Robust State Estimators Based on Maximum Correntropy and Weighted Least Median of Squares Criteria: A Comparative Study

This paper presents a comparative study of statistically robust state estimators for power system state estimation. Estimators based on the Maximum Correntropy Criterion (MCC) are evaluated using different Parzen window adjustment strategies, including the proposed Adaptive Kernel Spreading (AKS) method and two kernel reduction approaches, namely Safety-Margin Kernel Reduction (SMKR) and Hessian-Bounded Kernel Reduction (HBKR). These methods are compared with the Weighted Least Median of Squares (WLMS) estimator and the Weighted Least Squares estimator with residual analysis (WLSr). Computational simulations are conducted on the IEEE 14-bus and IEEE 30-bus test systems under Gaussian and non-Gaussian noise conditions, as well as in the presence of gross errors (GEs). The results reveal significant performance differences among the estimators, particularly under non-Gaussian noise and reduced measurement redundancy. MCC-based estimators demonstrate high robustness to GEs without requiring measurement removal, while the proposed AKS method achieves performance comparable to WLSr under Gaussian noise and superior performance in non-Gaussian scenarios with limited redundancy. These findings provide practical insights into the selection and application of robust state estimation techniques in power systems.