Aplicación de la técnica de Machine Learning para la estimación del potencial eólica en el sector de Lasso a partir de mediciones de una estación meteorológica

Wind energy is renewable that does not pollute, it allows to replace the use of fossil fuels, it is an alternative to generate electricity that can supply the demand in sectors of difficult access of the electrical networks for which the design of the algorithm has been carried out applying techniques of Machine Learning allowing to predict the speed of the wind to determine the wind potential based on the measurements made with the weather station, it should be noted that when there were null values, a data filtering was carried out, the record obtained was analyzed using the language of Python programming was used time series analysis using LSTM Network, which is a type of Recurrent Neural Network that allows analyzing historical data, 80% of data was used for training and 20% for validation, with the RMSPROP optimizer better results were generated from training, optimizing the prediction with the real values, in addition to its validation, the mean absolute percentage error (MAPE) was applied, obtaining as a result a 4% value that is within the allowed limit for a correct prediction validation, with the predicted values of the wind speed was applied the Weibull distribution determining the average speed of 2.72 m/s with which it allowed to select the Enair 30Pro wind turbine that has been made in strict accordance with the IEC 61400-2 Standard, with which the extractable average wind potential of 51.53 W was determined and the energy produced of 522 kWh/year.

• 1. Verfasser: Basantes Tisalema, Juan Carlos (author)
• Format: masterThesis
• Sprache: spa
• Veröffentlicht: 2022
• Schlagworte: MACHINE LEARNING; RED NEURONAL LSTM; ELECTRICIDAD
• Online Zugang: http://repositorio.utc.edu.ec/handle/27000/8811