Abstract
This paper revisits the non-Markovian state vector of an Ornstein-Uhlenbeck process-driven power system dynamics in non-linear filtering framework. The Fokker-Planck setting accounts for the process noise correction term and ignores the observation noise correction term in the analysis of stochastic systems. This paper introduces the notion of the ‘Kushner-Stratonovich setting’, which accounts for the process noise as well as observation noise correction terms in the conditional moment evolution equation. The Kushner-Stratonovich setting is the cornerstone formalism of nonlinear filtering problems of stochastic control systems. We wish to estimate the rotor angle from given observations using two non-linear filters: (i) a Gaussian non-linear filter, and (ii) the extended Kalman filter. This paper develops two non-linear filters for a filtering model for the machine rotor angle, in which the Ornstein-Uhlenbeck process is the process noise and the Brownian noise process is the observation noise. The filter efficacy is examined by utilizing quite extensive numerical experimentations with two sets of data.