Intelligent Control of Wastewater Treatment using Model Predictive Control & Reinforcement Learning

Jørgensen, Malthe Peter Højen ; Gehlert, Lise Bech ; Koch, Christoffer Brejnholm

4. term

Computer Science, Master

2025

2025-06-13

92

This thesis explores the use of Model Predictive Control (MPC) combined with Reinforcement Learning (RL) to optimize wastewater treatment plant (WWTP) operations, aiming to reduce energy consumption/cost and improve effluent quality. The report proposes a realistic implementation pipeline using a high-fidelity digital twin in WEST (by DHI) and a low-fidelity optimizer in UPPAAL Stratego, connected via the STOMPC MPC interface. Experiments under dry, rain, and storm weather scenarios show that MPC can improve both energy usage and effluent quality simultaneously. For dry weather conditions, MPC reduced energy consumption by up to 24% depending on the scenario.

Energy Efficiency ; BSM1 ; MPC ; UPPAAL ; UPPAAL Stratego ; Wastewater Treatment Plant ; Water Quality ; Optimization ; Reinforcement Learning

Download
View record in AAU Student Projects