A KPCA and SVR Based Dynamic State Estimation Method for Power System

Dynamic State Estimation (DSE) for power system considers statistical characters of systemic state variables in past period, has functions of state estimation and forecasting. This paper proposes a new method for state estimation problem in power systems based on Kernel Principle Component Analysis (KPCA) and Support Vector Regression (SVR). Firstly, the KPCA can extract the nonlinear relationship between original inputs from SCADA system to make data compression and feature extraction. KPCA is closely related to methods applied in Support Vector Regression (SVR). Then, the extracted principal data are used as inputs of SVM in order to forecast systemic state variables. Applying proposed system to IEEE14 data, the experiment results show that KPCA-SVR features high learning speed, good approximation and generalization ability compared with SVR.