Data-Driven Small-Signal and N-1 Security Assessment Considering Missing Data

Abstract

This paper proposes a modulated data-driven method to assess the small-signal and N−1 security status in large power systems. To do so, a three-module data-driven framework is designed, including: i) auto-encoder embedded feature selection to reduce the measurement dataset dimension and enhance the computational efficiency; ii) modified generative adversarial networks to improve the robustness against incomplete data and partial observability, which preserves the interpretability of the data measured by phasor measurement units (PMUs) using a reformulated loss function and a new noise generation process; and iii) a convolutional neural network (CNN) as a strong classifier for the assessment of the small-signal and N−1 security status. The proposed method is implemented on a 162-bus NESTA benchmark system. The results show the performance of the designed network for different cases and in comparison with several state-of-the-art methods in terms of accuracy, reliability, and computational burden.