Geometry-Guided Optimization for Offshore Wind Farm Collection System Routing via Adaptive Circuit Partitioning

Offshore wind farm collection system routing requires cost-effective networks that satisfy coupled electrical and geometric constraints. Existing methods often improve topology search or path feasibility separately, relying on restricted candidates, predefined routes, or post-hoc repair to handle crossings, obstacles, irregular boundaries, and export-corridor conflicts. This paper proposes a geometry-guided optimization method based on adaptive circuit partitioning. Circuit membership is used as the search representation, while each accepted candidate is evaluated as a complete routed collection network with feasible paths, electrical admissibility, cable sizing, and the objective value used in this study. Geometry and capacity information therefore guide the search without fixing a predetermined edge set or replacing final routed-network evaluation. The method is tested on three offshore wind-farm cases ranging from 60 to 71 turbines with increasingly difficult geometries, and compared with a genetic algorithm (GA), particle swarm optimization (PSO), differential evolution (DE), and a mixed-integer linear programming (MILP) reference. The proposed method obtains the lowest mean feasible costs in all cases, with values of 86.438, 176.985, and 209.754 million CNY, and shows faster early convergence over the first 25-50 generations. In the strongly concave case, all generic stochastic counterparts fail to return feasible routed networks, whereas the proposed method is feasible in every run. The results indicate improved convergence, feasible objective performance, and practical tractability for geometrically constrained collection system routing.