CN117518298A - Tidal combination forecasting method with self-adaptive model structure - Google Patents

Tidal combination forecasting method with self-adaptive model structure Download PDF

Info

Publication number
CN117518298A
CN117518298A CN202311846303.0A CN202311846303A CN117518298A CN 117518298 A CN117518298 A CN 117518298A CN 202311846303 A CN202311846303 A CN 202311846303A CN 117518298 A CN117518298 A CN 117518298A
Authority
CN
China
Prior art keywords
data
prediction
tidal
neural network
model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202311846303.0A
Other languages
Chinese (zh)
Inventor
孟祥坤
章文俊
杨雪
周翔宇
白伟伟
张国庆
吕红光
尹建川
曹亮
吴中岱
韩冰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Dalian Maritime University
Guangdong Ocean University
Shanghai Ship and Shipping Research Institute Co Ltd
Original Assignee
Dalian Maritime University
Guangdong Ocean University
Shanghai Ship and Shipping Research Institute Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Dalian Maritime University, Guangdong Ocean University, Shanghai Ship and Shipping Research Institute Co Ltd filed Critical Dalian Maritime University
Priority to CN202311846303.0A priority Critical patent/CN117518298A/en
Publication of CN117518298A publication Critical patent/CN117518298A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01W—METEOROLOGY
    • G01W1/00—Meteorology
    • G01W1/10—Devices for predicting weather conditions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
    • G06F18/256—Fusion techniques of classification results, e.g. of results related to same input data of results relating to different input data, e.g. multimodal recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Biophysics (AREA)
  • Computing Systems (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Evolutionary Biology (AREA)
  • Biomedical Technology (AREA)
  • Pure & Applied Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Environmental & Geological Engineering (AREA)
  • Databases & Information Systems (AREA)
  • Atmospheric Sciences (AREA)
  • Algebra (AREA)
  • Probability & Statistics with Applications (AREA)
  • Operations Research (AREA)
  • Biodiversity & Conservation Biology (AREA)
  • Ecology (AREA)
  • Environmental Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a tide combination forecasting method with a self-adaptive model structure, which comprises the following steps: s1, collecting actual measurement data of a target point position and a nearby point position; s2, establishing a harmonic constant model by using a harmonic analysis method; s3, calculating a first forecasting result output by the harmonic constant model, subtracting the first forecasting result from the actual tidal value of the measured data to obtain a difference value, and obtaining a subsequence; s4, establishing a variable structure neural network prediction model by using a dynamic orthogonal model selection algorithm, determining a subsequence prediction model structure by using a Lipschitz entropy method, and identifying and predicting a subsequence by using the variable structure neural network prediction model to obtain a second prediction result; and S5, adding the first forecasting result and the second forecasting result to obtain a final forecasting result. The method and the device reduce the defects of long time consumption, strong randomness, incapability of reflecting system dynamics and incapability of achieving the optimal performance caused by manually determining the model structure, and improve the prediction precision of the obtained model and the stability of an algorithm.

