Experimental model validation and predictive control strategy for an industrial fire-tube boiler

This paper presents a flexible control structure for fire-tube boilers based on a suitable integration of the typical decentralized PI control structure and model predictive control technique. First, a dynamic nonlinear reference model of the fire-tube boiler is developed combining models available in the technical literature, based on first principle laws. The overall system model is considered as a gray-box model, and it has been validated with real data. Then, a suitable control-oriented model is derived out of the nonlinear reference model, in order to design a hybrid cascade MPC-PI control structure capable of guaranteeing stability, improving performances and enforcing real-time constraints. The flexibility of such a structure can be exploited to impose different types of functional behavior to the boiler itself, from the performance-related ones to the efficiency increase ones. While the reference non-linear model is large and detailed, the control-oriented one is simplified so that a few process parameters are identified to reduce dramatically the implementation in the MPC controller hardware and software framework. A sensitivity analysis with respect to these process parameters highlights the robustness and easy implementation of such a strategy. Two MPC configurations have been developed and tested in simulation over the validated boiler model. Then, a customized algorithm has been developed to understand a massive quantization phenomenon on the boiler pressure measurement. Finally, a test session conducted on a real fire-tube boiler quantifies the performance benefits of one configuration of the MPC-PI control structure with respect to the PI control.