Optimising bat-friendly curtailment algorithms for wind turbines

Wind turbines worldwide cause high mortality in long-lived bats through collisions. To mitigate these impacts, operators commonly apply blanket curtailment by stopping turbine blades under low wind speed and mild temperature, when bat activity is often high and energy production is low. Nevertheless, this approach shows highly variable effectiveness. More advanced curtailment tools exist, but evidence for their efficiency and practical implementation remains limited. Using acoustic monitoring data collected over 25,588 nights at 123 wind turbine nacelles in France, we trained algorithms to predict bat presence probability from weather conditions, the time of year and time of night. We evaluated algorithm performance in simulated curtailment scenarios by estimating the proportion of bat acoustic activity protected and energy loss when turbines were stopped above a given predicted-probability threshold. We compared several implementation scenarios of algorithms, including whether site-specific data were included in model training and whether curtailment thresholds were defined at national, seasonal, or site levels. Simulations indicated high efficiency across year-round applications. The simplest scenario, which uses no site-specific training data and a national curtailment threshold, produced results similar to more restrictive approaches. For example, to protect 90% of bat activity through algorithm-based wind turbine shutdowns, mean estimated annual energy losses ranged from 5.2% to 7.6%. Furthermore, restricting curtailment to mid-April to mid-November reduced these losses to 3.5-4.9% while maintaining the same level of bat protection. Algorithms could provide an effective tool for reducing bat exposure to wind turbines, although field validation is needed to confirm real-world performance.