Description

Tidal combination forecasting method with self-adaptive model structure
Technical Field
The invention relates to the technical field of tide forecasting, in particular to a tide combination forecasting method with a self-adaptive model structure.
Background
Tidal phenomenon refers to periodic fluctuation motion of sea water under the action of the tidal force of celestial bodies, and the flow of sea water in the horizontal direction is called tidal current. The motive forces of earth tide, sea tide and qi tide are all caused by different gravitation of the earth on the day and month, and the three have mutual influence. The main reason for the tidal phenomenon is that the resultant force of the lunar attraction and the centrifugal force is the tidal force that causes the sea water to fluctuate. Tidal forecast occupies an important place in the exploitation and utilization of marine resources.
The traditional method of tidal prediction is harmonic analysis. According to the method, parameters of each tide in the model are obtained through statistics and analysis of long-term tide data of the tide station, and long-term tide forecast is obtained based on a mathematical model of the established tide. Since tides are affected by various factors, cyclic factors such as the tide guiding force, the inclination angle of the moon orbit, etc.; non-periodic factors such as wind, air pressure, precipitation, etc. The forecasting accuracy of the traditional harmonic analysis method is influenced by the quantity of receipts and the quantity of divided moisture, and the influence of non-periodic factors cannot be analyzed, so that the forecasting accuracy is low.
Tidal forecast with harmonic analysis has the following problems: (1) Because of the large number of tidal fractions required for harmonic analysis, the method requires long-term measured tidal data at one point to determine parameters of each tidal fraction, and the reality is that many tidal data blank areas exist in the current coastal sea area, and tidal data of response is required in some areas due to ocean engineering and ship navigation. Tidal forecast is often required with limited observation. (2) The model can not reflect non-periodic time-varying factors, such as the influence of the hydrological factors on the tide, limits the forecasting precision of the method on the tide, and particularly can lead to larger forecasting errors under the condition that the hydrological factors are severely changed, and often can lead to serious influence on navigation safety and shipping efficiency.
Disclosure of Invention
Therefore, the invention aims to provide a tide combination forecasting method with a self-adaptive model structure, so as to solve the technical problem that the conventional tide forecasting method cannot forecast the tide under the premise of limited observation.
The invention adopts the following technical means:
a tidal combination forecasting method with an adaptive model structure, comprising the steps of:
s1, collecting actual measurement data of a target point position and a nearby point position;
s2, establishing a harmonic constant model by using a harmonic analysis method;
s3, calculating a first forecasting result output by the harmonic constant model, subtracting the first forecasting result from the actual tidal value of the measured data to obtain a difference value, and performing multi-scale decomposition on the difference value to obtain a subsequence;
s4, establishing a variable structure neural network prediction model by using a dynamic orthogonal model selection algorithm, determining a subsequence prediction model structure by using a Lipschitz entropy method, and identifying and predicting a subsequence by using the variable structure neural network prediction model to obtain a second prediction result;
and S5, adding the first forecasting result and the second forecasting result to obtain a final forecasting result.
Further, the measured data includes tidal data including measured data of air pressure, wind force, wind direction, and air temperature, meteorological data including measured data of water temperature, precipitation, and salinity, and hydrological data.
Further, in S2, the harmonic constant model is:
wherein,is the average sea level height during analysis, +.>And->Correction values of the average amplitude and phase angle of the split tide, respectively, are needed by the periodic change of the moon orbit, +.>Is an intersection factor->Is the intersection correction angle, ++>Is the average amplitude of the moisture +.>Is the initial phase angle of the zero-division of Greenning, < +.>Is the angular velocity of moisture +.>Is the green's angle of delay.
Further, in S3, the difference is decomposed in multiple scales by using an empirical mode decomposition method, and the formula is as follows:
wherein,IMFin order to resolve the components of the component,ris the difference between the tidal forecast values.
Further, in S4, the specific steps of establishing the variable structure neural network prediction model by using the dynamic orthogonal model selection algorithm are as follows:
