How edge AI helps prevent bird collisions with wind turbines

Each year in France, collisions with wind turbine blades result in the deaths of approximately 56,000 birds, posing not only a serious ecological challenge but also exposing operators to potential fines of up to €3,000 per day under species protection regulations. To address these issues, a project led by ALTEN and Schneider Electric leverages embedded computer vision and Green Edge AI technologies, enabling real-time detection and prevention of bird collisions. This initiative empowers wind farms to implement immediate and effective measures, protecting biodiversity while ensuring regulatory compliance and operational efficiency.

dBird is an edge computing solution that harnesses smart cameras and embedded AI to prevent bird collisions with wind turbines. By using computer vision, the system detects and classifies bird activity, enabling wind turbines to respond and regulate operations in real time.

One of the main technical challenges is achieving accurate, long-distance bird detection on a resource-limited microcontroller. To overcome this, we fine-tuned advanced AI models using a robust, high-quality dataset, and applied sophisticated model compression techniques. This ensures the solution delivers high detection performance while maintaining low energy consumption, fully aligned with GreenAI principles. A custom decision-making algorithm further enhances the system by optimizing turbine responses based on real-time environmental conditions.

Multi-sensor integration with data flow management and data fusion on iPC:

• Agricultural activity detection on STM32H7

• Bird activity detection and tracking on Jetson Nano

• Extension to the new STM32N6

The dBird project focuses on detecting and classifying various bird species, each with different levels of protection requirements.

RGB imaging sensor