基于组合方法的造纸过程二氧化碳排放区间预测模型的构建

Study on Interval Forecasting Model of Carbon Dioxide Emission from Papermaking Process Based on Hybrid Method

关键词:  造纸过程  二氧化碳排放  区间预测模型  建模与模拟

Key Words:papermaking process  carbon dioxide emissions  interval forecasting model  modeling and simulation

摘要:造纸行业是我国高碳排放的主要工业之一。精准的二氧化碳排放预测对开发以低碳排放为目标的工艺参数优化模型具有重要意义，有助于推动造纸行业的绿色可持续发展。鉴于造纸过程中碳排放的不确定性和波动性较大，本研究提出了基于组合方法的造纸过程二氧化碳区间预测模型，利用变分模态分解（VMD）对原始数据进行信号分解，采用贝叶斯优化反向传播神经网络（BO-BPNN）建立预测模型，最后通过分位数回归（QR）构建区间预测模型。本研究通过采集实际生产数据和建立基于VMD-BO-LSSVM-QR的对比模型进行模型验证。结果表明，所提出的预测模型准确度高，R²为0.993 6，预测区间置信概率为0.892 4，显著优于对比模型，对造纸工业生产具有较高的实际应用价值。

Abstract:The papermaking industry is a major source of carbon emissions in China. Accurate carbon dioxide （CO₂） emission forecasting is crucial for developing process optimization models aimed at minimizing emissions， thereby promoting green and sustainable development. Given the significant uncertainty and high volatility in carbon emissions of papermaking process， this study proposed a CO₂ interval forecasting model based on Variational Mode Decomposition （VMD）， Bayesian Optimization-Back Propagation Neural Network （BO-BPNN）， and Quantile Regression （QR）. First， VMD was applied to decompose the raw data signals. Then， a BO-BPNN based model was developed for CO₂ emission forecasting， followed by the construction of the interval forecasting model using QR. To evaluate the model’s performance， actual production data from a paper mill were collected， and a comparative model based on VMD-BO-Least Squares Support Vector Machine （LSSVM）-QR was established. The results indicated that the proposed model achieves superior accuracy， with R² of 0.993 6 and a prediction interval confidence probability of 0.892 4， outperforming the comparison model. This model demonstrated significant practical value for papermaking industry.