A Behaviour-Aware Federated Forecasting Framework for Distributed Stand-Alone Wind Turbines
This work provides a practical, privacy-friendly solution for accurate wind power forecasting for operators of heterogeneous distributed wind turbine fleets, addressing data centralization concerns.
This paper addresses the challenge of short-term wind power forecasting for distributed turbines without centralizing data, proposing a two-stage federated learning framework. It first clusters turbines based on long-term behavioral statistics using a novel Double Roulette Selection (DRS) initialization with recursive Auto-split refinement, then trains cluster-specific LSTM models via FedAvg. The framework achieves competitive forecasting accuracy on 400 stand-alone turbines in Denmark, outperforming geographic partitioning and matching strong k-means++ baselines.
Accurate short-term wind power forecasting is essential for grid dispatch and market operations, yet centralising turbine data raises privacy, cost, and heterogeneity concerns. We propose a two-stage federated learning framework that first clusters turbines by long-term behavioural statistics using Double Roulette Selection (DRS) initialisation with recursive Auto-split refinement, and then trains cluster-specific LSTM models via FedAvg. Experiments on 400 stand-alone turbines in Denmark show that DRS-auto discovers behaviourally coherent groups and achieves competitive forecasting accuracy while preserving data locality. Behaviour-aware grouping consistently outperforms geographic partitioning and matches strong k-means++ baselines, suggesting a practical privacy-friendly solution for heterogeneous distributed turbine fleets.