In power system with a high proportion of wind power, the strong volatility of wind power causes drastic variations in the short-term balance of supply and demand, which increases the uncertainty of electricity prices and the difficulty of electricity price forecasting. This paper analyzes the cyclical characteristics of electricity prices, focusing on the impact of wind power fluctuations on electricity prices. Using load and wind power, a new feature is constructed to characterize the impact of variations in electricity generation capacity of other high-cost power generation methods on electricity prices. By combining attention mechanism with time convolutional network, a double-layer multi-head self-attention time convolutional network is established to explore the temporal patterns of electricity prices and the impact of external factors on electricity prices. The forecasting effect is validated using actual data from the Nordic electricity market, and the results indicate that the proposed method reduces the MAE value by approximately 45% compared to existing electricity price forecasting methods.
