CN105956709A

CN105956709A - A GUI-Based Modular Support Vector Machine Tide Forecasting Method

Info

Publication number: CN105956709A
Application number: CN201610316444.5A
Authority: CN
Inventors: 尹建川; 柳成; 张泽国
Original assignee: Dalian Maritime University
Current assignee: Dalian Maritime University
Priority date: 2016-05-12
Filing date: 2016-05-12
Publication date: 2016-09-21
Anticipated expiration: 2036-05-12
Also published as: CN105956709B

Abstract

The invention discloses a GUI-based modular support vector machine tide prediction method, which comprises the following steps: s1, obtaining continuous sequence type tide level information monitored by the tide station and tide values predicted by using a harmonic analysis method, S2: subtracting the tide measured value obtained by the tide level station from the tide value predicted by the harmonic analysis method, performing data accumulation processing on the input tide level information and the tide value according to a gray model AGO algorithm by the obtained time sequence of the non-astronomical tide, and using the processed data as input quantity for regression prediction of a support vector machine; s3: predicting the tide through a support vector machine according to the tide forecast time information set in the S1, S4: and (4) carrying out IAGO reverse accumulation processing on the predicted result of the support vector machine to complete data reduction, wherein the reduced data is used for correcting the tide predicted value of the harmonic analysis method.

Description

A GUI-Based Modular Support Vector Machine Tide Forecasting Method

技术领域 technical field

本发明涉及潮汐预测领域，尤其涉及一种基于GUI的模块化支持向量机潮汐预测方法。 The invention relates to the field of tide prediction, in particular to a GUI-based modular support vector machine tide prediction method.

背景技术 Background technique

由于潮汐受到多种因素的影响，周期因素如引潮力、非周期因素如风力、气压、海岸特性、降水、月球轨道的倾角等等。传统的调和分析法通过对长期潮汐数据的统计和分析,计算得到模型中各分潮的参数,并基于建立潮的数学模型得到长期的潮汐预报，但是该方法的预报精度除了受分潮数量的影响，还无法分析非周期因素的影响。目前常用的神经网络预报方法是将影响潮汐的各要素,如天体位置参数、风、气压、降水等信息作为网络的输入，以潮汐信息作为网络输出建立神经网络预报模型，并通过对历史数据的学习确定网络的参数。根据确立的神经网络模型和输入进行潮汐预报，该方法虽然一定程度上弥补了调和分析无法预报非周期因素的缺憾，但学习训练的样本要求数据量大，涉及面广，能覆盖各种可能出现的情况，而具有非周期因素的台站历史数据一般却很少。 Since the tide is affected by many factors, periodic factors such as tidal force, non-periodic factors such as wind force, air pressure, coastal characteristics, precipitation, the inclination angle of the lunar orbit and so on. The traditional harmonic analysis method calculates the parameters of each tidal component in the model through the statistics and analysis of long-term tidal data, and obtains the long-term tidal forecast based on the establishment of a mathematical model of the tide. It is not yet possible to analyze the impact of non-cyclical factors. At present, the commonly used neural network forecasting method is to use various elements that affect the tide, such as celestial body position parameters, wind, air pressure, precipitation, etc. Learn to determine the parameters of the network. According to the established neural network model and input for tidal forecasting, although this method makes up for the shortcomings of harmonic analysis that cannot predict non-periodic factors to a certain extent, the samples for learning and training require a large amount of data, involve a wide range, and can cover various possible occurrences. However, the historical data of stations with aperiodic factors are generally scarce.

