Electricity Market Price Forecasting Model Based on an Improved Long Short-Term Memory Network

Accurate electricity price forecasting is crucial for the operation of power systems and the management of electricity markets. Accurate predictions help market participants formulate reasonable strategies and reduce operational risks. However, the high volatility and uncertainty of electricity market prices make it difficult for traditional time series forecasting methods to achieve ideal results. In recent years, deep learning methods, especially Long Short-Term Memory (LSTM) networks, have shown significant advantages in time series forecasting. This paper proposes an improved LSTM model. First, the model’s generalization capability is enhanced, and overfitting is reduced by incorporating multiple LSTM layers and dropout layers. Second, simulated data containing seasonal components, trend components, and random noise are used to mimic the complexity of the actual electricity market. The dataset is divided into training and test sets, with Mean Squared Error (MSE) as the loss function and the Adam optimizer for training. Finally, by introducing gradually decreasing noise, the model experiences higher and more fluctuating losses in the initial training phase, which gradually converge later, ensuring the model’s stability and accuracy. Case study results indicate that the improved LSTM-based electricity price forecasting model demonstrates high accuracy and robustness in electricity market price forecasting. The predicted values align well with the actual values, validating the model’s effectiveness and superiority.