Collaborative planning for interconnected offshore wind farms using mixed-variable Bayesian optimization

The expansion of offshore wind energy, a cornerstone of the global energy transition, is increasingly hampered by underdeveloped transmission infrastructure. Planning modern offshore grids presents a significant challenge, requiring the co-optimization of network topology and asset capacities across a mix of transmission technologies. To address this, this paper proposes a multi-objective collaborative planning model for interconnected offshore wind farms (OWFs) and their transmission networks. A novel Adaptive Multi-objective Mixed-variable Bayesian Optimization (AMMBO) algorithm is developed to solve this mixed-integer problem efficiently. The key innovation within AMMBO is a Pareto-Guided Variable Freezing (PGVF) mechanism, which accelerates convergence by identifying and fixing promising network topologies before refining their continuous capacity variables in a lower-dimensional space. The proposed algorithm is validated on a modified IEEE 14-bus system. Results demonstrate that AMMBO significantly outperforms benchmark methods, including NSGA-II and an ablation variant without PGVF, by identifying a superior Pareto front with greater diversity and better convergence within a comparable computational time. The PGVF mechanism is shown to dramatically improve computational efficiency, leading to the discovery of higher-quality solutions. The proposed framework provides an effective and efficient tool for planners to navigate the complex trade-offs between economic and sustainability objectives, facilitating the strategic development of robust offshore power grids.