Physics-based modelling and coordinated control of pump interaction in distributed variable-speed pumping systems

Chilled water pumping contributes substantially to building energy consumption, where conventional valve-based regulation leads to significant throttling losses. Distributed variable-speed pumping (DVSP) systems can reduce these losses by replacing valves with pumps; however, parallel operation introduces severe pump interaction, which limits stable and energy-efficient operation. Because the hydraulic mechanisms underlying this interaction remain insufficiently understood, existing control strategies often rely on data-driven approaches with limited generalizability, thereby constraining the energy-saving potential of DVSP systems. Therefore, this study developed a physics-based hydraulic model of DVSP systems using graph theory. The model was calibrated and validated by experiments, achieving a flow-rate prediction error below 10%. The physics-based model enabled rapid predictions of real-time flow rates under variable-speed conditions, while accurately capturing pump interactions. The results show that speed variation of a single pump induced disproportional and opposing flow disturbances in other pumps, while the actively regulated one faced a pronounced risk of reverse flow. Pumps located farther downstream were both more prone to generating and more susceptible to its propagation. Furthermore, an interaction-aware model-predictive control framework was developed by integrating a genetic algorithm with physics-based simulations. The proposed controller enabled demand-responsive regulation while effectively suppressing pump interaction and maintaining stability. Finally, by varying terminal numbers, a strong negative linear correlation (r = −0.88) between passive response intensity and flow rate was observed. These findings provided physics-based insights into pump interaction and can be applied to improve flow stability and energy-efficient demand-responsive control in DVSP systems.