Performance analysis of data-driven techniques for detection and identification of low frequency oscillations in multimachine power system

Abstract

In power systems, identification and damping of low-frequency oscillations(LFO) is very crucial to maintain the small signal stability. Hence the computation of eigenvalues, eigenmode shapes, participation factors, and coherency of generators are essential parameters of critical LFO modes. The existing data-driven approaches explore either one or two aspects of modal parameters from the dynamic pattern of the measurement data. In the present work, two approaches i) Iterative Approach(IA) ii) Non-Iterative Subspace(SS) method of data-driven techniques are used to estimate the state-space model of the system under study from the measurement data in a holistic framework. Based on the estimated system model, the eigenvalues of LFO, eigenmode shapes, participation factor, and coherency of associated generators participating in electromechanical oscillations are computed. Finally, from the estimated participation factors for the Inter-area oscillation mode (IAM), the Static Synchronous Compensator (STATCOM) damping controller is designed and placed at the generator with the highest participation factor for damping of inter-area oscillation. The enhancement of damping ratio of inter area mode with STATCOM damping controller is estimated and verified using IA SS data driven approaches for the first time. In this work, IA uses prediction-error minimization algorithm (PEM) Parallel computing techniques and SS method uses Multivariable Output Error State Space (MOESP) algorithm for the estimation of Hankel matrix from the measured data. The effectiveness of data-driven approaches are demonstrated by the simulation of a IEEE 4-machine,10-bus power system using MATLAB/Simulink. IA SS methods incorporating wavelet based denoising techniques are very effective in identifying the LFO modes even with noisy measurement. The efficacy of the denoising to suppress the effect of noise is demonstrated by comparing with noiseless environment. The results of data-driven approaches indicate their high degree of accuracy and efficiency in being consistent with Eigenvalue analysis (EA) performed on the system.