Abstract With the increase in global environmental problems and the energy crisis, the oscillating wave surge converter has been extensively studied in the past decades owing to its simple geometry and direct energy capture mechanism. However, systematic optimization of this converter is yet to be achieved. Consequently, in this study, a scaled oscillating wave surge converter under regular waves is numerically investigated using the smoothed particle hydrodynamics method, which is validated against experimental data. With the random changes in nine typical design parameters (i.e., the wave period, wave height, water depth, width of bottom border of the flap, width of top border, flap height, hinge height, flap density, and damping of the power take-off system), a total of 379 cases are generated and simulated. Subsequently, the capture factors corresponding to each case are calculated to quantitatively describe the energy conversion efficiency. With the design parameter combinations as input and the capture factors as output, a radial basis function neural network is trained as the prediction model of capture factors of the converters, which performs satisfactorily. Finally, this prediction model is used with the genetic algorithm to optimize the converters corresponding to different wave periods, wave heights, and water depths. By interpolating the open-source optimization results, a converter with high performance can be easily designed. The optimization method used in this study includes a radial basis function neural network based prediction model and genetic algorithm based optimization model, which can not only optimize oscillating wave surge converters but also has the potential to solve other scientific and technical optimization problems.

中文翻译：

---

使用机器学习技术预测和优化振荡波浪涌变换器

摘要 随着全球环境问题和能源危机的加剧，振荡波浪涌转换器由于其简单的几何形状和直接的能量捕获机制在过去的几十年中得到了广泛的研究。然而，该转换器的系统优化尚未实现。因此，在本研究中，使用平滑粒子流体动力学方法对规则波下的缩放振荡波浪涌转换器进行了数值研究，并根据实验数据进行了验证。随机变化9个典型设计参数（即波浪周期、波浪高度、水深、襟翼下缘宽度、上缘宽度、襟翼高度、铰链高度、襟翼密度、取力器阻尼） -off system)，共生成和模拟了 379 个案例。随后，计算每种情况对应的捕获因子以定量描述能量转换效率。以设计参数组合为输入，捕获因子为输出，训练径向基函数神经网络作为转换器捕获因子的预测模型，效果令人满意。最后，将该预测模型与遗传算法结合，针对不同的波周期、波高、水深，对转换器进行优化。通过插入开源优化结果，可以轻松设计出高性能的转换器。本研究中使用的优化方法包括基于径向基函数神经网络的预测模型和基于遗传算法的优化模型，