Smart electricity grids are key to a successful transition towards a future sustainable energy system. Network topology processing (NTP) is an important and frequently reoccurring task in most power grid analysis and simulation applications, including power flow calculations, short circuit and contingency analysis. NTP is applied to a detailed node-breaker representation of the power grid, as provided by the IEC Common Information Model (CIM). It is performed in order to obtain a reduced bus-branch representation, which serves as input to downstream analysis and simulation applications. The present paper introduces a novel approach to NTP that utilises a semantic knowledge base for network model management and exploits semantic reasoning over the formal CIM ontology and instance data. The proposed system is tested on publicly available ENTSO-E data, validated by converting the generated bus-branch model to Matpower and running power flow calculations in a power grid modelling and simulation framework.