基于元学习的风电场短期风电功率组合预测  

Short-term Wind Power Portfolio Prediction for Wind Farms Based on Meta-learning

作　　者：仲铭 李子承 ZHONG Ming;LI Zi-cheng(School of Business,Nanjing Audit University,Nanjing 210029,China)

机构地区：[1]南京审计大学商学院,江苏南京210029

出　　处：《电脑与信息技术》2024年第2期6-9,共4页Computer and Information Technology

摘　　要：针对单一的模型受风电功率高度的不确定性和波动性影响无法很好地匹配不同风电涡轮机的序列特征从而导致预测精度低的问题,提出了使用一种基于深度卷积神经网络的元学习方法。该方法首先从原始数据中自动学习一个特征表示,然后将学习到的特征用神经网络与一组权值联系起来,并将这些权值分配给一组基预测模型,最终形成组合预测。实验结果表明,在真实的数据上的预测评估中,基于元学习的组合方法可以有效提高风电功率的预测精度。A meta-learning method using a deep convolutional neural network-based approach is proposed to address the problem that a single model is affected by the high uncertainty and volatility of wind power and the inability to match well the sequential characteristics of different wind turbines,thus resulting in low prediction accuracy.The method first automatically learns a feature representation from the original data,and then associates the learned features with a set of weights using a neural network,and assigns these weights to a set of base prediction models,resulting in a combined prediction.The experimental results show that the combined meta-learning-based method can effectively improve the prediction accuracy of wind power in the prediction evaluation on real data.

关 键 词：风电功率预测 深度学习 多步预测 元深度学习 组合预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]