Sensor fault estimation in the framework of model predictive control. Boiler case study

This paper deals with a nonlinear model predictive control designed for a boiler unit. The predictive controller is realized by means of a recurrent neural network which acts as a one-step ahead predictor. As the automatic control system can hide faults from being observed, the control system is equipped with a fault diagnosis block. It is realized by means of the binary diagnostic matrix, which is determined on the ground of residuals calculated using a set of partial models. Each partial model was designed in the form of the recurrent neural network. The paper proposes also a methodology of compensating sensor faults. When a sensor fault is isolated, the system estimates its size and based on this information the controller is fed with an estimated, close to real boiler level value. The proposed methodology is tested on the example of the boiler unit simulator. Copyright © 2012 IFAC