Online Transient Stability Margin Estimation Using Improved Deep Learning Ensemble Model

Abstract

This paper addresses a novel deep learning (DL) approach for online estimating the transient stability margin (TSM) in power grids. The TSM is characterized by a functional relationship between power system variables and the critical clearing time (CCT). To enhance the accuracy of TSM estimation, an improved DL ensemble (iDLE) model, which incorporates the dynamic error correction (DEC) and the multi-objective ensemble learning (MOEL), is proposed. The iDLE model is formulated as an evolutionary multi-objective framework and optimized using the non-dominated sorting genetic algorithm (NSGA-II) along with fuzzy decision analysis to derive the optimal solution. The proposed model is applied to a classical test system and a practical power system, followed by a discussion of the results.