Wind Turbine Control Using Load Sensors for Wind Inflow Estimation

The patent creates a virtual copy of the physical LiDAR measurement system by training a neural network model to replicate LiDAR's wind inflow measurement capabilities using only inexpensive mechanical load sensors. The model learns to map load measurements to wind conditions, producing estimates that copy LiDAR's functionality without requiring the complex optical hardware.

The patent replaces the optical LiDAR measurement system with a mechanical sensing approach using load measurement sensors on the rotor shaft and tower. These mechanical sensors measure forces and moments that are then processed through a neural network to infer wind conditions, substituting direct optical measurement with indirect mechanical measurement combined with computational modeling.

The patent makes the measurement system self-adaptive by training the neural network on site-specific data that captures the unique characteristics of each location's wind patterns and turbine response. The model learns to compensate for local variations and weather conditions automatically, making the system reliable without requiring external calibration or adjustment for different weather scenarios.

The neural network model serves as an intermediary that translates mechanical load measurements into wind inflow estimates. This intermediate computational layer can learn and adapt to different weather conditions and turbulence patterns, providing stable and reliable measurements even when the physical environment changes, unlike direct LiDAR measurements that are immediately affected by weather.

The patent makes the existing mechanical load measurement sensors serve multiple functions: they continue to provide structural monitoring data for safety and maintenance purposes while simultaneously providing input data for the neural network to estimate wind conditions and enable active power optimization. This eliminates the need for dedicated additional sensors and reduces overall system complexity.

The patent implements a feedback control loop where the neural network continuously estimates wind inflow conditions based on load measurements, and these estimates are fed back to the pitch control system to adjust blade angles for optimal power extraction. This closed-loop feedback enables real-time power optimization using only the existing sensor infrastructure combined with intelligent processing.