风电叶片双点疲劳加载系统同步控制研究
Research on Synchronous Control of Double-Point Fatigue Loading System for Wind Turbine Blades
张兴杰 1黄雪梅 1张磊安 1文永双 1李建伟 1于良峰2
作者信息
- 1. 山东理工大学机械工程学院,淄博 255049
- 2. 中车山东风电有限公司,济南 250101
- 折叠
摘要
为解决风电叶片全尺寸双点疲劳测试中两激振器振动不同步问题,采用GA-Adam-BP神经网络与传统PID混合控制策略,并引入切换边界值判断控制权的归属.基于遗传算法的全局搜索能力对BP神经网络进行权值和阈值的初始化筛选,该方法利用适应度选择、交叉和变异遗传操作,从初始种群中筛选出高质量的个体作为网络的初始权值和阈值,避免神经网络陷入局部最优解.引入Adam算法计算参数的指数加权移动平均值,实现神经网络学习率的动态更新,避免了梯度集中与消失问题,有效减少学习路线的震荡,使收敛时间缩短.仿真与试验结果表明,相比BP神经网络,混合控制下的电机转速误差在3%以下,主-从激振器相位差范围为±1.3°,实现了叶片双点疲劳测试激振器间更优的同步控制.
Abstract
In order to solve the problem of unsynchronized vibration of two shakers in the full-size double-point fatigue test of wind turbine blades,a hybrid control strategy of GA-Adam-BP neural network and tra-ditional PID is used,and the switching boundary value is introduced to judge the attribution of control pow-er.Based on the global search ability of genetic algorithm for BP neural network to initialize the screening of weights and thresholds,the Adam algorithm is used to realize the dynamic learning rate of BP neural net-work,which effectively reduces the oscillation of the learning route and makes the convergence time shor-ter.Simulink simulation model is established and test platform is built for validation.The simulation and ex-perimental results show that the motor speed error under hybrid control is below 3%,the master-slave excit-er phase difference range is±1.3° compared to BP neural network.The hybrid control exhibits higher ac-curacy and online operation capability with good robustness and anti-interference capability,and achieves better synchronization control between the blade dual-point fatigue test shakers.
关键词
风电叶片/疲劳加载/GA算法/Adam算法/神经网络/混合控制Key words
wind turbine blade/fatigue loading/GA algorithm/Adam algorithm/neural network/hybrid control引用本文复制引用
出版年
2024
NETL
NSTL
维普
万方数据