基于DCGCN模型的海上风电场超短期功率预测

Ultra-short-term Power Prediction for Offshore Wind Farms Based on Dual Channel Graph Convolution Network Model

图卷积网络(GCN)具有很强的数据关联挖掘能力,近年来在风电功率预测领域获得了广泛关注。然而,传统的基于GCN模型的超短期风电功率预测难以同时处理影响风电功率的两大核心因素(风速与机组状态信息)的双模态问题,基于此,提出了一种基于双通道图卷积网络(DCGCN)的海上风电场超短期功率预测模型。首先,建立以理论功率曲线为基准的机组状态指标模型,定量表征机组状态变化对其发电能力的影响;其次,构建海上风电场图拓扑,建立基于风速和状态邻接矩阵的风电场各机组捕获的风速与机组状态信息的关联关系模型;最后,建立基于DCGCN的风电场超短期功率预测方法。算例结果表明,所提模型有助于提高风电场功率预测模型的训练效率和预测精度。

Graph convolution network(GCN)has strong data correlation mining capabilities and has gained widespread attention in the field of wind power prediction in recent years.However,traditional ultra-short-term wind power prediction based on GCN model is difficult to deal with the dual-mode problem of the two core factors(wind speed and unit state information)that affect wind power simultaneously.Based on this,an ultra-short-term power prediction for offshore wind farms based on dual channel graph convolution network(DCGCN)model is proposed.Firstly,a unit state index model based on the theoretical power curve is established to quantitatively characterize the impact of unit state changes on its power generation capacity.Secondly,the graph topology of the offshore wind farm is constructed,and the correlation relationship model between the wind speed captured by each unit of the wind farm and the unit state information is established based on the wind speed and state adjacency matrices.Finally,an ultra-short-term power prediction method of wind farm based on DCGCN is established.Case study results indicate that the proposed model is helpful to improve the training efficiency and prediction accuracy of the wind farm power prediction model.