A short-term voltage stability online prediction method based on graph convolutional networks and long short-term memory networks

Abstract

Due to complex dynamic characteristics and large scale of power systems, it is a great challenge to predict short-term voltage stability (STVS) online. To address this challenge, a STVS online prediction method based on graph convolutional networks (GCN) and long short-term memory networks (LSTM) is proposed in this paper. Firstly, we propose a novel machine learning framework, GCN-LSTM, which is a combination of GCN and LSTM. Specifically, the GCN is used to capture spatial features of power grids, the LSTM is used to capture temporal features of power grids. Secondly, a STVS online prediction method based on the GCN-LSTM model is proposed. The proposed method can capture multiplex spatial–temporal STVS evolution trends and predict STVS results. Finally, case studies are carried out in two testing systems, including a modified 39-bus system and a 68-bus system. The training and testing data is generated by Power System Simulator / Engineering (PSS/E). Simulation results illustrate the high performance of the proposed method.