Improving accuracy of wind speed forecasting

As a sustainable and renewable source of power, wind energy has been rapidly expanding in the last decade, playing an increasingly significant role in the urgent transition to clean energy.

At over 93 gigawatts, the capacity of wind energy infrastructure installed globally in 2020 more than doubled that of the preceding year.

Still, there remains much room for improvement in terms of efficiently harnessing the energy source, including enhancing the accuracy in forecasting wind speeds–a considerable challenge given the variable nature of wind.

Jia et al. created a methodology to incorporate large-scale atmospheric information into short-term wind speed forecasts over a large geographical area, namely some 435,000 square kilometers in Alberta, Canada.

“We present a method to forecast short-term [up to 6 hours ahead] wind speed over a large region that served as our study area in Alberta,” said author Tianxia Jia. “We found that modeling multiple locations together produces more accurate forecasts than modeling locations separately.”

The research employed two publicly accessible datasets and 23 weather stations and compared predictive performance for atmospheric clustering methods and forecasting methods. For the purposes of short-term forecasting, the study presents time series regime-switching and mixture models incorporating data gleaned from clustering.

“We found that large-scale atmospheric information, such as the meteorological conditions over the entire region of Alberta, can be incorporated into our statistical wind speed forecasting models, and is quite valuable for short-term wind speed forecasting,” said Jia. “Together with the spatial information, such as the wind speeds from the multiple sites, we can reduce forecasting errors in the future.”

The work may help streamline the growing wind energy industry.

Source: “Short-term wind speed forecasting with regime-switching and mixture models at multiple weather stations over a large geographical area,” by Tianxia Jia, Deniz Sezer, and David Wood, Journal of Renewable and Sustainable Energy (2022). The article can be accessed at http://doi.org/10.1063/5.0098090 .