Presentations

Abstract

Click here for more info For large fleets of similar structures (i.e. a wind farm), the cost of installing a high-end multi-sensor structural monitoring system on each member of the fleet far outweighs any future potential gains. As such only a limited number of well-chosen structures (typically only 10% of the farm) are instrumented with a monitoring set-up. These locations are referred to as “fleetleaders”, where it is assumed that their behaviour is representative for the entire fleet. The behaviour of the other turbines can be derived from these fleet leaders by using trained machine learning models, together with limited low-cost IoT instrumentation in the rest of the farm . Within this project we developed and validated in real operational conditions methods that allow accurately method monitor fatigue progression over the entire windfarms by combining the standard SCADA data, data coming from low-cost accelerometers and making use of state-of-the are machine learning models.