Real-time predictive maintenance for wind turbines using Big Data frameworks

摘要

This work presents the evolution of a solution for predictive maintenance toa Big Data environment. The proposed adaptation aims for predicting failures onwind turbines using a data-driven solution deployed in the cloud and which iscomposed by three main modules. (i) A predictive model generator whichgenerates predictive models for each monitored wind turbine by means of RandomForest algorithm. (ii) A monitoring agent that makes predictions every 10minutes about failures in wind turbines during the next hour. Finally, (iii) adashboard where given predictions can be visualized. To implement the solutionApache Spark, Apache Kafka, Apache Mesos and HDFS have been used. Therefore, wehave improved the previous work in terms of data process speed, scalability andautomation. In addition, we have provided fault-tolerant functionality with acentralized access point from where the status of all the wind turbines of acompany localized all over the world can be monitored, reducing O&M costs.