Data‐driven disturbance source identification for power system oscillations using credibility search ensemble learning

Abstract

Low-frequency oscillations in power system degrade power quality and may trigger blackouts. This study identifies the source location of these oscillations using measurements from phasor measurement unit (PMU), offline credibility estimation and classification models. The performance of these classification models is ranked for each reported feature to use highly ranked models during the online stage. This proposed framework named as credibility search ensemble learning was tested and validated with promising results using western interconnection power system in North America (WECC-179). The reliability and robustness of the proposed framework were checked against measurement errors in PMUs as well as for practical topology change scenarios. Experimental results and performance comparison with average weight-based approach proved that the proposed approach is capable enough to predict the source location of oscillations with good accuracy. An interfacing tool, for MATLAB-WEKA, was developed and employed in this work for validation and testing of the proposed approach.

Power system wideband oscillation estimation, localization, and mitigation

“…Ref. [92] uses the confidence search ensemble learning to localize the low‐frequency oscillation source under the consideration of PMU measurement errors and system topology changes. Ref.…”

A Review of Machine Learning Approaches in Synchrophasor Technology

“…In localising FOs where PMUs and µPMUs are deployed, ML and DL algorithms function admirably. [163] proposes ensemble learning, a data mining-based ML approach that improves the localization of fault identification. One of the method's possible downsides is that it requires complete system observability.…”

“…Table 9 shows the ML applications in forced oscillation localization of PS network. Ensemble learning [163] MTS, DTW, k-NN [164] Multivariate classification [165] Deep Learning LSTM [166] Two-stage deep transfer learning [167] Transformerbased DL approach [168] CNN models [169]…”

Investigation the influence of the parameters of the computational scheme on detecting the source of low-frequency oscillations

“…Various ways of combining alternative methods are proposed [8,9], ensembles of methods are being built [10], computing platforms are being created for comparative experiments [10,1], test data sets are being accumulated [7], typical simulation models of power systems are being formed [12] for testing analysis methods, etc.…”