发明内容 Contents of the invention

根据现有技术存在的问题，本发明公开了一种基于GUI的模块化支持向量机潮汐预测方法，包括以下步骤： According to the problems existing in the prior art, the invention discloses a GUI-based modular support vector machine tide prediction method, comprising the following steps:

S1:获取潮汐站监测到的连续序列式潮位信息和利用调和分析法预测的潮汐值，设置潮汐预报的时间值； S1: Obtain the continuous sequential tide level information monitored by the tide station and the tide value predicted by the harmonic analysis method, and set the time value of the tide forecast;

S2：将潮位站获取的潮汐实测值与调和分析法预测的潮汐值做差，得到的非天文潮的时间序列根据灰色模型AGO算法对输入的潮位信息和潮汐值进行数据累加处理，处理后的数据作为输入量用于支持向量机的回归预测； S2: The difference between the tide measured value obtained by the tide station and the tide value predicted by the harmonic analysis method is made, and the obtained non-astronomical tide time series is processed according to the gray model AGO algorithm for the input tide level information and tide value. The processed The data is used as input for the regression prediction of the support vector machine;

S3：根据S1中设置的潮汐预报时间信息通过支持向量机对潮汐进行预测，在预测过程中对支持向量机的惩罚系数c和核函数半径g分别选用交叉验证、粒子群和遗传算法进行优化，并选择支持向量机训练过程中产生误差最小值的方法作为优化算法进行潮汐预测； S3: According to the tidal forecast time information set in S1, the tide is predicted by the support vector machine. During the prediction process, the penalty coefficient c and the kernel function radius g of the support vector machine are respectively optimized by cross-validation, particle swarm optimization and genetic algorithm. And choose the method that produces the minimum value of the error during the training process of the support vector machine as the optimization algorithm for tide prediction;

S4：支持向量机预测的结果经过IAGO反向累加处理完成数据还原，还原后的数据用于修正调和分析法的潮汐预测值。 S4: The results predicted by the support vector machine are processed by IAGO reverse accumulation to complete the data restoration, and the restored data are used to correct the tide prediction value of the harmonic analysis method.

将潮汐站获取的潮汐实测值定义为非天文潮的时间序列，表示为实测潮汐序列y₀:x₀(1),x₀(2),…,x₀(n)； The tide measured value obtained by the tide station is defined as the time series of non-astronomical tide, expressed as the measured tide sequence y ₀ : x ₀ (1), x ₀ (2),...,x ₀ (n);

调和分析法预测的潮汐数据，它在表示某地实际潮位高度H(t)时，计算方法如下所示： The tide data predicted by the harmonic analysis method, when it represents the actual tide height H(t) of a certain place, the calculation method is as follows:

$H h ((t t)) = = {α α}_{00} + + {Σ Σ}_{j j = = 11}^{m m} {R R}_{j j} c c o o s the s (({σ σ}_{j j} t t - - {θ θ}_{j j})) = = {α α}_{00} + + {Σ Σ}_{j j = = 11}^{m m} (({α α}_{j j} {cosσ cosσ}_{j j} t t + + {b b}_{j j} {sinσ sinσ}_{j j} t t))$

式中α₀为平均海面高度，R_j为分潮振幅，θ_j为分潮的初位相，σ_j为分潮的角速度，α_j＝R_jcosθ_j，b_j＝R_jsinθ_j，m为分潮的个数，是正整数；把调和分析法的预报部分看作天文潮部分，经过计算后得到的调和分析时间序列记为y₁:x₁(1),x₁(2),…,x₁(n)；将实测潮汐序列y₀与调和分析时间序列做差时: where α ₀ is the mean sea surface height, R _j is the tidal amplitude, θ _j is the initial phase of the tidal, σ _j is the angular velocity of the tidal, α _j =R _j cosθ _j , b _j =R _j sinθ _j ,m is the number of tides, which is a positive integer; the forecast part of the harmonic analysis method is regarded as the part of the astronomical tide, and the time series of the harmonic analysis obtained after calculation is recorded as y ₁ : x ₁ (1), x ₁ (2),… ,x ₁ (n); When making the difference between the measured tide series y ₀ and the harmonic analysis time series:

计算公式为: The calculation formula is:

y₂＝y₀-y₁ y ₂ =y ₀ -y ₁

新的序列为: The new sequence is:

y₂:x₃(1),x₃(2),…,x₃(n) y ₂ :x ₃ (1),x ₃ (2),…,x ₃ (n)

