Extracting data-driven insights from LiDAR and SCADA data for improved analysis of offshore wind farms

Abstract

Offshore wind energy is essential for the global transition to renewable energy. Wind farms, such as those at Anholt and Westermost Rough, are crucial in providing clean electricity. LiDAR (Light Detection and Ranging) and SCADA (Supervisory Control and Data Acquisition) data are particularly valuable for understanding turbine behavior and predicting energy output. Recent advancements in machine learning (ML) and artificial intelligence (AI) have opened new possibilities for analyzing complex wind farm data. This study employs data-driven techniques, specifically XGBoost and Long Short-Term Memory (LSTM) networks, to enhance the analysis of LiDAR and SCADA data from the Anholt and Westermost Rough offshore wind farms. Data-driven filters were applied to enhance the quality of input data, thereby improving the accuracy of the models and reducing noise. XGBoost demonstrated computational efficiency, training faster than Bi-LSTM while achieving an R2 of 0.97 for Anholt and 0.86 for Westermost Rough. While Bi-LSTM successfully captured temporal dependencies, it required significantly longer training times. RMSE and MSE results indicate that XGBoost outperformed Bi-LSTM at Anholt by 6.3% and 12.2%, respectively, whereas both models showed higher errors at Westermost Rough, likely due to data dependency. The residual analysis confirmed tighter error distribution for Anholt, whereas Westermost Rough exhibited higher prediction uncertainties. Wind speed loss analysis revealed that turbines in the selected rows experienced variations, highlighting the impact of local turbulence on wind flow characteristics.