A robust WKNN-TLS-ESPRIT algorithm for identification of electromechanical oscillation modes utilizing WAMS
- PDF / 440,842 Bytes
- 6 Pages / 595.276 x 790.866 pts Page_size
- 15 Downloads / 178 Views
Ó Indian Academy of Sciences
Sadhana(0123456789().,-volV)FT3](012345 6789().,-volV)
A robust WKNN-TLS-ESPRIT algorithm for identification of electromechanical oscillation modes utilizing WAMS SHEKHA RAI Department of Electrical Engineering, National Institute of Technology Rourkela, Rourkela, Orrisa, India e-mail: [email protected]; [email protected] MS received 11 May 2020; revised 12 August 2020; accepted 16 September 2020 Abstract. This paper proposes a robust WKNN-TLS-ESPRIT algorithm that takes into account the effect of the unavailability of phasor measurement unit (PMU) data for identifying the low-frequency oscillatory modes in power systems. The main contribution of the proposed work is to create an enhanced autocorrelation matrix using a weighted K nearest neighbours (WKNN)-based predictive model to deal with such issues. In the present work, a Bayesian approach is utilized to determine the empirical number of neighbourhood parameters. The improved autocorrelation matrix is then used by total least square estimation of signal parameters via rotational invariance technique (TLS-ESPRIT) algorithm to provide a robust estimate of the modes. Robustness of the proposed method over the other methods is validated with a simulated test signal with missing data through Monte Carlo simulations at different SNRs. The effectiveness of the proposed approach is further verified on real data derived from PMU located in Western Electricity Coordinating Council grid. Keywords.
Modal analysis; missing data; WAMS; K nearest neighbours (KNN).
1. Introduction The increasing capacity of generation through renewable sources of energy has caused uncertainty in generation and subsequently, its integration into the electric grid has posed severe operational challenges concerned with stability of large-scale power systems. Maintaining adequate damping for small-signal oscillatory modes is highly essential. Conventional techniques for identifying the dominant modes use an off-line approach that fails to observe the real-time system dynamics. In the recent times, with the advancement of wide area monitoring systems (WAMSs) that employ phasor measurement units (PMUs) and Global Positioning System (GPS) to provide highly accurate timestamped phasors [1], the viability for on-line extraction of modal information has increased. Numerous measurementbased online estimation algorithms have been proposed like Kalman filter [2], sparsity [3], variable projection [4], Prony [5] and ESPRIT [6]. The Kalman filter uses an iterative approach and hence, has the disadvantage of being numerically unstable. A variable projection algorithm involving orthogonal projection to extract the modal parameters from the signal space is cited in [4]. Some identification methods use ESPRIT [6] to create an autocorrelation matrix from the observed data. Methods based on ESPRIT show higher noise immunity than Prony.
PMUs may have incomplete measurements due to high congestion in communication network, malfunctioning of PMUs or PDCs, malicious attack, etc. [7].
Data Loading...
