Data Driven Optimization of Curtailment Strategies Under Turbine Operating Constraints
10 Sep 19
5:00 PM
-
6:00 PM
Tracks:
Data Science and Digitalization
The collection of large meteorological data sets, when combined with on-site sensing, offers unique opportunities to increase power production through real-time optimal control. This talk will address scenarios where a wind site (or portion of it) must adhere to curtailment restrictions due to environmental or wildlife concerns. The case considered here involves a budget on shadow flicker hours, where turbines are constrained to limit flicker to a certain maximum number of hours per year. An optimal control strategy is developed such that the flicker hour budget is expended strategically during portions of the year when power generation is likely to be high. This strategy may curtail turbine operation when flicker occurs but power production is low, thereby saving flicker hours to expend later when power production is likely to be higher. The algorithm described here is based on dynamic programming, a well-known technique used in software such as Google Maps. Long-term met data for a wind site is used to predict cloud cover and wind strength over the course of the year, then combined with on-site met sensor data to create an optimal control strategy. The algorithm is not limited to shadow flicker and can be used anytime a curtailment budget is enforced due to external factors (such as wildlife concerns). The talk will present the problem motivation, mathematical methodology for combining met data with real-time sensing to form an optimal controller, and example results.