Large inertia, serious delay, nonlinear parameters and complex structure are current problems of process of medium frequency induction furnace, the traditional PID control is difficult to meet the control requirements. As the development of intelligent control theory has been made tremendous results, adaptive PID control program is proposed based on the RBF neural network, using neural network identifier online identification system model to get PID controller parameters auto-adjusting real-time and on-line. So the control system would be able to maintain the indicators in ideal range, having a better balance between the robustness and controllability and improvement of performance between dynamic heating process and static. What's more, the system's capability of anti-jamming has made enhancement. And MATLAB simulation is based on sample data from some experiments, resulting to achieve the desired goal and obtain good control effect.