Abstract: Aiming at the shortages of effective historical data in actual manufaction， a carbon emission prediction and multi-objective optimization model driven by small sample data was proposed. The Box-Behnken experimental design was used to collect processing data， and then the back propagation neural network was used to establish a prediction model for carbon emissions and processing efficiency， which ensured the prediction accuracy with less historical sample data. Aiming to optimize the total carbon consumption and makespan， the improved gray wolf optimization algorithm and entropy-TOPSIS comprehensive evaluation were used for determining the optimal processing parameters. Finally， the effectiveness of the proposed method was verified by machining experiments. 

Key words: low carbon optimization, small sample drive, improved gray wolf optimization algorithm, entropy-（TOPSIS） comprehensive evaluation technology

摘要： 针对实际生产历史数据不足的情况，提出一种基于小样本数据驱动的碳排放预测和多目标优化模型。通过Box-Behnken实验设计收集加工数据后，采用反向传播神经网络建立面向碳排放和加工效率的预测模型，在保证预测精度的同时有效减少模型对数据量的需求。以总碳耗和总时长为优化目标，采用改进的多目标灰狼算法和熵权-逼近理想解排序综合评价法进行了最优工艺参数决策。加工实验验证了提出方法的有效性。

关键词: 低碳优化, 小样本驱动, 改进灰狼优化算法, 熵权-逼近理想解排序综合评价法