共查询到17条相似文献，搜索用时 156 毫秒
1.
为实现对海面风速精确的短期预测,提出了一种基于长短期记忆(LSTM,longshort-termmemory)神经网络的短期风速预测模型,选取OceanSITES数据库中单个浮标站点采集的风速历史数据作为模型输入,经过训练设置最佳参数等步骤,实现了以LSTM方法,对该站点所在海区海面风速在各季节性代表月份海面风速的24 h短期预测。同时通过不同预测时长的实验以及与BP(back propagation)神经网络神经网络和径向基函数神经网络(radialbasisfunctionneuralnetwork,RBF)的预测效果对比实验,证明了LSTM预测方法相比上述两种神经网络预测方法,在海表面风速预测应用中的优越性。最后通过多个海域对应的站点风速数据预测实验,证明了LSTM神经网络模型的普遍适用性,由相关系数和预测误差的分析可知该方法具备应对急剧变化数据的预测稳定性,可以作为海洋表面风速短期预测的一种可靠方法。 相似文献
2.
3.
精确的海浪有效波高（简称浪高）预测对于海上生产生活具有重要意义。针对现有海浪浪高预测模型对不同海洋要素间关联信息考虑不足,以及长时序浪高数据本身存在非平稳性的问题,本文设计了一种考虑物理约束与差值约束的海浪浪高时间序列预测方法。该方法基于风速与浪高之间的物理关联,设计物理约束,并通过提取差分信息设计差值约束,结合现有基于深度学习的时间序列预测模型,实现浪高预测。采用黄海和东海的6个不同站点浮标数据进行了大量实验。实验结果表明,本文提出的方法可以利用海洋要素间的物理关联,有效提高浪高预测精度,并避免因不同要素间融合造成的信息间干扰;同时,利用差值约束,限制时间序列预测结果的变动范围。本文方法可以与不同类型的时间序列预测模型相结合,显著提升原有模型的性能,并在长时间序列的预测中体现出很好的鲁棒性,为海洋要素预测中物理与数据驱动模型的有效结合提供了思路和验证。 相似文献
4.
海浪直接影响海上活动和航行安全,同时也蕴藏着巨大的可再生能源,对海浪核心参数之一波高预测至关重要。基于2015年7月～2022年6月山东小麦岛(36°N,120.6°E)站点实测的波高数据,利用反向传播神经网络(back-propagation neural network,BPNN)、长短记忆网络(long short-term memory, LSTM)和支持向量机回归(support vector regression, SVR)三种机器学习模型对波高进行预测,并分析了瑞利参数的引入对预测结果的影响。结果显示,模型输入项引入瑞利参数后,对1 h和6 h波高预测提升效果有限,预测值与测试集的相关性提升不超过0.02,均方根误差的降低不超过0.01 m;在12h和24h的预测中,BPNN和LSTM模型预测结果相关性提升0.03～0.07,均方根误差降低0.02～0.03m,而SVR模型预测结果变化不显著。说明瑞利参数有助改善BPNN和LSTM模型中长期海浪预报。此外,特征扰动方法(机器学习中特征重要性的计算方法之一)验证了瑞利参数在波高预测中的重要性,瑞利参数的引入为波高的机器学习预... 相似文献
5.
本文根据相干斑噪声的时间快变特征和非海浪纹理现象的时间缓变特征，基于交叉谱提出了一种对相干斑噪声和大尺度非海浪纹理的抑制的方法，进而结合SAR图像谱和海浪谱之间的准线性映射关系，基于SAR数据对海浪参数进行了反演。在反演过程中，首先仿真分析了不同海况下准线性近似法的海浪反演能力，结果表明：风浪引起的方位向截断效应会显著影响反演精度，因此该方法在低风速时的涌浪反演精度更高。通过将基于Sentinel-1卫星2020年的波模式SAR数据的反演结果与欧洲中期天气预报中心(European Centre for Medium-Range Weather Forecasts, ECMWF)提供的再分析数据进行对比，发现高海况海浪有效波高反演结果明显偏低，而且该反演误差与风速、方位向截断波长之间存在显著相关性。为了提高有效波高的反演精度，本文进一步给出了海浪有效波高反演误差与风速、方位向截断波长之间的经验校正函数模型，结果显示，通过该模型修正后的海浪有效波高反演结果与ECMWF数据和浮标测量数据具有良好一致性。 相似文献
6.
