The influence of different modelling techniques for stratified tank thermal storage systems is analyzed based on the computational time and simulated energy use of the system, when integrated in a district heating network. The different modelling techniques include a fully-mixed tank and two multi-layer models with four or twenty nodes. A district heating network with 35 individual dwellings is considered. The influence of several boundary conditions is investigated: central and decentral storage systems are compared and the network supply temperature is varied (55 ℃ and 75 ℃). The simulations are performed with the Modelica programming language using the IDEAS and MoSDH library. The results show that the simulated energy use is significantly higher when using a fully-mixed model compared to a multi-node model, due to the inherent simplification of uniform temperature and its effect on the tank heat losses in the fully-mixed model. The computational time increase when switching from a fully-mixed to a multi-layer tank with four layers is negligible in case of central storage, while the required time doubles for decentral storage. Additionally, an increase of 16–18% is found in case of a central storage system with twenty nodes and an eight times increase in case of a decentral storage system with twenty nodes when compared to using the fully-mixed model. Overall, a multi-node tank model with a limited amount of nodes gives the best trade-off between accuracy and computational load, albeit with a large increase in computation time for decentral storage systems.