Comparative Analysis of ARIMA and SARIMA Models in Electrical Load Forecasting: Insights for Long and Short-Term Projections

An accurate load forecast model is very important aspect in power system planning. It helps the decision makers to cut the expenses, enhance the operational performance and avoid unnecessary investments in power generation, transmission lines and distribution. This paper examines two popular time series forecasting models; the Autoregressive Integrated Moving Average (ARIMA) and the Seasonal Autoregressive Integrated Moving Average (SARIMA) on two different time series data. The US annual net generation data from 1949 to 2007 as well as the daily peak load of 11 kV distribution feeder between 2019 to 2023 was used in the analysis. The effectiveness of each model for long-term load forecasting was examined through error metrics, offering insights into their advantages and practical consequences. The findings show that the use of ARIMA or SARIMA is strongly depend on the nature of data. It is shown that while SARIMA is suitable for the daily peak load data obtained from 11kV distribution feeder as the consumption pattern repeats itself and the seasonality is obvious in the data, ARIMA was better for US net generation data where the seasonality is not obvious.