LAUSR

This study proposes a deep learning-based framework to improve the efficiency and sustainability of LCSCM. Firstly, a multi-scale time series decomposition LSTM (MS-TDLSTM) model is proposed, which combines empirical mode decomposition (EMD) and attention mechanism to capture multi-scale characteristics of carbon emission data. Secondly, a multi-objective optimization model based on deep reinforcement learning (DRL) is designed. Through soft constraint multi-objective reinforcement learning, the prediction and optimization processes are integrated into a unified system, and intelligent decision-making of LCSCM is realized.