Predictive maintenance on wind and photovoltaic assets

As part of its operations and maintenance of wind and photovoltaic farms, our client wished to set up a set of tools and analyses to monitor the health of various mechanical and electrical systems with predictive maintenance purposes.

Alongside the Operations, Diagnostics and Maintenance teams, our consultants contributed to the development of fifteen use cases related to the condition-based and predictive maintenance of wind and photovoltaic systems. Therefore, multiple approaches were used (Statistical analysis, Mathematical Modeling, Machine Learning) to design and develop the following analytical and algorithmic tools:

• Anomaly detection: identification and characterization of abnormal patterns in production and operational data assets (e.g., frost detection on wind turbine blades, production system underperformance)
• Assets failure prediction models: prediction of blade breakage risks, prediction of rotor and generator bearing degradation)
• Digital Twins on electrical systems and subsystems (e.g., Generator Thermal diagnostic)
• Key components remaining useful life estimation: stop/start cycles study in order to anticipate bearing degradation)

The integration was carried out in the existing client’s infrastructure: PI by OSISOFT, for some algorithms and in RESTFUL microservices using the PI web API for others.