基于改进花授粉算法和代理模型的Savonius型风机叶片形状高效智能优化方法研究

Savonius wind turbine has numerous advantages and broad application prospects, however, the inherent defect of low energy capture efficiency has severely limited its development. Optimizing the shape of its blades can significantly improve its energy capture efficiency, while the current design methods still have some deficiencies, such as relying too much on design experience, low efficiency, and poor results. This research intends to study the optimal design method for the blade shape of Savonius wind turbine aims to increase its power coefficient by comprehensively using the knowledges of computer graphics, CFD, surrogate models, and computational intelligence. Firstly, the blade shape is parameterized using Bessel curves. Then, some sample points are determined in the decision space using experimental design methods, and their responses are accurately calculated using CFD numerical simulation. Next, surrogate models with less computational complexity, higher solving efficiency, and enough fitting accuracy are built using these sample points and their responses and based on radial basis function model and Kriging model. After, improved flower pollination intelligent optimization algorithms are adopted to solve these surrogate models to obtain the best shape parameters. Finally, the above design procedures are integrated together based on Isight platform to form an efficient and reliable intelligent optimization method for the blade shape of Savonius wind turbine. This study can break through the limitations of existing design methods and improve the energy capture efficiency and the design level of Savonius wind turbine, and thus has important theoretical value and practical significance.

Savonius型风机具有诸多优点、应用前景广阔，但捕能效率低的固有缺陷严重限制了其发展。优化其叶片形状可显著提高其捕能效率，但现有设计方法普遍过于依赖经验、效率低、结果差。本研究拟综合运用计算机图形学、计算流体动力学、代理模型技术和计算智能等多学科知识，以提高风机功率系数为目标，研究该类型风机叶片形状的优化设计方法。首先基于贝塞尔曲线参数化表达叶片形状，其次基于试验设计方法确定决策空间内的样本点并利用计算流体动力学数值模拟技术获得各样本点的响应，然后基于径向基函数模型、Kriging模型建立计算量小、求解效率高的代理模型，并利用改进花授粉智能优化算法寻优代理模型获取优化结果，最后基于Isight平台集成整个设计过程，从而建立一套面向该类型风机叶片形状设计的高效可靠的智能优化方法。本研究可突破现有设计方法的局限，提高Savonius型风机的捕能效率和设计水平，具有重要的理论价值和现实意义。

本项目针对Savonius型风机捕能效率较低的固有缺陷，以叶片这一关键零部件的形状优化设计方法为研究对象展开研究。首先运用三阶贝塞尔曲线参数化表征叶片形状，然后采用拉丁超立方采样法获取设计样本并基于CFD评估各样本的力矩系数值，进而利用径向基函数模型、Kriging模型、支持向量回归模型拟合叶片形状设计参数与叶片力矩系数值之间的非线性关系，最后以代理模型作为评价工具、以花授粉算法等智能优化算法作为寻优工具对设计方案进行寻优并利用CFD对优化结果进行验证。通过上述过程，构建了面向Savonius型风机叶片形状优化设计问题的高效、可靠、智能的优化方法。仿真结果表明，在风速为7m/s，叶尖速比为1时，研究获得的新型叶片相比于传统半圆形叶片其平均力矩系数可提高7%以上，同时制造材料可节约10%以上。本研究突破了现有设计方法过于依赖设计人员的既有经验、设计效率较低的局限，提高了Savonius型风机的捕能效率，具有一定的理论价值和现实意义。基于项目研究情况，目前已发表SCI论文3篇，录用SCI论文1篇，授权发明专利1项，培养研究生3名。