摘要
针对光伏发电功率的随机性、波动性和非线性问题,提出了一种结合经红尾鵟(RTH)算法优化的变分模态分解(VMD)、核主成分分析(KPCA)和经RTH算法优化的门控循环单元(GRU)神经网络的光伏发电功率预测模型。首先,使用RTH算法对VMD和GRU神经网络的5个超参数进行优化;接着,应用优化后的VMD方法分解原始数据,以减少光伏数据的波动性和随机性;然后,采用KPCA方法降低数据维度,消除冗余;最后,利用经RTH优化的GRU神经网络模型进行时序建模。通过分析新疆某光伏电站的历史发电数据,并与GRNN、LSTM、GRU以及OVMD-GRU、OVMD-KPCA-GRU模型相比较,本模型的拟合优度高达98.96%,显示出更高的预测精度。
This study proposes a photovoltaic power generation forecasting model to solve the inherent randomness,volatility and non-linearity of photovoltaic power generation.The model combines with the Variational Mode Decomposition(VMD)optimized by the Red-tailed Hawk(RTH)algorithm,the Kernel Principal Component Analysis(KPCA),and the Gated Recurrent Unit(GRU)neural network refined by the RTH algorithm.Initially,the RTH algorithm is used to optimize the five hyperparameters within the VMD and GRU neural network.Then,the optimized VMD method is applied to process the original data for significantly mitigating the volatility and randomness associated with photovoltaic data.Following this,the KPCA technique is utilized to streamline data dimensions and reduce redundancy.Finally,a time-series model is established using the GRU neural network optimized by RTH.Through the analysis of the historical power generation data of a photovoltaic power station in Xinjiang,the prediction accuracy of this model reaches 98.96%,which is obviously better than that of GRNN,LSTM,GRU,OVMD-GRU,and OVMD-KPCA-GRU models.
作者
王红徐
严新军
夏庆成
刘佳琪
王雪虎
WANG Hongxu;YAN Xinjun;XIA Qingcheng;LIU Jiaqi;WANG Xuehu(College of Water Conservancy and Civil Engineering,Xinjiang Agricultural University,Urumqi 830052,Xinjiang,China;Xinjiang Water Conservancy Project Safety and Water Disaster Prevention and Control Key Laboratory,Urumqi 830052,Xinjiang,China)
出处
《水力发电》
CAS
2024年第9期98-103,共6页
Water Power
基金
自治区重点研发任务专项项目资助(2022B03024-3)
新疆水利工程安全与水灾害防治重点实验室研究项目(ZDSYS-YJS-2022-09)。
关键词
变分模态分解
核主成分分析
红尾鵟优化算法
门控循环神经网络
光伏功率预测
Variational Mode Decomposition(VMD)
Kernel Principal Component Analysis(KPCA)
Red-tailed Hawk(RTH)algorithm
Gated Recurrent Unit(GRU)neural network
photovoltaic power prediction
