we believe that probabilistic forecasts will become increasingly important as input to operational decisions. Physics-based ensemble simulations [2] and statistical quantile regression [3] are commonly used methods to estimate forecast uncertainty. We are also investigating the use of kernel density estimation for this purpose. Furthermore, we are analyzing different methods for generation of representative wind power scenarios [4] [5] as input to the decision problems in the electricity market.
III. UNIT COMMITMENT AND MARKET OPERATIONS
The commitment of generating units to provide energy and operating reserves is a key procedure for system operators to ensure that sufficient capacity is available to handle unexpected supply and demand deviations in real-time. In ISO/RTO markets in the United States the system operator runs a centralized UC as part of the market clearing procedure. WPF information could enter the UC process either through the bids from the market participants or directly from the system operator’s own forecast [6]. Currently, the market bids are used to clear the day-ahead market, whereas the system operator typically relies on its own forecast for adjustments in unit commitments closer to real-time. Stochastic UC formulations [7]-[12] are being proposed as a means to reduce operating costs in systems with a high share of wind power generation. We are investigating the use of probabilistic forecasts for this purpose. It is clear that the quality of the WPF is of high importance for the robustness of the resulting commitment decisions. An alternative to a stochastic UC is to use probabilistic WPF to dynamically estimate operating reserve requirements [13]-[15]. In comparing the different commitment approaches it is important to carefully consider the implications for the electricity market. A stochastic UC will commit additional resources in an implicit manner. This may distort prices for energy and reserves, and thereby increase the need for uplift payments for generators to cover their full operating cost. In contrast, the use of dynamic reserves would be better aligned with current market designs which have explicit operating reserve requirements and prices. Furthermore, it is important to consider the frequency of commitment decisions and the overall timeline for bidding and scheduling of resources in the electricity market. In general, moving operational decisions closer to real-time will facilitate the use of better forecasts among market participants as well as the system operator.