Wind Variability Analysis for the Kuwait Region Using Self-Organizing Maps

Electrical system operators utilizing wind energy production need accurate wind power forecasts to prepare for future changes in power production. Building a wind power forecasting system requires an understanding of predominant atmospheric causes of seasonal, synoptic, and diurnal patterns, resulting in varying wind speeds. These patterns can be identified with machine learning techniques such as self-organizing maps (SOMs). Here, SOMs are created from Weather Research and Forecasting (WRF) model output centered on the northern Arabian Peninsula to identify weather regimes responsible for wind patterns observed at the Shagaya wind farm in Kuwait. WRF output within these regimes are then compared to observations to determine which regimes best replicated the synoptic and mesoscale conditions that lead to seasonal and diurnal peaks in wind speed such as the summer shamal and the nocturnal low-level jet determined from a previous wind climatology from observations at Shagaya, Kuwait. These SOMs also indicate which regimes correlate with operationally significant changes in wind power production, otherwise known as wind power ramps. This provides insight to how WRF models the boundary layer processes in each regime and how accurate WRF is at predicting wind ramps in each regime.