As the cost of operations and maintenance (O&M) makes up about 30% of the total life-cycle cost, machine learning algorithms have been used to optimize O&M of wind turbines in order to reduce the cost.
With weather measurements and operating records of wind turbines, machine learning algorithms like neural networks can recognize patterns of electricity output in different weather and operating conditions, and then use the patterns to find the optimal settings in various weather scenarios. When similar weather conditions arise, the turbine can adjust its operation to the optimal setting.
Regarding maintenance, machine learning algorithms can help diagnose and predict failures of wind turbines, and further minimize downtime for maintenance and extend turbines’ lifespan. This development can shift maintenance from a calendar-based to condition-based schedule.
Earlier this year, Alphabet’s AI firm announced that it can now generate more accurate wind power forecast 36 hours ahead by applying machine learning techniques to weather forecast and turbine data. Electricity produced by wind at a set of time helps determine optimal time to deliver. Since uncertainty of wind power production decreases, wind power can be scheduled in the energy market and therefore its value in the grid system increases.