Main Article Content
State estimation is a major problem in industrial systems. The accurate estimation of states leads to effective monitoring of system, fault diagnosis and good control performance. The particle filter is potentially suited for better estimation of highly nonlinear and non-Gaussian system. The selection of a suitable importance proposal density is a crucial step in the design of particle filter. The unscented Kalman filter (UKF) provides better state estimates for a nonlinear system than the well known extended Kalman filter (EKF). The particle filter using an UKF to generate proposal density is referred as unscented particle filter (UPF). The potential advantage of UPF is that the UKF allows the particle filter to incorporate the latest measurements in to a prior updating routine. This paper proposes an application of UPF in the field of electrical engineering, with special emphasis on highly nonlinear Van der Pol oscillator (VPO). Simulation tests were carried out on VPO system to assess the state estimation performance of the sampling importance resampling particle filter (SIR-PF) and UPF under various conditions such as initial state estimate mismatch, large measurement noise and model error. The results indicate that the UPF is highly robust and provides accurate estimation of states than the SIR-PF.