基于VMD-IAVOA-GRU的超短期光伏功率预测模型  

Ultra-Short-Term PV Power Prediction Based on VMD-IAVOA-GRU

作　　者：郭鹏[1] 田敏[1] 刘祥 GUO Peng;TIAN Min;LIU Xiang(Shihezi University,Shihezi Xinjiang 832003,China)

机构地区：[1]石河子大学,新疆石河子832003

出　　处：《计算机仿真》2025年第3期111-116,共6页Computer Simulation

基　　金：国家自然科学基金(61962053)。

摘　　要：针对分布式光伏功率时间序列预测精度低的问题,提出一种基于变分模态分解(Variational mode decomposition,VMD)、改进的非洲秃鹫优化算法(Improved African vulture optimization algorithm,IAVOA)和门控循环单元(Gate Recurrent Unit,GRU)的超短期光伏功率组合预测模型。首先,使用VMD将光伏功率历史时间序列进行分解,得到若干个规律性较强的分量;然后,通过改进的AVOA算法优化GRU网络的超参数,并对各个分量建立IAVOA-GRU预测模型;最后,叠加各个分量预测值得到最终预测结果。仿真结果表明:所提出的VMD-IAVOA-GRU模型的预测精度优于其它模型。To address the problem of low accuracy of distributed PV power time series prediction,an ultra-shortterm PV power combination prediction model based on Variational mode decomposition(VMD),Improved African vulture optimization algorithm(IAVOA)and a Gate Recurrent Unit(GRU)model.Firstly,the historical time series of PV power is decomposed using VMD to obtain a number of more regular components.Then,the hyperparameters of the GRU network are optimized by the improved AVOA algorithm,and the IAVOA-GRU prediction model is developed for each component.Finally,the individual component predictions are superimposed to obtain the final prediction.The simulation experimental results show that the prediction accuracy of the VMD-IAVOA-GRU model proposed in this paper is better than that of other models.

关 键 词：光伏功率预测 变分模态分解 改进非洲秃鹫算法 门控循环单元神经网络 

分 类 号：TM743[电气工程—电力系统及自动化]