Anomaly detection for wind turbine damaged due to lightning strike

The development of wind energy is now underway globally as a renewable source and has attracted much attention as a cost-effective power generation method. However, blade damage caused by lightning strikes on wind turbines has occurred in Japan, and there is pressure to take measures to mitigate this problem. When blade lightning damage occurs, the damage is magnified by the continued rotation of the damaged blades. To stop the wind turbines before this secondary damage occurs, wind farms, in the regions of Japan vulnerable to winter lightning, are required to introduce lightning detection systems. Typically, the wind turbine stops automatically due to lightning detection and is restarted after the soundness of the blades is confirmed by visual inspection. However, it is often difficult to check the soundness of the blade visually due to bad weather, and the resulting delay in the restart process prolongs the downtime and reduces the availability of the wind turbine. In this paper, we report the results of using a machine learning model and supervisory control and data acquisition system data to check the soundness of the blades after a lightning strike to enable operations to resume sooner, thereby increasing up-time.