基于CBAM-DSC-UNet模型的时空风速预测算法

SPATIO-TEMPORAL WIND SPEED PREDICTION ALGORITHM BASED ON CBAM-DSC-UNet MODEL

针对时空风速预测任务通常使用的卷积神经网络(CNN)和循环神经网络(RNN)联合建模方法中空间信息损失的问题,提出一种基于CBAM-DSC-UNet模型的时空风速预测算法,用于提升空间信息利用率与模型预测精度。该算法将时空风速预测问题视为视频预测问题,在提取时空相关性的同时保持空间信息,进而直接输出未来多步的空间风速矩阵。以美国怀俄明州某风电场实际数据为算例进行实验,结果表明,相比其他对比算法,基于CBAM-DSC-UNet模型的时空风速预测算法的平均绝对误差下降8.4%~15.9%,精度有较大提升。

In response to the problem of spatial information loss in the joint modeling methods of convolutional neural networks(CNN)and recurrent neural networks(RNN)commonly used for spatial-temporal wind speed prediction tasks,we propose a spatial-temporal wind speed prediction algorithm based on the CBAM-DSC-UNet model.This algorithm aims to enhance the utilization of spatial information and improve the accuracy of model predictions.We treat the spatial-temporal wind speed prediction problem as a video prediction problem in order to preserve spatial information while extracting spatial-temporal correlations,thereby directly outputting the spatial wind speed matrix for multiple future steps.We conducted a calculating using actual data from a wind farm in Wyoming,USA as a case study.The results show that to other algorithms,the average absolute error of the spatial-temporal wind speed prediction algorithm based on the CBAM-DSC-UNet model reduces by 8.4%to 15.9%,demonstrating a significant improvement in prediction accuracy.