A Machine Learning-Based Method for Identifying Critical Distance Relays for Transient Stability Studies

TL;DR

This paper introduces a machine learning approach using random forest classifiers to identify critical distance relays in power systems, simplifying stability studies by focusing only on essential relays.

Contribution

It presents a novel method that leverages early-terminated stability study results to efficiently determine which distance relays are critical for accurate modeling.

Findings

High accuracy in identifying critical relays

Modeling only critical relays suffices for stability analysis

Method validated on WECC system data

Abstract

Modeling protective relays is crucial for performing accurate stability studies as they play a critical role in defining the dynamic responses of power systems during disturbances. Nevertheless, due to the current limitations of stability software and the challenges of keeping track of the changes in the settings information of thousands of protective relays, modeling all the protective relays in bulk power systems is a challenging task. Distance relays are among the critical protection schemes, which are not properly modeled in current practices of stability studies. This paper proposes a machine learning-based method that uses the results of early-terminated stability studies to identify the critical distance relays required to be modeled in those studies. The algorithm used is the random forest (RF) classifier. GE positive sequence load flow analysis (PSLF) software is used to perform stability studies. The model is trained and tested on the Western Electricity Coordinating Council (WECC) system data representing the 2018 summer peak load under different operating conditions and topologies of the system. The results show the great performance of the method in identifying the critical distance relays. The results also show that only modeling the identified critical distance relays suffices to perform accurate stability studies.