Wind power forecasting and its applications to the power system

The goal of research in this dissertation is to bring more wind resources into the power grid by mitigating the uncertainty of the current wind power, by developing a new algorithm to respond to the fluctuation of the future wind power, and by building additional transmission lines to bring more wind resources from a remote area to the load center. First, in order to overcome the wind power uncertainty, the probabilistic and ensemble wind power forecasting is proposed to increase the forecasting accuracy and to deliver the probability density function of the uncertainty. Accurate wind power forecasting reduces the amounts and cost of ancillary services (AS). As the mismatch between the bid and actual amount of delivered energy decreases, the imbalance between supply and demand also decreases. If the forecasting ahead is increased up to 24 hours, accurate wind power forecasting can also help wind farm owners bid the exact amount of wind power in the day ahead (DA) market. Furthermore, wind power owners can use the parametric probabilistic density of error distributions for hedging the price risk and building a better offer curve. Second, a novel algorithm to generate many wind power scenarios as a function of installed capacity of wind power is proposed based on an analysis of the power spectral density of wind power. Scenarios can be used to simulate the power system to estimate the required amount of AS to respond to the fluctuation of future wind power as the installed capacity of wind power increases. Scenarios have statistical characteristics of the future wind power that are regressed as a function of the installed capacity of wind power from the statistical characteristics of the current wind power. This algorithm can generate many possible scenarios to simulate the power system in many different situations. Third, optimal transmission expansion by simulating the power system with the multiple load and wind power scenarios in different locations is planned to prepare the preliminary result to bring more wind resources in remote areas to the load center in Texas. In this process, the geographical smoothing effects of wind power and the stochastic correlation structure between the load and wind power are considered. Furthermore, the generalized dynamic factor model (GDFM) is used to synthesize load and wind power scenarios to keep their correlation structure. The premise of the GDFM is that a few factors can drive the correlated movements of load and wind power simultaneously, so the scenario generation process is parsimonious.