Optimal water quality control in dynamically adaptive distribution networks
File(s)
Author(s)
Jenks, Bradley
Type
Thesis
Abstract
Water utilities face significant challenges in maintaining water quality across distribution networks, with current practices being largely manual and reactive. This thesis investigates control strategies to proactively manage water quality within the operational framework of dynamically adaptive networks. Specifically, it focuses on applying optimization methods to advance the modelling and control of discolouration risk and disinfectant residuals.
To control discolouration risk, this thesis formulates an optimization model to maximize network self-cleaning flow velocities. Flow velocities are controlled by jointly optimizing the placement and settings of pressure control and automatic flushing valves. A heuristic algorithm based on convex optimization is developed to solve the resulting mixed-integer nonlinear program. Simulation results using a large-scale operational network in the UK indicate potential self-cleaning improvements of up to 20%. The optimization model is then extended to coordinate self-cleaning with existing pressure management objectives over a daily control horizon. Distributed optimization techniques are leveraged to enable near real-time valve scheduling in complex, large-scale networks.
To better manage disinfectant residuals, this thesis advances water quality modelling through a computationally efficient Bayesian parameter estimation framework. High-resolution water quality data from a real-world distribution network is leveraged to formulate and solve a Bayesian inverse problem, quantifying disinfectant decay uncertainty under varying sensor noise levels. The resulting posterior distributions of decay parameters enable probabilistic water quality predictions and support the formulation of robust optimization problems for control.
This thesis proposes a holistic water quality control strategy that combines the new self-cleaning objective and probabilistic disinfectant decay modelling with existing pressure control schemes. The resulting joint quantity-quality optimization model demonstrates the potential for proactive water quality control in dynamically adaptive network operations. The thesis concludes by outlining a solution framework for the joint optimization model, methods to incorporate uncertainty into the model, and practical considerations for implementation in real-world settings.
To control discolouration risk, this thesis formulates an optimization model to maximize network self-cleaning flow velocities. Flow velocities are controlled by jointly optimizing the placement and settings of pressure control and automatic flushing valves. A heuristic algorithm based on convex optimization is developed to solve the resulting mixed-integer nonlinear program. Simulation results using a large-scale operational network in the UK indicate potential self-cleaning improvements of up to 20%. The optimization model is then extended to coordinate self-cleaning with existing pressure management objectives over a daily control horizon. Distributed optimization techniques are leveraged to enable near real-time valve scheduling in complex, large-scale networks.
To better manage disinfectant residuals, this thesis advances water quality modelling through a computationally efficient Bayesian parameter estimation framework. High-resolution water quality data from a real-world distribution network is leveraged to formulate and solve a Bayesian inverse problem, quantifying disinfectant decay uncertainty under varying sensor noise levels. The resulting posterior distributions of decay parameters enable probabilistic water quality predictions and support the formulation of robust optimization problems for control.
This thesis proposes a holistic water quality control strategy that combines the new self-cleaning objective and probabilistic disinfectant decay modelling with existing pressure control schemes. The resulting joint quantity-quality optimization model demonstrates the potential for proactive water quality control in dynamically adaptive network operations. The thesis concludes by outlining a solution framework for the joint optimization model, methods to incorporate uncertainty into the model, and practical considerations for implementation in real-world settings.
Version
Open Access
Date Issued
2025-09-14
Date Awarded
01/02/2026
License URL
Advisor
Stoianov, Ivan
Sponsor
Imperial College London
Natural Sciences and Engineering Research Council of Canada
Bristol Water (Firm)
Analytical Technology (Firm)
Grant Number
PGSD577767-2023
Publisher Department
Department of Civil and Environmental Engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)