数字孪生技术在风力机叶片结冰预测中的应用  

Application of Digital Twin Technology in Wind Turbine Blade Icing Prediction

作　　者：孙永朋 关新 刘传宝 王哲 SUN Yongpeng;GUAN Xin;LIU Chuanbao;WANG Zhe(School of Energy and Power,Shenyang Engineering University;School of New Energy,Shenyang Engineering University)

机构地区：[1]沈阳工程学院能源与动力学院 [2]沈阳工程学院新能源学院

出　　处：《上海节能》2025年第3期447-453,共7页Shanghai Energy Saving

摘　　要：根据风力机叶片结冰时风电机组的各项运行参数的变化特征,发现当风力机叶片结冰时,机组输出功率和风轮转速会显著下降,因此,在外界环境不变、无变桨和偏航异常时,如风电机组的输出功率和风轮转速同时下降,即可判定风力机叶片结冰。采用功率和风轮转速多参数模型来提高风力机结冰预测的准确度,由于多参数模型随机变量较多,为此引入高斯过程回归以提高模型的预测能力。利用BP神经网络对监测数据进行多维度筛选,筛选后数据输入到多参数模型得到预测初始值,再与相似条件下的历史数据进行比对,修正结果,最后利用数字孪生体3D可视化模型显示最终结果,实现风力机叶片结冰现象精准的预测。According to the changing characteristics of various operating parameters of the wind turbine when the wind turbine blades freeze,it is found that when the wind turbine blades freeze,the output power of the unit and the speed of the wind turbine will decrease significantly.Therefore,if the external environment remains unchanged and there are no abnormalities in pitch and yaw,if the output power of the wind turbine and the speed of the wind turbine decrease at the same time,it can be determined that the wind turbine blades are frozen.Multi-parameter models of power and rotor speed are used to improve the accuracy of wind turbine icing prediction.Since the multi-parameter model has many random variables,Gaussian process regression is introduced to improve the prediction ability of the model.BP neural network is used to screen the monitoring data in multiple dimensions.The filtered data is input into the multi-parameter model to obtain the predicted initial value.It is then compared with historical data under similar conditions to correct the results.Finally,the digital twin 3D visual model is used to display the final result.As a result,accurate prediction of wind turbine blade icing phenomenon can be achieved.

关 键 词：风力发电 结冰预测 数字孪生 多参数模型 

分 类 号：TK8[动力工程及工程热物理—流体机械及工程]