Autonomic features, i.e. the capability of systems to manage themselves, are necessary to control complex systems, i.e. systems that are open, large scale, dynamic, comprise heterogeneous third-party sub-systems and follow multiple, sometimes conflicting objectives. In this thesis, we aim to provide generic reusable supports for designing complex autonomic systems. We propose a formalisation of management objectives, a generic architecture for designing adaptable multi-objective autonomic systems, and generic organisations integrating such autonomic systems. We apply our approach to the concrete case of smart micro-grids which is a relevant example of such complexity. We present a simulation platform we developped and illustrate our approach via several simulation scenarios.