The Importance of Environmental Factors in Forecasting Australian Power Demand

TL;DR

This paper presents a hybrid SARIMA-regression model incorporating environmental factors like temperature and solar exposure to accurately forecast weekly peak power demand in Australia, significantly outperforming traditional models and machine learning approaches.

Contribution

The study introduces a hybrid SARIMA-regression model that effectively integrates environmental variables, demonstrating substantial improvements over basic SARIMA and machine learning methods in power demand forecasting.

Findings

Hybrid model achieves 3.41% MAPE.

Environmental factors improve forecast accuracy by 46.3%.

Model outperforms state-of-the-art machine learning methods.

Abstract

We develop a time series model to forecast weekly peak power demand for three main states of Australia for a yearly time-scale, and show the crucial role of environmental factors in improving the forecasts. More precisely, we construct a seasonal autoregressive integrated moving average (SARIMA) model and reinforce it by employing the exogenous environmental variables including, maximum temperature, minimum temperature, and solar exposure. The estimated hybrid SARIMA-regression model exhibits an excellent mean absolute percentage error (MAPE) of 3.41%. Moreover, our analysis demonstrates the importance of the environmental factors by showing a remarkable improvement of 46.3% in MAPE for the hybrid model over the crude SARIMA model which merely includes the power demand variables. In order to illustrate the efficacy of our model, we compare our outcome with the state-of-the-art machine learning methods in forecasting. The results reveal that our model outperforms the latter approach.