Abstract:A dynamic model for predicting SCR inlet NOx concentration based on Dung Beetle Optimizer (DBO) ensemble model weighted fusion is proposed to address the issue that a single prediction model cannot maintain prediction accuracy under different operating conditions. Firstly, a hybrid model of CatBoost and LightGBM is used to filter auxiliary variables, obtain the delay time and order information of the auxiliary variables, and determine the input variables of the prediction model based on the above information. Then, an integrated model consisting of LightGBM, XGBoost, and CatBoost was established, and the prediction results were weighted and fused using the dung beetle optimization algorithm. Finally, the DBO integrated weighted fusion dynamic prediction model was compared with three single models and the two weighted fusion prediction models optimized by the dung beetle algorithm. The evaluation indicators of the DBO integrated weighted fusion dynamic prediction model are superior to other models, with higher prediction accuracy, real-time performance, and adaptability, which can better meet the requirements of NOx concentration prediction under different working conditions.