Posters

Abstract

We will talk about a new condition monitoring platform for wind farms, which is preconfigured to be used as SAAS (software as a service). The platform has a complex system of deep learning algorithms that allows evaluating the condition of any monitored component in a wind turbine. The algorithms detect incipient faults of rapid and slow evolution, being extremely precise in the detection and designed to avoid false positives. These algorithms are accompanied by other indicators of the probability of failure of the equipment and an estimate of its remaining life from a historical database of wind turbine failures. This database is a database of failures in equipment shared anonymously by all users of the platform, in such a way that all users get better locations of the status of their equipment. The use of the platform can be done under demand, no data scientists or a complex knowledge are needed. The platform allows the user add more windfarms, more equipment to monitor, select the indicators that wants to use, configure them to detect some specific failures and much more. The other key fact is the failures data base. When a gearbox is replaced, all the reliability data (Time to failure, and others) are recorded in a general data base without recording the origin of the data. It is only recorded the model of the equipment failed and other features. This data is used to evaluate de failure probability of all the fleets.