A Near-real-time Estimation Method for Carbon Emissions from High-emission Industries Based on Electricity-Energy-Carbon Linkage Model

• Main Authors: Li, Shangze, Kong, Xiangyu, Gao, Bixuan, Liu, Ziyti, Wang, Shuo
• Format: Conference Proceeding
• Language: English
• Subjects: Carbon dioxide; Couplings; electricity-energy-carbon linkage model; Estimation; Feature extraction; Industries; long and short term memory networks; Near real-time carbon emission estimation; Predictive models; supervised clustering; Time-frequency analysis
• ISSN: 1944-9933
• DOI: 10.1109/PESGM52003.2023.10252275

With the aggravation of carbon emissions and global warming, the environmental threat has raised low-carbon awareness in all countries. Due to the time lag of energy statistical data, the current carbon emission accounting methods have limitations such as low frequency and poor real-time ability. Based on the Electric-Energy-Carbon Linkage Model (EEC), this paper proposes a carbon emission estimation method for industries with high carbon emissions, which improves estimation accuracy and real-time performance. Firstly, Supervised Clustering of Variables around Latent Variables (SCV-LV) is used to cluster and analyze the power consumption data of industrial users. In this process, the critical variable group influencing the model's predictive performance can be identified, and the behavior feature data is extracted by Ensemble Empirical Mode Decomposition (EEMD). Then, the Long and Short Term Memory Network algorithm (LSTM) is introduced to construct the EEC model, which increases the efficiency of energy-carbon emission's near-real-time estimation in high-emission industries. The actual dataset of Tianjin was used to verify the accuracy and reliability of the method in estimating real-time carbon emissions. The impact of COVID-19 on the carbon emissions of various industries in Tianjin was also analyzed.