New needs and emerging societal constraints have put the emphasis on the inadequacy of the actual electrical grid. Indeed, it is impossible, or at least very hard, to 1) integrate renewable energy sources at a great scale within the actual electric grid, 2) enable communications between the various power suppliers and consumers, 3) design several different services that meet the needs of a wide range of end users. A key solution to these issues consists in using Smart Grids (SG). SG can efficiently control power flows by means of Information Technology (IT). Technically, a SG consists of a power system and a bidirectional communication system. Multi-Agent Systems (MAS) constitute a possible technology that can be applied to control and monitor the operation of power grids. Moreover, MAS exhibit distribution, adaptive and intelligent features. The goal of this paper is to propose a framework of qualification and evaluation for comparison SG approaches. First, a set of features of importance for smart grids definition is identified. Then, in a second step, some criteria are given to evaluate the impact of SG on the society. Finally, these features are applied to existing MAS approaches addressing SG in order to understand and compare their different contributions.