In the industrialized cultivation process of fungi, CO2 concentration control system is a nonlinear, time-delay and time-varying system, which is difficult to establish a precise mathematical model. Considering the situation, CO2 concentration prediction model that based on neural network was built, and a fuzzy controller was proposed further based on the prediction model. Finally, matlab/labview based online forecast model was finished, and it is verified that the prediction system has higher prediction accuracy with robust character. It also provides a new approach to control key environmental factors under more favorable conditions for mushroom growth.