Real-time control of combined sewer systems: Risks associated with uncertainties

• Published in: Journal of hydrology (Amsterdam) Vol. 617; p. 128900
• Main Authors: van der Werf, J.A., Kapelan, Z., Langeveld, J.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.02.2023
• Subjects: case studies; Combined sewer overflows; hydrology; Model predictive control; Netherlands; Operational optimisation; pollution; Risk analysis; sewer systems; Uncertainty; Netherlands; Operational optimisation; Model predictive control; Uncertainty; Risk analysis; Combined sewer overflows
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2022.128900

•Model Predictive Control can reduce pollution loads when subject to uncertainties.•Relative performance loss is associated with various sources of uncertainty.•Model uncertainties can reduce the MPC performance significantly.•Precipitation Forecast uncertainty causes only minor performance deterioration.•Unanticipated dynamic system capacity changes cause significant operational risks. Model Predictive Control (MPC) of combined sewer systems can reduce environmental degradation caused by uncontrolled overflows. However, practical uncertainties are often neglected when assessing the potential of MPC strategies. This paper aims to understand the risks associated with using a non-perfect internal MPC-model, real precipitation forecast, and realistic dynamic system capacity fluctuations. An MPC with the objective to reduce the total combined sewer overflow (CSO) volume was implemented in the case study of Eindhoven in the Netherlands where highly sensitive waterways receive the sewer overflows. Two types of risks were identified: relative system performance loss and operative deterioration. The former entails a practical decrease in efficacy of controlling CSO spills compared to the theoretical situation, whereas the latter describes the aggravation of environmental pollution compared to a static form of system operation. The results obtained demonstrate that precipitation forecast uncertainty is associated with a small relative system performance loss. Opposite to this, significant performance loss was observed as a consequence of uncertainties in the internal MPC model and the actual sewer system capacity available. The latter caused additional combined sewer overflows compared to a statically optimised control for smaller precipitation events.