该序列首先需要经过AGO操作后用于支持向量机的回归预测: The sequence first needs to be used for the regression prediction of the support vector machine after the AGO operation:

经过AGO处理的序列表示为x⁽¹⁾(1),x⁽¹⁾(2),…,x⁽¹⁾(n),计算公式为: The sequence processed by AGO is expressed as x ⁽¹⁾ (1), x ⁽¹⁾ (2), ..., x ⁽¹⁾ (n), and the calculation formula is:

${x x}^{((11))} ((k k)) = = {Σ Σ}_{i i = = 11}^{k k} {x x}_{33} ((k k)),, k k = = 11,, 22,, ... ...,, n no$

在经过支持向量机预测后得到的预测序列为: The prediction sequence obtained after the support vector machine prediction is:

${\overset{^^}{x x}}^{((11))} ((11)),, {\overset{^^}{x x}}^{((11))} ((22)),, ... ...,, {\overset{^^}{x x}}^{((11))} ((n no))$

该序列经过IAGO的算法获取非天文潮部分最终预测序列： The sequence obtains the final forecast sequence of the non-astronomical tide part through the algorithm of IAGO:

${\overset{^^}{x x}}^{((00))} ((k k + + 11)) = = {\overset{^^}{x x}}^{((11))} ((k k + + 11)) - - {\overset{^^}{x x}}^{((11))} ((k k)),, k k = = 11,, 22,, ... ...,, n no$

其中表示非天文潮部分预测后还原的数据，最终潮汐数据通过非天文潮部分的预测值与天文潮部分相加得到最终预测结果； in Indicates the restored data after the prediction of the non-astronomical tide part, and the final tide data is obtained by adding the predicted value of the non-astronomical tide part to the astronomical tide part to obtain the final prediction result;

得到的时间为k+1的潮汐值表示为 The obtained tide value at time k+1 is expressed as

$y the y ((k k + + 11)) = = {\overset{^^}{x x}}^{((00))} ((k k + + 11)) + + {x x}_{11} ((k k + + 11)) . .$

由于采用了上述技术方案，本发明提供的一种基于GUI的模块化支持向量机潮汐预测方法，不仅可以将风向、降雨、风暴增水、海岸特性等非周期影响潮汐的因素融合到该预测方法中，而且小样本数据也可以达到较精确的结果。建立一个基于支持向量机SVM预测模型，首先在MATLAB 7.8中导入一个SVM工具箱，然后利用svmtrain函数对训练样本数据进行训练，再用测试样本svmpredict函数对形成的模型进行测试，经过训练和测试后的数据才能对同一验潮站的潮汐进行预测。 Due to the adoption of the above technical scheme, a GUI-based modular support vector machine tide prediction method provided by the present invention can not only integrate wind direction, rainfall, storm water increase, coastal characteristics and other non-periodic factors that affect tides into the prediction method In addition, small sample data can also achieve more accurate results. To build a prediction model based on support vector machine SVM, first import an SVM toolbox in MATLAB 7.8, then use the svmtrain function to train the training sample data, and then use the test sample svmpredict function to test the formed model, after training and testing Only the data of the same tide gauge station can be used to predict the tide.

附图说明 Description of drawings

为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请中记载的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。 In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments described in this application. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

图1为本方法的流程图； Fig. 1 is the flowchart of this method;

图2为潮汐预测过程的流程图。 Figure 2 is a flowchart of the tide prediction process.

