Posters

Abstract

As automated shut-down on-demand (ASDoD) systems become a popular solution for mitigating at-risk bird fatalities on wind farms, the methods they use should be constantly re-evaluated to drive the technologies further. While many ASDoD systems are currently on the market, one of the common elements they share is their curtailment strategy, in which a buffer zone is established around a turbine. If a protected species enters the buffer zone, the turbine is automatically shut down by the system. To balance species protection and power generation requirements, this buffer zone must be carefully fine-tuned to ensure optimal performance at the wind farm. In this work, we propose an adaptive method for establishing this buffer zone that seeks to further improve this tradeoff. Borrowing from the fields of Bayesian inference and stochastic control theory, we develop a model that tracks and learns an individual bird’s behavior to predict its probability of turbine collision. Seen as an improvement over standard collision risk modeling, this approach enables fewer necessary turbine curtailments, maintains low fatality rates, and can be run on systems in real time. We conclude by simulating this adaptive strategy on real data gathered from IdentiFlight, an ASDoD system that has observed hundreds of different species around the world.