摘要
为提取和整合多种影响风电功率的因素,减少单一因素波动带来的预测误差,基于图卷积网络(Graph Convolution Networks,GCN)提出了一种历史风速和功率数据相结合的模型。该方法以风机为结点,通过构建风机间的图结构,对风场风速与功率两种因素特征提取,该模型可捕捉风速和风功率之间的复杂相互作用,以及风场内不同风机之间的协同效应,有效提取局部和全局特征,适应风电功率的非线性和时变特性。通过在实际风电场数据上的实验结果表明,与传统方法相比,基于GCN的模型在短期风电功率预测上具有更高的准确性和稳定性,为风电场的调度和运行提供了可靠的参考,使模型在面对不同的环境条件时表现得更为稳定。
To extract and integrate various factors influencing wind power generation,and to reduce the prediction error caused by fluctuations in single factors,this paper proposes a model that combines historical wind speed and power data based on Graph Convolutional Networks(GCN).The method treats wind turbines as nodes and constructs a graph structure among turbines to extract the features of both wind speed and power.The model captures the complex interactions between wind speed and power generation,as well as the synergistic effects between different turbines within a wind farm.By effectively extracting both local and global features,the model adapts to the nonlinear and time-varying characteristics of wind power.Experimental results on real wind farm data show that,compared with traditional methods,the GCN-based model achieves higher accuracy and stability in short-term wind power forecasting.This provides a reliable reference for wind farm scheduling and operations,ensuring greater stability under varying environmental conditions.
作者
陈万康
张红英
刘振
高雁辉
颜义鹏
CHEN Wankang;ZHANG Hongying;LIU Zhen;GAO Yanhui;YAN Yipeng(Beijing FRP Institute Tengzhou Composite Materials CO.,LTD,Tengzhou 277500,China)
出处
《齐鲁工业大学学报》
2025年第5期56-60,共5页
Journal of Qilu University of Technology
