Modelo híbrido de predicción de fallas para los aerogeneradores de la Central Eólica Villonaco utilizando inteligencia artificial y datos del sistema SCADA
Wind energy has emerged as a promising alternative to the use of fossil fuels, becoming one of the fastest-growing sources of renewable energy worldwide. Despite its advantages, the efficiency of wind energy is limited by the costs associated with operation and maintenance (O&M), which represent...
Guardat en:
| Autor principal: | |
|---|---|
| Format: | bachelorThesis |
| Idioma: | spa |
| Publicat: |
2024
|
| Matèries: | |
| Accés en línia: | https://dspace.unl.edu.ec/jspui/handle/123456789/29974 |
| Etiquetes: |
Afegir etiqueta
Sense etiquetes, Sigues el primer a etiquetar aquest registre!
|
Afegir etiqueta | Sumari: | Wind energy has emerged as a promising alternative to the use of fossil fuels, becoming one of the fastest-growing sources of renewable energy worldwide. Despite its advantages, the efficiency of wind energy is limited by the costs associated with operation and maintenance (O&M), which represent a significant portion of the total expenditure in a wind farm. The objective of this curricular integration project (TIC) is to implement a hybrid model that combines artificial intelligence (AI) techniques and data from the SCADA (Supervisory Control and Data Acquisition) system to classify faults in the Central Eolica Villonaco (CEV). It is necessary to determine whether the recorded data corresponds to fault data or not. Therefore, the main result is a hybrid model based on Artificial Neural Networks (ANN), Random Forest (RF), and the Blending Ensemble technique called Blending-ANN-RF, whose execution was carried out under the adaptation of the CRISP-DM methodology, which was applied to guide experiments with various algorithms in conjunction with oversampling techniques and different combination techniques, where the Blending technique emerges with special relevance. This strategy achieved outstanding performance in specific evaluations, such as those derived from the confusion matrix, cross-validation, and the Wilcoxon test, in addition to demonstrating reduced training time. These results support the effectiveness and efficiency of the implemented hybrid model, confirming the suitability of the Blending technique to enhance the individual capabilities of the ANN and RF algorithms. Furthermore, promising areas for future research have been identified, including the exploration of new models that leverage SCADA data and more studied components of a wind turbine. These lines of future research promise to further contribute to the field of wind turbine failure prediction. Keywords: machine learning algorithms, fault prediction, oversampling techniques, algorithm combination techniques. |
|---|