s41, establishing a sliding data window;
establishing a sliding data window to observe the state of the ship motion, and dynamically adjusting a fitting model based on a radial basis function neural network by utilizing input and output data updated in real time;
the sliding window is a fixed-width first-in first-out data sample sequence, when a new input-output data group is received, the new data group is added into the sliding window, and the earliest data group is moved out of the sliding window, so that the data group is stored in the sliding windowtSliding window of time of dayW SD Expressed as:
wherein,Lis the width of the sliding window; using input-output data sets within sliding windows, i.e. using input matrices respectivelyPAnd corresponding output vectorQTo represent the real-time dynamics of the mapping relationship:
in the method, in the process of the invention,n p the dimension of the input matrix;
using input matrices, respectivelyPAnd corresponding outputQAs the input and output of the radial basis function neural network, training and dynamically adjusting the neural network;
s42, after receiving new data samples in each step, updating the sliding data window, adding the latest samples into the window, deleting the earliest samples from the window, and directly adding the new data samples into the hidden layer to serve as a new hidden node;
s43, calculating response matrix of hidden layerWherein
Wherein,c j is the firstjThe center of the individual hidden node is the center,p i is the firstiA number of samples of the sample were taken,indicating Euclidean distance, ">Is the width of the basis function;Mto be the number of hidden nodesΦOrthogonal decomposition of vectors in (a) using Gram-Schmidt's lawΦ=WAObtaining；
Calculating an error drop rate:
；
normalized error rate of decrease:
；
selecting hidden nodes with sum of output contributions smaller than the set value until the sum of normalized error reduction rates of the selected hidden nodesSelect k 1 , ..., k S Constructing a set of pre-deletion hidden nodesS k ={/>}；
Take the continuous in the pastM S Intersection of step-selected set of hidden nodesIAnd delete atIIs a hidden node in (a):
s44, after each step of hidden node determination, updating the connection weight from the hidden layer to the output layer by using a least square method;
。
further, in S4, the step of obtaining the second prediction result specifically includes the following steps:
for each componentIMF j All utilize neural network to identify and forecast it, forIMF j Proceeding withqStep advanced prediction, while the prediction model input order is determined as follows by Lipschitz entropy methodpThenqThe structure of the step advance forecast is thatpInput-1-output, whereas neural network recognition process inputThe output is:
wherein,p j is to the firstjThe number of the IMF component neural network inputs is determined by a Lipschitz entropy value method according to the data, so that the adaptivity of the method is improved;
the inputs and outputs in the prediction process are respectively:
neural network predictor for residualFrom the predicted value reorganization for each component:
wherein,nthe decomposition order is automatically determined for EMD decomposition.
Further, in S4, the formula of the final forecasting result is as follows:
wherein,y H (t+q) As a result of the first forecast,and the second forecast result.
The invention also provides a storage medium comprising a stored program, wherein the program, when run, performs any of the methods of tidal combination forecasting with an adaptive model structure described above.
The invention also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes any tide combination forecasting method with the adaptive model structure through the computer program.
Compared with the prior art, the invention has the following advantages:
the invention automatically determines the order of decomposing the tidal harmonic analysis forecast remainder, namely the number of sub-sequences generated by decomposition by using an empirical mode decomposition method; the input order of a prediction model of the neural network prediction method is automatically determined according to the observation data by using a Lipschitz entropy value method; the number of hidden nodes of the network of the variable structure neural network is automatically determined by utilizing the standard error drop rate. The method for adaptively determining the model structure greatly reduces the defects of long time consumption, strong randomness, incapability of reflecting system dynamics and incapability of achieving the optimal performance caused by manually determining the model structure, and improves the prediction precision of the obtained model and the stability of an algorithm.
