Utilização de técnicas de deep learning para predição de desbalanceamento de massa em rotores de turbinas eólicas

Abstract

A source of clean, renewable and permanently abundant energy, wind electric energy has been increasingly sought after all over the world, both in academia and in the business sector, due to its very low environmental impact, it has become increasingly an excellent option. alternative energy production to conventional modes of energy generation. Brazil has an enormous potential for expansion thanks to its geography that is very favorable to the use of winds, in the Northeast region, for example, it has the capacity to produce up to 75GW of wind energy, equivalent to half of all energy produced by the entire country. According to Abeeólica, Brazil supplies 11.5% of the country's total energy matrix, reaching 20% on days of high production. Despite all its advantages, it still leaves a lot to be desired in relation to its efficiency, many researches have been carried out driven by the great increase in interest in this type of energy production. In this work I seek to add to these researches carried out in the field a study developed with artificial intelligence using two different neural networks: a convolutional neural network (CNN) and a short and long term memory (LSTM), for an investigation of the blade mass imbalance. of wind turbine rotors. The final purpose of this work is to bring a way capable of enabling a categorical and competent classification of mass unbalance in the rotor speed, for the better functioning of the system, acting in a more precise, constant and less harmful way to the operation.