Low Complexity Distributed SDP Approach for General OPF Problems with Reactive Power Cost

TL;DR

This paper introduces a low-complexity distributed SDP approach for solving large-scale optimal power flow problems with reactive power costs, achieving near-global solutions efficiently by relaxing subproblems instead of the entire network.

Contribution

It proposes a novel distributed SDP relaxation method that scales linearly with network size and provides conditions for exact relaxation, improving computational efficiency for large OPF problems.

Findings

Complexity grows approximately linearly with network size.

Proposed algorithms find near-global optima efficiently.

Numerical results show shorter computational times than existing methods.

Abstract

Optimal power flow (OPF) problem is a class of large-scale and non-convex optimization problem. Various algorithms are proposed to solve the challenging OPF problem. Recent studies show that semidefinite programming (SDP) can either provide an exact or near global optima for many existing OPF problems. However, SDP-based approaches usually have the complexity growing exponentially with respect to the network size, which may not be suitable for large-scale OPF problem. In this paper, we rewrite the OPF problem as a combination of several non-convex subproblems. We then consider SDP convex relaxation on the subproblems instead of the relaxation on the centralized formulation commonly found in the literature. The formulation leads to new conditions of exact SDP convex relaxation that generalize some existing results. Based on the distributed formulation, we also develop algorithms to find a near global optimum for general OPF problems. A bound on the difference between the near global solution and the optimum is also established. An important feature of the proposed SDP algorithms is that the complexity only grows approximately linearly with respect to the network size. Numerical studies further demonstrate that the computational time of the proposed algorithms need noticeable shorter time than the existing SDP-based OPF solvers.