Maneuver-Bounded Optimal Reconfiguration of Distribution Networks via Mixed-Integer Convex Programming: A Screening Approach with AC Validation

This paper addresses the optimal distribution feeder reconfiguration (DFR) problem with an explicit and practical consideration of switching maneuver limitations. While extensive research has focused on losses minimization by optimizing the network topology, most of the existing formulations overlook the operating costs and equipment wear associated with switching actions, creating a gap between academic solutions and industrial practice. Therefore, a novel mixed-integer convex programming (MICP) formulation is proposed which incorporates a constraint limiting the number of switching maneuvers to 2m, where m represents the maximum allowable number of open–close pairs. By solving the model for increasing values of m=0,1,2,…, system operators can systematically trace the Pareto frontier between loss reduction and switching effort, enabling rational decision-making based on marginal benefit analysis. The proposed model combines a quadratic objective, linear constraints, and second-order cone constraints into a computationally tractable framework. Extensive validations on benchmark distribution test systems, including the 14-, 24-, 33-, 69-, 84-, and 136-bus networks, demonstrates that the proposed MICP formulation matches the best-known loss values reported in the specialized literature, achieving 466.43 kW, 318.04 kW, 139.55 kW, 99.59 kW, 469.88 kW, and 280.19 kW in the aforementioned feeders. The maneuver-bounded analysis reveals strikingly different behaviors across networks: the 69-bus system achieves a dramatic 41% loss reduction with just one switching pair, the 136-bus system resolves all voltage concerns and delivers a 10.49% reduction with a single maneuver, and the 84-bus network requires ten maneuvers to obtain a modest 11.7% reduction. The gap analysis confirms that the convex approximation consistently underestimates losses by a bounded margin of 5–9%. Nevertheless, after AC power flow validation, the identified topology reproduces the best-known loss values reported in the specialized literature for all test systems. It should be emphasized that the optimality guarantee applies to the approximated convex model only, and not to the original AC power-flow-based reconfiguration problem. The solution summary further reveals that the 24-, 69-, 84-, and 136-bus systems admit reconfiguration solutions satisfying the 0.95 p.u. voltage constraint, while the 33-bus system reveals that topological changes alone are insufficient to resolve its voltage issues. These results underscore that the potential benefit of reconfiguration depends critically on network topology, loading patterns, and tie switch placement. By allowing operators to identify the most attractive interventions, i.e., those that deliver the greatest loss reduction with the fewest switching actions, the proposed framework bridges the gap between academic optimization and real-world operational constraints. It should be emphasized that the proposed method is conceived as a two-stage screening framework rather than a rigorous reconfiguration optimizer: the mixed-integer convex model generates, for each maneuver budget, a single deterministic candidate radial topology under a convex loss approximation, and each candidate is subsequently validated through an exact AC power flow analysis to establish operational admissibility, particularly with respect to voltage limits. Consequently, the optimality guarantee reported by the solver applies to the approximated convex problem only, and the validity of the identified topologies is established empirically through AC power flow validation rather than theoretically. The computational cost of the two-stage process is dominated by the MICP solve, since only one power flow evaluation per maneuver budget is required, making the framework suitable for practical, daily or hourly reconfiguration studies.