Aplicación de Técnicas de Inteligencia Artificial para la predicción de fallas en un aerogenerador de eje horizontal utilizando los datos del Sistema SCADA

In this work, the prediction of failures in a generator of horizontal axis wind turbine is performed by binary classification using Artificial Intelligence techniques, specifically Machine Learning and Deep Learning. The main objective of this research is to generate a fault prediction model by analyzing the data collected by the SCADA (Supervisory Control and Data Acquisition) system. First of all, the data analysis was performed with descriptive statistics and data visualization techniques, which allowed a good selection of the variables, the preparation of the database, and the use of different resampling techniques to eliminate the problem of class imbalance. As a result of the analysis, it showed that the random subsampling presents better results for the classification with the Decision Trees algorithm, so the balanced set in its failure classes and normal state is applied in the next part of the study. Next, in order to effectively apply Machine Learning and Deep Learning techniques, it is necessary to properly select those that are best suited to the classification task, and then, once their hyperparameters have been adjusted, compare their performance with respect to the one that best predicts failures, transcending the Extra Trees algorithm and convolutional neural networks (CNN). although the latter works with a different data set, reaching a Recall value of 0.79, which is not very efficient. Finally, the evaluation of the models is performed with metrics derived from the confusion matrix, especially Recall and the area under the ROC curve (AUC). Key words: Machine Learning, Deep Learning, imbalanced classification, confusion matrix, wind turbine.

• Autor principal: Puchaicela Castillo, Jorge Alexander (author)
• Format: bachelorThesis
• Idioma: spa
• Publicat: 2023
• Matèries: MACHINE LEARNING; DEEP LEARNING; CLASIFICACIÓN DESEQUILIBRADA; MATRIZ DE CONFUSIÓN; AEROGENERADOR.
• Accés en línia: https://dspace.unl.edu.ec/jspui/handle/123456789/27657