Electricity power systems are a major source of carbon dioxide emissions and are thus required to change dramatically under climate policy. Large-scale deployment of wind power has emerged as one key driver of the shift from conventional fossil-fuels to renewable sources. However, technical and economic concerns are arising about the integration of variable and intermittent electricity generation technologies into the power grid. Designing optimal future power systems requires assessing real wind power capacity value as well as back-up costs.

This thesis develops a static cost-minimizing generation capacity expansion model and applies it to a simplified representation of the U.S. I aggregate an hourly dataset of load and wind resource in eleven regions in order to capture the geographical diversity of the U.S. Sensitivity of the optimal generation mix over a long-term horizon with respect to different cost assumptions and policy scenarios is examined.

I find that load and wind resource are negatively correlated in most U.S. regions. Under current fuel costs (average U.S. costs for year 2002 to year 2006) regional penetration of wind ranged from 0% (in the South East, Texas and South Central regions) to 22% (in the Pacific region). Under higher fuel costs as projected by the Energy Information Administration (average for the period of 2015 to 2035) penetration ranged from 0.3% (in the South East region) to 59.7% (in the North Central region). Addition of a CO2 tax leads to an increase of optimal wind power penetration. Natural gas-fired units are operating with an actual capacity factor of 17% under current fuel costs and serve as back-up units to cope with load and wind resource variability. The back-up required to deal specifically with wind resource variations ranges from 0.25 to 0.51 MW of natural gas-fired installed per MW of wind power installed and represents a cost of $4/MWh on average in the U.S., under current fuel costs.