1.1
Applications of the NCEP Climate Forecast System Model (v2) to Monthly and Seasonal Prediction of Wind Power over the Continental United States

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 3 February 2014: 11:00 AM
Room C114 (The Georgia World Congress Center )
J. Craig Collier, Garrad Hassan America, Inc., San Diego, CA

Seasonal variability of renewable energy resource has significant impacts on transmission system operations and reliability, balancing and reserve allocation by utilities, as well as hedging and long-term transaction decisions for energy market participants. This study addresses the needs for seasonal time-scale predictability of wind resource and its variability at the regional level and its applicability to economic decisions made by the system operator, the utility, and the power marketer. The main engine used for such predictions is the NCEP Climate Forecast System (CFS) Model, an operational climate model predicting all major atmospheric variables and features at the seasonal and inter-seasonal time scales. Energy application is considered one of its intended uses, among other agricultural applications.

The CFS was originally implemented into NCEP Operations in mid-2004, is a quasi-global fully coupled atmosphere-ocean-land model for seasonal climate prediction. It atmospheric model, borrowed from the NCEP Global Forecast System (GFS) for medium- and long-range forecasts, solves the equations of motion for the atmosphere at T126 truncation, with 64 sigma-pressure hybrid layers vertically. Its sea-ice and land models are multi-layer and fully interactive. The second generation model (CFSv2) improves many aspects of data assimilation enhancing predictability on seasonal and sub-seasonal time scales. The CFSv2, with its 4-member ensemble, and predictions out to 3-9 months, offers a rich, operationally-supported, and robust set of seasonal predictions of all major atmospheric, oceanic, and land surface variables. In this study, the model's predictions for wind resource have been statistically downscaled and applied to physical wind power plant models for regional power predictions across the continental U.S. The resulting predicted anomalies in power are evaluated against actual wind power generation records from major wind development regions and stratified diurnally for on-peak and off-peak load. Additionally, model ensemble variance is used to estimate seasonal forecast uncertainty and statistical significance of anomaly prediction. Finally, we speculate on causes of prediction error and, in its context, estimate overall benefits to stakeholders.