基于布谷鸟算法优化独立循环神经网络深度学习的超短期风电功率预测 |
点此下载全文 |
引用本文:邓亚平1,段建东1,贾颢2,王璐3,同向前1.基于布谷鸟算法优化独立循环神经网络深度学习的超短期风电功率预测[J].电网与清洁能源,2021,37(9):18~26 |
摘要点击次数: 170 |
全文下载次数: 0 |
|
基金项目:陕西省自然科学基础研究计划项目(2019JQ-329) |
|
中文摘要:风电功率数据具有强烈的时序特性,其序列数据的特征提取,是进行风电功率准确预测的重要前提。为此,引入了更长、更深层次的多隐层独立循环神经网络来最大程度上提取可反映输入风电功率序列数据的本质特征量,进而建立起特征量与风电功率之间的非线性关系。然而,在建立深层独立循环神经网络时,存在模型超参数设置与优化困难的问题。为此,进一步提出结合布谷鸟算法对独立循环神经网络关键超参数进行优化设计的方法。最终,结合某风电场实际数据,将模型预测结果与实测数据进行对比,验证所提方法能够有效提高预测精度。 |
中文关键词:布谷鸟算法 独立循环神经网络 深度学习 风电功率 功率预测 |
|
Ultra-Short-Term Wind Power Prediction Based on Deep Learning with Independent Recurrent Neural Network via Cuckoo Algorithm Optimized |
|
|
Abstract:The wind power data possess strong serial characteristics and it is the prerequisite for accurate wind power prediction to extract the characteristics of wind power data with sequence feature. Therefore, in this paper, a longer and deeper neural network, namely independent neural network (IndRNN), is adopted to obtain the characteristics to the maximal possible, which can reflect the essential features of wind power data with sequence feature. And then, the nonlinear relationship between extracted features and wind power is established. However, the fact that many hyper-parameters needs to be designed during establishing a multi-layer IndRNN network makes it difficult for the design and optimization of all these hyper-parameters. Therefore, a method based on IndRNN with Cuckoo Algorithm Optimized is further proposed to design and optimize hyper-parameters. Finally, by using the actual wind power data from a real wind farm, the results between predicated data and measured data are compared, which shows that this method can effectively improve the prediction accuracy. |
keywords:Cuckoo algorithm independent recurrent neural network deep learning wind power power prediction |
查看全文 查看/发表评论 下载PDF阅读器 |