Multi-head attention-based probabilistic CNN-BiLSTM for day-ahead wind speed forecasting

Yi-Ming Zhang and Hao Wang

Energy, 2023, vol. 278, issue PA

Abstract: Wind energy is one of the most widely used and fastest-growing renewable energy. Wind speed prediction is an efficient way to rationally dispatch wind power generation and ensure the stability of the power system. Typical deep learning algorithms, such as the convolution neural network (CNN), long short-term memory (LSTM), and their hybrid model (CNN-LSTM), have been extensively used for time-series prediction in various fields. The CNN-LSTM model exhibits superior forecasting performance by integrating the advantages of CNN and LSTM, but it fails to quantify the forecasting uncertainty that is critical for optimal management. This study proposes a probabilistic model to forecast day-ahead wind speed based on the CNN-bidirectional LSTM (BiLSTM) and deep ensemble strategy. Unlike purely statistical methods, the hybrid physical-statistical model that combines the numerical weather prediction (NWP) model and onsite measurements is employed to improve long-term forecasting accuracy. Specifically, the probabilistic CNN-BiLSTM model is developed by adjusting the network structure, and the uncertainty is optimized through a proper scoring rule. A combination of ensembles is then used to improve the robustness of probabilistic prediction. The spatial and temporal correlations of NWP data are both considered. The new probabilistic model is applied to forecast wind speed in measurements collected from the outdoor competition venues in the 2022 Winter Olympics. Nine probabilistic benchmark methods are used to compare the performance of the proposed deep ensemble model. The results indicate that the proposed model exhibits the highest forecasting accuracy and the best ability in uncertainty estimation.

Keywords: Wind speed; Day-ahead probabilistic forecasting; Deep ensemble; Numerical weather prediction; Onsite measurements (search for similar items in EconPapers)
Date: 2023
References: View references in EconPapers View complete reference list from CitEc
Citations: View citations in EconPapers (10)

Downloads: (external link)
http://www.sciencedirect.com/science/article/pii/S0360544223012598
Full text for ScienceDirect subscribers only

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:eee:energy:v:278:y:2023:i:pa:s0360544223012598

DOI: 10.1016/j.energy.2023.127865

Access Statistics for this article

Energy is currently edited by Henrik Lund and Mark J. Kaiser

More articles in Energy from Elsevier
Bibliographic data for series maintained by Catherine Liu ().