Learning-based Predictive Control for Acid Flue Gas Abatement in Waste to Energy Plant

This work describes the application of a Model Predictive Control (MPC) strategy to the acid abatement process in a Waste to Energy (WtE) plant. By exploiting real closed-loop operational data, collected from an Italian WtE plant, black-box methods are applied to identify suitable prediction and simulation models for the development and validation of the learning-based MPC strategy. Simulation results under real operation conditions show an increment in cost-effectiveness of reagent usage in the abatement process, with potential savings of up to 24.5 tons/year. Furthermore, it also allows increasing the effectiveness in reference tracking and the compliance with stricter emission limits.