Short Term Voltage Stability Assessment with Incomplete Data Based on Deep Reinforcement Learning in the Internet of Energy

Abstract

Short-term voltage stability (STVS) assessment is one of the important challenges in the power system. With the wide application of phasor measurement units (PMUs), STVS assessment based on data-driven has attracted more and more attentions. In fact, some of the PMU data may be lost when used for STVS assessment, and it may affect the accuracy of STVS assessment. In this paper, we focus on assessing the STVS with incomplete PMU data based on data-driven method. Firstly, long short term memory (LSTM) network is adopted to fill the missing PMU data. Secondly, double deep Q-Learning (DDQN) is used for STVS assessment. The effectiveness of the proposed intelligent evaluation model is verified on the New England 39-bus system. As the experimental results shows, the deviation degree between the predicted PMU data and the true PMU data is small with a 20% data missing rate. Moreover, proposed STVS assessment model can achieve good performance when using the predicted PMU data.