基于改进的信赖域模型管理技术识别风电转子系统不对中载荷

Identification of misalignment load of wind turbine rotor system based on improved trust region model management technology

针对求解耗时的风电转子系统不对中载荷识别问题,提出基于改进的信赖域模型管理技术的识别算法。该算法将整个先验分布空间的不对中载荷识别问题转化为一系列信赖域上的近似优化问题,通过区域遗传智能采样技术采集样本,加强径向基函数构建代理模型,再采用遗传算法进行近似优化。通过每个信赖域上的最小目标函数和近似优化结果确定信赖度和下代域的中心、半径,进而不断地缩放、平移信赖域,来保证获得与真实模型一致的不对中载荷。通过四种方法对比表明该方法样本遗传策略,遗传落在下代信赖域空间上的样本,减少实验设计样本个数而提高效率;最小目标函数作为信赖中心调整提高了关键区域代理模型的精度而加快收敛,降低了对代理模型精度的依赖。

Here,to solve time-consuming identification problems of misalignment load of wind turbine rotor system,an identification algorithm based on improved trust region model management technology was proposed.The algorithm could convert the misalignment load identification problem in entire prior distribution space into a series of approximate optimization problems in trust region.Samples were collected using the regional genetic intelligent sampling technology,and radial basis functions were enhanced to build an agent model.Then,genetic algorithm was used for approximate optimization.The minimum objective function and approximate optimization results on each trust region were used to determine the trust level,and the next generation region’s center and radius,then the trust region was continuously scaled and translated to ensure obtaining misalignment load consistent with that of the real model.Through comparing 4 methods,it was shown that the sample genetic strategy of the proposed method inherits samples falling on the next generation trust region space,and reduces number of experimental design samples for improving efficiency;the minimum objective function as the trust center can adjust and improve the accuracy of agent model in key areas for accelerating convergence and reducing dependence on accuracy of agent model.