Electricity price forecasting is a critical aspect of modern energy markets, with far-reaching implications for both industry stakeholders and consumers. Accurate and timely price predictions are essential for optimizing energy trading, demand-side management, and investment decisions in the power sector. Traditional forecasting models often struggle to capture the intricate patterns and dependencies in electricity price time series data, particularly in the presence of volatile market dynamics and the integration of renewable energy sources. This research introduces a novel approach to electricity price forecasting leveraging the power of deep learning and attention-based transformer models. We propose an architecture that combines the strengths of long short-term memory (LSTM) networks and attention mechanisms within a transformer framework. This attention-based transformer model not only captures temporal dependencies but also learns to focus on the most informative historical data points, making it exceptionally suited for handling complex and non-linear electricity price patterns. Through extensive experimentation on real-world electricity market data, we demonstrate the superiority of our attention-based transformer over traditional time series forecasting models, such as autoregressive models and recurrent neural networks. Our model achieves higher accuracy, offering insights into price trends and volatility on various time scales, from intraday to long-term forecasting. Furthermore, we conduct a comprehensive sensitivity analysis to examine the influence of hyperparameters, dataset variations, and market conditions on the model's performance. These findings provide valuable insights for fine-tuning the model for specific market environments and forecasting horizons.

Deep learning, Electricity price forecasting, LSTM, Transformer.