Control and coordination of a distribution network via decentralised decision making

Active Network Management is the use of IT, automation and control to manage voltage and power flow constraints in distribution networks. A multi-agent system (MAS) is proposed to perform distributed decision making on the dispatch of distributed energy resources to control network voltages within prescribed limits. Agents at each node estimate the effect of their local actions on their node voltage and choose the best solution according to a cost function. Three voltage change approximations were implemented, one used only local information whilst the other two had varying degrees of knowledge about the network. Agents apply the algorithms iteratively so the approximate solutions are refined over time. The voltage control algorithms were tested on a Java based MAS controlling a distribution network simulation. The results show that approximations based on strictly local data require hysteresis to avoid hunting. Approximations with upstream network data achieve a stable solution at all nodes.