A K-Means-ANN Based Technique for Identification of Oscillatory Modes in Power System

Due to heavy communication network congestion, malicious attacks, malfunction of phasor measurement units (PMUs) or Phasor data concentrators (PDCs), and other factors, PMUs data collected at PDC may comprise of nonlinear treads with missing values and outliers which results in degradation in the performance of estimator, thereby resulting in system stability and sustainability issues. So, in the proposed work, a novel K-Means-ANN based technique has been proposed to overcome this detrimental effect. So, in the proposed technique, an unsupervised K-Mean clustering algorithm has been implemented for the detection and removal of the outliers and, thereby the Artificial Neural Network (ANN) is exploited for missing data imputation. Finally, total-least square-estimation-of-signal-parameters via rotational invariance technique (TLS-ESPRIT) is applied for mode estimation. To validate the robustness of the proposed approach over the other approaches, comparative study is carried out on synthetic signal with different modes of oscillation and noise levels through Monte Carlo simulations. This technique is also validated on two area data and real probing data obtained from Western Electricity Co-ordinating Council (WECC).