When investigating physical systems, there is a growing need to perform increasingly complex and cross-domain computer simulations, which require suited methods to describe hybrid simulation models with both continuous and discrete-time dynamics. For example, a simulation model of a production facility should be able to incorporate production entities (discrete) as well as energy flows (continuous). However, implementing heterogeneous hybrid simulation models is still a challenge. Two alternative approaches for this problem, both of which promise different advantages and drawbacks, are investigated and evaluated in this thesis. The first and more common approach pursues coupling of several simulation tools as part of a co-simulation. This is compared to a novel approach that uses a formal model description based on DEVS (Discrete Event System Specification). Two comprehensive case studies in the context of interdisciplinary simulation of production facilities for energy efficiency investigations are implemented to demonstrate both modeling approaches in practical application. Both case study models include discrete as well as continuous aspects, reflecting components for production equipment, energy system, building services and the building hull. For implementation, modern state-of-the-art simulation tools from research literature are employed: BCVTB (Building Controls Virtual Test Bed) provides a middleware solution for co-simulation, while the MatlabDEVS Toolbox implements a hybrid DEVS simulator. Based on the case studies, relevant modeling aspects are examined and compared how these can be implemented using co-simulation and the DEVS-based approach, such as modeling of discrete persistent entities, communication between components and handling of differential equations. A subsequent evaluation provides a direct comparison of both approaches with regard to relevant criteria derived from state-of-the-art literature, including reusability of model components, modularity, support of simulation algorithms and overall modeling effort. Compared to co-simulation, the DEVS-based approach is able to provide integration of continuous and discrete modeling aspects not just on the data level, but also on the modeling level, which entails several major benefits for model development, including improved modularity of hybrid components, model maintainability and ultimately better reusability. However, DEVS-based modeling currently lacks support for high-level specialized modeling features, thus requiring more effort from model developers for initial implementation.