基于小样本神经网络与多约束的风力机翼型快速优化设计

Rapid Optimization Design of Wind Turbine Airfoil based on Small Sample Neural Network and Multiple Constraints

针对神经网络模型可以基于现有数据快速准确地预测风力机翼型的气动性能,但大量学习样本的构建需要较高的时间成本的问题,建立基于小样本集的风力机翼型神经网络模型,提出了多约束条件下的翼型气动性能优化设计方法,解决了训练数据过少所造成的学习不充分问题。基于建立的优化设计模型,应用粒子群算法完成了NACA4415翼型的优化设计,将新翼型与原始翼型进行气动特性对比分析。结果表明:新翼型在主要工作攻角范围内最大升力系数提高了6.96%,最大升阻比提高了7.37%,气动性能明显改善;该方法的优化效率远远高于传统方法,从而验证了该方法的可行性。

Neural network model can quickly and accurately predict the aerodynamic performance of wind turbine airfoil based on existing data,but the construction of a large number of learning samples requires a high time cost.To solve this problem,a neural network model of wind turbine airfoil based on small sample set was established,and an optimization design method for aerodynamic performance of airfoil under multiple constraints was proposed,which solved the problem of insufficient learning caused by too little training data.Based on the established optimization design model,the optimization design of NACA4415 airfoil was completed by using particle swarm optimization algorithm.The aerodynamic characteristics of the new airfoil and the original airfoil were compared and analyzed.The results show that the maximum lift coefficient and the maximum lift-drag ratio of the new airfoil in the main angle of attack are increased by 6.96%and 7.37%,and the aerodynamic performance of the new airfoil is significantly improved.Moreover,the optimization efficiency of this method is much higher than that of the traditional method,which verifies the feasibility of this method.