Uso del análisis de series de tiempo para pronosticar la producción de energía eléctrica de una planta solar fotovoltaica
Files
Date
Authors
Subject
ARIMA
autocorrelation
data science
renewable energy
solar irradiance
ARIMA
autocorrelación
ciencia de datos
energía renovable
irradiancia solar
autocorrelation
data science
renewable energy
solar irradiance
ARIMA
autocorrelación
ciencia de datos
energía renovable
irradiancia solar
Language:
Journal Title
Journal ISSN
Volume Title
Publisher
Instituto Tecnológico de Santo Domingo (INTEC)
Existen varios factores que justifican el uso de fuentes de energía renovable en la producción de electricidad, como por ejemplo la reducción de emisiones contaminantes al medio ambiente y el costo nulo de la fuente primaria. Por tales razones, el uso de plantas solares fotovoltaicas para la generación de energía eléctrica ha crecido continuamente en los últimos años, y estas podrían estar conectadas a la red eléctrica externa o desconectadas de la red. El pronóstico de la producción de energía eléctrica de este tipo de plantas es importante para su gestión, operación y mantenimiento, por lo que en esta investigación se propone un estudio del pronóstico de la generación eléctrica de plantas solares fotovoltaicas utilizando el análisis de series de tiempo con modelos ARIMA, en escala semanal y mensual, haciendo uso de los datos reales de una planta solar fotovoltaica del Laboratorio Nacional de Energías Renovables de los Estados Unidos. Aplicando la metodología Box-Jenkins1, se consiguen cuatro modelos de pronóstico, dos para los datos semanales y dos para los datos mensuales, para un mismo período de tiempo. Con el fin de evaluar el desempeño de los modelos se obtuvieron las métricas MAE, RMSE y MAPE. Se encontró que desde el punto de vista del MAPE, los modelos con datos mensuales fueron los de mejor desempeño, al ser su valor menor al 10 % para los dos modelos.
There are several factors that justify the use of renewable energy sources in the production of electricity, such as the reduction of polluting emissions into the environment and the zero cost of the primary source. For such reasons, the use of photovoltaic solar plants for the generation of electrical energy has grown continuously in recent years, and these could be connected to the external electrical grid or disconnected from the grid. The forecast of the production of electrical energy of this type of plants is important for its management, operation, and maintenance, so in this research a forecast study of the electrical generation of photovoltaic solar plants is proposed using the analysis of series of time with ARIMA models, on a weekly and monthly scale, and making use of real data from a photovoltaic solar plant from the United States National Renewable Energy Laboratory. Applying the Box-Jenkins methodology, four forecast models are obtained, two for weekly data and two for monthly data, for the same period. To evaluate the performance of the models, the MAE, RMSE, and MAPE metrics are obtained. It was obtained that from the MAPE point of view, the models with monthly data were the best performing, since their value was less than 10% for both models.
There are several factors that justify the use of renewable energy sources in the production of electricity, such as the reduction of polluting emissions into the environment and the zero cost of the primary source. For such reasons, the use of photovoltaic solar plants for the generation of electrical energy has grown continuously in recent years, and these could be connected to the external electrical grid or disconnected from the grid. The forecast of the production of electrical energy of this type of plants is important for its management, operation, and maintenance, so in this research a forecast study of the electrical generation of photovoltaic solar plants is proposed using the analysis of series of time with ARIMA models, on a weekly and monthly scale, and making use of real data from a photovoltaic solar plant from the United States National Renewable Energy Laboratory. Applying the Box-Jenkins methodology, four forecast models are obtained, two for weekly data and two for monthly data, for the same period. To evaluate the performance of the models, the MAE, RMSE, and MAPE metrics are obtained. It was obtained that from the MAPE point of view, the models with monthly data were the best performing, since their value was less than 10% for both models.
Description
Type
info:eu-repo/semantics/article
info:eu-repo/semantics/publishedVersion
Art.
info:eu-repo/semantics/publishedVersion
Art.
Source
Science, Engineering and Applications; Vol. 6 No. 1 (2023): Science, Engineering and Applications; 9-32
Ciencia, Ingenierías y Aplicaciones; Vol. 6 Núm. 1 (2023): Ciencia, Ingenierías y Aplicaciones; 9-32
2636-2171
2636-218X
10.22206/cyap.2023.v6i1
Ciencia, Ingenierías y Aplicaciones; Vol. 6 Núm. 1 (2023): Ciencia, Ingenierías y Aplicaciones; 9-32
2636-2171
2636-218X
10.22206/cyap.2023.v6i1