Neural networks applications for fault detection on wind turbines

Wind energy is the renewable energy source with a higher growth rate in the last decades. The huge proliferation of wind farms across the world has arisen as an alternative to the traditional power generation and also as a result of economic issues which necessitate monitoring systems in order to optimize availability and profits. Tools to detect the onset of mechanical and electrical faults in wind turbines at a sufficiently early stage are very important for maintenance actions to be well planned, because these actions can reduce the outage time and can prevent bigger faults that may lead to machine stoppage. The set of measurements obtained from the wind turbines is enormous and the use of neural networks may be useful in understanding if there is any important information that may help the prevention of serious failures. The training of the Neural Networks however is not easy because the measurement set used for training must represent a period of time with no faults in the equipment of the turbine that is being monitored.