A data-driven analytical model for wind turbine wakes using machine learning method

• Published in: Energy conversion and management Vol. 252; p. 115130
• Main Authors: Nai-Zhi, Guo, Ming-Ming, Zhang, Bo, Li
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 15.01.2022; Elsevier Science Ltd
• Subjects: Actual wind farm; administrative management; Analytical model; Cooperative control; Data analysis; energy conversion; Feature extraction; Inflow; Learning algorithms; Machine learning; Mathematical analysis; Mathematical models; Optimization; prediction; SCADA data; Turbines; wind; Wind farms; Wind measurement; Wind power; Wind turbine wake; Wind turbines; Analytical model; Actual wind farm; SCADA data; Machine learning; Wind turbine wake
• ISSN: 0196-8904; 1879-2227
• DOI: 10.1016/j.enconman.2021.115130

[Display omitted] •A data-driven analytical model for wind turbine wakes is proposed.•The machine learning method is used to model the wake expansion.•Two actual wind farm cases are used to verify the proposed model.•The wake prediction performance of the new model has improved significantly. To reduce the wake effect by means of layout optimization or cooperative control, it is significant to modeling wind turbine wakes in an accurate and efficient way. However, existing analytical wake models still have large errors in actual wind farms due to the inadequate consideration of various inflow factors and local environmental characteristics. To satisfy this accuracy requirement, a data-driven analytical wake model is proposed in this paper. In the model, the local inflow information and wake expansion feature are extracted from measured data of wind farms, and a machine learning model is trained to establish the relationship between the two. In this way, the model can be well adapted to the local environment and inflow conditions. Verifications in two actual wind farm cases illustrate that there is a good agreement with the measured velocity and power data. Compared with traditional analytical models, the wake prediction performance of the new model has improved more than 20%. Therefore, the proposed model can serve as a reliable tool for wind farm control and optimization.