A comparative Study of Multiple Deep Neural Network Architectures and XGBoost Model in Electricity Price Forecasting

In addressing the complexity of electricity market price fluctuations, this study reviews various electricity price forecasting methods and their application effectiveness. By employing time series analysis and principles of machine learning, a comparison is made between traditional methods such as ARIMA and VAR, and machine learning algorithms including XGBoost and LSTM. This approach theoretically enhances the ability of electricity price forecasting to capture nonlinear data and multidimensional complex interactions. Experiments focused on electricity price prediction demonstrate that machine learning methods outperform traditional techniques in terms of accuracy. Specifically, XGBoost achieves efficient data fitting, LSTM excels at handling long time series data, and the CNN-LSTM combined model is particularly adept at capturing both local and global features of price changes, leading to further improvements in predictive accuracy. These findings indicate that the proposed methods exhibit significant advancements in forecasting precision.