Comparativa del modelo autorregresivo y media móvil para el pronóstico de la demanda eléctrica del Ecuador considerando el criterio de Akaike

Abstract

The objective of electrical load forecasting is to satisfactorily and accurately predict the demand that could increase or decrease in the future. Many engineering applications have accurate and reliable prediction models for electricity demand. Accurate load forecasting assists in planning the capacity and operation of different utilities to reliably supply power to consumers. With the above mentioned, this study focuses on the comparison of autoregressive and moving average models, mainly in the SARIMA regression model that considers the seasonality of load data, providing an evaluation of the model based on the Akaike information criterion. The input data has been divided into two data sets, one annual and one monthly, to build the forecast model. A comparison is made between the estimated and actual data on an annual and monthly basis to check the margin of error. The accuracy of the prediction models has been evaluated using the different error matrices. For the prediction analysis, RStudio and Matlab, software was used, where the final results in the data comparison between actual and estimated data were with a minimum error value of less than 5%.