Abstract : Scientific applications are a real challenge for programmers due to their complexity and the amount of data they manipulate. Their execution requires the use of complex and varied hardware resources such as supercomputers, computing clusters and grids. Software component models and specifically those that offer composition schemas dedicated to high performance computing offer an interesting approach for their development. They make it possible for applications to abstract themselves from hardware resources and to obtain high performances on a vast range of architectures. These models do however remain restricted to a limited set of composition schemas. This leads to a multiplication of models as soon as new composition forms are required. This is a problem for code reuse that is stated as one of the motivations behind software component models. The complexity of modifying existing software component models is also limiting these efforts. This thesis defines a software component model called HLCM in which new composition operators can be introduced without modifying the model. This is possible thanks to the introduction in HLCM of four concepts: hierarchy, connectors, genericity and implementation choice. In order to increase code reuse, this component model is abstract and it is made concrete in specialisations that are based on primitive elements of existing models. During this thesis, we have studied how the concept of genericity can be applied to software component models and we have proposed an approach to implement it. We have studied the problem that arises when combining the concepts of connectors and hierarchy and we have proposed a solution based on the novel approach for the description of interactions amongst components. We have also described an approach for the implementation of HLCM based on concepts originating in the field of model-based software engineering. These propositions have been validated by the development of a prototype implementation of HLCM/ CCM, a specialisation of HLCM based of the CORBA Component Model (CCM). This implementation has been used to describe complex interactions between components (shared memory and parallel method calls) and to ensure an efficient execution on various hardware resources such as the Grid'5000 experimental platform.