We present a framework for maximising the power output of a wind farm considering wake effects. A parameter representing the total power coefficient of a wind farm is introduced, which quantifies the wind speed deficit due to wake. For any given wake model, this parameter is a function of the blade pitch angle and the tip speed ratio of the individual turbines, and is independent of the wind speed. Thus, the variables associated with each turbine can be optimised offline, given the wind farm layout, in order to determine reference set points that maximise power production. An MPC controller is designed for individual turbines to track the optimal reference set points. The performance is illustrated with the turbine simulator FAST and the wind farm simulator Aeolus SimWindFarm.