Wind Direction Numerical Simulation and Application Based on Wind Direction Vector Correlation and Two-dimensional Spatial Neural Networks

• Published in: 2023 8th International Conference on Power and Renewable Energy (ICPRE) pp. 1287 - 1297
• Main Authors: Bingkun, Xu, Xiaoyu, Wang, Haojiang, Bai, Yuangui, Zhou, Xiaowei, Jia, Pengfei, Tian
• Format: Conference Proceeding
• Language: English
• Published: IEEE 22.09.2023
• Subjects: Abnormal identification; Analytical models; Correlation; Correlation coefficient; Kernel canonical correlation analysis; Numerical simulation; Power measurement; Recurrent neural networks; SCADA wind direction data; Time measurement; Two-dimensional spatial neural networks; Wind direction numerical simulation; Wind direction vector correlation
• ISSN: 2768-0525
• DOI: 10.1109/ICPRE59655.2023.10353694

A wind direction numerical simulation algorithm based on wind vector correlation and two-dimensional spatial neural networks is proposed to achieve accurate numerical simulation of the target unit wind direction using the measured wind directions of wind turbine network. The study firstly proposes a calculation method combining wind direction vector and included angle vector, which provides a new perspective for solving the "circular characteristic" problem of wind direction data. Secondly, the wind direction data between wind turbines is verified to possess strong correlation according to kernel canonical correlation analysis(KCCA). Then, two-dimensional spatial back propagation neural network(BP) and recurrent neural network(RNN) are adopted to establish the wind direction vector numerical models, which are trained by SCADA wind direction data of wind farms in different terrains respectively. The results show that the numerical simulation error angle of the wind direction in the offshore and plain areas is only 1.1°~2.6°, and the abnormal identification of wind direction measurement is realized effectively.