快速的地面沉降是一种地质灾害,它关系到社会的可持续发展,甚至威胁人类的生命财产安全。InSAR技术可以获取地表长时间、大范围的形变数据,可用于分析潜在的地面沉降问题,为预防地质灾害提供了一种可靠手段。如何基于InSAR数据对地面沉降进行预测,一直是研究人员关注的重点方向和难题。为此,本文在前人对地面沉降预测研究的基础上,提出了一种将差分移动平均自回归(ARIMA)模型与深度学习中的长短期记忆单元(LSTM)模型相结合的地面沉降预测方法,即利用InSAR得到的形变量数据与ARIMA模型预测结果作差,然后利用LSTM对该差值进行训练与预测。以杭州湾2017—2019年InSAR监测数据为例验证了该方法,结果表明,与传统的单一预测算法相比,本文方法的均方根误差至少减小了2.23 mm,平均绝对误差至少减小了0.98 mm,平均预测精度至少提升了15.19%,验证结果证实了本文方法的可行性,为地面沉降预警工作提供了新的思路和方法。 相似文献
7.
本文引入RNN的升级算法LSTM神经网络技术,建立了一套海量数据、高精度的自动拾取地震资料初至流程。相比于其他神经网络方法,LSTM神经网络能够提取数据的时序特征,加强学习初至前噪音时序特征,从而提高初至拾取的精度,为地震资料的初至拾取提供一套新的思路。首先设计样本制作过程并建立、训练模型,通过模拟资料验证方法的有效性,应用于胜利油田浅海与西部山地地震勘探资料的初至拾取,取得理想效果,证明LSTM神经网络初至拾取具有较高的精度与适用性。 相似文献
8.
《海洋技术学报》2023,(2)
本文基于唐山近海海域1#、2#浮标2017年4月至11 月实时海浪观测数据及部分风速风向数据, 对唐山近海海域波浪有效波高、有效波向、有效波周期等波参数特征进行了统计分析, 并利用origin 软件对波参数与风速、风向相关性进行了研究。研究结果表明： 1#、2# 浮标海域常浪向为SSW、SW、SSE, 常浪向有效波高均以0.2 ～ 0.4 m 小浪及3 ～ 4 s 短周期为主,有效波高1 m 以上较大波浪极少出现; 该海域波浪以风浪为主, 波浪破碎速度较快, 有效波高与风速相关性较强, 相关系数r 为0.71, 风向与波向、有效波高与周期基本无相关性, 该研究资料可为海上活动及防灾减灾提供技术依据。 相似文献
9.
介绍采用谱分析原理提取海浪信息的方法。根据GPS浮标测量得到的海浪高度数据,先通过滑动平均的方法分离海浪信息和潮流潮位信息,然后对海浪信息采用AR模型法进行功率谱估计,并对海浪信息进行连续小波变换,对AR模型法功率谱估计结果和小波变换结果进行综合研究,提取海浪特征参数———周期。最后通过一个GPS浮标试验对上述方法进行了验证。 相似文献
10.
静态集合样本的构造及其在全球海浪滤波同化中的应用 总被引：4，自引：2，他引：2
本文提出一种最佳静态集合样本的构造方法,利用不同时段内海浪有效波高的模拟偏差构造静态集合样本,并将其与由模拟结果和观测资料统计的模式误差进行概率密度分布及时空相关性分析,结果表明24h间隔有效波高偏差与后者的相关性最好,称为最佳静态集合样本,可用于近似背景误差。将所构造的静态集合样本应用于滤波同化调整过程,采用MASNUM海浪模式,利用Jason-1卫星高度计数据,对2008年全球海域开展海浪同化实验,实验结果表明,基于最佳静态集合样本的海浪同化调整,可以有效地改善海浪模式的模拟效果。 相似文献
11.
To explore new operational forecasting methods of waves, a forecasting model for wave heights at three stations in the Bohai Sea has been developed. This model is based on long short-term memory(LSTM) neural network with sea surface wind and wave heights as training samples. The prediction performance of the model is evaluated,and the error analysis shows that when using the same set of numerically predicted sea surface wind as input, the prediction error produced by the proposed LSTM model at Sta. N01 is 20%, 18% and 23% lower than the conventional numerical wave models in terms of the total root mean square error(RMSE), scatter index(SI) and mean absolute error(MAE), respectively. Particularly, for significant wave height in the range of 3–5 m, the prediction accuracy of the LSTM model is improved the most remarkably, with RMSE, SI and MAE all decreasing by 24%. It is also evident that the numbers of hidden neurons, the numbers of buoys used and the time length of training samples all have impact on the prediction accuracy. However, the prediction does not necessary improve with the increase of number of hidden neurons or number of buoys used. The experiment trained by data with the longest time length is found to perform the best overall compared to other experiments with a shorter time length for training. Overall, long short-term memory neural network was proved to be a very promising method for future development and applications in wave forecasting. 相似文献
12.
