Natural gas hydrates are often considered as a hazard in hydrocarbon production and pipeline transport, but are also recognized as a potential energy resource. Naturally occurring hydrate deposits exist under suitable conditions of low temperature and high pressure, typically in shallow marine sediments and in and below the permafrost, and may host large quantities of recoverable natural gas. Besides, the process of exchanging hydrate guest molecules potentially offers the added benefit of storage of carbon dioxide, as \ce{CO2}-hydrates are, under typical conditions, thermodynamically more stable than \ce{CH4}-hydrates. Several short-term field pilots have been undertaken, but the longer term dynamics of natural gas hydrate reservoirs remain unclear. Core-scale experiments performed at the University of Bergen were designed to provide essential data for field-scale numerical simulation of hydrate formation, dissociation and guest molecule exchange processes. Implementation of experimental findings into a simulator for hydrate-bearing formations may lead to a better understanding of the dynamics of hydrate-bearing geologic media and, in turn, to more effective strategies for commercial use of hydrate reservoirs. This thesis describes the development of an extension of the DARTS general-purpose reservoir simulator for hydrate-bearing reservoirs. The complexity of such systems can be effectively captured using the operator-based linearization approach and experimentally observed behaviour of hydrate systems can be implemented and matched with relative ease. This enables the combination of conventional conservation equations with the use of complex physics and empirical models. The existing DARTS-framework has been extended to include a kinetic description of the hydrate formation and dissociation reaction. Ultimately, a non-isothermal model has been tested for 1-D formation and dissociation cases. Comparison of results with literature shows similar behaviour and trends. The effect of water salinity has also been captured by the model, which has to be validated with experimental results. However, further development of the model is required in order for the model to be able to match measured data.