基于神经网络的大型风力机动力学仿真预测分析

DYNAMIC SIMULATION PREDICTION ANALYSIS OF LARGE WIND TURBINE BASED ON NEURAL NETWORK

单一仿真分析方法不能完整反映风力机各部分的耦合关系,影响结果准确性。针对这一问题,采用联合仿真技术,利用空气动力学理论计算气动力,运用多柔体动力学分析理论和软件实现整机的气弹相互耦合。该多柔体建模方法能较好地模拟风力机耦合振动特性,仿真结果更接近实际工作情况。由于多柔体模型考虑的自由度多,精度要求高,导致仿真时间增长,计算成本加大。因此引入人工神经网络方法对风电机组动力学性能进行分析预测。结果表明,采用联合仿真与神经网络相结合的方法,在保证预测精度的同时还能减少动力学仿真时间,弥补单独使用仿真分析方法的局限性。

The single simulation analysis method could not completely reflect the coupling relationship between the various parts of wind turbine, and the accuracy of the results was affected. Using co-simulation technology to solve this problem, the aerodynamic force was calculated theoretically using the aerodynamics theory; the flexible multi-body dynamics analysis theory and software were used to achieve aero-elastic mutual coupling control of whole wind turbine. The flexible multi- body modeling method could better simulate coupling vibration characteristics of wind turbine, the simulation results was more close to the actual work situation. Because more freedom and high precision of flexible multi- body model needed increasing simulation time and calculation cost, therefore the artificial neural network method was introduced to analyze and forecast the structural dynamics performance of wind turbine. The results show that the method combining neural network and co-simulation can ensure prediction accuracy and reduce dynamics simulation time, make up the limitations of using simulation analysis alone.