A successive linear/convex programming framework under a voltage–current representation for loss-minimizing optimal power flow

A contraction-certified successive linear/convex programming (SCP) framework is proposed for loss-minimizing optimal power flow (OPF) under a voltage–current representation. At each iterate, a linearized current-injection model is embedded as equality constraints, and a closed-form contraction analysis establishes sufficient conditions for convergence to the exact AC solution; the damping parameter is shown to tighten the contraction certificate, and the resulting subproblems reduce to linear or quadratic programs. On six standard IEEE/pandapower test systems — radial feeders up to 4876 buses and meshed networks up to 1888 buses — the proposed SCP is benchmarked against five baselines (Lin-DistFlow/DOPF, SOCP, SDP, LM-OPF, NLP) and is shown to be the only method that remains feasible and Newton–Raphson–consistent across all cases, while matching the SOCP/NLP optimum to within kilowatt-level accuracy. An on-load tap changer extension reduces NR-validated losses by up to 12%. Solver times range from s on small feeders to under 4 s on systems of several thousand buses.