具体实施方式 detailed description

为使本发明的技术方案和优点更加清楚，下面结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚完整的描述： In order to make the technical solutions and advantages of the present invention more clear, the technical solutions in the embodiments of the present invention are clearly and completely described below in conjunction with the drawings in the embodiments of the present invention:

如图1所示的一种基于GUI的模块化支持向量机潮汐预测方法，具体包括以下步骤： As shown in Figure 1, a GUI-based modular support vector machine tide prediction method specifically includes the following steps:

S1:获取潮汐站监测到的连续序列式潮位信息和利用调和分析法预测的潮汐值，设置潮汐预报的时间值，该时间为假如用户想预测10天后的潮汐值，则该时间为10。 S1: Obtain the continuous sequence tide level information monitored by the tide station and the tide value predicted by the harmonic analysis method, and set the time value of the tide forecast. If the user wants to predict the tide value after 10 days, the time is 10.

S2：将潮位站获取的潮汐实测值与调和分析法预测的潮汐值做差，得到的非天文潮时间序列根据灰色模型AGO算法对输入的潮位信息和潮汐值进行数据累加处理，处理后的数据作为输入量用于支持向量机的回归预测。 S2: The difference between the actual tide value obtained by the tide station and the tide value predicted by the harmonic analysis method is made, and the obtained non-astronomical tide time series is processed according to the gray model AGO algorithm for the input tide level information and tide value. The processed data It is used as an input for the regression prediction of the support vector machine.

模块化预报程在运行过程中，首先需要获取两个输入，其一为用户载入的潮汐站实测数据，这些数据来源于潮汐站对每天潮位信息的记录，在表示的形式上是一个时间序列，该序列记为y₀:x₀(1),x₀(2),…,x₀(n)。其二为调和分析法预测的潮汐数据，它在表示某地实际潮位高度H(t)时，计算方法如下所示： During the operation of the modular forecasting process, two inputs first need to be obtained. One is the measured data of the tide station loaded by the user. These data come from the tide station’s records of the daily tide level information, and the representation form is a time series , the sequence is denoted as y ₀ :x ₀ (1),x ₀ (2),…,x ₀ (n). The second is the tide data predicted by the harmonic analysis method. When it represents the actual tide height H(t) of a certain place, the calculation method is as follows:

$H h ((t t)) = = {α α}_{00} + + \overset{m m}{Σ Σ} {R R}_{j j} c c o o s the s (({σ σ}_{j j} t t - - {θ θ}_{j j})) = = {α α}_{00} + + \overset{m m}{Σ Σ} (({α α}_{j j} {cosσ cosσ}_{j j} t t + + {b b}_{j j} {sinσ sinσ}_{j j} t t))$

式中α₀为平均海面高度，R_j为分潮振幅，θ_j为分潮的初位相，σ_j为分潮的角速度，α_j＝R_jcosθ_j，b_j＝R_jsinθ_j，m为分潮的个数，是正整数。由于调和分析法在预测潮位时主要考虑的是天体引潮力对潮汐形成的作用，所以把调和分析法的预报部分看作天文潮部分，经过计算后得到的调和分析时间序列记为y₁:x₁(1),x₁(2),…,x₁(n)。 where α ₀ is the mean sea surface height, R _j is the tidal amplitude, θ _j is the initial phase of the tidal, σ _j is the angular velocity of the tidal, α _j =R _j cosθ _j , b _j =R _j sinθ _j ,m is the number of tides, which is a positive integer. Since the harmonic analysis method mainly considers the effect of celestial body tidal force on the formation of tides when predicting the tide level, the forecast part of the harmonic analysis method is regarded as the astronomical tide part, and the harmonic analysis time series obtained after calculation is recorded as y ₁ :x ₁ (1),x ₁ (2),...,x ₁ (n).

以上两部分的差值，则被我们看作受环境等其他因素影响的非天文潮部分，计算方法为:y₀-y₁,该序列表示为x₃(1),x₃(2),…,x₃(n)处理后的序列在图2中表示为“非天文潮潮汐序列”，这个序列作为原始输入用于非天文潮部分即支持向量机的预测，在进行预报之前首先通过灰色模型的AGO算法来对输入序列进行处理以提高预测精度。AGO对时间序列的处理方法如下： The difference between the above two parts is regarded as the non-astronomical tide part affected by other factors such as the environment. The calculation method is: y ₀ -y ₁ , and the sequence is expressed as x ₃ (1), x ₃ (2), …,x ₃ (n) The sequence after processing is represented as "non-astronomical tidal sequence" in Figure 2. This sequence is used as the original input for the prediction of the non-astronomical tide part, that is, the support vector machine. Before forecasting, it first passes through the gray The AGO algorithm of the model is used to process the input sequence to improve the prediction accuracy. AGO processes time series as follows:

