Abstract: Under the national “carbon peaking” and “carbon neutrality” policy, emission reduction in the coal-fired power generation sector is imperative. Improving the quality of carbon emission data and strengthening regulatory oversight are essential measures to ensure effective decarbonization. As a key parameter in carbon emission accounting, the elemental carbon content of as-fired coal is critical for verifying the accuracy of reported data by coal-fired power plants. To address this need, an intelligent verification method for as-fired coal elemental carbon content data is proposed. First, a dataset comprising approximately 1 000 samples of proximate analysis and elemental analysis of typical coals was compiled. Then, a regression-based machine learning model was developed using Gaussian Process Regression （GPR） combined with a heuristic optimization algorithm, trained on the US coal quality dataset. The model achieved high predictive performance, with R² values of 0.989 8 and 0.987 7 on the training and testing sets, respectively, enabling accurate prediction and verification of elemental carbon content. Subsequently, the generalizability of machine learning model was validated using Chinese standard coal samples and coal data from typical Chinese coal-fired units. The model yielded a mean relative error of only 1.68% on standard coal samples and an R² of 0.987 7 on unit data, demonstrating its accuracy and applicability. Finally, the model was deployed in a 600 MW coal-fired power plant in China, where it achieved a mean relative error of 0.79% between predicted and measured values, enabling timely and accurate monitoring of as-fired coal carbon elemental content at the shift level, and supporting reliable elemental carbon content data reporting for emission accounting.