基于卫星图像的厂区尺度温室气体排放核算模型：以制浆造纸企业为例

Satellite Imagery-based Greenhouse Gas Emission Accounting Model at the Factory Scale: A Case Study of Pulp and Paper Enterprises

关键词:  卫星图像  机器学习  厂区尺度  温室气体排放  制浆造纸企业

Key Words:satellite imagery  machine learning  factory scale  greenhouse gas emission  pulp and paper enterprises

摘要:本研究以制浆造纸企业为例，构建了一种融合了卫星图像和机器学习方法的厂区尺度温室气体排放核算模型。首先通过卫星图像将制浆造纸企业分成原生纤维制浆企业、高定量纸张生产企业、低定量纸张生产企业和特种纸企业4类；然后采用Lasso算法，将厂区面积与碳排放进行拟合，4种不同类型企业的R²均在0.74以上。采用碳排放核算模型对全国重点排放单位名单中的112家造纸企业，进行了厂区尺度的温室气体排放核算。结果表明，112家造纸行业重点排放单位之间温室气体排放差异较大，最小排放量为4.45万t CO₂eq，最大排放量达426.01万t CO₂eq，该方法也为厂区尺度的碳排放核算提供了一种新思路。

Abstract:Taking the pulp and paper enterprises as examples， this study constructed a greenhouse gas emission accounting model at the factory scale that integrated satellite images and machine learning methods. Firstly， pulp and paper enterprises were divided into four categories， namely， virgin fiber pulping enterprises， high basis weight papers enterprises， low basis weight papers enterprises， and specialty paper enterprises through satellite images. Then， the Lasso algorithm was adopted to fit the plant area with carbon emissions， and the R² of the four different types of enterprises were all above 0.74. The carbon emission accounting model was used to conduct greenhouse gas emission accounting at the factory scale for the 112 paper enterprises in the list of key emission units in China. The results showed that there were significant differences in greenhouse gas emissions among the 112 key emission units in the paper industry. The minimum emission was 44 500 t CO₂eq， and the maximum emission reached 4 260 100 t CO₂eq. This method also provided a new idea for carbon emission accounting at the factory scale.