| 引用本文: | 王荣涛,赖喜德,陈小明.基于人工神经网络模型的混流式水轮机转轮多目标优化[J].灌溉排水学报,2023,42(9):46-52. |
| WANG Rongtao,LAI Xide,CHEN Xiaoming.基于人工神经网络模型的混流式水轮机转轮多目标优化[J].灌溉排水学报,2023,42(9):46-52. |
|
| 摘要: |
| 【目的】探究一种同时提升混流式水轮机运行效率、空化性能及运行稳定性的优化方法，为混流式水轮机转轮的多目标优化提供技术途径。【方法】以转轮叶片进出口安放角、安装角为优化变量，通过对叶片几何参数随机离散抽样获取样本数据库，基于CFD数值计算获取各样本的性能参数，进而建立同时考虑混流式水轮机转轮效率、出口旋流数以及空化系数的多目标函数；基于人工神经网络建立优化变量与多目标函数的映射关系，最后采用遗传算法对转轮叶片的18个几何参数进行全局寻优，并对优化前后的转轮叶片性能进行对比分析。【结果】在导叶开度为112°且运行水头分别为160、175、180 m的3个工况下，优化后的转轮效率相较优化前分别提高了0.22%、0.56%、0.60%；叶片压力分布情况得到有效改善；转轮无叶区与尾水管锥管段处压力脉动幅值显著降低。【结论】叶片进口安放角的优化程度越大，混流式水轮机综合性能的提升幅度越大。 |
| 关键词: 混流式水轮机；转轮；多目标优化；人工神经网络；遗传算法 |
| DOI：10.13522/j.cnki.ggps.2022588 |
| 分类号: |
| 基金项目: |
|
| Using Artificial Neural Network to Solve Multi-objective Optimization of Francis Turbine Runner |
|
WANG Rongtao, LAI Xide, CHEN Xiaoming
|
|
Xihua University, School of Energy and Power Engineering, Chengdu 610039, China)
|
| Abstract: |
| 【Background】Technical renovation of turbine equipment in hydropower plant has attracted increased attention due to its low investment, quick return and high economic benefit. The purpose of modifying turbine equipment in hydropower station is to increase its capacity or improve its operation performance. Generally, it is impossible to make large modification to embed components in turbine modification. Most modifications aimed at the runners or the guide vanes. It is important to study effective multi-objective runner optimization design method to ensure efficiency, cavitation performance and hydraulic stability of the unit, as well as the comprehensive performance of the unit under different operating conditions.【Objective】This paper is to explore an optimization method which can simultaneously improve the operation efficiency, cavitation performance and operation stability of the Francis turbine, and provide a technical approach for multi-objective optimization of the Francis turbine runner.【Method】Taking the position angle and installation angle of runner blade inlet and outlet as optimization variables, the sample database was obtained by random discrete sampling of blade geometry parameters. The performance parameters of each sample were obtained based on CFD numerical calculation. A multi-objective function was then established, considering runner efficiency, swirl numbers at the outlet and cavitation coefficient of the Francis turbine. The mapping relationship between optimization variables and the multi-objective function is established based on artificial neural network. Runner blades with18 geometric parameters are optimized by genetic algorithm, and the performance of the runner blades before and after optimization is compared and analyzed. 【Result】For guide vane opening 112 degrees with running head being 160 m, 175 m or 180 m, the optimized runner efficiency was increased by 0.22%, 0.56% and 0.60% respectively compared with that without optimization. The optimization also improved the pressure distribution in the blades, and reduced pressure fluctuation amplitude at the vaneless area of the runner and the conical section of the draft tube.【Conclusion】The greater the optimized blade inlet placement angle was, the greater the improvement of comprehensive performance of Francis turbine would be. |
| Key words: francis turbine; runner; multi-objective optimization design; artificial neural network; genetic algorithm |
