For variable speed pump-controlled hydraulic cylinder system, the nonlinear change of hydraulic system parameters is brought in by large-scale change of speed or load. It causes the control system, which is designed by the linear model, to have the problems such as difficult correction of control parameters, unstable precision or even control instability. In this paper, a multi-model adaptive PID (MMA-PID) control method is proposed by analyzing the state space of a typical variable speed pump-controlled hydraulic cylinder system. According to the nonlinear change of the bulk elastic modulus of oil caused by the change of the system pressure, the system behavior is described by using multiple linear sub-models. A reasonable controller is designed for each sub-model. During the control process, the output weight coefficient of each sub-model is estimated separately through the Kalman filter, and the weighted fusion of all the sub-models control output is used as the final control input of the system. The simulation and experimental results demonstrate that when the working conditions are vary widely, the MMA-PID can adapt to the nonlinear change of system parameters better than the traditional PID, and it owns better control effect and dynamic performance.