Regresar

Decision-support tools for renewables-rich power systems: A stochastic futures approach

Abstract:

The growing penetration of intermittent renewables (primarily wind and solar generation) in deregulated electric power systems is introducing significant challenges in forecasting generation and scheduling units. At the same time, the pervasive integration of cyber- tools in the control room provides unique opportunities for leveraging data sources like weather forecasts, computational resources, and visualization tools for real-time decision-making. Here, we introduce a framework and algorithm set for day-ahead generation scheduling, or unit commitment, that takes advantage of the close tie between cyber- and physical- resources in the electric power grid. First, we use a class of stochastic automata models known as influence models to forecast relevant spatio-temporal environmental parameters (wind speeds/direction, cloud cover), and in turn simulate probabilistic wind and solar generation futures across a wide area. These models can be parameterized in real time to statistically match publicly-available ensemble forecast products, yet can be tailored to provide generation futures at appropriate spatial and temporal resolutions for scheduling. The models also permit rapid selection of representative renewable-generation futures, and are able to capture local variability and spatial/temporal correlation in the generation profiles. Second, a new method for unit scheduling for the day-ahead market, which uses the probabilistic wind/solar generation futures, is proposed and developed in a preliminary way. A novelty in this approach is a pre-selection step that can provide operators with situational awareness of critical (sensitive) units. The generation-scheduling and unit-commitment tools are demonstrated on a small-scale example, which is concerned with wind generation in the Columbia River Gorge of Washington State on a historical weather day.