Abstract: The reliability of flue gas flowrate data is a critical prerequisite for achieving accurate carbon emission measurement. However, issues such as the inhomogeneity of the flow field, the selection of measurement points and the clogging of the sampling tubes contribute to challenges to the accurate monitoring of flue gas flowrate. To address these issues in online flue gas flowrate monitoring, this study proposes to establish a multivariate time series Transformer model to achieve online prediction and data diagnosis of flue gas flowrate. Firstly, 25 d of flue gas flowrate data and corresponding operating condition parameters were collected from a typical 660 MW coal-fired unit in China to establish the Transformer flue gas flowrate prediction model. The model demonstrates excellent performance in predicting flue gas flowrate, with R² exceeding 0.9 on both the training and test sets. Its performance was significantly superior to that of non-time series machine learning models such as random forest and artificial neural network, as well as recurrent neural network models such as long short-term memory network. These results indicate that the Transformer-based flue gas flowrate prediction model can accurately establish a dynamic mapping relationship between unit operating condition parameters and flue gas flowrate. Then, a transfer learning approach was adopted to transfer the pre-trained Transformer source model from the 660 MW coal-fired unit to the generalized application of flue gas flowrate prediction in another 630 MW coal-fired unit. Using only 0.5 d of continuous operating data from the 630 MW unit for transfer training, the model accurately predicted the subsequent 24.5 d of flue gas flowrate, achieving R² of 0.84. Results indicate that the transferred Transformer model demonstrates superior performance in flue gas flowrate prediction, while the transfer learning method significantly improves the model’s training efficiency. Finally, based on the favorable performance of Transformer model in the prediction of flue gas flowrate, the model was further used to implement the intelligent diagnosis on flue gas flowrate data of coal-fired units.