[Paper Review] Feature Selection for Transient Stability Assessment Based on Improved Maximal Relevance and Minimal Redundancy Criterion

Yang Li, Xueping Gu|arXiv (Cornell University)|Jan 23, 2019

Smart Grid and Power SystemsEngineering4 references17 citations

TL;DR

This paper proposes an improved maximal relevance and minimal redundancy (mRMR) feature selection method for transient stability assessment in power systems, enhancing feature correlation and redundancy measurement via a weight coefficient. Evaluated on the New England 39-bus and IEEE 50-generator systems, the method identifies optimal feature subsets using incremental search and SVM classification, achieving high classification accuracy with real-time PMU data.

ABSTRACT

A new feature selection method based on an improved maximal relevance and minimal redundancy (mRMR) criterion was proposed for power system transient stability assessment. First, the standard mRMR was improved by introducing a weight coefficient in the evaluation criteria to refine the measurement of the features correlation and redundancy. Then, the possible real-time information provided by phasor measurement unit (PMU) considered, a group of system-level classification features were extracted from the power system operation parameters to build the original feature set, and the improved mRMR was employed to evaluate the classification capability of the original features for feature selection. A group of nested candidate feature subsets were obtained by using the incremental search technique, and each candidate feature subset was tested by a support vector machine classifier to find the optimal feature subset with the highest classification accuracy. The effectiveness of the proposed method was validated by the simulation results on the New England 39-bus system and IEEE 50-generator test system.

Motivation & Objective

• Address the challenge of selecting relevant, non-reduundant features from high-dimensional PMU data for transient stability assessment.
• Improve the standard mRMR criterion by introducing a weight coefficient to refine correlation and redundancy measurement.
• Develop a real-time capable feature selection framework using incremental search and SVM-based evaluation.
• Optimize classification accuracy for transient stability prediction using system-level features derived from PMU measurements.
• Validate the method’s effectiveness on standard power system test systems under various contingency scenarios.

Proposed method

• Enhance the mRMR criterion by incorporating a weight coefficient in the evaluation function to better balance relevance and redundancy.
• Extract system-level classification features from PMU measurements, including voltage, current, and angle quantities, to form the original feature set.
• Apply an incremental search technique to generate nested candidate feature subsets, progressively adding features based on the improved mRMR score.
• Use a support vector machine (SVM) classifier to evaluate the classification performance of each candidate subset.
• Select the optimal feature subset that maximizes classification accuracy while minimizing redundancy.
• Integrate real-time PMU data into the feature selection pipeline to ensure practical applicability in online transient stability assessment.

Experimental results

Research questions

• RQ1 How can the standard mRMR criterion be improved to better capture feature relevance and redundancy in transient stability assessment?
• RQ2 What is the impact of introducing a weight coefficient on the selection of optimal feature subsets from PMU data?
• RQ3 Can the proposed method achieve higher classification accuracy than conventional mRMR in transient stability prediction?
• RQ4 How effective is the incremental search strategy combined with SVM in identifying the best feature subset from a large feature space?
• RQ5 To what extent does the method maintain low redundancy while preserving high relevance in real-time power system applications?

Key findings

• The improved mRMR criterion with a weight coefficient significantly enhances the measurement of feature relevance and redundancy, leading to better feature selection performance.
• The proposed method achieved high classification accuracy in transient stability assessment on both the New England 39-bus and IEEE 50-generator test systems.
• The incremental search strategy efficiently explored nested feature subsets, enabling systematic evaluation and selection of the optimal subset.
• The SVM classifier demonstrated strong generalization performance when used to validate candidate feature subsets, confirming the robustness of the selected features.
• The integration of real-time PMU data into the feature selection framework ensures practical feasibility for online transient stability monitoring.
• The method outperformed standard mRMR in identifying a compact, high-performing feature subset, reducing computational load without sacrificing accuracy.