The aim of this study was to develop a load prognosis model for Uppsala district heating system to be used as a tool for heat production optimization. The methodwas to build three models for the different customer types; housing, industry andoffices and then scale them for the total system using data from Uppsala districtheating system. The heat load consists of two parts, one that is temperaturedependent and one that is dependent of the social behavior of the customers. Thetemperature part was modelled with an ARX model using an outdoor temperatureprognosis as input signal. The social behavior part was modelled using the mean ofthe social behavior from some days before and additionally by distinguishing betweenweekdays and weekends. The outcome was a model that would produce a prognosisfor the heat load for each customer type. The total model for the whole districtheating system was less accurate, but still usable. All models developed are howeverrelying on the quality of the available weather prognosis. The benefit of a precise loadprognosis is to facilitate production planning and optimization. Accurate predictions ofthe heat demand, especially in the case of peak load, will result in better productionplanning and thus cost efficiency.