Wind Speed and Direction Measurement Method Based on WMNet With Arc Ultrasonic Array

The gap between measurement accuracy and computational complexity is an important problem of wind parameter measurement algorithms for real-time applications. In this article, a wind speed and direction measurement method based on wind measure network (WMNet) is proposed, applying the convolution neural network (CNN) to the array wind measurement algorithm, to narrow the gap. The arc array structure is used as the receiving array of ultrasonic signals, and the array output vector is used as the neural network training set. A WMNet model is built based on CNN, and the wind parameter features are extracted by WMNet. Hence, the wind estimation problem is transformed into a wind feature classification problem, avoiding the process of decomposition of eigenvalues, construction of spatial spectra, and 2-D spectral peak search in the traditional multiple signal classification (MUSIC) algorithm. An experimental platform for wind speed and direction measurement is built, and the effectiveness of the proposed method is verified by wind tunnel experiments. The success rate and real-time performance of wind parameter estimation are analyzed through simulation calculations and statistical performance experiments. Compared to the traditional wind measurement algorithm, the proposed method not only ensures the accuracy of wind speed and direction estimation but also reduces the estimation time.