PMU-based voltage stability prediction using least square support vector machine with online learning

Abstract Composite load model comprising an impedance-current-power (ZIP) load and an induction motor (IM) is widely used for dynamic voltage stability analysis. With the load model, this paper proposes a new PMU-based method to predict short-term voltage instability (STVIS). Using synchronous measurements, the method firstly performs online contingency analysis to obtain a set of three-element look-up tables comprising the presumed contingency, the corresponding post-fault stable and unstable equilibrium point (SEP and UEP) for each IM load individually. Next, when a fault really occurs, the time-series of IM slip is computed by an Euler algorithm based on local PMU measurements, and then a new time-series prediction method is proposed for rolling prediction of IM slip trajectory by introducing least square support vector machine (LSSVM) with online learning. Finally, from the view of IM stability mechanism, the STVIS status can be detected in advance by monitoring that the predicted slip trajectory reaches the IM’s UEP in the look-up table. The effectiveness of the proposed method is verified on the New England 39-bus system.