摘要
安装在寒冷地区的风力机叶片不可避免地发生覆冰现象,叶片覆冰位置和覆冰质量的不同导致其固有频率发生不同的改变。基于覆冰叶片的固有频率利用人工神经网络技术,实现预测叶片的位置和质量信息。将2kW风力机叶片划分为叶尖、叶中和叶根,分别进行自然环境覆冰试验和覆冰叶片模态试验,得到不同位置覆冰质量和前4阶固有频率的对应关系,用其训练RBF神经网络模型,实现以模态频率为输入,覆冰位置和质量为输出的叶片覆冰状态预测。研究表明.RBF神经网络预测的整体覆冰质量相对误差率均在6.5%以内,叶尖、叶中和叶根处覆冰质量预测相对误差率分别为8.31%、4.41%、6.96%,相对误差率随冰层质量的增加而降低。RBF神经网络可以实现基于叶片固有频率的变化对覆冰位置和质量信息预测,为除冰系统精确除冰提供了理论依据。
The wind turbine blades installed in cold regions are inevitably icing.The natural frequency of blade will change due to different icing position and icing quality.The artificial neural network technology based on the natural frequency of icing blades was used to achieve the purpose of predicting blade position and quality information.The blade of 2kW wind turbine was divided into tip,blade and root.The natural environment icing test and icing blade modal test were carried out respectively.The corresponding relationship between icing mass at different positions and the first four natural frequencies was obtained.The RBF neural network model was used to train the icing state prediction with modal frequency as input and icing quality as output.The results show that the relative error rates of the overall icing quality predicted by RBF neural network are all within 6.5%,the relative error rates of icing quality prediction of tip,middle and root are 8.31%,4.41%and 6.96%,and the relative error rate decreases with the increase of ice mass.The RBF neural network can predict the location and quality of icing based on the change of blade natural frequency,which provides theoretical basis for accurate deicing of deicing system.
作者
李飞宇
崔红梅
苏宏杰
马志鹏
LI Fei-yu;CUI Hong-mei;SU Hong-jie;MA Zhi-peng(College of Electromechanical Engineering,Inner Mongolia Agricultural University,Hohhot 010018,China)
出处
《水电能源科学》
北大核心
2021年第6期171-174,共4页
Water Resources and Power
基金
国家自然科学基金项目(11262015)
内蒙古农业大学优秀青年科学基金项目(2014XYQ-9)。
关键词
风力机叶片
覆冰检测
人工神经网络
模态参数
固有频率
wind turbine blade
icing detection artificial neural network
modal parameters
natural frequency
