Ensuring the microbiological safety of large, heterogeneous water distribution systems (WDS) typically requires the maintenance of appropriate disinfectant residuals such as chlorine. However, direct experimentation on real-world WDS infrastructure is infeasible, and traditional control algorithms face challenges due to its noisy, nonlinear dynamics and complex, coupled fluid-species interactions. To address this, we propose a novel, evolutionary framework which integrates surrogate-assisted prescription, curricular multi-objective optimization, and Neuroevolution of Augmenting Topologies (NEAT). Our approach evolves NEAT controllers via a surrogate-informed NSGA-II instance and periodically fine-tunes the surrogate on data collected by evaluating the best agents on the hydraulic simulator, EPANET. Explicitly, four objectives are optimized, including chlorine bound violations, spatial homogeneity of chlorine concentration in the WDS, temporal homogeneity of chlorine injections by the agent, and operational cost in terms of the total amount of chlorine injected. We show that student-teacher distillation enables modeling of the complex WDS state. Moreover, evaluated agents using the proposed framework produces a diverse range of Pareto-optimal policies, outperforming a variety of policies, including Proximal Policy Optimization (PPO) algorithms. These findings highlight the potential of evolutionary, multi-objective optimization in chlorination control systems and offer a pathway toward efficient deployment for real-world urban, water systems.

PDF

Technical Report, Department of Computer Sciences, The University of Texas at Austin, Austin, Texas, December 2025.

@techreport{monsia:hthesis25, title={Leveraging Evolutionary Surrogate-Assisted Prescription in Multi-Objective Chlorination Control Systems}, author={Rivaaj Monsia}, month={December}, address={Austin, Texas}, institution={Department of Computer Sciences, The University of Texas at Austin}, type={Undergraduate Honors Thesis}, url="http://nn.cs.utexas.edu/?monsia:hthesis25", year={2025} }