Neural-adaptive control of waste-to-energy boilers

This paper presents a design of an advanced adaptive controller for coker-off-gas boilers. We aim to achieve a high level of performance and stability, even in the presence of poorly-modeled nonlinear effects and unmeasured input waste-fuel gas. The proposed control uses adaptive parameters in the nonlinear control, and a neural network to compensate for the unknown fuel. Standard Lyapunov-stable techniques provide on-line updates for the adaptive parameters and neural-network weights such that all signals remain uniformly ultimately bounded. Simulation results show the proposed control can maintain stability in a utility boiler when faced with disturbances (unknown fuel input) that would make a PID control go unstable. The proposed technique is applicable for all types of waste-to-energy systems that utilize boilers and steam turbines.