基于PSO-SVR的海缆刚度预测模型研究  

STUDY ON PREDICTION MODEL OF SUBMARINE CABLE STIFFNESS BASED ON PSO-SVR ALGORITHM

作　　者：苏凯[1] 赵鑫蕊 朱洪泽 程永光[1] Su Kai;Zhao Xinrui;Zhu Hongze;Cheng Yongguang(State Key Laboratory of Water Resources Engineering and Management,Wuhan University,Wuhan 430072,China)

机构地区：[1]武汉大学水资源工程与调度全国重点实验室,武汉430072

出　　处：《太阳能学报》2024年第8期458-465,共8页Acta Energiae Solaris Sinica

基　　金：国家自然科学基金(51379159)。

摘　　要：海缆刚度是表征海缆截面力学特性的重要指标。精细数值模拟方法可考虑层间接触、摩擦等非线性行为,实现刚度的精确获取。但在海缆的截面初步设计中需进行不同几何参数下刚度的对比分析,这一过程需投入大量的人力物力以完成多个对比模型的建模与计算。依托Nysted海上风电工程,建立海缆的有限元模型,采用正交试验法确定影响海缆抗拉刚度、逆时针抗扭刚度和顺时针抗扭刚度的主要因素;将主要影响因素作为粒子群优化算法-支持向量回归模型(PSO-SVR)的特征输入,分别建立海缆抗拉刚度、逆时针抗扭刚度和顺时针抗扭刚度的预测模型,并对比分析PSO-SVR模型与GRNN神经网络模型、BP神经网络模型的预测性能。计算结果表明:导体直径、钢丝直径、钢丝节距和铠装层数对海缆刚度的影响较大,而钢丝弹模对其影响较小;PSO-SVR模型的决定系数高于0.95且误差较低,预测效果均优于GRNN神经网络模型和BP神经网络模型,该预测模型可为海缆结构初步设计提供技术支撑。Submarine cable stiffness is an important index to characterize the cross-sectional mechanical properties.Fine numerical simulation methods can consider nonlinear behaviors such as interlayer contact and friction to obtain the exact stiffness.However,in the preliminary design of the submarine cable cross-section,it is necessary to conduct a comparative analysis of the stiffness under different geometric parameters,which requires a significant amount of manpower and material resources to complete the modeling and calculation of multiple comparative models.This study establishes a finite element model of the submarine cable based on the Nysted offshore wind farm project,and uses the orthogonal test method to determine the main factors influencing the tensile stiffness,counterclockwise torsional stiffness and clockwise torsional stiffness of the submarine cable.Then,the main influencing factors are used as feature inputs of the Particle Swarm Optimization-Support Vector Regression(PSO-SVR)model to establish prediction models for tensile stiffness,counterclockwise torsional stiffness and clockwise torsional stiffness of the submarine cable respectively.Besides,the prediction performance of the PSO-SVR model is compared with the GRNN neural network model and the BP neural network model.The results show that the conductor diameter,steel wire diameter,steel wire pitch and the number of armor layers have more influence on the submarine cable stiffness,while the steel wire elastic modulus has less influence on it.The determination coefficient of the PSO-SVR model is higher than 0.95 and the error is relatively low,and the prediction performance is better than that of the GRNN neural network model and the BP neural network model.This prediction model can provide technical support for the preliminary design of submarine cable structures.

关 键 词：海上风电 海缆 刚度 粒子群优化 支持向量回归 正交试验 

分 类 号：P751[交通运输工程—港口、海岸及近海工程]