Nonlinear Spatial Characterization and Interpolation of Wind Data

Abstract

High fossil fuel (oil and gas) prices; concerns about the stability of the Arabian oil supply; and the adverse effects of externalities generated by fossil fuel production, distribution, and consumption are fueling the move towards sustainable energy sources like wind generated electricity. Wind energy is becoming one of the preferred substitutes to fossil fuels because of the widespread availability, sustainability, and renewability of wind resources. However, the economic viability of a wind energy generation project is strongly dependent on the accurate characterization and estimation of the wind resource (wind speed) and its associated uncertainty. Analyzing wind resources is a complicated process due to the spatial variability, uncertainty and the complexity of the meteorological processes underlying the formation and behavior of wind. The objective of this work is to employ the two most common nonlinear kriging algorithms (indicator kriging and probability kriging) and five variogram models (spherical, exponential, circular, Gaussian, and hole effect) to characterize and interpolate wind data (in both vector and raster format). The ten sets of variograms and nonlinear kriging algorithms were compared to discover which set is best suited for the characterization and interpolation of each type of wind data used in wind power generation projects. The kriged data was subjected to leave-one-out cross-validation and the resulting statistics were employed in the ranking and comparison of the ten sets of algortihms. The research used ten combinations of nonlinear interpolation methods and variograms and determined that the best nonlinear kriging algorithm for characterizing and interpolating the vector wind data was indicator kriging and exponential variogram. Using indicator kriging with either the spherical or the exponential variogram would result in the same estimates for the raster data.