Distributed Photovoltaic Fault External Characteristics

作　　者：范荣奇 李宽 王安宁 高帅 黄涛[4] FAN Rongqi;LI Kuan;WANG Anning;GAO Shuai;HUANG Tao(State Grid Shandong Electric Power Company,Jinan 250001,China;State Grid Shandong Electric Power Research Institute,Jinan 250003,China;Shandong Smart Grid Technology Innovation Center,Jinan 250003,China;Nanjing NR Electric Co.,Ltd.,Nanjing 211102,China)

机构地区：[1]国网山东省电力公司,山东济南250001 [2]国网山东省电力公司电力科学研究院,山东济南250003 [3]山东省智能电网技术创新中心,山东济南250003 [4]南京南瑞继保电气有限公司,江苏南京211102

出　　处：《山东电力技术》2025年第4期69-80,共12页Shandong Electric Power

基　　金：国网山东省电力公司科技项目“高密度、高比例电力电子化区域电网仿真建模及控保协同技术研究”(52062623000W)。

摘　　要：提出一种基于LSTM神经网络的分布式光伏故障外特性聚合等值建模方法。该方法可以输出任意出力下等值光伏系统并网点的I-V曲线,在并网点电压跌落后能够预测等值系统的故障特性。相比传统机理建模,该方法无须对复杂物理系统进行具体建模,能够精确映射配电网的强非线性输入输出。所建立的LSTM模型首先使用一维卷积层对光伏出力系数进行特征提取,然后利用两层隐藏层处理序列数据,在全连接层中将向量映射为外特性曲线序列。基于传统光伏电源模型搭建典型的配电网网络,选择大量不同出力组合进行仿真,为LSTM模型训练提供有效的训练集和验证集数据,同时建立独立的测试集测试最优模型的准确性。最后利用最佳模型建立等值配电网系统,在并网点设置不同程度电压跌落,将故障特性与完整模型的故障特性进行对比,仿真结果可以证明所提方法的可靠性和实用性。In this paper,we propose a method for modelling distributed PV power fault output aggregation based on the LSTM neural network.The method can provide the I-V curves of equivalent PV system parallel points under any output and can predict the fault characteristics of the equivalent system after a voltage drop at the parallel point.Compared to traditional mechanism modelling,this method does not require specific modelling of complex physical systems and is able to accurately represent the highly non-linear inputs and outputs of distribution networks.The LSTM model established in this paper first uses a one-dimensional convolutional layer for feature extraction of PV power coefficients,and then two hidden layers are used to process the sequence data,and the vectors are mapped into a sequence of external characteristic curves in the fully connected layer.In this paper,a typical distribution network is constructed based on the traditional PV power model,and a large number of different output combinations are selected for simulation to provide effective training set and validation set data for the LSTM model training,and at the same time,an independent test set is established to test the accuracy of the optimal model.Finally,the optimal model is used to construct an equivalent distribution network system,and different degrees of voltage drops are set at the network connection points to compare the fault characteristics with those of the full model,and the simulation results can prove the reliability and practicality of the proposed method.

关 键 词：分布式光伏 LSTM 神经网络 聚合建模 故障特性 

分 类 号：TM743[电气工程—电力系统及自动化]