Sequential Second-Order Cone Programming for AC Load Maximization Problems

Abstract

AC load maximization problems are challenging to solve in their nonconvex form. Second-order cone programming relaxations facilitate an efficient solution but often result in infeasible solutions especially for meshed networks. This paper proposes a sequential convex programming procedure to recover feasibility by augmenting the relaxed problem with linearized branch and voltage angle constraints. Systematic experiments on radial and meshed networks are designed to identify the effective formulations of constraints. The quantitative results certify the efficacy of the proposed procedure in retrieving near-global solutions for a wide range of test cases. Comparison with established nonlinear solvers reveals the computational superiority of the proposed procedure, which is especially important in making timely maximal load delivery decisions.