Electricity Price Forecasting in Power Markets Based on Machine Learning

Short-term electricity price forecasting is a critical task in the context of market reform, crucial for the stable operation of the market. With the increasing share of renewable energy in the energy mix, traditional fossil fuel power generation is being gradually replaced, which may have long-term impacts on electricity prices. The high volatility of electricity prices further underscores the importance of selecting a reasonable and effective model for forecasting. This paper uses the electricity price dataset from the German power market and introduces a hybrid forecasting model based on machine learning algorithms. The effectiveness of the VMD-SSA-LSTM combined forecasting model is verified through various experiments. Variational Mode Decomposition (VMD), an advanced signal processing method, decomposes highly volatile price series into components with different frequency characteristics. The Sparrow Search Algorithm (SSA) optimizes the hyperparameters of Long Short-Term Memory (LSTM) networks, thereby improving the model’s prediction accuracy. Empirical analyses are conducted for different seasons, and comparisons are made among single LSTM, VMDLSTM, and SSA-LSTM models. Results show that the proposed combined model effectively forecasts electricity prices across seasons, reducing seasonal interference, with an average Mean Absolute Percentage Error (MAPE) of only 2.75% across the four seasons. The model also performs well in extreme price months, with a MAPE of only 3.05%. Thus, after a series of experimental validations, it is concluded that the VMD-SSA-LSTM model is not only feasible but also exhibits high robustness in short-term electricity price forecasting.