Locational Marginal Price Prediction in Electricity Markets Based on CNN-LSTM-Attention

Accurate pricing is essential for effective trading and grid optimization in electricity markets. The Locational Marginal Price (LMP) plays a key role, and its precise forecasting is crucial for optimal decision-making. This paper presents a combined model of Convolutional Neural Network - Long ShortTerm Memory (CNN-LSTM) and several attention mechanisms, including Squeeze-and-Excitation (SE), Efficient Channel Attention (ECA), Convolutional Block Attention Module (CBAM), and Hard-Weighted Attention (HWA), to achieve highly accurate forecasting of LMPs. First, the CNN is employed to extract local features and capture interactions among the three key influencing factors. The LSTM captures long-term dependencies, while the attention mechanisms improve the model’s focus on critical features and key time points. Experimental results demonstrate that the proposed approach significantly outperforms traditional statistical models and standalone neural networks, highlighting its potential for LMP forecasting in electricity markets. This study offers a robust and efficient solution to the complex problem of electricity price prediction.