1 Abstract—System identification methods have been widely used for the study of low frequency electromechanical oscillations and the development of low order dynamic models. This paper introduces a hybrid frequency/time-domain approach to estimate the dominant modes contained in ringdown responses of power systems. Practical issues and solutions encountered in the application of the hybrid method are discussed. The performance of the proposed technique is evaluated by applying the Monte Carlo method to synthetic signals and simulated responses from a large-scale power system, as well as to measurements recorded in a microgrid laboratory test facility. Results in all cases proved to be very accurate, verifying the robustness of the proposed method. Index Terms—Dynamic equivalencing, least-squares methods, mode estimation, ringdown analysis, system identification, system measurements. I. INTRODUCTION HE identification of the oscillatory modes contained in a post-disturbance or “ringdown” response can provide vital information to assess the dynamic performance of the power system and plan measures to mitigate possible small-signal stability problems and reliability impacts [1]. Traditionally, power system modes are obtained by applying eigenanalysis on a detailed linearized model. However, this approach lacks of usability in cases of large power systems and real-time monitoring, due to the significant computational burden and the difficulties to keep the developed dynamic models updated over time and operating conditions. Other drawbacks also include the difficulty to limit the identification only to the modes of interest, neglecting possible surplus artificial modes, as well as limitations in cases of strongly nonlinear problems [2]-[4]. To overcome these weaknesses, measurement-based approaches have been proposed as supplementary solutions to directly estimate the dominant system modes from measured Manuscript received December 1, 2015. The work of T. A. Papadopoulos is supported by the ‘Ι Fellowships of Excellence for Postgraduate Studies in Greece – Siemens Program’. T. A. Papadopoulos is with the Power Systems Laboratory, Dept. of Electrical & Computer Engineering, Democritus University of Thrace, Xanthi, Greece, GR 67100, (corresponding author’s e-mail: thpapad@ee.duth.gr). A. I. Chrysochos, E. O. Kontis, P. N. Papadopoulos, and G. K. Papagiannis are with the Power Systems Laboratory, School of Electrical & Computer Engineering, Aristotle University of Thessaloniki, Thessaloniki, Greece, GR 54124, (e-mail: grigoris@eng.auth.gr). responses [5]. This technique gains significant interest considering the continuously increasing installation of phasor measurement units (PMUs), supported by global position system (GPS) and high-speed communication infrastructure in the form of wide-area measurement systems (WAMS) [4], [6]. In this context, real-time monitoring of power systems as well as the development of online dynamic equivalents [7] is possible using system identification techniques [4], [8]. The measurement-based approach also includes cases where model revalidation and calibration of power system components is required, providing a safer and more cost-effective alternative to staged tests [1]. Measurement-based methods can be generally classified into two main categories, in terms of the used type of data: • Ambient-mode estimation methods that analyze responses excited by small load variations. Ambient data are continuously available through WAMS and intrude the least possible into power systems [9]-[11]. • Methods that perform analysis of ringdown responses occurred after a major disturbance, such as large load step-up, line-tripping, etc. The advantages of these methods are the high accuracy of the mode estimates and the fast convergence to the true values, since ringdown responses contain higher level of mode information density compared to ambient data [3], [12]. Several identification methods have been proposed to estimate the oscillatory modes from ringdown responses in power systems. The majority of them are based on the direct identification of model parameters from time-domain (TD) responses, with the most known being Prony method, originally proposed in [13], and later extended and improved to include transfer function applications and multiple output models [5], [14]-[17]. Other popular ringdown techniques include the minimal realization algorithm [18], the eigenvalue realization algorithm (ERA) [19], the matrix pencil method [20], the numerical algorithm for sub-space state-space system identification (N4SID) [21], and the prediction error method (PEM) [22]. Alternatively, in [23]-[25] the dominant modes are extracted in frequency-domain (FD) using the fast Fourier transform (FFT), combined also with the sliding window method for the estimation of the mode damping factor, while in [26] Hilbert-Huang transform is applied. Most of these methods use high-order models that contain additional artificial modes apart from the dominant ones. This is done to Measurement-Based Hybrid Approach for Ringdown Analysis of Power Systems Theofilos A. Papadopoulos, Member, IEEE, Andreas I. Chrysochos, Student Member, IEEE, Eleftherios O. Kontis, Student Member, IEEE, Panagiotis N. Papadopoulos, Member, IEEE, and Grigoris K. Papagiannis, Senior Member, IEEE T