An intelligent plant-wide decision-support framework for waste valorization: Optimizing hydrochar production and energy recovery

Abstract

This study presents an intelligent plant-wide decision-support framework, MIRA (Multi-objective Integrated Resource Allocation), which integrates deep learning and thermodynamic process modeling with particle swarm optimization (PSO) to optimize hydrochar production and energy recovery from diverse waste streams. Its hybrid architecture leverages artificial neural networks (ANNs), trained on experimental data but unable to enforce mass-energy conservation, coupling with thermodynamic simulation to ensure mass and energy conservation and thermodynamic consistency. The framework models two major waste valorization pathways: (1) direct combustion with energy recovery, as demonstrated by Thailand's Phuket waste-to-energy plant, and (2) hydrothermal carbonization (HTC) followed by electricity generation. MIRA simultaneously optimizes environmental and economic outcomes by adjusting HTC temperature and hydrochar routing fraction. Scenario-based optimization was applied to three representative feedstocks, organic household waste digestate (OHWD), municipal solid waste (MSW), and agricultural residue (AGR), under CO₂-focused, revenue-focused, and balanced objectives. AGR demonstrated the highest responsiveness, achieving up to 3.14 MWh of electricity and $274.2 in revenue per ton of wet feed when prioritizing energy recovery. OHWD showed moderate potential, while MSW performance was limited by high ash and moisture. Overall, MIRA offers a scalable, accurate tool for waste-to-energy optimization, with future extensions to broader thermochemical and infrastructure systems.