Machine learning-based simplified methods using shorter wind measuring masts for the time ahead wind forecasting at higher altitude for wind energy applications

Abstract

Since wind is a fluctuating resource, the integration of wind energy into the electricity grid necessitates precise wind speed forecasting to maintain grid stability and power quality. Machine learning models built on different algorithms are widely used for wind forecasting. This requires a vast quantity of past wind speed data collected at the hub levels of the wind electric machines employed. Tall met masts pose a variety of practical issues in terms of installation and long-term maintenance, which will grow more challenging as next-generation wind turbines come with larger capacities and higher hub heights. In this paper, we propose four non-conventional methods for the time ahead forecasting of wind speed at a higher height by utilizing the wind speed data collected with relatively shorter wind measuring masts. We employ machine learning-based models and rely on the principle of interrelation between wind speeds at different altitudes in our investigations. Wind speed forecasts generated by the new methods at an altitude of 80 m above the ground level using wind speed data measured at lower altitudes of 50 m and 20 m are of industrially acceptable accuracy. The simplified physical requirements such methods demand far outweigh the marginal fall in prediction accuracy observed with these methods.