Energy efficient fault detection and classification using hyperparameter-tuned machine learning classifiers with sensors

Abstract

This paper presents a novel approach in which hyperparameter-tuned Machine Learning (ML) classifiers with Optuna is used for fault detection and classification over a power transmission line. In this paper, popular ML models Random Forest (RF), Decision Tree (DT), XGBoost (XGB), and Light Gradient Boosting Machine (LGBM) are used for fault detection and classification. This study uses a two-layer approach for fault detection and classification. An electrical utility and data simulation provided the PMU measurement and recorded data from a simulated grid. The faults had varied impedances and included various fault classes at distinct line locations. The optimal feature from 3-phase current and voltage signals is extracted using Pearson correlation, recursive feature elimination, and univariate feature (t-test) methods. The synthetic minority class oversampling technique (SMOTE) was used to address the issue of imbalanced data. Hyperparameters of the evaluated LGBM classifier is trained with the Optuna. The performance of the proposed classifier is measured in terms of the accuracy, precision, Recall and F1-score metrics. The proposed method outperformed the conventional ML methods.