【作者】 常青； 宋冬然； 蒋珊珊； 董密； 杨建； 赵燃；

【Author】 Chang Qing;Song Dongran;Jiang Shanshan;Dong Mi;Yang Jian;Zhao Ran;School of Automation,Central South University;

【机构】 中南大学自动化学院；

【摘要】 近年来,激光雷达在成本、紧凑型和可靠性方面都有了很大的提升,为提前获得叶轮面等效风速提供了新途径。叶轮面等效风速的相关研究主要基于机理建模,依赖于机理模型参数,难以实现对叶轮面等效风速的准确预测。针对这一问题,本文提出一种基于经验模态分解和门控循环单元神经网络的混合模型,对每个被分解得到的内涵模态分量进行预测,最后进行聚合计算得到最终的叶轮面等效风速。通过数据驱动挖掘激光雷达多点测风数据与叶轮面等效风速估计值之间存在的时空依赖特性,该模型可实现叶轮面等效风速可靠预测。更多还原

【Abstract】 In recent years, the cost, compactness and reliability of lidar have been greatly improved, which provides a new way to obtain the rotor effective wind speed in advance. The research on the rotor effective wind speed is mainly based on the mechanism modeling, which depends on the mechanism model parameters, so it is difficult to accurately predict the rotor effective wind speed. To solve this problem, a hybrid model based on empirical mode decomposition and gated cyclic element neural network is proposed to predict each decomposed connotative mode component. Finally, the final equivalent wind speed on the impeller surface is obtained by aggregation calculation. Through data-driven mining of the time-space dependence between lidar multi-point wind measurement data and the estimated value of equivalent wind speed on the impeller surface, the model can reliably predict the rotor effective wind speed.更多还原