8%和5%。Based on the physical characteristics of the intake system in a coal-bed gas engine equipped with two-valve type gas mixer, a hierarchical model for steady state air fuel ratio feed-forward control was established using mean value and polynomial regression analysis as well as fuzzy-neural network modelling methods. A 3D original control MAP graph was generated from the model. By means of an identified adaptive neural fuzzy inference system (ANFIS) model of the excess air factor, the simulation of static state air fuel ratio feed-forward control was carried out. Subsequently, the actual engine control experiment was performed with data produced by the feed-forward control model. The results show that the feed-forward control model has good predication ability. Maximal and average prediction errors of the model are less than 8% and 5%, respectively.