6th International Conference Energy & Meteorology: Abstract Submission

Renewable Energy Forecasting for Kuwait: A Progress Report (783)

Jared A. Lee 1 , Sue Ellen Haupt 1 , Branko Kosovic 1 , Gerry Wiener 1 , Pedro A. Jimenez 1 , Majed Al-Rasheedi 2
  1. National Center for Atmospheric Research, Boulder, CO, United States
  2. Kuwait Institute for Scientific Research, Shuwaikh, Kuwait

As part of the goals by the Government of Kuwait to achieve 15% renewable energy production by 2030, the Shagaya Renewable Energy Park has been commissioned in western Kuwait. Phase 1 of Shagaya has been completed, and comprises 10 MW wind power, 10 MW photovoltaic (PV) solar power, and 50 MW concentrated solar power (CSP). Future phases will add thousands of megawatts of wind and solar power capacity to Shagaya. 

Addressing the need for good forecasting in order to reliably integrate that wind and solar power into Kuwait’s national electrical grid, the National Center for Atmospheric Research (NCAR), in a 3-year project funded by the Kuwait Institute for Scientific Research (KISR), is in the midst of building and delivering an operational Shagaya Renewable Energy Prediction System to KISR.

This wind and solar power prediction system, which operates both on intra-day/nowcasting and days-ahead lead times, incorporates various global numerical weather prediction (NWP) models, a specialized high-resolution configuration of the Weather Research and Forecasting (WRF) regional NWP model tailored for wind and solar applications (WRF-Solar-Wind), satellite observations, statistical machine learning models (StatCast), and meteorological and power observations from the wind and solar plants. All of this data is used as input to NCAR’s Dynamic Integrated Forecasting (DICast®) system to generate updated power forecasts every 15 minutes using machine learning. Probabilistic power forecasts, including confidence intervals, are also generated using an analog ensemble (AnEn) technique. A web-based operator display presents the resultant DICast and AnEn power predictions, overlaid with observed power when viewing historical cases. In this presentation we highlight results from the performance of the prediction system, both over extended periods and for individual cases, and outline plans for ongoing development and refinement of the system.