基于人工神经网络的高频雷达风速反演

风速是重要的海洋状态参数之一，对海面风速的准确提取是实现海洋环境监测和沿海工程应用的重要保证。目前，作为新兴海洋环境监测设备，高频雷达在风速提取方面仍然存在挑战。本文提出了一种基于人工神经网络的风速提取方法，利用历史浮标测量海态数据训练风速提取网络，实现风速与有效波高、波周期、风向及时间因素之间的非线性映射。测试结果表明了这一网络在时间和空间上的稳定性；进而将已训练的网络应用到便携式高频地波雷达OSMAR-S的风速反演中，得到的风速与浮标测量风速间的相关系数达到0.849，均方根误差为2.11 m/s。这一结果明显优于常规由浪高反演风速的SMB方法，验证了该方法在高频雷达风速反演中的可行性。

Wind speed inversion of high frequency radar based on artifical neural network

Wind speed is one of the important ocean state parameters. Accurate extraction of sea surface wind speed is an important guarantee for achieving marine environmental monitoring and coastal engineering applications. At present, as an emerging marine environment monitoring device, high frequency radar still has challenges in wind speed extraction. This paper proposes a wind speed extraction method based on artificial neural network which can be trained by historical sea state data measured by buoys to achieve non-linear mapping among wind and effective wave height, wave period, wind direction, and time. The test results show the stability of the trained network both in time and space and the trained network was applied to the wind speed inversion of the high frequency surface wave radar, OSMAR-S. The correlation coefficient between the inversion wind speed and the measured wind speed of the buoy reaches 0.849, and the root mean square error is 2.11 m/s. This result is significantly better than the conventional SMB method which inverts the wind speed from wave height, and verifies the feasibility of this method in high frequency radar wind speed inversion.