Power Trading Price Forecasting using an LSTM-GRU Ensemble and an XGBoost Residual Correction Model

Key Points

• The aim is to enhance the accuracy of power price predictions to stabilize energy markets.
• Developed a hybrid model combining LSTM-GRU ensemble and XGBoost for residual correction.
• Predicted main trends first and then corrected the final predictions by modeling residual errors.
• Compared performance against traditional time series and other hybrid models.
• Achieved RMSE of 10.9843 and R² of 0.9282, indicating strong prediction accuracy.
• Outperformed traditional models and complex hybrid models with significant reductions in prediction error.
• Reduced prediction error by approximately 5.21% compared to a single LSTM model.

Abstract