The Indonesian government announced that the first phase of carbon trading will be implemented in 2023, starting with coal-fired power plants (PLTU). One of the PLTUs in Indonesia, a case study in this research, still identifies greenhouse gas (GHG) emissions manually and has not provided information supporting the efficiency of future carbon trading activities. Based on this, the research focuses on developing a predictive model that can provide information on carbon emission predictions, coal fuel consumption, and gross electricity as a strategy for carbon emission reduction efforts using The Cross Industry Standard Process for Data Mining (CRISP-DM) methodology. The development of the predictive model was built by combining time series models to predict each variable as an input to a regression model to predict carbon emissions. The evaluation used the MAE, RMSE, and MAPE metrics. Model training used five machine learning models: Linear Regression, Support Vector Regression, Decision Tree Regression, Random Forest Regression, and LightGBM. The time series model testing results for PLTU Units 3 and 7 produced the best model, Linear Regression, while PLTU Unit 8 had the best model, Random Forest Regression. Subsequently, the regression model testing results for all units produced the best model, Linear Regression. With this study, PLTU can know the prediction of carbon emissions and plan carbon emission reduction strategies by estimating gross electricity outcomes and considering the predicted results of coal fuel consumption and carbon emission predictions, thus allowing PLTU to assess the impact from both environmental and financial perspectives.INDEX TERMS coal-fired power plants, emissions prediction, greenhouse gases emissions, machine learning.