Compared with the tide forecasting by using the traditional harmonic analysis method, the method fully utilizes the capability of realizing accurate nonlinear fitting by using the radial basis function neural network, and can obtain more accurate tide forecasting. While the traditional forecasting method based on harmonic analysis can reflect the influence of the celestial body tide guiding force and give stable tide forecasting, the influence of time-varying aperiodic hydrological weather and other factors cannot be reflected, so that the situation of low forecasting precision occurs, and larger forecasting errors are easy to occur under the condition of large influence of external environment factors.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained according to the drawings without inventive effort to a person skilled in the art.
FIG. 1 is a flow chart of the prediction of the present invention.
FIG. 2 is a flow chart of the model training of the present invention.
FIG. 3 is a graph of the predicted result and the predicted error of the measured tidal and harmonic analysis method of the present invention.
FIG. 4 is a graph of IMF and residual sequences obtained from empirical mode decomposition of the present invention.
FIG. 5 is a graph of the input order of each sub-sequence prediction model obtained by Lipschitz entropy method according to the present invention.
FIG. 6 is a graph of the predicted result of the present invention using a variable structure neural network to obtain the residual value.
Fig. 7 is a graph of the prediction result of each component using the neural network according to the present invention.
Detailed Description
In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
As shown in fig. 1, the present invention provides a tidal combination forecasting method with an adaptive model structure, comprising the steps of:
s1, collecting actual measurement data of a target point position and a nearby point position;
the method comprises the steps of collecting actual measurement tide and hydrological element data of a target point position and a nearby point position as standby information, wherein the actual measurement tide and hydrological element data comprise (1) tide data comprising actual measurement tide data of the target point position and the nearby point position; (2) meteorological data including measured data of air pressure, wind force, wind direction and air temperature; (3) hydrologic data including measured data of water temperature, precipitation, salinity.
S2, establishing a harmonic constant model by using a harmonic analysis method;
and under the condition that the data volume can support to establish an accurate harmonic analysis forecasting model, establishing a harmonic constant model by using a harmonic analysis method.
The actual tide level at a certain location can be expressed as
Wherein the method comprises the steps ofIs the average sea level height during analysis, +.>And->The correction values of the average amplitude and phase angle of the split tide are respectively made according to the period change of the moon orbit, wherein +.>Called intersection factor->Called the intersection correction angle. />Is the average amplitude of the moisture +.>Is moisture separation when GreenningAngle of beginnings, 10>Is the angular velocity of moisture +.>Is the green's angle of delay. The wind power generation system is a non-astronomical tide part, and the factors causing the change of the part are meteorological factors such as wind power, air pressure and the like, so that the change of the part has extremely strong randomness and nonlinearity, and can be regarded as noise in physics.
Average amplitude of moistureAnd Greennine delay angle->Known as the harmonic constants of the actual tidal split, are a response of the ocean to the effects of periodically varying external forces, determined by the kinematic nature of the ocean itself. The value of the harmonic constant varies from sea to sea, but for a fixed location, the harmonic constant has extremely high stability because of the long period of variation of the marine overall environment, and can be regarded as a constant approximately over a long period of time.
The harmonic constant can be obtained by a harmonic analysis method according to the observation data of the actual tide, and the prediction of the local tide can be realized after the harmonic constant of the tide in a certain place is mastered.
To achieve tidal forecast, we require the harmonic constants of the divided tidesAnd->For calculation of the harmonic constants, a least square method is generally employed. If not taking into consideration non-astronomical tidesr(k) Influence of the fraction and average sea surface +.>Regarded as->Is a special moisture, the tidal height expression can be written as
Order the，/>，
Taking out，/>。
Wherein,the above formula can be expressed as:
by analyzing the measured tide level data, each tide can be obtainedA，BFurther calculate the moisture-separating amplitudeRAnd initial phase angleθFinally, the harmonic constant is obtained. The specific expression is as follows
WhileH=R/f，。
S3, calculating a first forecasting result output by the harmonic constant model, subtracting the first forecasting result from the actual tidal value of the measured data to obtain a difference value, and performing multi-scale decomposition on the difference value to obtain a subsequence;
