基于改进RBF网络的潮汐预报系统的设计与实现

发布时间：2018-10-10 06:45

【摘要】：潮汐是海洋环境中最重要的组成部分之一。潮汐预测在海上交通、港口建设和潮汐能利用等领域都具有重要意义,随着航运业的不断发展,以及对航行安全和航运效率的要求,对潮汐数值预测的精度也提出了更高的要求。目前,利用调和常数计算潮汐分潮是潮汐预测的主要方法。也有一些人利用潮汐的历史数据采用非线性的数学预测手段对潮汐进行预测,例如混沌理论、神经网络、支持向量机等。传统分析方法进行潮汐预测时,由于仅考虑了潮汐天文潮部分的影响,导致其在复杂环境因素影响下的海区预测精度明显下降。将神经网络应用于潮汐预报领域是近年来出现的一种新的研究方向。径向基(RBF)神经网络在模式识别和系统预测领域应用广泛,本文将RBF神经网络用于潮汐预测,并对结果进行了探讨。同时,传统的RBF神经网络缺少必要的推理过程和依据,部分参数需要依据具体问题来进行确定,针对以上问题本文运用粒子群(PSO)优化算法优化了RBF神经网络的权值、径向基函数的中心和宽度值,建立了PSO-RBF神经网络直接进行潮汐预测的模型。该模型基于实测潮汐数据进行实时的潮汐预测,并与其它常用的优化算法进行了比较,结果体现了较高的预测精度。本文的具体内容和结论如下:1.建立了利用粒子群优化的RBF神经网络预测潮汐的模型,利用影响潮汐的最重要的两个天体(太阳和月球)的十个参数对潮汐进行预测;2.搭建了潮汐预报与显示系统,利用系统对潮汐进行实时的预测和查看;3.选取某港口三个月整点监测的潮汐数据作为历史数据进行分析和预测,预测出了未来一个月内的整点潮汐值,经检验结果较为准确。
[Abstract]:Tide is one of the most important components of marine environment. Tidal forecasting is of great significance in the fields of maritime traffic, port construction and tidal energy utilization. With the continuous development of the shipping industry and the requirements of navigation safety and efficiency, A higher requirement for the accuracy of tidal numerical prediction is also put forward. At present, the calculation of tidal component by harmonic constant is the main method of tidal prediction. Some people use the historical data of tide to predict the tide by nonlinear mathematical methods, such as chaos theory, neural network, support vector machine and so on. When the tidal prediction is carried out by the traditional analysis method, the accuracy of the sea area prediction under the influence of complex environmental factors is obviously decreased because the influence of the tidal astronomical tide is only taken into account. The application of neural network to tidal prediction is a new research direction in recent years. Radial basis function (RBF) neural network is widely used in pattern recognition and system prediction. In this paper, RBF neural network is applied to tidal prediction and the results are discussed. At the same time, the traditional RBF neural network lacks the necessary reasoning process and basis, and some parameters need to be determined according to the specific problems. Aiming at the above problems, the paper optimizes the weight of RBF neural network by using particle swarm optimization (PSO) (PSO) optimization algorithm. Based on the center and width of radial basis function, a PSO-RBF neural network model for tidal prediction is established. The model is based on the measured tidal data to carry out real-time tidal prediction, and compared with other commonly used optimization algorithms, the results show that the prediction accuracy is high. The specific contents and conclusions of this paper are as follows: 1. A tidal prediction model using RBF neural network based on particle swarm optimization is established. Ten parameters of the two most important celestial bodies (the sun and the moon) which affect the tide are used to predict the tide. A tidal prediction and display system is set up, which can be used to predict and view the tide in real time. In this paper, the tidal data of a port which is monitored for three months is analyzed and forecasted as historical data, and the tidal value of the whole point in the next month is predicted, and the results are more accurate.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TH766

【参考文献】