A gossip-like distributed optimization algorithm for reactive power flow control

Abstract

AbstractWe considered the problem of minimizing reactive power flows in a smart microgrid. First we modeled this problem as a linearly constrained quadratic optimization, in which the decision variables are the amount of reactive power that compensators inject into the network. Then, we designed a distributed algorithm in which agents are clustered into overlapping subsets according to a given communication graph; at each time, agents belonging to a randomly chosen subset update their states in order to minimize the reactive power flows on the grid. We showed that, by sensing the network at their points of connection, agents can perform this minimization with just the data that they can gather from the other agents belonging to the subset. We characterized the convergence of this algorithm and we studied its rate of convergence.We finally analyzed some specific grid topologies and clustering choices.