Diseño de una metodología basada en redes neuronales artificiales para la predicción de vida útil acumulada del sistema hidráulico de un aerogenerador G52/850 instalado en el Parque Eólico Los Santos

Abstract

In this project, it is developed the design of a methodology based on artificial neural networks, in order to predict the useful life accumulated of the hydraulic system of a G52/850 wind turbine developed at the Parque Eólico Los Santos. It begins with a study of the components that make up the hydraulic system, as well as, the analog signals obtained from it, corresponding to the condition variables that can be monitored thanks to the software SGIPE. Then, Inspections and testing of the different elements of the hydraulic system were carried out, in order to determine the current condition of the components of this system, and based on the variation of the condition variables when a fault occurs, it will determine which of these best represent the useful life of a component. Data are extracted from the selected condition variables from November 2014 to September 2016. The neural network of the hydraulic system was programmed, using the Matlab software, programming in this training and validation of the algorithms. Finally, 80% of the data from the condition variables selected for the training of the network are used and 20% for the validation of the same, obtaining as a result the Neural Network of the Hydraulic System ready for its implementation.