Abstract: Because of the insufficient stability and high time cost of single data-driven models Genetic Algorithm-Back Propagation Neural Network（GA-BPNN）, Genetic Algorithm-Support Vector Machine（GA-SVM）, and Extreme Learning Machine（ELM） to establish linear fusion models of information entropy, a fusion modeling method of information entropy Stacking is proposed. Using actual production data, the gasifier load, feed pressure and flow rate, and cooling water flow rate were used as inputs, while the gasifier outlet temperature, syngas flow temperature and rate, and syngas composition at the outlet of the washing tower were used as outputs. An information entropy Stacking fusion model of the gasifier was established. The Mean Relative Errors (MRE) of the predictions of the information entropy Stacking fusion model on the gasifier outlet temperature, syngas outlet temperature and flow rate, syngas CO content, and H₂ content were determined to be 1.89.%, 0.17%, 0.78%, 0.95%, and 0.71%, respectively. These results are more stable than those obtained by the single data driven model. The fitting speed was about 19% higher than that of the linear information entropy fusion model. Combined with the optimization algorithm, our model can be applied to the online optimization of operating conditions such as oxygen coal ratio in the gasification process, the gasification temperature of the gasifier, and the effective gas yield during the process.