Future industrial energy systems have to adapt to new requirements, caused by a flexible production environment, the usage of volatile renewable energy sources, and the integration into energy markets. Consider these energy systems as Industrial Cyber-Physical Systems (ICPSs), Digital Twins (DTs) are the key enabling technology for optimizing the operation by providing services for advanced monitoring, diagnosis, prediction, and control. Although the concept of DTs is not new, there are still missing architectural patterns and methods for creating them and their services in the domain of industrial energy systems efficiently. As an outcome, the thesis provides architectural guidelines for creating DTs and their services in the domain of industrial energy systems, considering RAMI 4.0 and utilizing existing OPC UA infrastructures to provide context information and run-time data stored in a federated knowledge graph. Therefore, a generic DT architecture is presented and a method is shown, how existing OPC UA information models can be used to instantiate domain-specific ontologies to provide context information for a DT. Also, an ontology- based data access method for OPC UA run-time data is developed to integrate time series data into a knowledge graph. The query performance of the developed approach is evaluated in comparison with other semantic sensor data integration methods. To showcase the applicability of the stored context information, the information is used to identify a data-driven simulation model automatically. In the end, functional and non- functional requirements and a service framework architecture for DTs are presented. The results of the thesis show the benefits of combining OPC UA and Semantic Web technology in the context of an architectural DT framework by making the DT adaptable to its environment and providing semantic interoperability between the DT’s services.