Wind turbine wake reconstruction based on physical information guided neural network (PINN)

It is of great significance for the real-time control of wind turbine and the intelligent monitoring of wind farm to restore the velocity field of the whole wake region from the limited wind measurement data. In this paper, the physical information guided neural network (PINN) is utilized to reconstruct the flow field in the wake region of the wind turbine by leveraging limited lidar measurement data, with the Navier-Stokes equations serving as constraint conditions for the neural network, and the dynamic development law of the wake can be predicted. By analyzing the velocity field of the reconstructed wake, the wake centerline and the velocity on the vertical section, the results show that the PINN method performs well in wake reconstruction and prediction, and has high accuracy and reliability. The maximum velocity deviation of the entire wake reconstruction area does not exceed 1.5 m/s, and the maximum RMSE of the velocity deviation on the vertical section is 0.255 m/s. From the distribution frequency distribution of the wake center, the difference between the wake meander amplitude predicted by the PINN model in the 1D-3D distance and the actual result is small, but the prediction accuracy will decrease with the increase of the downstream distance.