Neural networks for wind turbine supervision

Wind energy converters are generally working under very variable conditions which require high maintenance costs. Therefore, it is important to supervise their behavior in order to provide an accurate and reliable forecasts of the energy production. The aim of this paper is to demonstrate a monitoring system for wind turbines which will result in a cost reduction of maintenance and an improvement of the cost/benefit ratio of wind energy. As it is very difficult to define the faulty behavior, we propose to use autoassociators networks in order to process the information which is available under operating conditions. These networks are used to detect any modification of the behaviour of the wind converter. This system complements the standard monitoring equipment of the wind energy converters to yield detailed on-line information on the state of the machines.