Short-Term Electricity Load Forecasting Model Based DSARIMA

Abstract

Forecasting short-term electrical load is very important so that the quality of the electrical power supplied can be maintained properly. The study was conducted to measure the results of electrical load forecasting based on parameter estimates and the presentation of time series data. It is important to manage stationary data, both in terms of mean and variance. Data presentation is done by determining the value of variance through the Box-Cox transformation method and the mean value based on the ACF and PACF plots. This study considers the pattern of electricity consumption which contains double seasonal patterns. The results of previous studies show the electric power prediction model, the DSARIMA model with a MAPE of 2.06%. The condition of the model used to predict the electrical load still has a tendency not to be normally distributed and it is estimated that there are outliers. Improvements to the AR and MA parameters that meet the standard error tolerance value of 5 percent are increased in this study. The results showed improvement of parameters to predict electrical load with DSARIMA model. The significance of this study was obtained by the MAPE value of 1.56 percent when compared to the actual data.

Short-Term Forecasting of Electricity Consumption Using Fuzzy Logic

“…Experimental results show that this is a practical approach that can significantly improve forecasting accuracy compared to benchmark models. The research of (Mado et al, 2022), proposed the Double Seasonal Autoregressive Integrated Moving Average (DSARIMA) method to predict or forecast the power demand model at PT. PLN Gresik Indonesia is based on three years of training and load testing data (daily data every half hour).…”