Model Predictive Control of a Sewer System
Authors
Pilgaard, Thomas Holm ; Pedersen, Jacob Naundrup
Term
4. term
Education
Publication year
2018
Submitted on
2018-06-07
Pages
123
Abstract
Large industrial discharges in the city of Fredericia cause substantial fluctuations in the influent to the wastewater treatment plant (WWTP), challenging the biological processes that depend on a more stable load. This thesis investigates whether model predictive control (MPC) can be used to stabilize the sewer outflow to the WWTP. To this end, a simulation environment is built to mimic a real sewer system: pipe hydraulics are modeled with the nonlinear Saint-Venant equations, and storage is represented by a linear tank model. The environment also allows side inflow disturbances to increase realism. The aim is to apply MPC to this model to reduce flow variations and thereby support optimal WWTP performance. However, difficulties with linearization and the size and complexity of the system were encountered during the MPC design, and satisfactory control performance was not achieved in this study.
Store industrielle udledninger i Fredericia skaber betydelige variationer i tilførslen til renseanlægget, hvilket udfordrer de biologiske processer, der kræver et mere stabilt belastningsniveau. Denne afhandling undersøger, om modelprædiktiv regulering (MPC) kan bruges til at stabilisere kloakens udløb mod renseanlægget. Til formålet konstrueres et simuleringsmiljø, der efterligner et virkeligt kloaksystem: strømningen i kloakrør beskrives med de ikke-lineære Saint-Venant-ligninger, og magasinering modelleres med en lineær tankmodel. Miljøet kan desuden indføre sideforstyrrelser for at øge realismen. Målet er at anvende MPC på modellen for at reducere flowvariationer og derved understøtte optimal drift af renseanlægget. Under designet opstod der dog udfordringer med linearisering og systemets størrelse og kompleksitet, og derfor blev der ikke opnået tilfredsstillende resultater med MPC i denne undersøgelse.
[This apstract has been generated with the help of AI directly from the project full text]