Motif Analysis in Internet of the Things Platform for Wind Turbine Maintenance Management

Abstract

Wind energy is one of the most competitive renewable energy sources. Supervisory control and data acquisition system provides alarm activations in case of failure, and also signals of the system. Due to the volume and different type of data, these systems require advanced analytics to ensure a suitable maintenance management. Several methods are employed, mainly based in artificial intelligence that involve advanced trainings and elevated computational costs with high possibilities to detect false positives. The novelty proposed in this work is based on motif analysis using an Internet of the Things platform to analyze large time series data for wind turbine monitoring. It is presented an approach considering personalized motifs in specific periods of the signal dataset with more influence in the alarm activation. A real case study is presented analyzing periods before historical alarm activation to forecast relevant trends in time series data. The results obtained with the proposed method provide high accuracy, where this information can be implanted in the maintenance management plan.