Flicker source detection including fixed speed wind turbines using empirical mode decomposition

Abstract

This study proposes an approach to identifying multiple icker sources at the Point of Common Coupling (PCC). The voltage signals of di erent icker sources such as the electric arc furnace, the xed-speed wind turbine, and the diesel-engine driven generator were recorded at the PCC. For this purpose, various aerodynamic and mechanical faults of a wind turbine such as wind shear and tower shadow, gearbox tooth-breaking, blade crash, pitch angle error and various mechanical faults of diesel-engine driven generator such as mis ring, exciter, and governor error are considered. After acquiring voltage signals of various faults, the Empirical Mode Decomposition (EMD) as a robust signal processing technique for extracting useful features was used. Then, to reduce required memory space and computational burden, the Minimal-Redundancy-Maximal-Relevance (MRMR) and the Symmetric Uncertainty (SU) as the feature selection methods were applied. Also, to increase the efficiency of feature selection methods, the cooperative game-theoretic method was utilized. Afterward, two classi ers based on the Naïve-Bayes and the Support Vector Machine (SVM) were employed to detect the faults. Simulation results are presented to validate the effctiveness of the proposed method.