algorithm: empirical mode decomposition
The Empirical Mode Decomposition (EMD) is a data-driven multi-scale signal processing method, which decomposes signals according to time scale characteristics of the signals, does not need to perform priori verification on the signals according to signal characteristics, and has strong self-adaptability. EMD will original signalx(k) Decomposition into eigenmode functions (IMFs)A residual signalr(k)：
Giving an original signal, the standard EMD algorithm is calculated as follows:
1) Locate allx(k) Is defined by the extreme values of (a).
2) At all minima as the lower signal envelopee min (k) And all maxima as upper signal envelopee max (k) Respectively performing interpolation.
3) Calculating the average value of the upper and lower envelopesm(k) =(e min (k) +e max (k))/2。
4) Subtracting the average extracted details from the original signal:s(k) =x(k)−m(k)。
5) If it iss(k) Meets the stop standard of calculation, then setd(k) =s(k) For IMF, otherwise setx(k) =s(k) And iterating the residual error of S1.
After initial IMF validation, iterate to residualTo obtain the remaining IMFs.
S4, establishing a variable structure neural network prediction model by using a dynamic orthogonal model selection algorithm, determining a subsequence prediction model structure by using a Lipschitz entropy method, and identifying and predicting a subsequence by using the variable structure neural network prediction model to obtain a second prediction result;
and carrying out multi-scale empirical mode decomposition on the difference value output by the measured data and the harmonic analysis method.
Tidal changes are affected by environmental factors such as various hydrologics, which are superimposed on each other, so that tidal movements exhibit complex movement characteristics. By empirical mode decomposition, the original ship motion time sequence can be adaptively changedx(t) Divided into sub-sequences, in this case, the differences between measured tidal data and tidal forecast values obtained or obtained using harmonic analysisrAnd decomposing by using an empirical mode method to obtain an approximate component and a detail component.
And determining the order of the time sequence forecast of each component by calculating Lipschitz coefficients and the like, and after determining the input/output order, carrying out the time sequence forecast of each component by using a variable-structure radial basis function neural network.
The method comprises the following steps of establishing a variable structure neural network prediction model.
(1) Establishing a sliding data window
The motion of the ship at sea has the characteristic of dynamic time variation, in order to reflect the latest ship motion state, a sliding data window is established to observe the ship motion state, and a fitting model based on a radial basis function neural network is dynamically adjusted by utilizing input and output data updated in real time;
the sliding window is a fixed-width first-in first-out data sample sequence, when a new input-output data set is received, the new data set is added into the sliding window, and the earliest data set is moved out of the sliding window. Will betSliding window of time of dayW SD Expressed as:
wherein the method comprises the steps ofLIs the width of the sliding window; using input-output data sets within sliding windows, i.e. using input matrices respectivelyPAnd corresponding output vectorQTo represent the real-time dynamics of the mapping relationship:
in the method, in the process of the invention,n p the dimension of the input matrix;
using input matrices, respectivelyPAnd corresponding outputQAs an input and an output of the radial basis function neural network, training and dynamic adjustment are performed on the neural network.
(2) After each step receives a new data sample, the sliding data window is updated, the latest sample is added to the window, and the earliest sample is deleted from the window. The new data sample is directly added into the hidden layer to serve as a new hidden node.
(3) Calculating response matrix of hidden layerWherein
。
Wherein the method comprises the steps ofc j Is the firstjThe center of the individual hidden node is the center,p i is the firstiA number of samples of the sample were taken,indicating Euclidean distance, ">Is the width of the basis function;Mis the number of hidden nodes. Will beΦOrthogonal decomposition of vectors in (a) using Gram-Schmidt's lawΦ=WAObtaining。
Calculating an error drop rate:
。
normalized error rate of decrease:
。
those hidden nodes that contribute to the output and that are less than the set point are selected. Up to the sum of the normalized error rate of the selected hidden nodes. Selection ofk 1 , ...,k S Constructing a set of pre-deletion hidden nodesS k ={/>}。
Take the continuous in the pastM S Intersection of step-selected set of hidden nodesIAnd delete atIIs a hidden node in (a):
。
