Automating machine learning pipelines to reduce cost in renewable energy production

Abstract

Despite the strong world wide growth of the wind power industry, cost-effectiveness re-mains crucial for turbine operators. Un expected wind turbine failures contribute a high share to the total cost for operation & maintenance. By analysing failure data of over 1800 wind turbines, this study identifies the most cost contributing on components and evaluates the possibility of cost reduction with the application of predictive maintenance strategies. The results suggest a high potential of cost savings for maintenance and an even higher potential for down time cost. Further more, this work includes a practical example of ananomaly detection algorithm as a tool for wind turbine failure prediction.