A Transformer-BiLSTM based Hybrid Deep Learning Approach for Day-Ahead Electricity Price Forecasting
Abdullah Al Ahad Khan, Md Habib Ullah, Ruchira Tabassum, Md Faisal Kabir
Abstract
Accurate electricity price forecasting in smart grids is critical for operational risk management and optimal decision-making in bidding strategies in the day-ahead (DA) electricity markets. However, it is highly challenging due to volatile characteristics, seasonality, rapid spikes, and other nonlinear factors of the price signals. In the given context, deep learning (DL) has gained attention in recent years due to its high potential in nonlinear approximation, but each model has its strengths and limitations. Therefore, this paper proposes a hybrid DL approach for time-series DA electricity price forecasting based on the Transformer-BiLSTM model. Hybrid deep neural networks give the freedom to strategically combine several components to extract patterns and further improve the sequence processing task. Our proposed model utilizes transformer architecture to capture patterns, temporal dynamics, and BiLSTM networks to forecast electricity price fluctuations. Through comprehensive testing against a range of standard models, our approach demonstrated superior forecasting precision. It achieved a Mean Absolute Error of 2.7818 and an R-squared value of 0.9393. These metrics validate the model’s capability to acquire patterns and efficiency in predicting electricity prices, representing an advancement in energy economics.