基于相空间重构的门控循环网络短期风速预测模型  

Short⁃term wind speed prediction model based on phase space reconstruction and gated recurrent unit network

作　　者：葛宽宽 王晓晶[1] 代怡重 肖群 GE Kuankuan;WANG Xiaojing;DAI Yizhong;XIAO Qun(School of Electric Engineering,Xinjiang University,Urumqi 830047,China)

机构地区：[1]新疆大学电气工程学院,新疆乌鲁木齐830047

出　　处：《现代电子技术》2022年第15期93-98,共6页Modern Electronics Technique

基　　金：新疆维吾尔自治区自然科学基金(2020D01C031)。

摘　　要：高精度的短期风速预测在一定程度上可以提高风资源评估的准确度,但因风速高度随机性导致预测难度较大,为了提升风速预测的精度,首先对风速混沌特性进行分析,验证其是否具有混沌特性,在此基础上提出一种基于变分模态分解(VMD)、相空间重构(PSR)和门控循环神经网络(GRU)相结合的短期风速预测模型。将原始风速数据通过VMD分解为若干子序列并对各子序列进行PSR技术优化处理,可以使样本更加符合实际风速的变化,起到对风速降噪的作用;接着用处理过的各子序列对GRU预测模型训练并预测,最后求和得出风速预测结果。以上海某地为例,将预测结果与实际风速值进行对比分析,分析结果表明:所建模型能有效提高风速预测精度,在预测准确度上优于单一预测模型,验证了所提模型的有效性和合理性。High⁃precision short⁃term wind speed prediction can improve the accuracy of wind resource assessment to a certain extent.However,due to the high randomness of wind speed,it is difficult to obtain a satisfied prediction.In order to improve the prediction accuracy of wind speed,the chaotic characteristics of wind speed is analyzed firstly to verify whether the chaotic characteristics exist.On this basis,a short⁃term wind speed prediction model based on the combination of variational modal decomposition(VMD),phase space reconstruction(PSR)and gated recurrent unit(GRU)neural network is proposed.The original wind speed data is decomposed into several sub⁃sequences by VMD,and the PSR technology optimization processing is performed on each sub⁃sequence,which can make the samples more in line with the actual wind speed changes and furthermore play a role in reducing wind speed noise.And then,the processed sub⁃sequences are used to train and predict the GRU prediction model,and summation is performed to get the wind speed prediction result,finally.A place in Shanghai is taken as an example.The prediction results were contrasted with the actual wind speed value.The analysis results show that the built model can improve the prediction accuracy of wind speed effectively,and its prediction accuracy is better than that of a single prediction model,which verifies the validity and rationality of the proposed model.

关 键 词：风速预测 混沌特性 风资源评估 变分模态分解 相空间重构 门控循环网络 数据预处理 预测模型 

分 类 号：TN919.5-34[电子电信—通信与信息系统] TM614[电子电信—信息与通信工程]