Production planning in the sense of balancing inventory costs against the cost of making production rate changes is most often accomplished with linear programming applied to aggregate planning models. However, in continuous processing systems, such as those found in the chemical processing industries, linear models are often inadequate because production rates are related to the operating conditions through non-linear models or plant simulations. The costs of production rate changes are, therefore, non-linear. To fully minimise costs a large scale non-linear optimisation must be performed, but, in a long term rolling horizon with a complex process the computing costs of such an optimisation become prohibitive. In order to incorporate variable time intervals that allow for lower costs and decrease the computing time an approach is developed where the optimisation is broken down into three levels involving; the number of rate changes, the production switching times and the production rates. The levels are optimised based on minimum inventory and a model of the cost of production rate changes to obtain an optimum schedule optimum rates are then refined with a NLP using the found schedule.