Smart grids as distributed learning control

The topic of smart grids has received a lot of attention but from a scientific point of view it is a highly imprecise concept. This paper attempts to describe what could ultimately work as a control process to fulfill the aims usually stated for such grids without throwing away some important principles established by the pioneers in power system control. In modern terms, we need distributed (or multi-agent) learning control which is suggested to work with a certain consensus mechanism which appears to leave room for achieving cyber-physical security, robustness and performance goals.