A Learning-Driven Stochastic Hybrid System Framework for Detecting Unobservable Contingencies in Power Systems

TL;DR

This paper introduces a learning-driven stochastic hybrid system framework that enhances detection and classification of hidden contingencies in power systems, outperforming traditional methods in speed and accuracy.

Contribution

It proposes a novel hybrid system model combined with machine learning for unobservable contingency detection in power systems, addressing limitations of existing monitoring schemes.

Findings

Improved detection speed over traditional methods

Higher accuracy in classifying contingencies

Effective on IEEE 5-bus and 30-bus systems

Abstract

This paper presents a new learning based Stochastic Hybrid System (LSHS) framework designed for the detection and classification of contingencies in modern power systems. Unlike conventional monitoring schemes, the proposed approach is capable of identifying unobservable events that remain hidden from standard sensing infrastructures, such as undetected protection system malfunctions. The framework operates by analyzing deviations in system outputs and behaviors, which are then categorized into three groups: physical, control, and measurement contingencies based on their impact on the SHS model. The SHS model integrates both system dynamics and observer-driven state estimation error dynamics. Within this architecture, machine learning classifiers are employed to achieve rapid and accurate categorization of contingencies. The effectiveness of the method is demonstrated through simulations on the IEEE 5-bus and 30-bus systems, where results indicate substantial improvements in both detection speed and accuracy compared with existing approaches.