海床风电桩通常安装在工程性质较差的近海软土地层中，在外部复杂荷载的作用下容易发生较大偏转甚至失稳，影响整个风电系统的正常运转。在现有的风电桩稳定性研究中，对偏转的监测和预测是最经济有效的方法之一。鉴于传统监测方法的不足以及风电桩偏转变化的非线性，该文提出了一种基于超弱光纤光栅（UWFBG）和机器学习（ML）的海床风电桩偏转监测和预测新方法，并将其应用于山东半岛的一个海床风电桩案例研究中。利用UWFBG成功获得沿风电桩的连续应变数据，计算出风电桩的最大偏转角0.35°；分析了风速、风向和潮汐等荷载影响因素与桩顶转角之间关系，发现在盛行风向下，桩顶转角与风速呈正相关，与潮汐振幅呈负相关；在此基础上建立了 EEMD-PSO-SVR 预测模型并成功对风电桩偏转进行了预测，与实测值相比，预测结果均方根误差和平均绝对误差分别为分别为0.0438°和0.0358°，验证了所提出预测模型的准确性。

Seabed wind monopiles are usually installed in offshore soft clay layers with poor engineering properties, and are prone to large deflections even destabilization under complex external loads, affecting the normal operation of the wind power system. Among the existing offshore wind monopile stability studies, monitoring and predicting deflection is one of the most costeffective methods. In view of the shortcomings of the traditional monitoring methods and the nonlinearity of monopile deflection changes, this study proposes a new method for monitoring and predicting the deflection of seabed wind monopiles based on Ultra Weak Fiber Bragging Grating (UWFBG) and Machine Learning (ML), and applies it to a case study of seabed wind monopiles in Shandong Peninsula. The continuous strain data along the monopile were successfully obtained by UWFBG, and the maximum deflection angle of the monopile was calculated to be 0.35°; The load influencing factors of top deflection angle such as wind speed, wind direction and tide were analyzed, and it was found that the top deflection angle was positively correlated with the wind speed and negatively correlated with the amplitude of the tides under the prevailing wind direction; The EEMD-PSO-SVR prediction model was established on this basis and successfully predicted the monopile deflection, compared with the measured values, the root-mean-square error and the mean absolute error of the prediction results were 0.0438° and 0.0358°, which verified the accuracy of the proposed prediction model.