Advancing Spatiotemporal Renewable Energies and Air Quality Forecasting with Deep Learning and Quantum Machine Learning - Thesis

Aug 5, 2026

Upload by Aurelia

This thesis addresses limitations in spatiotemporal wind speed, solar irradiance, and air quality forecasting that hinder rapid early warning for poor air quality management and decision support in renewable energy generation. Existing approaches are computationally expensive, lack real-time capability, and suffer from long run times and outdated datasets that reduce forecast accuracy. New deep learning architectures rapidly evaluate time-sensitive spatial-temporal management problems by updating and filling temporal and spatial data gaps, enabling near-instantaneous real-time input updates using distributed sensor and monitoring networks.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages262

Reading Time11 h 0 min

AccessFree to download

1 Introduction

1.1 Air Quality Management in Major Urban Centers

1.2 Renewable Energies as an Alternative for Reducing Atmospheric Pollution

1.3 Research Motivation

1.4 Research Objectives

1.5 Organization

Advancing Spatiotemporal Renewable Energies and Air Quality Forecasting with Deep Learning and Quantum Machine Learning by Victor Oliveira Santos A Thesis presented to The University of Guelph In partial fulfilment of requirements for the degree of Doctor of Philosophy in Engineering Guelph, Ontario, Canada © Victor Oliveira Santos, June , 2025 Abstract Advancing Spatiotemporal Renewable Energies and Air Quality Forecasting with Deep Learning and Quantum Machine Learning Victor Oliveira Santos University of Guelph, 2025 Advisor(s): Dr. Bahram Gharabaghi Dr. Jesse Van Griensven Thé This thesis addresses the limitations of the existing spatiotemporal methods for rapid wind speed , solar irradiance, and air quality forecasting to provide timely early warning for poor air quality management and decision support in renewable energy generation applications. Current methods are computationally expensive and are not able to provide real-time forecasts to meet emergency management requirements for early warning given the long run time and lack of updated dataset for improved forecast accuracy. Therefore, this research developed new deep learning-based architectures to allow for the rapid evaluation of complex and time-sensitive management problems requiring spatial and temporal data. The results of this research address the local spatiotemporal requirements that time-series estimation models need by updating and filling data gaps in both the temporal and the spatial data inputs to rapidly construct more accurate forecast estimates. The resulting methodologies can provide near-instantaneous model input updates for real-time forecasts, employing and integrating measurements from networks of distributed sensors, monitoring stations, and web-based resources, among others. The application of the deep learning approach highlighted the capacity of deep learning methodology to accurately predict wind speed and air quality well in advance , proving it to be a reliable tool for environmental monitoring. The quantum-based machine learning approach demonstrated QML’s competitive advantage compared to classical machine learning applications, surpassing accuracy for longer predictive horizons. The final step in this research work was the development and validation for a new state-of-the-art methodology, which combined the classical deep learning approach with the quantumbased one , forming a new hybrid framework for air quality forecasting. The hybrid classical-quantum framework managed to return results comparable to published classical deep learning models and surpassed them for longer forecasting horizons. Acknowledgements I must reckon that I could not have reached this far alone. I need to express my deepest gratitude to my parents, who always fostered my education. Thank you to my wife, Amanda Rodrigues, who agreed to stay by my side during this journey. Thank you for being my cornerstones. I would also like to thank my advisor, Prof. Bahram Gharabaghi, for his guidance and patience during this doctoral course. Thank you for being always approachable and present, always offering important and creative insights aiming to improve this research. I would like to express my gratitude to Prof. Jesse Van Griensven Thé, for also offering words of wisdom which helped me during this research. Thank you for believing in my potential. Thank you, Prof. Paulo Alexandre Costa Rocha, for also being part of my journey in this doctoral course. You have offered me guidance, understanding, and time when I needed. I would like to express my gratitude to the University of Guelph, which offered all the structure I needed to develop top-notch research. Also, there is gratitude towards the Natural Sciences and Engineering Research Council of Canada (NSERC) Alliance, and the Lakes Environmental Software Inc. , which provided the financial resources for this research. Finally, thank you all to my relatives and

FAQs

What problem does the thesis target in air quality forecasting?

It targets the lack of rapid, real-time spatiotemporal forecasting methods for wind speed, solar irradiance, and air quality that are needed for early warning and decision support.

How do the proposed deep learning methods improve forecast accuracy and timeliness?

They use architectures that update and fill data gaps in both temporal and spatial inputs, enabling near-instantaneous model input updates for real-time forecasts.

What role does quantum machine learning play in the research?

The quantum-based approach is shown to have competitive advantages over classical machine learning for longer prediction horizons, and the work develops a hybrid classical-quantum framework for air quality forecasting.

Aug 5, 2026

Upload by Aurelia