Outlier-Detection-Based EKF for Power Systems

This paper investigates the state estimation problem for the power systems suffering from outliers, and a mixed outlier detection (MOD) method in extended Kalman filter (EKF) is proposed. Firstly, the vulnerability of traditional EKF method suffering from outliers is analyzed based on a generator model, as well as the motivation of using outlier detection method. Then, an outlier-detection-based EKF (ODB-EKF) is obtained by applying the proposed MOD to the recursive process of EKF. In this case, a time-varying detection matrix is established based on the parts of predictions and innovations where projection statistics and least squares methods are used to detect the outliers in two parts respectively. Subsequently, the data will be updated according to the results of the MOD method. Finally, the numerical simulation results based on the parameters of the IEEE 39 bus test system show the good performance of the proposed ODB-EKF against the outliers.