Análisis del potencial de irradiación solar en la hacienda "La Campiña" de la parroquia de Mulaló. Propuesta para aplicación de un algoritmo de optimización basado en machine learning en la predicción de generación fotovoltaica.

The research presents a photovoltaic generation prediction model using Machine Learning and one of its main learning machines "Random Forest", this research took place at La Hacienda "La Campiña" located in the Mulaló parish that belongs to the Latacunga canton, where made the actual data collection of solar irradiation and temperature in the period of the year 2020 with a measuring instrument called Solar Power Meter SM206 pyranometer. The prediction model was created in Python and the variables with the highest incidence. This model makes a prediction of the Power variable, it was trained with the Solar Irradiation and Temperature variables. The process consists of entering the variables first of any day to predict a selected variable of another day, here we take data from January 1 to its training and we predict the 20th day of the same month with the largest amount of variable data to have an error that is between the permissible margins and guarantee the efficiency of the model in the same way the prediction of a full month was made where it was trained with data for the month of March to predict May. The efficiency in the daily prediction is 91.41% and the monthly one is 94.47%, the monthly prediction becomes more effective because it contains a greater number of data for each variable, therefore the training performance is more profitable. , it should be noted that this model uses 1000 Decision Trees.

• Autor principal: Núñez Verdezoto, Marlon Daniel (author)
• Format: masterThesis
• Idioma: spa
• Publicat: 2022
• Matèries: SISTEMA FOTOVOLTAICO; MACHINE LEARNING; ÁRBOL DE DECISIÓN; VARIABLE ESTOCÁSTICA; ELECTRICIDAD
• Accés en línia: http://repositorio.utc.edu.ec/handle/27000/9788