WOA-VMD-SCINet: Hybrid model for accurate prediction of ultra-short-term Photovoltaic generation power considering seasonal variations

Accurate prediction of photovoltaic (PV) power generation has become a key technical approach to improve the efficiency and stability of photovoltaic power plants. This paper proposes an ultra-short-term photovoltaic power prediction method based on a hybrid model of whale optimization algorithm, va...

Full description

Saved in:

Bibliographic Details
Published in:	Energy reports Vol. 12; pp. 3470 - 3487
Main Authors:	Zhao, Yonghui, Peng, Xunhui, Tu, Teng, Li, Zhen, Yan, Peiyu, Li, Chao
Format:	Journal Article
Language:	English
Published:	Elsevier Ltd 01.12.2024
Subjects:	Deep learning Photovoltaic power Forecasting Sample convolution and interactive neural network Variational mode decomposition Whale optimization algorithm Deep learning Variational mode decomposition Photovoltaic power Forecasting Whale optimization algorithm Sample convolution and interactive neural network
ISSN:	2352-4847, 2352-4847
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Accurate prediction of photovoltaic (PV) power generation has become a key technical approach to improve the efficiency and stability of photovoltaic power plants. This paper proposes an ultra-short-term photovoltaic power prediction method based on a hybrid model of whale optimization algorithm, variational mode decomposition and sample convolutional interactive neural network. Firstly, the Pearson correlation coefficient method was used to select the meteorological features with high correlation as input feature variables. Secondly, Variational mode decomposition (VMD) was used to smoothly decompose the original photovoltaic power generation data. Then, the Whale Optimization Algorithm (WOA) was used to obtain each decomposition sequence containing the best PV power time series feature information, and it was input into the Sample Convolutional Interactive Neural Network model (SCINet), which was first applied to the PV short-term power prediction. Finally, the proposed model was analyzed on a dataset containing historical time series of photovoltaic power generation. The accuracy of this model was then compared with Long Short-Term Memory network (LSTM), Temporal Convolutional Network (TCN), SCINet, VMD-LSTM, VMD-TCN and VMD-SCINet, WOA-VMD-LSTM, WOA-VMD-TCN and WOA-VMD-SCINet for different seasons, time periods, and climatic conditions The SCINet model is compared, and the results show that the proposed model is better than the other eight models in terms of prediction performance. Compared with LSTM, TCN, SCINet and VMD-SCINet models. The Mean Absolute Error (MAE) of spring, summer, autumn and winter were reduced by 9%, 6.9%, 3.4% and 2%, respectively. 7.7%, 5.5%, 3.2%, 2.1%; 11.4%, 5.9%, 3.4%, 2.2%; 11.4%, 5.9%, 3.4%, 2.2%. In July and August, when PV power generation fluctuates the most, the RMSE (Root Mean Square Error)value of the model is 1.243, which is 24% higher than the prediction accuracy of the other eight models on average.
ISSN:	2352-4847 2352-4847
DOI:	10.1016/j.egyr.2024.09.025