作者：梁瑞庆1, 邓燕飞2, 冯玮3
作者单位：1. 国家能源集团广东电力有限公司，广东 广州 510000;
2. 哈尔滨工业大学（深圳），广东 深圳 518000;
3. 电子科技大学（深圳）高等研究院，广东 深圳 518000
关键词：系泊张力;数据解算;fcn模型;lstm模型;系泊倾角
摘要：
针对浮式风机系泊倾角监测数据如何准确解算为系泊张力响应的问题，提出一种基于长短时记忆（lstm）网络与全连接网络（fcn）相结合的系泊张力解算模型。首先，针对在役的“扶摇号”浮式风机，建立风-浪-流耦合时域仿真模型，用以获取不同工况下系泊张力与倾角数据，从而形成神经网络的训练与验证数据集。在keras框架下，开发fcn与lstm的组合网络模型，选取adam优化器进行训练，获得验证集损失函数最小的网络参数。最后，分别采用传统悬链线方程与神经网络模型对系泊张力响应进行预测，并与实际结果进行比对分析。结果表明：神经网络能够考虑到不同系泊倾角对应的导缆孔平均高度变化以及系泊运动的动态过程，在系泊张力时程曲线、均值、标准差及最大值等预测上比传统悬链线方程具有更高的精度，两者对于系泊张力最大值的预报误差分别为6.4%和17.1%。
calculation method for mooring monitoring data of floating wind turbines using lstm model
liang ruiqing1, deng yanfei2, feng wei3
1. chn energy guangdong power co., ltd., guangzhou 510000, china;
2. harbin institute of technology (shenzhen), shenzhen 518000, china;
3. shenzhen institute for advanced study, uestc, shenzhen 518000, china
abstract: addressing the challenge of accurately converting the monitoring declination of floating wind turbine into mooring tension responses, this study introduces a hybrid neural network model that integrates long short-term memory (lstm) networks with fully connected networks (fcns). initially, a coupled time-domain simulation model considering wind, wave and current interactions was constructed for the in-service "fuyao" floating wind turbine to obtain mooring tension and declination data under various conditions, forming the dataset for neural network training and validation. within the framework of keras, a combined network model of fcn and lstm was developed and trained using the adam optimizer, and the network parameters with the minimal validation loss were ultimately preserved. comparative analysis with traditional catenary equation predictions revealed that the proposed neural network model demonstrated enhanced accuracy in forecasting mooring tension in terms of time series, mean, standard deviation and peak values. the prediction errors for the maximum mooring tension of the two are 6.4% and 17.1% respectively.
keywords: mooring tension;data calculation;fcn model;lstm model;mooring declination
2024, 50(8):28-33 收稿日期: 2024-05-07;收到修改稿日期: 2024-05-27
基金项目: 国家自然科学基金（52301317）；广东省自然科学基金（2024a1515011587）；广东省海洋六大产业专项（粤自然资合[2023]48）
作者简介: 梁瑞庆（1987-），男，广东高州市人，工程师，本科，研究方向为海上风电开发。
参考文献
[1] 李达, 吕柏呈, 武文华, 等. 海洋平台现场监测技术现状及展望[j]. 中国海上油气, 2022, 34(2): 165-179.
li d, lyu b c, wu w h, et al. status and prospect of offshore platform on-site monitoring technology[j]. china offshore oil and gas, 2022, 34(2): 165-179.
[2] 锁刘佳, 唐达, 马书义. 深水半潜式平台锚链倾角监测的自容式传感器的设计[j]. 海洋技术, 2012, 31(2): 77-80.
suo l j, tang d, ma s y. design of self-contained sensors used in monitoring the deep water semi-submersible platform mooring anchor inclination[j]. ocean technology, 2012, 31(2): 77-80.
[3] 杜宇, 武文华, 王延林, 等. 基于自容式技术的系泊缆水下监测方法[j]. 哈尔滨工程大学学报, 2016, 37(8): 1003-1008.
du y, wu w h, wang y l, et al. an underwater monitoring method for mooring lines based on self-contained technique[j]. journal of harbin engineering university, 2016, 37(8): 1003-1008.
[4] 包建平. 悬链线式系泊系统测量技术研究与实测分析[d]. 大连: 大连理工大学, 2014.
bao j p. research on measurement technique for catenary mooring system and analysis for measurement result[d]. dalian: dalian university of technology, 2014.
[5] 杨健, 廖晨茜, 蔡晋辉, 等. 基于改进对抗迁移学习的滚动轴承故障诊断研究[j]. 中国测试, 2022, 48(5): 96-101.
yang j, liao c x, cai j h, et al. rolling bearing fault diagnosis based on improved adversarial transfer learning[j]. china measurement & test, 2022, 48(5): 96-101.
[6] 史业照, 郭斌, 郑永军. 基于 lstm 网络的 igbt 寿命预测方法研究[j]. 中国测试, 2024, 50(2): 54-58.
shi y z, guo b, zheng y j. research on igbt life prediction based on lstm network[j]. china measurement & test, 2024, 50(2): 54-58.
[7] 范玉莹, 何赟泽, 孙高森, 等. 无人机图像风力发电机轮毂中心检测与跟踪[j]. 中国测试, 2022, 48(7): 90-96.
fan y y, he y z, sun g s, et al. wind turbine hub center detection and tracking based on uav images[j]. china measurement & test, 2022, 48(7): 90-96.
[8] 李琰, 肖龙飞, 魏汉迪, 等. 半潜式平台波浪爬升在线预报神经网络模型对比研究[j]. 中国造船, 2023, 64(4): 28-40.
li y, xiao l f, wei h d, et al. comparative study of two neural network models for online prediction of wave run-up on semi-submersible[j]. shipbuilding of china, 2023, 64(4): 28-40.
[9] 肖裕程, 李亮, 徐铭泽. 基于长短期记忆神经网络的涡激振动快速预报[j]. 中国造船, 2023, 64(5): 130-145.
xiao y c, li l, xu m z. efficient prediction for vortex induced vibration based on long short-term memory neural network[j]. shipbuilding of china, 2023, 64(5): 130-145.
[10] 范高杰. 基于深度学习的半潜平台极短期运动和系泊张力预报研究[d]. 大连: 大连海事大学, 2023.
fan g j. prediction of short-term motion and mooring tension of semi-submersible based on depth learning[d]. dalian: dalian maritime university, 2023.
[11] 靳丛林. 基于lstm-am神经网络的水下软钢臂动态系泊力估算[d]. 哈尔滨: 哈尔滨工程大学, 2024.
jin c l. study on dynamic tension estimation for the underwater soft yoke mooring system with lstm-am neural network[d]. harbin: harbin engineering university, 2024.
[12] 杨洁. 基于深度学习的单点系泊系统动态张力预测方法研究[d]. 哈尔滨: 哈尔滨工程大学, 2022.
yang j. research on the dynamic tension prediction method of single mooring system based on deep learning[d]. harbin: harbin engineering university, 2022.
[13] qiao d, li p, ma g, et al. realtime prediction of dynamic mooring lines responses with lstm neural network model[j]. ocean engineering, 2020, 219(4): 108368.
[14] saad a m, schopp f, barreira r a, et al. using neural network approaches to detect mooring line failure[j]. ieee access, 2021, 9: 27678-27695.
