Energy performance improvement of mixed flow pump based on coupling of geometric and hydrodynamic parameters

Mixed flow pumps, as the core power components in energy system, play a crucial role in determining overall system efficiency. A novel redesign approach of mixed-flow pump blade is proposed by coupling an artificial neural network with a multi-island genetic algorithm, in which geometric and hydrodynamic parameters are used as decision variables. During optimization, the weighted efficiency at 0.8Qdes, 1.0Qdes, and 1.2Qdes was adopted as the optimization objective to expand the high efficiency zone, while the head at 1.0Qdes was imposed as the constraint to maintain a constant specific speed. The efficiency of the optimized model is improved by 1.14 %, 2.71 %, and 6.48 % at low, design and high flow rates, respectively, compared with the original model, and the change in head at design flow rate is less than 3 %, which satisfies the constraints. Furthermore, the underlying mechanism of performance improvement is investigated through the lens of entropy production theory, focusing on energy dissipation characteristics. The results show that the enhanced impeller performance results from reduced turbulence dissipation and lower velocity gradients near the wall, whereas the diffuser performance improvement is attributed to mitigated turbulence dissipation due to an improved jet-wake structure at the impeller outlet.