Correlations between meteorological conditions and power loss in wind turbines under icing conditions

Abstract

Wind energy has become a significant renewable energy source, playing a crucial role in the global transition to a sustainable future. Recently, the use of this renewable energy has been expanding, particularly in cold-climate regions, where strong winds and higher air density at lower temperatures are prevalent. However, wind farm operators face a challenge: icing. Icing occurs when sub-zero temperatures cause ice to form on wind turbine components, including blades, the nacelle, and the tower. The presence of ice on wind turbines can have serious repercussions on their aerodynamic properties and compromise their ability to capture wind energy efficiently. Ice formations increase the blades' weight and alter their shape, potentially reducing rotational speed and energy production. As a result, wind turbines operating under icing conditions produce less electricity, limiting their contribution to the power grid and reducing their overall efficiency. It is essential for wind farm operators to closely monitor and manage icing conditions to ensure the safety of personnel and the public. Addressing the influence of icing on wind turbines necessitates a multifaceted approach. The variability of temperature, humidity, and wind speed predicts challenging icing events. Advanced weather monitoring systems, including sensors and meteorological models, improve the accuracy of forecasts for icing events. By monitoring these conditions, operators can anticipate and prepare for potential icing events. This study aims to establish correlations for estimating the expected power loss of wind turbines during icing conditions. The correlations proposed here were developed using dimensional analysis, also known as the “ᴨ theorem,” and validated using three machine learning methods. This solution enables operators to improve the efficiency of wind turbine operations under icing conditions by systematically assessing icing severity. The originality of the proposed methodology lies in its ability to estimate power losses directly from meteorological, geometrical, and operational parameters of wind turbines. A database has been created using results from literature-based icing simulation studies, data collected at a research center, and results from a CFD simulation study conducted as part of this thesis. The correlations obtained were validated using multiple machine learning methods and demonstrated consistency across different scenarios. The proposed methodology facilitates the utilization of dimensionless parameters that establish correlations among meteorological, geometrical, and operational parameters. Recommendations are presented for the utilization of machine learning methods and dimensionless analysis to improve the efficiency of wind turbines under icing conditions. A review paper has been published and included as a chapter in the thesis. This paper reviews various methods for modeling power loss due to icing in wind turbines. A second research paper has been published, examining the impact of surface roughness modeling on the aerodynamics of an iced wind turbine airfoil. This paper is also included as a chapter in the thesis. This project will contribute to advancing knowledge on icing and to optimizing wind turbine performance in northern climates.

Date	8 Sept 2026
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Adrian Ilinca (Supervisor) & Santiago Lain (Co-supervisor)