State Estimation of a Nonlinear System by Neural Extended Kalman Filter

Online estimation of state variables that are difficult or expensive to measure has been a widely studied problem. But those measurements are needed in a variety of engineering applications such as condition monitoring, fault diagnosis and process control. An observer can be designed to produce an estimate xcapped(k) of the state x(k) by making use of relevant process inputs, outputs and process knowledge in the form of mathematical model. The design of any good state estimator necessitates the development of a nonlinear model of the plant. In this paper, an approach to design a neural network based extended Kalman filter (NNEKF) with a recurrent neural model to estimate the state of a noisy dynamic system has been attempted. The effectiveness of the proposed state estimator has been demonstrated on a three tank benchmark system