Faculty Publications
Permanent URI for this communityhttps://idr.nitk.ac.in/handle/123456789/18736
Publications by NITK Faculty
Browse
4 results
Search Results
Item A comparison of power system signal detrending algorithms(Institute of Electrical and Electronics Engineers Inc., 2018) Rao, K.; Shubhanga, K.N.Wide Area Measurement Systems (WAMS) have facilitated tracking of oscillations in power system response signals. This has provided an impetus for application of signal measurement-based modal detection methods such as matrix pencil and Prony analysis. Detrending, which means removal of trend in a signal, is a pre-requisite for effective functioning of these modal detection methods. In this work, performance of three methods of detrending viz., Center-of-Inertia (COI)-based detrending, MATLAB function-based detrending and 'Zhou' detrending are compared with particular reference to power system signals. It is indicated that COI-based detrending is better suited to detection of modes from slip signals of generators. © 2017 IEEE.Item A Comparative Study between Prony and Eigensystem Realization Algorithm for Identification of Electromechanical Modes(Institute of Electrical and Electronics Engineers Inc., 2018) Sarkar, N.; Rao, K.; Shubhanga, K.N.Two measurement-based ring-down electromechanical mode identification algorithms, namely Prony and Eigensystem Realization Algorithm (ERA) are taken up for a comparative study. Since the number of excited modes might vary in a practical power system, it is not easy to determine the model order. This requires an iterative procedure in case of Prony whereas a Singular Value Decomposition (SVD)-based technique achieves this directly in case of ERA as demonstrated through two case studies. It is further shown that ERA estimates the signal better and generates less number of trivial modes as compared to Prony. © 2018 IEEE.Item EMD based Detrending of Non-linear and Non-stationary Power System Signals(Institute of Electrical and Electronics Engineers Inc., 2021) Aalam, M.K.; Shubhanga, K.N.In electromechanical modal analysis of power systems using Wide Area Measurement System (WAMS) based setup, signal processing is complex as the signals are non-stationary and non-linear in nature. In order to get accurate modal parameters, as a first step, it is required to remove the non-linear trend of the signal. In the literature, many conventional methods such as filtering, averaging and peak detection techniques are employed for removing trend. In this paper, Empirical Mode Decomposition (EMD) method, an iterative algorithm is presented to detrend a signal. The EMD method and its variant are compared with another popularly used peak detection method referred to as the Zhou's detrending algorithm to find the efficacy of the EMD methods. To test the algorithms, a four machine, two-area power system with three-wind farms is modeled and simulated to generate the power system signals which bring out non-linear and non-stationary nature. Further, the modal characterization is carried out employing Prony analysis. © 2021 IEEE.Item MAPE-An Alternative Fitness Metric for Prony Analysis of Power System Signals(De Gruyter peter.golla@degruyter.com, 2018) Rao, K.; Shubhanga, K.N.Phasor Measurement Units have facilitated tracking of oscillations in power system response signals. This has provided an impetus for identifying unstable component modes directly from oscillatory signals. Prony analysis, the earliest method proposed for this purpose, throws up some trivial modes. These not only distract the analyzer but also prolong processing time thereby delaying corrective action. Hence the fitness metric chosen should serve to minimize the number of trivial modes. The conventional fitness metric is Signal-To-Noise Ratio (SNR), which is actually Signal-To-Estimation error Ratio (SER). This paper proposes that Mean Absolute Percentage Error (MAPE) can also serve well as a fitness metric. It is shown through case studies carried out on well-known four-machine power system that there are a few cases where MAPE performs better than SER while in some instances SER works better. This inference is verified even in the presence of measurement noise. Hence a novel fitness metric is proposed combining MAPE with SER. Case studies on simulated signals obtained from New England-power system prove that this novel metric can achieve considerable reduction in processing time. Besides, an exponential binary search has been suggested for determining the optimal model order in minimum number of iterations. © 2018 Walter de Gruyter GmbH, Berlin/Boston 2018.