Constrained Dynamic State Estimation Based on Extended Kernel Risk Sensitive Loss Unscented Kalman Filter

As an effective means for monitoring the dynamic change of the system, the dynamic state estimation is of great significance to the stable operation of the power system. However, time-varying non-Gaussian measurement noise and outliers often exist in the collected and transmitted data, which reduces the accuracy of traditional Kalman filter estimation methods based on the mean squared error criterion. To this end, the generalized kernel risk sensitive loss function is defined firstly, and it is introduced into the unscented Kalman filter framework for robust state estimation. Secondly, considering the different condition constraints of the synchronous generator and controller models, the constraints are introduced into the aforementioned estimation algorithm through the pseudo-measurement method to solve the problem that the estimated value is beyond the true value range and produces a large estimation deviation, so as to further improve the estimation accuracy. Finally, the simulation experiments are performed under different conditions to verify the effectiveness of the proposed algorithm in the New England 16-machine 68-bus network model. © 2023 Automation of Electric Power Systems Press. All rights reserved.