Structural health monitoring of offshore wind turbines is challenging due to the harsh marine environment, where physical sensors suffer from high salinity, strong currents, and maintenance difficulties. To address these issues, this study introduces a virtual strain sensor based on the PatchTST model, designed to capture both local and global patterns in multivariate time series data. The proposed model incorporates pre-training and fine-tuning with real-world strain data from the Alpha Ventus wind farm. Evaluation through performance metrics confirms the model’s superior accuracy and reliability. These findings highlight virtual sensors as a reliable and cost-effective alternative, enhancing the sustainability and performance of offshore wind energy systems.