Forecasting DK1 Electricity Prices: Comparison and Performance Evaluation of ARIMAX and XGBoost.

Prosad, Manik Chandra

4. semester

Economics and Business Administration (Finance), MSc.

2024

2024-12-01

70 pages

Electricity price forecasting has become increasingly critical in dynamic and volatile markets. Market like Denmark’s DK1 zone is one of them which is mostly driven by the rising integration of renewable energy sources and external geopolitical influences. This thesis investigates the comparative performance of two forecasting models- ARIMAX (AutoRegressive Integrated Moving Average with Exogenous Variables) and XGBoost (Extreme Gradient Boosting) on electricity elspot prices. The analysis integrates both statistical and machine learning approaches, leveraging a dataset containing hourly electricity spot prices DK1 and energy production variables. The analysis begins with pre-model testing for stationarity, autocorrelation and normality of time series data for ARIMAX modeling. Afterwards, the ARIMAX and XGBoost models are optimized through AIC, BIC, grid search, cross-validation, optimal parameter selection and so on. Later, this thesis examines the performance of both models for forecasting electricity spot prices. Focusing on short-term and long-term spot price predictions, this paper highlights the necessity of robust forecasting models to capture the volatile characteristics of electricity prices. The empirical findings reveal that ARIMAX performs highly accurate for in-sample predictions but struggles with out-of-sample generalization where XGBoost outperforms ARIMAX in forecasting accuracy for unseen data.

DK1, ARIMAX, XGBoost, Elspot forecasting

Download
View record in AAU Student Projects