Statistical Modelling of Individual Substations in a District Heating System

Abstract: In the rise of digitalization, new possibilities are being discovered for district heating in areas like demand side management and fault detection. For these purposes it is necessary to have reliable models describing the substations. In this thesis, the aim is to develop a group of mathematical models to describe measured quantities and their progress through time, for a well performing substation. For the models to be relevant in applications, they must apply also to other substations. The study explores the possibilities to use the models for fault detection and to track slow drifts in the substations' performance. The method is chosen based on earlier studies on heat load modelling on the system level and all combinations of the endogenous variables heat power and delta-T, and the exogenous variables outdoor and supply temperature are tested. The results show that the best suited model is a SARIMAX (0,1,1)x(0,1,1)_24, for any combination of variables. As heat load patterns of individual substations are random in nature it is impossible to create a model with high detail, but it fits the measurements reasonably well. The model for delta-T is applicable also to other substations than the the reference unit, but the heat power model does not perform as well. A sudden fault simulated on one of the substations could be detected as a deviation from the delta-T model and a slow change in the performance of another substation can be detected by re-estimating the model parameters over time. The results are discussed and some ideas for improvement are suggested.