OptiWindNet RouteSets

OptiWindNet RouteSets is a database comprising cable-routing solutions for 13,954 distinct problem instances derived from operating, planned, and procedurally generated wind-farm layouts. The database is designed to serve as an open benchmark for routing algorithms, a repository of challenging instances, a competitive baseline for reinforcement-learning approaches, and a supervised-learning dataset for the Wind-Farm Cable-Routing Problem (WFCRP).

The solutions were generated with the OptiWindNet package using three solver classes: exact mathematical optimization and two inexact meta-heuristics, namely hybrid genetic search for the capacitated vehicle routing problem (HGS-CVRP) and Lin-Kernighan-Helsgaun (LKH-3). The database covers two collection-system topologies: radial, path-based networks and branched, tree-based networks. Each solution includes both the feasible cable network and associated metadata, including the solver, cable capacity, method configuration, route length, detour overhead, runtime, and, for exact runs, the certified optimality gap. The instances span wind-farm layouts with 50 to 200 turbines, maximum cable capacities from 2 to 12, and diverse spatial geometries. A solution with an optimality gap below 1% is available for 63% of the instances, while 79% have a solution with a gap below 2%.