Enhancing Regional Wind Power Forecasting Through Advanced Machine-Learning and Feature-Selection Techniques

Sep 17, 2026

An in-depth study addresses forecasting aggregated regional wind power production using advanced machine-learning and feature-selection techniques. The central challenge is selecting wind speed measurement points across a large area when wind plant locations are treated as unknown. Main cities (province capitals) are modeled as candidate features, and four selection methods are tested: Pearson correlation, Spearman correlation, mutual information, and Chi-squared with Fisher score. Results show feature selection improves prediction for high-dimensional regional data, and LSTM outperforms other compared models in northern Italy.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages22

Reading Time55 min

AccessFree to download

1. Introduction

2. Problem Formulation and Methodology

3. Feature Selection Methods

4. Machine-Learning Models

5. Experimental Results

6. Conclusion

energies Article Enhancing Regional Wind Power Forecasting Through Advanced Machine-Learning and Feature-Selection Techniques Nabi Taheri and Mauro Tucci * Citation: Taheri, N.; Tucci, M. Enhancing Regional Wind Power Forecasting Through Advanced Machine-Learning and FeatureSelection Techniques. Energies 2024, 17, 5431. [https://doi.org/10.3390/](https://doi.org/10.3390/)[ ](https://doi.org/10.3390/)en17215431 Academic Editor: John Boland Received: 2 October 2024 Revised: 24 October 2024 Accepted: 27 October 2024 Published: 30 October 2024 Copyright: © 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license ([https://](https://)[ ](https://)[creativecommons.org/licenses/by/](creativecommons.org/licenses/by/)[ ](creativecommons.org/licenses/by/)[4.0/](4.0/)) . Department of Energy, Systems, Territory and Constructions Engineering, University of Pisa, L.go Lucio Lazzarino 1, 56122 Pisa, Italy; [nabi.taheri@phd.unipi.it](nabi.taheri@phd.unipi.it) * [Correspondence: mauro.tucci@unipi.it](Correspondence: mauro.tucci@unipi.it) Abstract: In this study, an in-depth analysis is presented on forecasting aggregated wind power production at the regional level, using advanced Machine-Learning (ML) techniques and featureselection methods. The main problem consists of selecting the wind speed measuring points within a large region, as the wind plant locations are assumed to be unknown. For this purpose, the main cities (province capitals) are considered as possible features and four feature-selection methods are explored: Pearson correlation, Spearman correlation, mutual information, and Chi-squared test with Fisher score. The results demonstrate that proper feature selection significantly improves prediction performance, particularly when dealing with high-dimensional data and regional forecasting challenges. Additionally, the performance of five prominent machine-learning models is analyzed: Long Short-Term Memory (LSTM) networks, Artificial Neural Networks (ANNs), Support Vector Machines (SVMs), Convolutional Neural Networks (CNNs), and Extreme-Learning Machines (ELMs) . Through rigorous testing, LSTM is identified as the most effective model for the case study in northern Italy. This study offers valuable insights into optimizing wind power forecasting models and underscores the importance of feature selection in achieving reliable and accurate predictions. Keywords: wind power forecasting; machine learning; feature selection; regional prediction; renewable energy 1. Introduction The increasing dependence on renewable energy sources is essential to meet the rising global energy demand, which is expected to exceed 10 billion by 2050 [1] . Renewable energy sources, such as wind, solar, and hydro-power, offer sustainable alternatives to depleting fossil fuels, significantly reducing environmental pollution and mitigating the adverse effects of climate change [2] . Recent technological advancements have made renewable energy technologies more economically viable and improved integration into the energy market [1] . Accurately predicting wind power generation is crucial for maintaining the balance between supply and demand in electricity systems, yet the inherent variability of wind resources complicates this task [3,4] . Fuzzy models [5], as demonstrated in various applications, provide an effective means for managing uncertainty and nonlinearity in complex systems, making them particularly advantageous for ultra-short-term wind speed and power forecasting. A two-stage fuzzy nonlinear fusion model, using empirical mode decomposition and T-S fuzzy models combined with IT2-based aggregation, has demonstrated superior accuracy in ultra-short-term forecasting for wind farms in Colorado, USA [6] . A data-driven probabilistic Wind Power Ramps WPR forecasting (p-WPRF) method has be

FAQs

What is the main forecasting task studied in the document?

Forecast aggregated wind power production at the regional level using machine-learning techniques. The focus is on improving regional predictions rather than single-site forecasting.

How does the study handle the uncertainty about wind plant locations?

It assumes wind plant locations are unknown and instead considers main cities (province capitals) as possible features for selecting measurement points.

Which feature-selection methods and machine-learning models are evaluated?

Four feature-selection methods are explored: Pearson correlation, Spearman correlation, mutual information, and Chi-squared with Fisher score. Five models are analyzed: LSTM, ANN, SVM, CNN, and ELM.

Sep 17, 2026

Uploaded by Evangeline