Effectiveness of an artificial intelligence-based system to curtail wind turbines to reduce eagle collisions

Area of Science:

• Wildlife conservation
• Renewable energy technology
• Artificial intelligence applications

Background:

• Wind power projects face challenges in protecting volant wildlife from turbine collisions.
• Automated systems using artificial intelligence (AI) aim to detect, identify, and track wildlife to mitigate risks.
• Effectiveness assessments of these AI-driven wildlife protection systems are limited.

Purpose of the Study:

• To evaluate the identification accuracy of an automated AI system for detecting eagles and non-eagles.
• To assess the rate of virtual curtailments triggered by the system for both target and non-target species.
• To understand the factors influencing the system's effectiveness in a real-world wind project.

Main Methods:

• An automated AI system was installed and monitored at a California wind project for one year.
• The system's performance was assessed based on identification accuracy for eagles and non-eagles.
• Rates of "virtual" curtailment orders were recorded for both intended (eagles) and unintended (non-eagles) targets.

Main Results:

• The system achieved 77% accuracy in identifying eagles and 85% in identifying non-eagles.
• Curtailment orders were significantly more frequent for non-eagle targets (5,439) than for eagle targets (850).
• Misidentification of abundant common ravens as eagles led to a substantial increase in unintended curtailments.

Conclusions:

• The AI system's effectiveness is influenced by species abundance and identification accuracy, particularly concerning common ravens.
• The cost-benefit analysis of such systems depends on biological context, curtailment speed, and operator objectives.
• Balancing energy generation with wildlife protection requires careful consideration of system performance and operational parameters.