Under the condition of considering weather factors, combined with the machine learning model, the multi-objective optimization method is used to optimize the production operation of cement rotary kiln, reduce the energy consumption of cement rotary kiln, and reduce pollutant gas emissions. The variables in the production process of rotary kiln are divided into operating quantity, non-operating quantity and non-control quantity, where the non-control quantity includes weather data. Firstly, the non-operand quantity was taken as the predicted quantity, and the operand quantity and the non-control quantity were used as the characteristic quantities, and the cross-validation recursive feature elimination method was used to select the features. Secondly, using 8 machine learning models, the GridSearchCV method was employed to optimize hyperparameters, and the normalized root mean square difference was used as the evaluation measure to select the optimal model. Finally, the constraints such as non-operands weights and monotonicity are set to form the optimization objective function, and the optimal operands are obtained by using the multi-adaptive operator differential evolution algorithm. The results show that considering the weather factors, the operation volume after optimization can reduce the current of the rotary kiln by 16%, the current of the high-pressure fan by 10.53%, and the concentration of SO2, NO and CO at the end of the kiln can be reduced by 13.79%, 7.65% and 16.12%, respectively.