A MFA-SM control scheme based on RBFNN disturbance observer is designed to solve the problem that it is difficult to accurately control the output shaft speed of the valve-controlled electro-hydraulic rotation system during the drilling of surrounding rock due to the non-linear factors such as uncertain parameters, unknown load and external disturbance. First, an improved dynamic linearization method is used to linearize the electro-hydraulic system to an incremental model only related to the system I/O data. Unknown loads and external disturbances are combined into an unknown nonlinear time-varying term. Then, an RBFNN disturbance observer is designed to estimate the nonlinear term on-line in real time, and the time-varying pseudo-gradient parameters are estimated from the system I/O data. Finally, the corresponding controller design is given. The simulation results show that the MFA-SM controller can effectively compensate unknown load and external disturbance. Compared with other methods, this scheme shortens the system adjustment time by 10 to 15 seconds, reduces the maximum overshoot by about 7.4%, and the speed tracking error can converge to zero.