Research On Dynamic State Estimation For Power System Based On Unscented Kalman Filter

State estimation is the heart of any approach to controlling and monitoring differentvariables in power systems. The purpose of the state estimation is using themeasurements and the grid topology information to get the real time state of the powersystem. Power system state estimators have been classically performed by a staticapproach, based on the weighted least square (WLS) method, in which a single set ofmeasurements is used to estimate the state of the system. The dynamic state estimationbased on the Kalman fiter. It’s calculation time can be faster than the WLS. Therefore itis very significant to study and solve the above problems in the field of dynamic stateestimation.This paper introduced the theory of the Kalman filter and other filters based on theKalman filter theory. The unscented Kalman filter is a nonlinear filter, it use theunscented transformation to propagate mean and covariance information throughnonlinear transformations, rather than linearization of the nonlinear system, and itperforms better than the extend Kalman filter. This paper based on the unscented Kalmanfilter, combine with the results of the phasor measurement units，to estimate the state ofpower system. The filters based on Kalman’s theory need to know the exact model of thesystems and the noises, but in the fact, the predict models and the system noises can’t beexactly the same with the real system, and the estimated parameter can be erroneous.Based on the adaptive theory, improve the unscented Kalman filter in order to estimatethe time-varying system noise and the error made by the predict model.We programmed at the Matlab platform and took advantage of the IEEE14, IEEE30, IEEE57and IEEE118test systems to compare the perfomance of the extendedKalman filter, unscented Kalman filter, cubature Kalman filter and the adaptiveunscented Kalman filter.