Square Root Unscented Kalman Filter With Modified Measurement for Dynamic State Estimation of Power Systems

Abstract

Dynamic state estimation of a power system provides essential information about its inherent dynamic change. Nonlinear Kalman filters (NKFs) have been identified as potential versatile tools in performing state estimations. One key challenge of using NKFs lies in the fact that the available observations are inaccurate due to the presence of non-Gaussian noise that degrades the filtering precision and robustness. Some robust NKFs are developed to correct the state estimation via using robust optimality criteria or approximating the posterior density. Considering that the essence in these robust NKFs is to weight the error covariance and noise variance, we propose a novel method that can directly act on measurement to weight error covariance and noise variance. The novel method is based on a square root unscented Kalman filter that can enhance the numerical stability. Then, a weighting factor is applied to the measurement model for alleviating the effect of abnormal measurement errors. The weighting factor is derived from consideration of an anisotropy covariance function and an upper envelope of the traditional “sinc” function for suppressing large measurement errors and retaining the information of very large errors. We call the new filter the square root unscented Kalman filter with modified measurement (SRUKF-MM). The proposed SRUKF-MM is applied to state estimation of the Western System Coordinating Council (WSCC) 3-machine system and the Northeastern Power Coordinating Council (NPCC) 48-machine system in a non-Gaussian noise environment. Simulation results show that the proposed method achieves significantly improved filtering performance compared to other related NKFs.