Abstract: For rapidity and strong anti-interference requirements of self-adaptive control of hydraulic support controller under stable support state, a kinematics model of hydraulic support and a transfer function of hydraulic system in valve-controlled hydraulic cylinder were built taking ZY10800/28/63 roof-standing and shielding hydraulic support as a research object, and a RBF neural network controller including speed compensation was designed. Based on the controller, a roof beam position and attitude control system of hydraulic support was proposed, which ensured stability of roof beam height and attitude angle of hydraulic support under supporting condition through controlling length of column cylinder and equilibrium jack. A simulation model of the system was built in Matlab/Simulink and experiments were taken out. The experimental results show that the RBF neural network controller has short response time to step signal and no over-shoot. When impact force is loaded on roof beam, the RBF neural network controller has faster response speed, smaller over-shoot and stronger anti-interference performance than PID controller and sliding mode controller, which can keep roof beam position and attitude in target value with good stability.