Identifying Representative Days of Wind Speed in Brazil Using Machine Learning Techniques

Wind power generation has attracted high investment levels in the past years in Brazil due to its relatively low construction costs and general short-time building construction, along with the worldwide incentive towards the establishment of a carbon-neutral power system. In this context, studies over the behaviour and future dynamic characteristics of its main fuel (wind speed) during the days and the years are proving to be extremely necessary. Therefore, in this work, we leverage on Machine Learning techniques to identify similar patterns on time-linked hourly-based wind speed data, aiming at identifying common representative days, crucial for planning, designing, operation, and assessment of renewable-based energy systems. A set of descriptive analytics is presented based on real data from locations with current high wind power installed capacity and potential production levels, namely, the two Brazilian regions: Northeast and South.