The Smart Grid vision has been sparked by the need for a more reliable, efficient, and sustainable energy network. However, new technologies and new policies intended to realize this vision may increase significantly the complexity of the power network. In particular, with greater consumer and supplier choice, and with the introduction of renewable energy resources that are often unpredictable, the range of possible system behaviors — that is, its entropy — may increase dramatically. With proper design, this entropy can be reduced, and the term ‘Smart Grid’ will be justified. A theoretical scaffolding able to deal with uncertainty and dynamics must complement the successful power and energy systems methodologies that are used today. It is necessary to have tools to understand and control complex dynamic systems subject to uncertainty, variability, and shared constraints. These tools are needed to create and evaluate new policies. We must also revisit our ways of measuring success — what do we mean by reliability in a system with increased user participation, and increased load shedding? Methodology from decision & control, and simulation & learning are promising sources of techniques and results to handle uncertainty and dynamics in complex systems, taking into account the cost and reliability constraints faced in energy networks. Several techniques from these fields are surveyed in the paper.