Wind velocity distribution reconstruction using CFD database with Tucker decomposition and sensor measurement

• Published in: Energy (Oxford) Vol. 167; pp. 1236 - 1250
• Main Authors: Qin, Li, Liu, Shi, Kang, Yi, Yan, Song An, Inaki Schlaberg, H., Wang, Zhan
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 15.01.2019; Elsevier BV
• Subjects: Computer simulation; Decomposition; Error detection; Forecasting; Mathematical analysis; Matrix algebra; Matrix methods; Noise levels; Reconstruction; Sensor measurement; Tensors; Third-order CFD database; Three dimensional models; Tucker decomposition; Velocity; Velocity distribution; Wind measurement; Wind power; Wind speed; Wind tunnel testing; Wind tunnels; Wind velocity distribution; Third-order CFD database; Tucker decomposition; Wind velocity distribution; Sensor measurement
• ISSN: 0360-5442; 1873-6785
• DOI: 10.1016/j.energy.2018.11.013

Wind forecasting holds the key to the management of wind power. Previous vector or matrix wind forecast methods may not best reflect the intrinsic inter relationship among the wind velocity components of a three-dimensional wind field. Alternatively, a tensor-based model is developed to reconstruct the wind velocity distribution within a short period of time, enabling a new way for wind forecasting. A third-order CFD database is established by CFD simulations and the Tucker decomposition is used to obtain the tensor basis off site. Then in real time, the tensor basis can be employed to rapidly reconstruct wind velocity distributions in any direction, which can also form a new way to reconstruct wind velocity distribution in 3-D spaces. A comparison of the maximum and relative reconstruction errors shows that the newly proposed method performs better than the authors' previously published wind field reconstruction method. The influences of sampling rate, noise level and sensor distributions on the reconstruction error are also discussed in this paper. Finally, a wind tunnel experiment is carried out to evaluate the accuracy of the proposed method, and in most cases, the experimental results show that relative errors drop around 0.03%-0.4% and maximum errors drop around 0.02%-1.7% when using the newly proposed method. •A new tensor-based method is proposed to predict wind velocity distributions.•Tucker decomposition is used to decompose CFD database to obtain tensor bases.•The proposed method can be applied to different geometrical models in 3-D domains.•Wind tunnel experiments are carried out to validate the proposed method.