Stochastic Modeling and Operational Optimization in District Heating Systems

Abstract: Operation of a district heating system is accomplished via a sequence of decisions by the operators controlling the system. These decisions are based on expectations of conditions in the system that are not known at decision time. The operators could be helped by a decision support system that computes predictions of future system variables and suggests appropriate control actions given the available information. This thesis presents a new model that gives a both physical and stochastic description of a district heating system. The model describes both technical and economical information of the system that are important for the control decisions. It is easy to calculate predictions based on this model as well as performing simulations. The ambient temperature is the single most important explanatory variable for the heat demand in a district heating network. A model that can be used to calculate reliable temperature predictions are presented where the full advantage of both local measurements and forecasts from a meteorological institute are utilized. A heuristic approach to the operational optimization problem is presented and it is shown in simulations to be superior to a traditional control, based on a priority scheme. The operational optimization problem is a complex stochastic optimization problem and the heuristic approach gives a solution that can be calculate instantly. An online computer program, EnerPlan, is developed where the described models are used to calculate predictions and simulate alternative future scenarios. The program is currently used in the control room at the Heleneholm power plant in Malmö, Sweden.