The control of electrohydraulic systems has been the focus of powerful research over the last decades. Inherent nonlinear behaviour of the hydraulic systems makes them ideal subjects for applying different types of sophisticated controllers. In this work, a simulink model is developed of a proportional valve controlled hydraulic cylinder system. This study considers the application of neural networks to the control of a hydraulic cylinder which is driven by a proportional valve, subjected to variable reference trajectory and variable load conditions. In industrial applications, these types of systems are used widely in robotics, milling machines, in aerotics field with G and motion simulators and military applications. The system model is fully implemented in Matlab's Simulink simulation package and model of each hydraulic component is developed and combined in a library for easy reuse. Bond graph model of the system is also developed. Applying neural network model predictive control (NNMPC) to the fourth order nonlinear system model, position control is performed. Performance of the NNMPC algorithm is also investigated by changing system parameters. Piston load and effective bulk modulus of hydraulic fluid are the most important system parameters which have valuable effects on the hydraulic system dynamics. Simulations show that, neural network model predictive control is successful for position control in spite of different values of piston load and different bulk modulus.