It is difficult to satisfy most of the performance targets by using the PID control law only, if the plants are the processes with uncertain time-delay, varying parameters and non-linearity. For this reason a genetic algorithm based neuro-fuzzy network adaptive PID controller is proposed in this paper. The neuro-fuzzy network is used to amend the parameters of the PID controller online, the global optimal parameters of the network are found with a high speed, and the improved genetic algorithm is introduced to overcome the local optimum defect of the BP algorithm. Finally, the simulation experiment of the control method on the tobacco-drying control process is performed. The simulation results demonstrate that this kind of control method is effective.