A knowledge-driven iterative heuristic with Braess-paradox reversal for multi-objective water distribution network planning

The Battle of the Water Futures (BWF) Challenge requires simulating a nation-wide, pressure-driven water supply system under different type of stresses (demographic, economic, climate-related and aging infrastructure) over a period of 75 years and optimize master plan choices relative to four conflicting objectives: service reliability, greenhouse gas emissions, financial viability defined as the net present value of the remaining bond debt by the end of the planning horizon, and affordability. We introduce a knowledge driven, iterative heuristic that splits the challenge into a constructive Stage-1 (from 2025 – 2049) with six predefined parameterized scenarios and a diagnostic feedback loop for Stage-2 (from 2050 – 2074). Each simulation result from a previous stage is used to select the new candidate designs for the current one. The innovative element in our approach is that we explicitly treat Braess's paradox as a design asset rather than a constraint: In an isolated network, strategically downgrading two pipelines (CG1213 and CG1221) resulted in a +0.158 increase in service reliability for the water utility WU10 (the worst case) and was superior to all other possible upgrade options. Using the corrected financial viability metric, this anti-Braess configuration weakly Pareto-dominates all other candidate solutions with respect to the four objectives. We also document a solar-farm reporting bug in the BWF v0.5.3 evaluator that propagated into the Stage-2 input data and quantify its ex-post correction. The final Stage 2 masterplan achieves a WU10 mean service reliability of 0.858, a national mean of 0.964, and €446M of remaining bond-debt NPV in 2074. The last stage of the challenge (up to 2099) is not detailed in the present paper.