【摘要】 预测纺织服装业碳排放值可使行业制定合理的减碳政策。为更精确地预测行业碳排放，提出基于改进鲸鱼优化算法(WOA)优化长短期记忆神经网络(LSTM)的预测模型，引入机器学习方法，为探索行业碳减排路径提供依据。首先，使用WOA对LSTM模型的关键参数寻优，采用混沌映射初始化种群和自适应权重2种方法改进算法并构建改进WOA-LSTM模型；然后，测算1990—2020年纺织服装业碳排放，使用STIRPAT模型筛选输入变量，通过对比分析验证模型性能，结合情景设计预测行业碳排放趋势。结果表明：模型预测精确度明显提升，测试集的MAE值为4.868,RMSE值为4.984,MAPE值为0.024;行业将主要依靠技术革新实现碳中和，为研究工业部门碳排放预测问题提供新思路。更多还原

【Abstract】 Carbon emissions prediction in the textile and garment industry makes the industry’s carbon reduction policies more practical. In order to predict the industry’s carbon emissions more accurately, this paper proposed a prediction model based on Long-Short Term Memory neural network(LSTM), which was optimized by an improved Whale Optimization Algorithm(WOA). The machine learning method was introduced to provide a basis for exploring the industry’s carbon reduction path. Firstly, the LSTM model used WOA to optimize the key parameters, and chaotic mapping initialization population and adaptive weights were taken to improve the algorithm. At the same time, an improved WOA-LSTM model was constructed. Secondly, the carbon emissions of the textile and garment industry were calculated from 1990 to 2020, and the STIRPAT model was used to screen the influencing factors of carbon emissions in the industry. Comparative analysis was used to confirm the model’s performance, and several scenarios were used to predict the industry’s carbon emissions trend. Experiments show that the prediction accuracy is significantly improved, and the MAE, RMSE, and MAPE values for the model test set are 4.868, 4.984, and 0.024, respectively. Meanwhile, the industry will primarily rely on technological innovation to achieve carbon neutrality. This study offers new perspectives to research on carbon emissions prediction in industrial sectors.更多还原