An Interdisciplinary Machine Learning Approach for Wind Speed Forecasting

Abstract

Multidisciplinary researchers have collaborated with industry to develop advanced high-fidelity simulation and optimization tools for wind power plants and turbine interactions with the atmosphere. These tools are capable of modeling the processes needed to predict plant interactions and provide state-of-the-art simulation and analysis capabilities that allow industry stakeholders to perform a wide variety of forecasting and optimization to lower the energy costs and mechanical impacts. Insights from machine learning and computational intelligence have the potential to transform nearly every aspect of the world as we know it. Today, these insights are being applied to accelerate the pace of discovery in a wide variety of areas including materials science, wind and solar energy, health care, national security, emergency response, and transportation. In order to provide effective wind speed forecasting, an interdisciplinary approach based on artificial intelligence (AI) by supervised machine learning with human judgment is presented in this work. An approach is proposed for a representative site in the Colonia Eulacio, Soriano Department, Uruguay. The statistical results are evaluated, and a quantitative interpretation given to choose the machine learning configuration that best forecasts the actual data. These machine learning methods have lower computational costs than other techniques such as numerical models for weather or climate prediction. The proposed method is a scientific contribution to reliable large-scale wind energy prediction and integration into existing grid systems in Soriano, Uruguay, and is a powerful tool that can help the UTE manage the national energy supply.

Keywords

• atmospheric science
• interdisciplinary communication
• machine learning
• wind speed forecasting
• system sciences and engineering