A safe-by-design approach based on safe Bayesian optimization for tuning individual pitch controllers on floating offshore wind turbines

Individual pitch control (IPC) is an effective strategy for mitigating structural loads on floating offshore wind turbines (FOWTs). However, tuning IPC controllers is challenging, as conventional optimization methods risk exploring unsafe parameters. To address this critical safety issue, this paper proposes a safe-by-design framework for the automated tuning of a PI-based IPC controller using safe Bayesian optimization (SafeBO). The framework ensures safety by restricting its search to regions that remain above a predefined safety threshold, guaranteeing zero safety violations across all tested scenarios and defined metrics—a feat impossible for standard optimization methods. The final controller tuned by SafeBO demonstrates exceptional performance. Under challenging sea states, it achieves a 20.3% reduction in platform pitch motion and a 12.8% reduction in tower fatigue loads compared to a conventionally tuned baseline controller, without compromising power production. These findings confirm that the SafeBO framework effectively resolves the safety-performance trade-off, offering a robust and practical solution for the safe, autonomous optimization of FOWT control systems.