Data-Driven Transient Stability Assessment of Power Systems with a Novel GHM-Enhanced CatBoost Algorithm

TL;DR

This paper presents a novel GHM-enhanced CatBoost algorithm for power system transient stability assessment, effectively handling class imbalance and noise, and optimizing sensor placement for cost-efficient operation.

Contribution

It introduces a GHM mechanism into CatBoost for improved stability assessment, addressing class imbalance and noise, and guides sensor placement for economical power system management.

Findings

Superior accuracy in stability prediction

Reduced application and maintenance costs

Effective handling of class imbalance and noise

Abstract

This study introduces an advanced transient stability assessment (TSA) method for power systems, addressing the challenges of sample class imbalance and data noise through a novel CatBoost algorithm framework. By implementing a Gradient Harmonizing Mechanism (GHM), this method adjusts the gradient norm distribution across samples by incorporating a coordination parameter for each, thus optimizing the gradient weights for various sample types. This enhancement enables more effective training of the CatBoost algorithm, reducing the negative impacts of class imbalance and noise, and enhancing algorithmic performance. Additionally, the feature importance functionality of the CatBoost framework guides the placement of phasor measurement units, promoting economical operation of the power system. Numerical results from the New England 10-machine 39-bus system demonstrate the superior versatility, reduced application cost, and lower maintenance expenses of the proposed method compared to existing techniques.