Wind energy forecasts in calculation of expected energy not served

The stochastic nature of wind energy generation introduces uncertainties and risk in generation schedules computed using optimal power flow (OPF). This risk is quantified as expected energy not served (EENS) and computed via an error distribution found for each hourly forecast. This thesis produces an accurate method of estimating EENS that is also suitable for real-time OPF calculation. This thesis examines two statistical predictive models used to forecast hourly production of wind energy generators (WEGs), Markov chain model, and auto-regressive moving-average (ARMA) model, and their effects on EENS. Persistence model is used as a benchmark for comparison. For persistence and ARMA models, both Gaussian and Cauchy error distributions are used to compute EENS via a closed-form solution that reduces computational complexity. Markov chain and ARMA both provide accurate forecasts of WEG power generation though Markov Chain model performs significantly better. The Markov chain model also produces the most accurate EENS estimate of the three models.