Article: Deep learning optimisation for spatial wind power forecasting: a data driven approach to grid stability enhancement Journal: International Journal of Automation and Control (IJAAC) 2025 Vol.19 No.2 pp.188 - 212 Abstract: While wind power has surged as a clean energy source in recent decades, its inherently unstable nature poses a challenge to grid stability. However, forecasting challenges remain, including inconsistent historical data for individual turbines and growing errors in multi-step predictions. This paper presents a novel solution to tackle the intricate problem of spatial dynamic wind power forecasting, leveraging the latest advancements in deep learning-based forecasting models. To achieve the best possible settings for the wind power forecasting model, we prepared the solution after exploring different dimensions including deep learning models, features selection, scaling methods, look-back window size, and optimisers. We selected 6 state-of-the-art forecasting models, 3 scaling methods, 8 optimisers, and a look-back window size ranging from 1 to 14 days. Our findings demonstrate the effectiveness of the proposed framework and establish a foundation for further advancements in wind power forecasting accuracy and grid stability. Inderscience Publishers - linking academia, business and industry through research

Title: Deep learning optimisation for spatial wind power forecasting: a data driven approach to grid stability enhancement

Authors: Nashwa Ahmad Kamal; Mohamed ElSobky; Ahmed M. Ibrahim; Zeeshan Haider

Addresses: Faculty of Engineering, Electrical Power and Machine Department, Cairo University, Giza, 12613, Egypt ' Faculty of Engineering, Electrical Power and Machine Department, Cairo University, Giza, 12613, Egypt ' Faculty of Engineering, Electrical Power and Machine Department, Cairo University, Giza, 12613, Egypt ' Automated System and Soft Computing Lab, Prince Sultan University, Riyadh, 11586, Saudi Arabia

Abstract: While wind power has surged as a clean energy source in recent decades, its inherently unstable nature poses a challenge to grid stability. However, forecasting challenges remain, including inconsistent historical data for individual turbines and growing errors in multi-step predictions. This paper presents a novel solution to tackle the intricate problem of spatial dynamic wind power forecasting, leveraging the latest advancements in deep learning-based forecasting models. To achieve the best possible settings for the wind power forecasting model, we prepared the solution after exploring different dimensions including deep learning models, features selection, scaling methods, look-back window size, and optimisers. We selected 6 state-of-the-art forecasting models, 3 scaling methods, 8 optimisers, and a look-back window size ranging from 1 to 14 days. Our findings demonstrate the effectiveness of the proposed framework and establish a foundation for further advancements in wind power forecasting accuracy and grid stability.

Keywords: wind power forecast; forecast; deep learning; SDWPF; spatially dynamic wind power forecasting; turbine.

DOI: 10.1504/IJAAC.2025.144724

International Journal of Automation and Control, 2025 Vol.19 No.2, pp.188 - 212

Received: 20 Feb 2024
Accepted: 20 Apr 2024
Published online: 28 Feb 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Deep learning optimisation for spatial wind power forecasting: a data driven approach to grid stability enhancement

Keep up-to-date