Coal Price Prediction with GOA-AE-LSTM-Attention Based Models
WANG Qi
ZHENG Xiaoliang
Abstract：As a key economic indicator in the energy market,the volatility of thermal coal prices exerts far-reaching impacts on the macroeconomy,energy structure regulation,and related industrial chains.To address issues such as over-reliance on historical data,insufficient consideration of key time points,and dynamic changes in coal price trends,this study proposes a combined prediction model:GOA-AE-LSTM-Attention.The model takes historical coal price data as input,extracts potential time-series features via the autoencoder(AE),captures long-term dependencies through LSTM,and incorporates the Efficient Additive Attention Mechanism to enhance feature perception at key time points.GOA is further used to optimize hyperparameters,improving prediction performance.Cross-regional tests were conducted using coal price data from Caofeidian Port(Hebei),Huainan(Anhui),and Baotou(Inner Mongolia Autonomous Region)to verify generalization ability.Results indicate that the proposed GOA-AE-LSTM-Attention model exhibits excellent prediction accuracy and robustness across the three regions,with stronger adaptability and promotional value compared to other models.
Keywords：thermal coalcoal price forecastingautoencoderEfficient Additive Attention Mechanism
Publication Date：2025-11-20
Online Publishing Date：2025-12-03(First online date of this platform, not the publication date of the document)
Pages：9( 65-73 )