Data-driven wide-area situation analyzer for power system event detection and severity assessment

Abstract

Real-time power system monitoring and assessment leads to two major concern, prediction and evaluation of security and stability of power system. This assists in determination of in-time probable anomaly of the system. However, at the same time it requires real—time technological applications to measure network data at all strategic geographical locations. Synchrophasor technology based wide-area situational awareness ensures power system real—time monitoring and assessment. The chapter proposes real—time data driven Wide-area Situation Analyzer (WASA). WASA first detects an event in the system using synchrophasor measurements and then assesses its vulnerability posed to power network. The vulnerability is measured as severity in terms of first swing transient instability. Level of severity index is developed in terms of generator going out of step. The bus voltage trajectories going away with rest of the system due to generator(s) transient instability are considered. The proposed new approach is based on Center of Frequency (COF) formulated from limited Phasor Measurement Unit measurements. To check for an event existence in the system, a new decision based COF concept is defined. In order to determine the severity of the identified event, a new Predictor Indices (PI) is proposed using COF and PMU measurements. These predictor indices are used in assessment methodology, based on Adaptive Boosting (AdaBoost) of decision estimators. Furthermore, comparative results of proposed wide-area situational analyzer with other machine learning algorithms are also shown. The proposed WASA is instigated on IEEE New England 39 Bus system, successfully validating analyzer performance. The different type of events considered are generation outage, bus outage, load outage and line events. Additionally, if any bus outage occurs due to line faults then it is considered as single severity. The results reflects the efficacy of the proposed analyzer in event detection and its assessment efficiently and effectively with very less computational burden. The ability of the proposed analyzer to identify events quickly and correctly makes it appropriate for real—time applications.