Estimating anthropogenic CO2 emissions from satellite observations contributes to transparency in CO2 emissions reporting. In this study, we proposed a method for calculating CO2 emissions using Orbiting Carbon Observatory-2 (OCO-2) XCO2 (the column-averaged CO2 dry-air mole fraction) observations. We identified local XCO2 plume enhancements on the OCO-2 track and retrieved CO2 emissions through minimizing the difference between the XCO2 enhancement identified from OCO-2 and the XCO2 enhancement simulated by the Gaussian plume model based on emission rates provided by Open-Data Inventory for Anthropogenic Carbon Dioxide (ODIAC). Among 473 cases successfully retrieved from the OCO-2 from September 2014 to June 2023, the average hourly estimated CO2 emissions (OCO-2 emissions) are 6166.52 tCO2/h with the uncertainty of 1604.51 tCO2/h. Comparison with ODIAC, EDGAR (Emissions Database for Global Atmospheric Research) and MEIC (Multi-resolution Emission Inventory for China) emission inventories shows that the OCO-2 emissions are slightly overestimated. Globally, OCO-2 emissions are approximately 1.25 times that of ODIAC and 1.22 times that of EDGAR, while in China, they are 1.39 times that of MEIC. By conducting comprehensive assessments encompassing case studies in California, Riyadh, and Xinjiang, and by comparing the results with relevant investigations, this discrepancy can be attributed to missing statistics on emission inventories, actual emissions varying with the time of day, or complex terrain and other factors. This study effectively quantifies CO2 emissions using OCO-2 satellite data and provides valuable insights for monitoring significant regional CO2 emissions.