| Summary: | The substantial electricity demand data necessitates modelling and forecasting using an extension of the Seasonal Autoregressive Integrated Moving Average (SARIMA), known as Double SARIMA (DSARIMA). DSARIMA is suitable since the electricity data exhibits two seasonalities. However, researchers often employed the multiplicative model to analyze seasonal data without considering the alternative models, additive and subset. Thus, this study incorporates all DSARIMA, additive, multiplicative, and subset models in forecasting electricity demand data while comparing different parameter estimation methods: Maximum Likelihood (ML) and least squares. The dataset from the United Kingdom for the years 2022 and 2023 was utilized in this study. The analysis accounted for seasonal patterns with half-hourly and weekly intervals, represented by 48 and 336 periods, respectively. The results reveal that the subset DSARIMA model with the least square estimation method produces the highest forecasting accuracy with the lowest Mean Absolute Percentage Error (MAPE) value compared to the other possible models for all forecasting horizons, ranging from one to four weeks. This study has demonstrated the significance of considering different DSARIMA models with alternative parameter estimation methods. This, ultimately, ensures more accurate and reliable predictions in double seasonal data.
|
|---|