由于非天文潮分的原始序列为x₃(1),x₃(2),…,x₃(n)，则经过AGO处理的序列表示为x⁽¹⁾(1),x⁽¹⁾(2),…,x⁽¹⁾(n),其中在处理之后，新的非天文潮部分的潮汐时间序列被应用于支持向量机的预报中。 Since the original sequence of non-astronomical tide is x ₃ (1),x ₃ (2),…,x ₃ (n), the AGO-processed sequence is expressed as x ⁽¹⁾ (1),x ⁽¹⁾ ( 2),...,x ⁽¹⁾ (n), where After processing, the new non-astronomical tidal part of the tidal time series is applied to the SVM forecast.

S3：根据S1中设置的潮汐预报时间信息通过支持向量机对潮汐进行预测，在预测过程中对支持向量机的惩罚系数c和核函数半径g分别选用交叉验证、粒子群和遗传算法进行优化，并选择支持向量机训练过程中产生误差最小值的方法作为优化算法进行潮汐预测。 S3: According to the tidal forecast time information set in S1, the tide is predicted by the support vector machine. During the prediction process, the penalty coefficient c and the kernel function radius g of the support vector machine are respectively optimized by cross-validation, particle swarm optimization and genetic algorithm. And choose the method that produces the minimum value of error in the training process of support vector machine as the optimization algorithm for tide prediction.

在预测的过程中，程序将自动选用三种优化算法对支持向量机进行优化并选取误差最小的算法作为当次预测的优化方法，这三种方法分别是交叉验证选择法，粒子群优化算法和遗传算法。其中，程序对误差的评判标准为训练过程产生的均方根误差RMSE，计算方法为： During the prediction process, the program will automatically select three optimization algorithms to optimize the support vector machine and select the algorithm with the smallest error as the optimization method for the current prediction. These three methods are cross-validation selection method, particle swarm optimization algorithm and genetic algorithm. Among them, the program’s criterion for judging the error is the root mean square error RMSE generated during the training process, and the calculation method is:

$R R M m S S E E. = = \sqrt{\frac{{Σ Σ}_{i i = = 11}^{L L} {((y the y - - \overset{^^}{y the y}))}^{22}}{L L}}$

其中L为训练样本数目，y和分别代表潮汐的训练数据和该部分的预报值。 where L is the number of training samples, y and Represent the training data of the tide and the predicted value of this part, respectively.

在支持向量机预报后就得到潮汐非天文潮部分的预测值，由于在预报之前对数据进行了AGO处理，所以在获得最终的预测结果前还需要进行数据的还原，即IAGO操作，在图2中显示为“获得预测结果并进行数据还原”，IAGO的计算方法为： After the prediction of the support vector machine, the predicted value of the non-astronomical part of the tide is obtained. Since the AGO processing is performed on the data before the prediction, it is necessary to restore the data before obtaining the final prediction result, that is, the IAGO operation, as shown in Figure 2 It is displayed as "obtain forecast results and perform data restoration", and the calculation method of IAGO is:

其中表示预测后还原的数据，代表预测后未还原的数据。 in Represents the restored data after prediction, Represents unrestored data after prediction.

将非天文潮部分的预报数据与天文潮部分调和分析法计算的数据做和得到最终的预报结果。即完成了整个预报程序的运行。 The final forecast result is obtained by summing the forecast data of the non-astronomical tide part and the data calculated by the harmonic analysis method of the astronomical tide part. That is to say, the operation of the whole forecasting program is completed.

