A Bi-Level Optimal Siting Method for PV Cluster DC Aggregation Points in Urban Distribution Networks

Urban PV is commonly connected through individual inverters, causing low utilization, uncoordinated reverse power flow, and overvoltage at high penetration. This paper addresses where to site DC aggregation points that collect nearby PV systems on a common DC bus and connect through centralized inverters. This paper formulates a bi-level model under urban land-use constraints as follows: the upper level jointly selects point locations, capacities, and PV allocations to minimize annualized cost and maximum voltage deviation; the lower level routes collectors along roads and verifies AC/DC power flows. PV uncertainty is represented by clustered scenarios, and an improved NSGA-II is coupled with SOCP DistFlow evaluation. On a modified IEEE 33-bus feeder at 122% penetration, the method reduces annualized cost by 8.3% versus two-stage siting and 14.6% versus K-means siting, with maximum voltage deviation reduced from 5.5% to 4.8%. Against conventional AC connection, it eliminates overvoltage and reduces active loss by 40.3% from the no-PV base. Tests on a 69-bus system plus sensitivity, extreme-scenario, and reliability analyses confirm scalability and robustness.