Real time estimation of carbon emissions for industrial users based on load monitoring in advanced metering infrastructure

To reduce carbon emissions, it is important to monitor carbon emissions by industrial users related to electricity. Current monitoring schemes have limited effect in real-time carbon emission monitoring while the development of Advanced metering infrastructure (AMI) and emission factors for carbon-related devices introduces a fresh outlook. Hence, a method based on carbon-related load monitoring in AMI is proposed. The proposed method comprises three essential components: the one-hot component, the random convolution component, and the grid search component. The one-hot component can transform multi-state and multi-device identification into multi-classification problems, making ${\text{CO}}_2$ emissions calculations easier for industrial users. The random convolution component effectively distinguishes the electricity characteristics to identify various states of carbon-related devices, while the grid search component optimizes hyperparameters to enhance recognition accuracy and decrease carbon emission monitoring errors. The effectiveness of the proposed approach is evaluated through experiments conducted on several industrial users. Comparative analysis with alternative methods demonstrates the superior performance of the proposed approach, indicating its effectiveness in accurately estimating carbon emissions for industrial users on these metrics including ACC, ${\mathrm{Recall}}_{\mathrm{micro}}$, ${\mathrm{Recall}}_{\mathrm{macro}}$, ${\mathrm{Precision}}_{\mathrm{micro}}$, ${\mathrm{Precision}}_{\mathrm{macro}}$, $F{1}_{\text{micro}}$ and $F{1}_{\text{macro}}$.