Neural-Adaptive Control and Nonlinear Observer for Waste-To-Energy Boilers

This paper looks at the problem of controlling an incinerator that burns waste gas to generate power. The system is modelled as a standard utility boiler using one known and one unknown (waste) fuel input. Standard linear controls have trouble dealing with large variations in the waste input, and in practice boiler shutdowns can occur. In this work, a nonlinear adaptive control design accounts for uncertainty in the plant parameters, and an adaptive neural‐network estimates the effect of the waste input. Since a linear observer design cannot guarantee convergence away from a set point, a novel nonlinear observer design provides estimates of the states. The observer design uses fictitious states to estimate nonlinear terms in the observer dynamics. The analysis guarantees Lyapunov stability, thus the observer bounds depend on the accuracy of the observer initial conditions. Simulation results show the proposed method can obtain accurate performance and stability, improving over results obtained withproportional–integral control.