Assessing the Contribution of SCADA-Derived Pump Operating Regimes to Leakage Assessment in Water Distribution Networks

SCADA residuals are widely used to assess abnormal hydraulic behavior in water distribution networks. However, their value for leakage detection depends on how empirical reference behavior accounts for changing operating conditions. This study evaluates whether pump-flow SCADA telemetry can be converted into temporally coherent and hydraulically interpretable operating regimes, and whether these regimes add value to leakage-oriented scoring of monitored pipe-flow and pressure signals. Pump-flow telemetry is transformed into timestamp-level operating states using descriptors of current flow level, recent changes, short-term variability, deviation from routine daily behavior, and activity status. These states define pump-state and joint pump-state plus time-of-day baselines for monitored pipe-flow and pressure signals. Global and time-of-day References are retained as comparators, and leakage labels are introduced only after score construction, allowing empirical baseline fit and leakage-oriented score informativeness to be evaluated separately. The framework is tested on the BattLeDIM/L-Town benchmark, using 2018 as the main evaluation period and 2019 for temporal replication. Results show that pump-flow telemetry can be represented by compact, stable, and transition regimes. The strongest contribution appears in the pressure branch: combined pump-regime and time-of-day conditioning produced pressure-deficit scores that more consistently separated lower- and higher-leakage periods across both years. In contrast, contextual baselines reduced the magnitudes of pipe-flow deviations, but leakage-oriented enrichment remained location-dependent and did not replicate consistently. Overall, SCADA-derived pump operating regimes are most useful as contextual reference information for pressure-deficit assessment, while their value for local pipe-flow scoring depends on sensor location and leakage structure.