(4) after each step of hidden node determination, the connection weight from the hidden layer to the output layer is updated by using a least square method for convenience.
。
Algorithm: lipschitz entropy method
The Lipschitz entropy method is a simple and reliable data-driven prediction model order determining method, is used for identifying input-output mapping of an unknown nonlinear dynamic system, does not need priori knowledge of any learning process, is suitable for determining the input order of a neural network prediction model, and is used for adaptively determining the input order of the model.
Consider a nonlinear input-output mapping that is continuously smooth over the region:
the Lipschitz entropy method is based on the persistence of a nonlinear persistence dynamic system, assuming that the partial derivatives of the mapping to its parameters are bounded.
Wherein the method comprises the steps ofMIs a positive number, and the Lipschitz quotient is defined as follows:
wherein,is the distance between two points in the input space, < >>Is the distance between two points in the output space.
Wherein,q (n) (r) Is thatrq (n) ij Middle (f)rThe large quotient value is obtained by the method,gis [0.01 ]N, 0.02N]A natural number in the interval of the two,Nfor the number of samples, the model input ordernAutomatically determining when the following condition is satisfied:
and S5, adding the first forecasting result and the second forecasting result to obtain a final forecasting result.
After forecasting by using a harmonic analysis method and a variable structure radial basis function neural network, overlapping the two forecasting resultsAnd obtaining a final modular ship motion forecast result.
The invention relates to a real-time tide prediction algorithm, which comprises the steps of firstly obtaining a prediction value of tide by using a harmonic analysis method in each step of predictiony H (t) Then utilize the measured valuey(t) Subtracting the predicted value obtained by the harmonic analysis method to obtain the residual value of tideR(t)：
The remainder still contains rich information, and decomposing the remainder is helpful for information mining. The Empirical Mode Decomposition (EMD) method can autonomously determine the decomposition order and can adaptively decompose the residual value:
for each componentIMF j The neural network is utilized to identify and forecast the same. In particular, forIMF j Proceeding withqStep advanced prediction, while the prediction model input order is determined as follows by Lipschitz entropy methodpThenqThe structure of the step advance forecast is thatp-input-1-output, whereas neural network recognition process input-output is:
wherein,p j is to the firstjThe number of the IMF component neural network inputs is determined by a Lipschitz entropy value method according to the data, so that the adaptivity of the method is improved.
The inputs and outputs in the prediction process are respectively:
neural network predictor for residualFrom the predicted value reorganization for each component:
wherein,nthe decomposition order is automatically determined for EMD decomposition.
After obtaining the predicted value of the neural network, the predicted value obtained by the harmonic analysis methody H (t+q) Residual value predicted value obtained by neural networkAnd adding to obtain a final predicted value.
Examples
And selecting actual measurement tidal data of a port to simulate and verify the algorithm. The results of the measured tides, the predicted results using the harmonic analysis and the resulting residual sequences are shown in fig. 3. And decomposing the residual value by using an empirical mode decomposition method, automatically determining that the decomposition order is 9, and decomposing to obtain 9 IMFs and 1 residual sequences, wherein the sequences are shown in figure 4. The Lipschitz entropy method is used to determine the input order corresponding to each IMF, as shown in FIG. 5. The prediction is performed by using the variable structure neural network by using the input-output map determined by the order, and a predicted value of the residual value is obtained, as shown in fig. 6. The predicted value is obtained by superimposing predictions for each component, and the predicted result for each component is shown in fig. 7. In order to verify the effectiveness of the algorithm, simulation verification is performed by using the same measured data by using a harmonic analysis method, a BP neural network method and a support vector machine method, and the results are shown in Table 1. Wherein the precision index adopts Root Mean Square Error (RMSE) and Correlation Coefficient (CC), and the operation efficiency index adopts single-step operation time.
Table 1 tidal forecast simulation test results table
Prediction algorithm RMSE(m) CC Single step time(s)
Harmonic analysis method 0.215287 0.862387 ――――
BP neural network 0.0681157 0.912931 0.738772
Support vector machine 0.0619586 0.913233 0.091719
Adaptive structure prediction model 0.0328030 0.977497 0.071602
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention.

