A dynamical-statistical atmospheric modeling system designed to simulate wind-generated power production rates

The concept of generating electrical power from the energy of the wind has received renewed interest during the last several years. This has been a result of (1) increased interest in reducing the emission of greenhouse gases by traditional electric generation plants fueled by oil, coal or natural gas; (2) major changes in the economic and organizational structure of the electric utility industry which allow customers who have a choice in the supplier of their electricity and how it is generated; and (3) improvements in wind power technology that has reduced the cost of generation to a level that is competitive with the traditional methods of generating electricity. Within the last year the total US capacity has doubled to approximately 2500 megawatts. In order to achieve optimal wind power generation and use, the wind power industry requires two types of meteorological information. First, there is a need for high resolution spatial maps of the wind climatology to provide assistance in optimally siting wind power generation facilities (i.e. arrays of wind turbines). Second, there is need for short-term (0-48 hr) forecasts of the hourly wind power output which in turn demands highly accurate localized forecasts of the wind speed and generation at the wind turbine height (usually about 50 m above the surface). In order to address these needs of the wind power industry, MESO, Inc. has developed a dynamical-statistical modeling system called FOREWIND.

FOREWIND has two major components: a high resolution boundary model and an adaptive statistics model. The dynamical model is based upon the basic physical principles of conservation of mass, momentum and energy and includes a Turbulent Kinetic Energy (TKE) boundary layer submodel. FOREWIND is designed to simulate the lowest 2-3 km of the atmosphere over a limited horizontal domain with a horizontal resolution between 1 and 5 km. The conditions at the vertical and lateral boundaries are specified from an external source. In non-real-time applications (such as constructing a wind climatology) this source is a grid point analysis of observational data. In forecast applications the boundary conditions are specified by a regional or mesoscale forecast model. FOREWIND forecasts are currently being executed on an experimental basis each day. The forecasts are being compared to wind forecasts generated by a variety of other real-time models including National Weather Service's NGM MOS and Eta models, MESO's implementation of the MM5 model and MESO's operational MASS model simulations. In addition, an extensive validation of the FOREWIND forecasts of winds at multiple levels is being done at a number of wind energy-related sites with extensive wind measurement instrumentation in Pennsylvania, New Hampshire, Massachusetts, Minnesota and California. Preliminary results indicate that the magnitude of wind forecast errors are highly correlated to the site which suggests that local factors (e.g. roughness, exposure etc.) are a major factor in wind simulation errors. In order to account for these factors which are below the grid scale of the dynamical model, an adaptive statistics module is used to refine the output of the dynamical model for a particular point. A description of the modeling system and a variety of validation results will be presented at the conference.