Hydraulic servo control systems pose significant challenges for controller design due to their nonlinear, time-varying of parameters, inertial force, and susceptibility to external disturbances. Traditional PID control methods struggle to set optimal control parameters, necessitating new approaches. In this study, we construct an Expanded State Observer(ESO) to estimate the external disturbance and uncertainty of modeling, a BP neural network to adaptively adjust PID control parameters and an RBF neural network to identify input and output Jacobian information. Our simulations demonstrate superior disturbance rejection and accurate dynamic tracking capabilities, as well as robust stability under inaccurate hydraulic system parameters and models. By regarding the controlled object as an open-loop basic type, we offer a new, effective approach for adaptive control of these challenging systems.