Integrating Uncertainty in Electricity Price Forecasting: A Probabilistic Model with Dynamic Trading Applications

As electricity spot markets continue to open and competition intensifies, accurately forecasting electricity prices in a volatile market environment has become crucial for market participants to optimize profits and manage risks. However, traditional point prediction models often struggle to capture the nonlinear characteristics and inherent uncertainties of electricity price fluctuations, limiting their practical effectiveness. To address this challenge, this study proposes an innovative probabilistic electricity price forecasting model, P-TiDE, which utilizes a Probabilistic Uncertainty Residual Block (P-ResNet). By integrating Bayesian regularization and Monte Carlo Dropout (MC Dropout) techniques, the model not only provides point predictions of electricity prices but also generates probabilistic prediction intervals that quantify uncertainty, offering more comprehensive decision support. Additionally, this study introduces a dynamic and adjustable trading strategy that incorporates risk premiums, the upper and lower bounds of prediction intervals, and market volatility, aiming to optimize profits while effectively controlling risks. Through empirical analyses conducted on multiple regional electricity spot markets, the robustness and broad applicability of the model are validated, demonstrating its significant practical value in enhancing forecasting accuracy and improving trading strategy adaptability under uncertain market conditions.