我国海上风力发电已进入平价并网时代，行业中所有零部件设计都将面临优化设计的挑战。弯曲限制器广泛应用于风电缆保护中，其弯曲刚度与Mises峰值应力都是结构设计的关键指标。目前，弯曲限制器结构优化设计多基于经验和有限元分析迭代，该方法效率较低，并且多目标优化较困难。本文针对该问题提出一种基于神经网络的风电海缆弯曲限制器多目标优化方法。先在给定的设计域内，利用正交试验设计法和有限元分析获得的样本，构建RBF神经网络代理模型；再采用非支配排序遗传算法Ⅱ（NSGA-Ⅱ）对弯曲限制器进行多目标优化，得到了Pareto最优解集。本文为限弯器结构结构设计提供了一种可行的多目标优化方法。

China's offshore wind power generation enter the era of affordable grid connection， and all component designs in the industry face the challenge of optimizing design. Bend restrictors are widely used in wind cable protection， and their bending stiffness and Mises peak stress are key indicators in structural design. At present， the optimization design of bend restrictors is mostly based on experience and finite element analysis iteration， which has low efficiency and is difficult to achieve multiobjective optimization. The paper proposes a multi-objective optimization method based on neural networks for the optimization of marine cable of wind farm bend restrictors to address this issue. Firstly， construct an RBF neural network proxy model using samples obtained from orthogonal experimental design and finite element analysis within the given design domain. Furthermore， the non-dominated sorting genetic algorithm Ⅱ （NSGA-Ⅱ） is used for multi-objective optimization of the bend restrictor， and the Pareto optimal solution set is obtained. A feasible multi-objective optimization method is provided for the structural design of the bend restrictor.