PSO-MSVR Based Prediction Approach for Unstable Equilibrium Points of Disturbed Power System

With expansion of power system and continuous proliferation of renewable energy generation, the transient stability characteristics of power system become more complicated, and online transient stability assessment faces severe challenges. The calculation of unstable equilibrium point (UEP) is a critical step in the direct method of power system transient stability assessment. This paper presents a new method for predicting transient UEPs of power systems based on PSOMSVR. Instead of complicated calculation, UEPs can be obtained by steady state measurement date. Firstly, UEPs are calculated by boundary of stability region based controlling UEP (BCU) method under different operating conditions of the system to provide sample data for the prediction model. Then the voltage amplitude and phase angle of the steady state operating conditions are taken as the sample characteristics, and the mapping relationship between the measured data and UEPs is constructed by multi support vector regression (MSVR). In the meantime, the penalty and kernel parameters in MSVR are optimized by particle swarm optimization (PSO) algorithm. Finally, fast prediction of the transient UEPs based on the steady state operating information of the system is achieved. Case study on IEEE 9-bus system shows that the proposed approach has high prediction accuracy through a small amount of training data.