This dissertation addresses the problem of optimally scheduling the hoists of a twin rock winder system in a demand side management context. The objective is to schedule the hoists at minimum energy cost taking into account various physical and operational constraints and production requirements as well as unplanned system delays. The problem is solved by first developing a static linear programming model of the rock winder system. The model is built on a discrete dynamic winder model and consists of physical and operational winder system constraints and an energy cost based objective function. Secondly a model predictive control based scheduling algorithm is applied to the model to provide closed-loop feedback control. The scheduling algorithm first solves the linear programming problem before applying an adapted branch and bound integer solution methodology to obtain a near optimal integer schedule solution. The scheduling algorithm also compensates for situations resulting in infeasible linear programming solutions. The simulation results show the model predictive control based scheduling algorithm to be able to successfully generate hoist schedules that result in steady state solutions in all scenarios studied, including where delays are enforced. The energy cost objective function is proven to be very effective in ensuring minimal hoisting during expensive peak periods and maximum hoisting during low energy cost off-peak periods. The algorithm also ensures that the hoist target is achieved while controlling all system states within or around their boundaries for a sustainable and continuous hoist schedule. Copyright
Dissertation (MEng)--University of Pretoria, 2010.