以上所述，仅为本发明较佳的具体实施方式，但本发明的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本发明揭露的技术范围内，根据本发明的技术方案及其发明构思加以等同替换或改变，都应涵盖在本发明的保护范围之内。 The above is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto, any person familiar with the technical field within the technical scope disclosed in the present invention, according to the technical solution of the present invention Any equivalent replacement or change of the inventive concepts thereof shall fall within the protection scope of the present invention.

Claims

1. a modularity support vector machine tide prediction method based on GUI, it is characterised in that: include with Lower step:

S1: obtain the continuous sequence formula tidal level information that tidal gaging station monitors and the tide utilizing system of harmonic analysis to predict Nighttide value, arranges the time value of tide prediction；

S2: the tide measured value obtained at tidal level station does difference with the tide value of system of harmonic analysis prediction, obtains The time series of non-astronomical tide according to gray model AGO algorithm to input tidal level information and tide value number According to accumulation process, the data after process are used for the regression forecasting of support vector machine as input quantity；

S3: tide is predicted by the tide prediction temporal information according to arranging in S1 by support vector machine, During prediction, the penalty coefficient c and kernel function radius g of support vector machine are selected respectively cross validation, Population and genetic algorithm are optimized, and select support vector machine to produce error minima during training Method carries out tide prediction as optimized algorithm；

The result of S4: SVM prediction completes data convert, after reduction through the reverse accumulation process of IAGO Data for revising the tide prediction value of system of harmonic analysis.

A kind of modularity support vector machine tide prediction method based on GUI the most according to claim 1, It is further characterized in that: the tide measured value that tidal gaging station obtains is defined as the time series of non-astronomical tide, represents For actual measurement tide sequence y₀:x₀(1),x₀(2),…,x₀(n)；

The tidal data of system of harmonic analysis prediction, it is when representing actual tide level H (t) in somewhere, computational methods As follows:

H (t) = α_{0} + Σ_{j = 1}^{m} R_{j} c o s (σ_{j} t - θ_{j}) = α_{0} + Σ_{j = 1}^{m} (α_{j} {cosσ}_{j} t + b_{j} {sinσ}_{j} t)

α in formula₀For mean sea level height, R_jFor amplitude of component tide, θ_jFor the initial phase of partial tide, σ_jFor partial tide Angular velocity, α_j=R_jcosθ_j, b_j=R_jsinθ_j, m is the number of partial tide, is positive integer；System of harmonic analysis Forecast part regard astronomical tide part as, after calculating, the harmonic analysis time series that obtains is designated as y₁: x₁(1),x₁(2),…,x₁(n)；Will actual measurement tide sequence y₀When doing poor with harmonic analysis time series:

Computing formula is:

y₂=y₀-y₁

New sequence is:

y₂:x₃(1),x₃(2),…,x₃(n)

This sequence is firstly the need of the regression forecasting being used for support vector machine after AGO operates:

The sequence table processed through AGO is shown as x⁽¹⁾(1),x⁽¹⁾(2),…,x⁽¹⁾(n), computing formula is:

x^{(1)} (k) = Σ_{i = 1}^{k} x_{3} (k), k = 1, 2, ..., n

At the forecasting sequence obtained after SVM prediction it is:

{\hat{x}}^{(1)} (1), {\hat{x}}^{(1)} (2), ..., {\hat{x}}^{(1)} (n)

This sequence is through the algorithm non-final forecasting sequence of astronomical tide part of acquisition of IAGO:

{\hat{x}}^{(0)} (k + 1) = {\hat{x}}^{(1)} (k + 1) - {\hat{x}}^{(1)} (k), k = 1, 2, ..., n

WhereinThe data reduced after representing non-astronomical tide fractional prediction, final tidal data is by non-astronomical tide The predictive value of part is added with astronomical tide part and is finally predicted the outcome；

The time obtained is that the tide value of k+1 is expressed as

y (k + 1) = {\hat{x}}^{(0)} (k + 1) + x_{1} (k + 1) .