This project will develop a comprehensive transfer learning strategy to bridge the simulation-reality gap for wind turbines. A Cyclic Generative Adversarial Network (CGAN) will be trained to learn underlying probability distributions from both simulated and real-world data, and will be tested to extrapolate data between these domains under significantly different testing conditions. The proposal will focus on two main objectives: first, to develop a virtual sensing scheme by training the CGAN to bridge the gap between the OpenFast 1.5MW model and the NREL 1.5MW test platform, and testing the trained models generalization capability against various wind inputs and turbines of the same rating. Second, to propose a transfer learning strategy for extreme domain shifts, validating the trained models generalization capability using data from higher-rated wind turbines (e.g., 5MW). By incorporating network pruning and fine-tuning techniques, the proposal aims to enhance the CGAN’s scalability and prediction accuracy for future wind turbine prototypes, e.g. IEA 22MW turbine. This research has the potential to reduce the reliance on costly and scarce real data by leveraging existing simulation models and extrapolating them into real wind turbines.