Dynamic State Estimation of Generator Based on Amended Singular Value Decomposition Cubature Kalman Filter

The noise inaccuracy issue in nonlinear dynamic systems leads to a reduction in the accuracy of the filtering algorithm, and the cubature Kalman filter (CKF) can effectively deal with this problem. Singular value decomposition (SVD) is used to replace Cholesky decomposition can avoid the case of negative definite prior covariance matrix and improve the stability of the filtering. The measurement innovation and noise covariance matrix are used to set the amending rules, and the rules directly amend the state prediction value or Kalman gain and adjust the proportion of the measured value in the filter. It can reduce the influence of the abnormal measured value on the state estimation value. Simulation results show that the amended singular value decomposition cubature Kalman filter algorithm proposed in this paper has higher accuracy and robustness.