Distributed constraint optimisation for flexible network management
Dimitrios Athanasiadis, Ivana Kockar, S.D.J. McArthur · 2013
This paper presents a network management approach formalised as a Distributed Constraint Optimization (DCOP) problem, in particular power flow management. In DCOP, a group of distributed agents work to assign variables' values that optimise a set of constraints applied to the variables. This provides a way to optimise a global function through message-passing between asynchronous agents, with certain performance guarantees. Max-sum algorithm, a message passing algorithm will be applied as it is the state of the art for solving DCOP problems. Max-sum algorithm gives a sufficiently good approximate solution which tends to reach optimality but with less computation and communication. To apply the max-sum algorithm the network model will be decomposed as a factor graph which maps directly onto it.