Kalman Filters With Generalized Correntropy Loss For Robust Power System Forecasting-Aided State Estimation

State estimation is a crucial core part of the energy,management system(EMS).Especially in today's increasingly large power grid.The main function of the state estimation filter is to provide accurate information for the EMS(such as the voltage amplitude and angle of all buses in the power systems).However,due to various abnormal conditions such as measurement noise,bad data,and sudden load changes in the power system,the results of state estimation will be disturbed,which may mislead dispatchers and cause unpredictable harm to the stability of the grid operation.Therefore,improving the robustness of the state estimation algorithm and the ability to suppress bad data is of great significance for ensuring the safe and stable operation of the power system.Based on the commonly used extended Kalman filter algorithm,this paper improves two forecasting-aided state estimation algorithms,and explores the impact of different non-Gaussian measurement noise on estimation accuracy.The main contents are as follows:First,based on the data acquisition and monitoring control(SCADA)measurement system,a mathematical model for power system auxiliary prediction state estimation is established.The state process and measurement equation are used to describe the quasi-steady process of the system,and a method to assist prediction state estimation is introduced.Basic algorithm.Second,on the basis of extended Kalman filter,an adaptive extended Kalman filter algorithm with maximum cross-correlation entropy is proposed and applied to auxiliary prediction state estimation.First,we use the maximum cross-entropy criterion(MCC).Instead of the minimum mean square error criterion in the traditional extended Kalman filter,the robustness of the algorithm is increased through the form of exponential weighting,so that the proposed algorithm maintains a good estimation accuracy in the face of non-Gaussian noise;secondly,considering that the algorithm is running At the time,the initial value of the noise covariance matrix was artificially set to cause the estimation accuracy to decrease.We furtheradded an adaptive mechanism about the noise covariance matrix,which uses the error of the observation data and measurement data while filtering,and continuously estimates online and Correct the statistical characteristics of the filter noise to improve the filtering accuracy and get the optimal value of the estimated state.Third,on the basis of the above algorithm,considering the influence of higher-order terms and free parameters of the power system on the estimation accuracy,this paper combines the generalized maximum cross-correlation entropy criterion(GCL)with unscented Kalman filtering,and proposes the Correlation Entropy Unscented Kalman Filtering Algorithm(GCL-UKF).First,a dynamic model of the power system is established to realize the auxiliary prediction state estimation based on the traditional unscented Kalman filter,and then the generalized large mutual entropy criterion(GCL)is used to replace the minimum mean square error criterion in the traditional unscented Kalman filter.In addition,considering the bad data will affect the information matrix of the system.We introduce an.exponential function to adjust the information matrix to propose an enhanced GCL-UKF(EnGCL-UKF)algorithm,which makes the proposed algorithm face different types of non-Gaussian noise and bad the data still maintains good estimation accuracy.Finally,in the IEEE 14-bus,IEEE 30-bus,and IEEE 57-bus test systems,a comparative analysis of the filtering performance of various methods for assisting predictive state estimation was performed,and the applicability of the algorithm was simulated in different scenarios.The analysis verified the feasibility of the algorithm and constructed a robust state estimator to provide a high-quality database for the energy management system.