To plan for proper mitigation measures, one should have an advanced knowledge of the phenomenon of tsunami propagation from the deep ocean to coastal waters. There are a few methods to predict tsunamis in the ocean waters; one method is the effective use of data buoy measurements. Although data buoys have been used along the Indian waters there has been a tremendous growth in the number of buoy deployment recently. Under the National Data Buoy Programme (NDBP) of India, the 2.2 m diameter discus data buoys were deployed along the east and west coasts of India for measuring meteorological and ocean parameters. It would be advantageous if these buoys could be efficiently used to measure rare events such as tsunamis. Understanding the dynamic behavior of the buoy is of prime importance if a tsunami warning system is to be successful. This may be accomplished through experimental or numerical studies. A comprehensive experimental study has been conducted to understand the dynamic behavior of a wave rider buoy exposed to a variety of waves. It is common that tsunami waves are represented in terms of shallow water waves, namely solitary and cnoidal waves. Hence, in the present study, the discus type data buoy is scale modeled and tested under the action of solitary and cnoidal waves in the laboratory. The time histories of wave elevations, as well as heave and pitch motions of the buoy model, were analyzed through a spectral approach as well as through wavelet transformations. The wavelet approach gives more detailed insight into the spectral characteristics of the buoy motion in the time scale. The harmonic analyses were performed for the cnoidal wave elevations and subsequent motion characteristics that give an insight into the energy variations. The details of the model, instrumentation, testing conditions and the results are presented in this paper. 相似文献
13.
14.
冰区四季通用灯浮标是一种为满足北方冬季冰冻港口一年四季的助航服务需求研制的新型灯浮标,二阶波浪力对其漂浮姿态和漂移运动有较大影响。文中研究了浮标受到的二阶波浪力的数值计算方法,计算了不同流速下罐形和锥形灯浮标的二阶波浪力。研究结果显示,罐形和锥形灯浮标受到的一阶波浪力相差不大,罐形的二阶波浪力明显小于锥形,具有一定的外形优势。 相似文献
15.
Tayeb Sadeghifar Maryam Nouri Motlagh Massoud Torabi Azad Mahdi Mohammad Mahdizadeh 《Marine Geodesy》2017,40(6):454-465
The prediction of wave parameters has a great significance in the coastal and offshore engineering. For this purpose, several models and approaches have been proposed to predict wave parameters, such as empirical, soft computing, and numerical based approaches. Recently, soft computing techniques such as recurrent neural networks (RNN) have been used to develop sea wave prediction models. In this study, the RNN for wave prediction based on the data gathered and the measurement of the sea waves in the Caspian Sea, in the north of Iran is used for this study. The efficiency of RNNs for 3, 6, and 12 hourly and diurnal wave prediction using correlation coefficients is calculated to be 0.96, 0.90, 0.87, and 0.73, respectively. This indicates that wave prediction by using RNNs yields better results than the previous neural network approaches. 相似文献
16.
海平面变化包含多种不同时间尺度信息，传统的预测方法仅对海平面变化趋势项、周期项进行拟合，难以利用海平面变化的不同时间尺度信号，使得预测精度不高。本文基于深度学习的预测模型，提出一种融合小波变换（wavelet transform，WT）与LSTM （long short-term memory，LSTM）神经网络的海平面异常组合预测模型。首先利用小波分解得到反映海平面变化总体趋势的低频分量和刻画主要细节信息的高频分量；然后通过LSTM神经网络对代表不同时间尺度的各个分量预测和重构，实现海平面变化的非线性预测。基于该模型的海平面变化预测的均方根误差、平均绝对误差和相关系数分别为12.76 mm、9.94 mm和0.937，预测精度均优于LSTM和EEMD-LSTM预测模型，WT-LSTM组合模型对区域海平面变化预测具有较好的应用价值。 相似文献
17.
This article uses a comparison of four different numerical wave prediction models for hindcast wave conditions in Lake Michigan during a 10-day episode in October 1988 to illustrate that typical wave prediction models based on the concept of a wave energy spectrum may have reached a limit in the accuracy with which they can simulate realistic wave generation and growth conditions. In the hindcast study we compared the model results to observed wave height and period measurements from two deep water NOAA/NDBC weather buoys and from a nearshore Waverider buoy. Hourly wind fields interpolated from a large number of coastal and overlake observations were used to drive the models. The same numerical grid was used for all the models. The results show that while the individual model predictions deviate from the measurements by various amounts, they all tend to reflect the general trend and patterns of the wave measurements. The differences between the model results are often similar in magnitude to differences between model results and observations. Although the four models tested represent a wide range of sophistication in their treatment of wave growth dynamics, they are all based on the assumption that the sea state can be represented by a wave energy spectrum. Because there are more similarities among the model results than significant differences, we believe that this assumption may be the limiting factor for substantial improvements in wave modeling. 相似文献