Mixed-integer optimization for Volt-VAr control in distribution networks.

Abstract

As the load demand persists to increase globally with the growing capacity of distributed generation, distribution systems witness various voltage violation problems. In this context, designing strategies to ensure voltage profile enhancement is a significant challenge in the current operation of distribution networks. Volt-VAr control (VVC) is a major function that is employed by distribution management systems to manage the voltage magnitudes throughout the distribution system. VVC also serves other secondary objectives such as the minimization of the real power loss in distribution networks. To address its objectives, VVC periodically adjusts capacitor switches, transformer taps, and the reactive power set-points of distributed generation. This thesis builds on solving the VVC problem using mixed-integer conic programming (MICP) in radial and meshed networks. To speed up computations and alleviate voltage violations in meshed networks, the thesis proposes solving the VVC problem using a discrete coordinate-descent algorithm, starting from a solution to the continuous relaxation of the VVC mixed-integer conic program. The optimality of such an approach is investigated by evaluating the gap relative to the MICP objective function value. Numerical results are reported on radial and meshed test distribution networks with up to 3146 nodes. The obtained results demonstrate the superior performance of discrete-coordinate descent (DCD), when initialized by solving a continuous relaxation of the MICP, against the DCD algorithm with the classical initialization from the current operating point. Although the DCD is a local search method, the proposed approach yields an acceptable VVC solution with improved computational performance in various distribution networks.