Claims (9)

1. A tide combination forecasting method with an adaptive model structure is characterized by comprising the following steps:
s1, collecting actual measurement data of a target point position and a nearby point position;
s2, establishing a harmonic constant model by using a harmonic analysis method;
s3, calculating a first forecasting result output by the harmonic constant model, subtracting the first forecasting result from the actual tidal value of the measured data to obtain a difference value, and performing multi-scale decomposition on the difference value to obtain a subsequence;
s4, establishing a variable structure neural network prediction model by using a dynamic orthogonal model selection algorithm, determining a subsequence prediction model structure by using a Lipschitz entropy method, and identifying and predicting a subsequence by using the variable structure neural network prediction model to obtain a second prediction result;
and S5, adding the first forecasting result and the second forecasting result to obtain a final forecasting result.
2. The tidal combination prediction method with adaptive model structure of claim 1, wherein the measured data comprises tidal data, meteorological data, and hydrological data, the meteorological data comprising measured data of barometric pressure, wind force, wind direction, and air temperature, the hydrological data comprising measured data of water temperature, precipitation, and salinity.
3. The tidal combination prediction method with adaptive model structure according to claim 1, wherein in S2, the harmonic constant model is:
wherein,is the average sea level height during analysis, +.>And->Correction values of the average amplitude and phase angle of the split tide, respectively, are needed by the periodic change of the moon orbit, +.>Is an intersection factor->Is the intersection correction angle, ++>Is the average amplitude of the moisture +.>Is the initial phase angle of the zero-division of Greenning, < +.>Is the angular velocity of moisture +.>Is the green's angle of delay.
4. The tidal combination prediction method with the adaptive model structure according to claim 1, wherein in S3, the difference is subjected to multi-scale decomposition by using an empirical mode decomposition method, and the formula is as follows:
wherein,IMFin order to resolve the components of the component,ris the difference between the tidal forecast values.
5. The tidal combination prediction method with adaptive model structure according to claim 1, wherein in S4, the specific steps of establishing the variable structure neural network prediction model using the dynamic orthogonal model selection algorithm are as follows:
s41, establishing a sliding data window;
establishing a sliding data window to observe the state of the ship motion, and dynamically adjusting a fitting model based on a radial basis function neural network by utilizing input and output data updated in real time;
the sliding data window is a fixed-width first-in first-out data sample sequence, when a new input-output data set is received, the new data set is added into the sliding window, and the earliest data set is moved out of the sliding window, so that the data set is stored in the sliding windowtSliding window of time of dayW SD Expressed as:
wherein,Lis the width of the sliding window; using input-output data sets within sliding windows, i.e. using input matrices respectivelyPAnd corresponding output vectorQTo represent the real-time dynamics of the mapping relationship:
in the method, in the process of the invention,n p the dimension of the input matrix;
using input matrices, respectivelyPAnd corresponding outputQAs the input and output of the radial basis function neural network, training and dynamically adjusting the neural network;
s42, after receiving new data samples in each step, updating the sliding data window, adding the latest samples into the window, deleting the earliest samples from the window, and directly adding the new data samples into the hidden layer to serve as a new hidden node;
s43, calculating response matrix of hidden layerWherein
Wherein,c j is the firstjThe center of the individual hidden node is the center,p i is the firstiA number of samples of the sample were taken,indicating Euclidean distance, ">Is the width of the basis function;Mto be the number of hidden nodesΦOrthogonal decomposition of vectors in (a) using Gram-Schmidt's lawΦ=WAObtaining；
Calculating an error drop rate:
；
normalized error rate of decrease:
；
selecting hidden nodes with sum of output contributions smaller than the set value until the sum of normalized error reduction rates of the selected hidden nodesSelect k 1 , ..., k S Constructing a set of pre-deletion hidden nodesS k ={/>}；
Take the continuous in the pastM S Intersection of step-selected set of hidden nodesIAnd delete atIIs a hidden node in (a):
s44, after each step of hidden node determination, updating the connection weight from the hidden layer to the output layer by using a least square method;
。
6. the method for tidal combination prediction with adaptive model structure according to claim 1, wherein in S4, obtaining the second prediction result specifically comprises the steps of:
for each componentIMF j All utilize neural network to identify and forecast it, forIMF j Proceeding withqStep advanced prediction, while the prediction model input order is determined as follows by Lipschitz entropy methodpThenqThe structure of the step advance forecast is thatp-input-1-output, whereas neural network recognition process input-output is:
wherein,p j is to the firstjThe number of the IMF component neural network inputs is determined by a Lipschitz entropy value method according to the data, so that the adaptivity of the method is improved;
the inputs and outputs in the prediction process are respectively:
neural network predictor for residualFrom the predicted value reorganization for each component:
wherein,nthe decomposition order is automatically determined for EMD decomposition.
7. The tidal combination prediction method with adaptive model structure according to claim 1, wherein in S5, the formula of the final prediction result is as follows:
wherein,y H (t+q) As a result of the first forecast,and the second forecast result.
8. A storage medium comprising a stored program, wherein the program, when run, performs the tidal combination forecasting method with adaptive model structure of any one of claims 1 to 7.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor is operable by the computer program to perform the tidal combination forecasting method with adaptive model structure of any one of claims 1 to 7.
CN202311846303.0A 2023-12-29 2023-12-29 Tidal combination forecasting method with self-adaptive model structure Pending CN117518298A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311846303.0A CN117518298A (en) 2023-12-29 2023-12-29 Tidal combination forecasting method with self-adaptive model structure

