Sliding-window neural state estimation in a power plant heater line

Power plant monitoring is addressed by means of a sliding-window neural state estimator. The complexity and the nonlinearity of the considered power plant application prevents us from using standard techniques such as Kalman filtering. The statistics of noises are assumed unknown and the estimator is designed by minimising a given best squares cost function (in general, non-quadratic) under very general assumptions on the state equation and the system measurement channel. The estimator has been designed off-line in such a way as to be able to process any possible measurement online. Extensive simulation of a state estimation problem in a model of a section of a real power plant are reported showing the effectiveness of the applied method as compared to the extended Kalman filter.