A Two-Stage transient stability prediction method using convolutional residual memory network and gated recurrent unit

Abstract

In order to improve the accuracy as well as the reliability of transient stability prediction (TSP), a two-stage TSP method using convolutional residual memory network (CRMN) and gated recurrent unit (GRU) is proposed. In the first stage, the underlying measurement data are directly used as input features to build a CRMN-based TSP prediction model for qualitative and quantitative analysis. In the second stage, the GRU-based generator rotor angle trajectory (GRAT) prediction model is firstly established. Subsquently, the unstable samples in qualitative analysis and the samples with a confidence interval of 99.66% in quantitative analysis are used for GRAT prediction. As a consequence, more reliable prediction results can be obtained by comprehensive judgment about the results from qualitative analysis, quantitative analysis and GRAT prediction. Case studies conducted on a modified New England 10-machine 39-bus system and an IEEE 50-machine 145-bus system demonstrate superior accuracy, stronger robustness of the proposed model than other traditional models involving LSTM, GRU and CNN. Furthermore, the results of numerical experiments also prove that the proposed two-stage TSP method improves the reliability of prediction results.