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311846303.0A CN117518298A (en) 2023-12-29 2023-12-29 Tidal combination forecasting method with self-adaptive model structure

Publications (1)

Publication Number Publication Date
CN117518298A true CN117518298A (en) 2024-02-06

Family

ID=89745966

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311846303.0A Pending CN117518298A (en) 2023-12-29 2023-12-29 Tidal combination forecasting method with self-adaptive model structure

Country Status (1)

Country Link
CN (1) CN117518298A (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110146293A (en) * 2019-06-04 2019-08-20 昆明理工大学 A kind of Fault Diagnosis of Roller Bearings based on PCA and ELM
CN112906955A (en) * 2021-02-04 2021-06-04 广东海洋大学 Tidal data fitting method based on multi-source data driving
CN113156815A (en) * 2021-03-10 2021-07-23 广东海洋大学 Data-driven marine ship motion attitude real-time forecasting method
CN116822336A (en) * 2023-06-01 2023-09-29 大连海事大学 Multi-model combined tide forecasting method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110146293A (en) * 2019-06-04 2019-08-20 昆明理工大学 A kind of Fault Diagnosis of Roller Bearings based on PCA and ELM
CN112906955A (en) * 2021-02-04 2021-06-04 广东海洋大学 Tidal data fitting method based on multi-source data driving
CN113156815A (en) * 2021-03-10 2021-07-23 广东海洋大学 Data-driven marine ship motion attitude real-time forecasting method
CN116822336A (en) * 2023-06-01 2023-09-29 大连海事大学 Multi-model combined tide forecasting method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
刘培国，覃宇建: "《电磁兼容现场测量与分析技术》", 30 June 2013, 国防工业出版社, pages: 110 - 114 *

Similar Documents

Publication Publication Date Title
Gao et al. Hourly forecasting of solar irradiance based on CEEMDAN and multi-strategy CNN-LSTM neural networks
Mendoza et al. An intercomparison of approaches for improving operational seasonal streamflow forecasts
CN107944648B (en) Large ship speed and oil consumption rate prediction method
US20170338802A1 (en) Actually-measured marine environment data assimilation method based on sequence recursive filtering three-dimensional variation
CN103743402B (en) A kind of underwater intelligent self adaptation Approach of Terrain Matching of topographic information based amount
CN113837454B (en) Ship three-degree-of-freedom hybrid neural network model prediction method and system
CN114330935B (en) New energy power prediction method and system based on multiple combination strategies integrated learning
CN112946784B (en) SuperDARN radar convection diagram short-term forecasting method based on deep learning
CN116595394A (en) Training method of wind speed correction model, wind speed prediction method, wind speed prediction equipment and medium
CN112668238A (en) Rainfall processing method, device, equipment and storage medium
Girimurugan et al. Application of Deep Learning to the Prediction of Solar Irradiance through Missing Data
CN116021981A (en) Method, device, equipment and storage medium for predicting ice coating faults of power distribution network line
CN115186923A (en) Photovoltaic power generation power prediction method and device and electronic equipment
CN116822336B (en) Multi-model combined tide forecasting method
CN114676621A (en) Method for improving accuracy of land water reserve abnormity based on deep learning weight load
Suparta et al. Modeling of precipitable water vapor using an adaptive neuro-fuzzy inference system in the absence of the GPS network
CN117526274A (en) New energy power prediction method, electronic equipment and storage medium in extreme climate
CN111811465A (en) Method for predicting sea wave effective wave height based on multi-sine function decomposition neural network
CN117518298A (en) Tidal combination forecasting method with self-adaptive model structure
CN111753469B (en) Typhoon storm surge situation simulation method and device
CN115204073A (en) Tide prediction method based on kriging interpolation
CN115034159A (en) Power prediction method, device, storage medium and system for offshore wind farm
Junaida et al. A method to predict heavy precipitation using the artificial neural networks with an application
Ma et al. A Method for Establishing Tropospheric Atmospheric Refractivity Profile Model Based on Multiquadric RBF and k-means Clustering
CN113449920A (en) Wind power prediction method, system and computer readable medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination