Renewable production assets, such as wind or solar farms or individual wind turbines or PV panels, ideally produce energy following a static mapping from weather to energy.
However, in real life, they do not because of reasons like: Some of the energy-producing units are out for maintenance. Export restrictions prevent full energy production, due to e.g. a general system failure on the export cable, due to grid restrictions or because the produced energy cannot be sold on the market. Wear and tear or system upgrades have caused the mapping to change.
In the project, CITIES ENFOR and Ørsted have worked with performance monitoring of renewable production assets, considering wind farms as an example.
The solution approach is a software solution PMON™ from ENFOR, which Ørsted uses for proactive monitoring in their production today.
PMON™ is a self-learning and self-calibrating software system based on a combination of physical models and advanced machine learning. This combines the best of artificial intelligence with relevant domain knowledge in order to provide a system which automatically can identify faulty assets.
For fault detection, PMON™ can use either a warranted power curve or build an expected power curve, based on historical data, which are then locked down for future fault detection. PMON™ will then identify and issue a warning if an asset starts producing less power than expected, given the actual weather conditions.