Resilient Dynamic State Estimation for Power System Using Cauchy-Kernel-Based Maximum Correntropy Cubature Kalman Filter

Accurate estimation of dynamic states is the key to monitoring power system operating conditions and controlling transient stability. The inevitable non-Gaussian noise and randomly occurring denial-of-service (DoS) attacks may, however, deteriorate the performance of standard filters seriously. To deal with these issues, a novel resilient cubature Kalman filter based on the Cauchy kernel maximum correntropy (CKMC) optimal criterion approach (termed CKMC-CKF) is developed, in which the Cauchy kernel function is used to describe the distance between vectors. Specifically, the errors of state and measurement in the cost function are unified by a statistical linearization technique, and the optimal estimated state is acquired by the fixed-point iteration method. Because of the salient thick-tailed feature and the insensitivity to the kernel bandwidth (KB) of Cauchy kernel function, the proposed CKMC-CKF can effectively mitigate the adverse effect of non-Gaussian noise and DoS attacks with better numerical stability. Finally, the efficacy of the proposed method is demonstrated on the standard IEEE 39-bus system under various abnormal conditions. Compared with standard cubature Kalman filter (CKF) and maximum correntropy criterion CKF (MCC-CKF), the proposed algorithm reveals better estimation accuracy and stronger resilience.