水泵水轮机转轮叶片尾缘型线多目标优化

Multi-objective Optimization of Trailing-edge Profile of Pump-turbine Runner Blades

摘要:

针对抽水蓄能电站水泵水轮机在双工况下效率提升与运行稳定性的工程需求,提出了一种结合径向基函数(RBF)代理模型与非支配排序遗传算法(NSGA-Ⅱ)的多目标优化方法,对转轮叶片尾缘的4个关键控制点(P1、P2、P3、P4)进行参数化设计。首先采用改进的最优拉丁超立方抽样生成60组样本,并在ANSYS CFX中完成数值模拟,建立以水泵效率与水轮机效率为输出的RBF预测模型,其总体预测误差控制在2%以内。随后利用NSGA-Ⅱ算法对目标函数进行优化寻优,获得水轮机工况效率95.67%与水泵工况效率92.28%的帕累托最优解,对比原型转轮分别提升约4.01个百分点与0.921个百分点。对优化前后叶片流场的分析结果表明,高压峰值削弱、低压涡核收缩且尖端泄漏涡显著抑制;研究结果表明,所提出的RBF – NSGA – Ⅱ多目标优化方法能够在有限样本下高效地探索设计空间,实现水泵水轮机双向效率的同步提升,为高效、稳定的抽水蓄能机组优化设计提供了有效的理论支撑与工程参考。

Abstract:

The engineering demand for improving efficiency and operational stability of pump-turbine units in pumped-storage power plants under both pump and turbine modes was addressed by proposing a multi-objective optimization approach that combined a radial basis function surrogate model with the non-dominated sorting genetic algorithm Ⅱ. Four key control points (P1, P2, P3, P4) of the runner-blade trailing-edge profile were parameterized. An improved optimal Latin hypercube sampling scheme was employed, 60 design cases were generated, and numerical simulations were performed in ANSYS CFX. A predictor with pump efficiency and turbine efficiency as outputs was established; its overall prediction error was maintained below 2%. The objectives were optimized with NSGA – Ⅱ, and a Pareto-optimal solution was obtained, yielding turbine-mode and pump-mode efficiencies of 95.67% and 92.28%, which represented increases of 4.01 and 0.921 percentage points over the prototype, respectively. Flow-field comparisons revealed attenuated high-pressure peaks, contracted low-pressure cavitation cores, and a marked suppression of tip-leakage vortices after optimization. These results demonstrated that the proposed RBF – NSGA – Ⅱ framework efficiently explored the design space with limited samples, achieved simultaneous efficiency gains in both operating modes, and provided robust theoretical support and practical guidance for the high-efficiency, stable design of pumped-storage pump-turbine units.