Time Series Analysis of Electricity Demand Forecasting Using Seasonal ARIMA and an Exponential Smoothing Model

Foresting demand for electrical power is essential for energy sector because it promotes efficient use of resources, a steady electricity supply, and grid stability. Effective electricity demand forecasting is essential for calculating day-ahead and current energy pricing too. The impact of several non-linear factors, such as calendar elements, seasonal patterns, and weather conditions play a crucial role in influencing electricity power demand. To predict the demand accurately based on time series data, it is important to employ effective statistical methods that can account for the influencing factors of electricity demand. In this research, we developed statistical models that could forecast future electricity demand using seven years of time series data. The seasonal and trend components of the available dataset employ Triple Exponential Smoothing (Holt-Winters) and Seasonal Autoregressive Integrated Moving Average (SARIMA). The accuracy of forecasting is measured using Mean Absolute Percentage Error (MAPE) matrix. The findings of this research work indicate that the Holt-Winters Exponential Smoothing technique, utilizing additive trend and additive seasonality based on the data type, is a suitable method for electricity demand forecasting since it yields the lowest MAPE value. Even though the data used to construct the model is specific to Thailand, the approach is universally applicable and can be used for predicting future electricity demand in any other context.