Improving Power Plant CO2 Emission Estimation with Deep Learning and Satellite/Simulated Data

2025-02-04

AI Summary

Accurate quantification of CO₂ emissions from power plants in data-scarce regions remains challenging due to limited ground-based monitoring. Method: This paper proposes a satellite remote sensing–deep learning inversion framework that integrates multi-source observations—Sentinel-5P NO₂ column densities, OCO-2/3 in situ XCO₂ measurements, and high-fidelity simulated XCO₂—to construct a spatiotemporally aligned dataset covering 71 data-deficient power plants. A customized U-Net architecture, designed to handle heterogeneous input resolutions, enables the first NO₂-assisted continuous XCO₂ mapping and multi-source synergistic inversion. Contribution/Results: The framework achieves significantly improved emission rate estimation accuracy (32% reduction in validation RMSE versus state-of-the-art methods), supports near-real-time, high-spatial-resolution identification, and has been successfully deployed across representative data-scarce power plants globally. It provides a scalable, reproducible technical solution for carbon regulation and third-party verification.

Technology Category

Machine Learning: Calibration & Uncertainty QuantificationComputer Vision: Remote Sensing / Geospatial AIConstraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

Systems and Infrastructure for Web, Mobile and WoT: Sustainability and carbon-aware systems for Web, mobile, and WoTSecurity and Privacy: Data transparency and provenanceSearch and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

CO2 emissions from power plants, as significant super emitters, contribute substantially to global warming. Accurate quantification of these emissions is crucial for effective climate mitigation strategies. While satellite-based plume inversion offers a promising approach, challenges arise from data limitations and the complexity of atmospheric conditions. This study addresses these challenges by (a) expanding the available dataset through the integration of NO2 data from Sentinel-5P, generating continuous XCO2 maps, and incorporating real satellite observations from OCO-2/3 for over 71 power plants in data-scarce regions; and (b) employing a customized U-Net model capable of handling diverse spatio-temporal resolutions for emission rate estimation. Our results demonstrate significant improvements in emission rate accuracy compared to previous methods. By leveraging this enhanced approach, we can enable near real-time, precise quantification of major CO2 emission sources, supporting environmental protection initiatives and informing regulatory frameworks.

Problem

• Estimating CO2 emissions from power plants accurately.
• Integrating satellite and simulated data for emission analysis.
• Developing a deep learning model for precise emission quantification.

Innovation

• Deep Learning U-Net model
• Sentinel-5P NO2 data integration
• OCO-2/3 satellite observations