Electricity Price Forecasting based on XGBooST and ARIMA Algorithms

Abstract

Accurate electricity price information is critical for wholesale electricity markets. Machine learning algorithms are thought to be most efficient methods as they successfully observe the dependencies between electricity price, historical data and other factors. In this study, two machine-learning models are purposed for electricity price forecasting. The historical prices and other important factors are processed. Then, the future values of the electricity prices are forecasted using properly fitted XGBoost and ARIMA models. To validate the results, some statistical error measurement methods are selected. Consequently, when comparing the results in terms of performance in detail, XGBoost model has become more efficient as the computation speed and lowest error.