Bayesian Inference for Thermal Model of Synchronous Generator Part—II: State Estimation

This paper is the Part II of the series on parameter and state estimation using the Bayesian inference for a thermal model of a synchronous generator. Part I is about Parameter Estimation. In this paper, state estimation of rotor copper and air-gap temperatures of the synchronous generator are performed using Bayesian inference. Estimators such as Unscented Kalman Filter (UKF), Ensemble Kalman Filter (EnKF), and Particle Filters (PFs) with different sampling algorithms, are compared based on the estimation accuracy and computational time. The inferences are drawn for the posterior distributions of the state, dispersion of particles, error convergence and particle realizations of the state estimator, choice, and the computational effort of the estimators. Results show that UKF has fair estimation accuracy with the fastest computational time as compared to other estimators.