CN117494057A - Ship motion prediction method based on hierarchical multi-scale decomposition - Google Patents
Ship motion prediction method based on hierarchical multi-scale decomposition Download PDFInfo
- Publication number
- CN117494057A CN117494057A CN202311846304.5A CN202311846304A CN117494057A CN 117494057 A CN117494057 A CN 117494057A CN 202311846304 A CN202311846304 A CN 202311846304A CN 117494057 A CN117494057 A CN 117494057A
- Authority
- CN
- China
- Prior art keywords
- neural network
- decomposition
- component
- data
- matrix
- 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
Links
- 230000033001 locomotion Effects 0.000 title claims abstract description 79
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 55
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000013528 artificial neural network Methods 0.000 claims abstract description 44
- 238000005259 measurement Methods 0.000 claims abstract description 11
- 230000008859 change Effects 0.000 claims abstract description 7
- 230000006798 recombination Effects 0.000 claims abstract description 5
- 238000005215 recombination Methods 0.000 claims abstract description 5
- 239000011159 matrix material Substances 0.000 claims description 39
- 230000006870 function Effects 0.000 claims description 17
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 230000004044 response Effects 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 9
- 230000036961 partial effect Effects 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 239000013598 vector Substances 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000012217 deletion Methods 0.000 claims description 3
- 230000037430 deletion Effects 0.000 claims description 3
- 238000010238 partial least squares regression Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 2
- 238000005312 nonlinear dynamic Methods 0.000 abstract description 2
- 238000013277 forecasting method Methods 0.000 description 4
- 238000004088 simulation Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000001556 precipitation Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000002829 reductive effect Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- 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
- 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
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Linguistics (AREA)
- Computational Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Operations Research (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a ship motion prediction method based on hierarchical multi-scale decomposition, which relates to the technical field of ships and ocean engineering and comprises the following steps: s1, collecting ship motion actual measurement data; s2, performing information multi-scale decomposition on ship motion actual measurement data by using an empirical mode decomposition method, and performing discrete wavelet change on a high-frequency component to obtain a decomposed component; s3, carrying out neural network identification and prediction by using each decomposed component, dynamically adjusting the scale, the number of hidden nodes and the positions of the neural network, and respectively predicting each component through the variable structure neural network; s4, carrying out information recombination on the forecasting result of each component to obtain a final forecasting model. According to the invention, by using the nonlinear dynamic fitting capacity of the variable structure radial basis function neural network based on the sliding window and the dynamic reflecting capacity of the variable structure radial basis function neural network to the system state, more accurate marine vessel motion forecast can be obtained.
Description
Technical Field
The invention relates to the technical field of ships and ocean engineering, in particular to a ship motion forecasting method based on hierarchical multi-scale decomposition.
Background
The real-time prediction of the ship in the marine attitude plays an important role in ship steering in high storms, take-off and landing of an aircraft on a deck, retraction and extension of an underwater vehicle, offshore refuting, cable retraction and other offshore construction operations. However, when the ship sails on the sea, the ship is influenced by environmental factors such as wind, waves, currents and the like, and the swinging motion of the ship has the complex characteristics of nonlinearity, dynamic time variation and the like, so that the forecasting difficulty is increased.
The existing real-time prediction technology of the marine attitude is designed aiming at the ship motion complexity, and the nonlinear characteristic and the adaptivity of a nonlinear prediction model represented by a neural network are directly utilized, and the dynamic characteristic of the ship motion is fitted by dynamically adjusting the input-output mapping of the neural network by adjusting the connection weight of the neural network or the network structure, so that the purpose of ship motion identification and prediction is achieved; another method is to preprocess the ship motion data to reduce the difficulty of subsequent data processing, such as decomposing the ship motion data first and then identifying the decomposed data. Both methods can effectively reduce adverse effects of ship motion complexity and improve the precision of ship motion prediction.
However, existing modular models of marine vessel motion prediction have the following problems:
(1) It is difficult to obtain an accurate linear model of the ship motion. Because the ship motion online data are acquired in real time, serious multiple collinearity exists among partial variables, the serious uncertainty exists in the model parameter estimation, and the variance of the parameter estimation is close to infinity under serious conditions, so that the application of the parameter model in ship motion prediction is directly limited.
(2) The non-parametric model alone tends to be difficult to reflect the complex effects of time-varying factors on vessel motion. For example, dynamic changes of time-varying factors such as hydrologic (wave, ocean current, tide, temperature, salinity and density) weather (wind power, wind direction, air temperature, air pressure and precipitation) and the like of a sea area where a ship is located influence the time-varying state of the ship motion dynamics, static nonlinear prediction models such as neural networks with fixed network scale, structure and weight are difficult to reflect the time-varying dynamics of the ship motion, and a single variable structure neural network is also difficult to reflect the time-varying dynamics of the ship motion, so that the precision of ship motion prediction is difficult to ensure.
Disclosure of Invention
In view of the above, the present invention aims to provide a ship motion prediction method based on a hierarchical multi-scale decomposition, so as to solve the problem that it is difficult to obtain an accurate ship motion linear model in the prior art.
The invention adopts the following technical means:
a ship motion prediction method based on hierarchical multi-scale decomposition comprises the following steps:
s1, collecting ship motion actual measurement data;
s2, performing information multi-scale decomposition on ship motion actual measurement data by using an empirical mode decomposition method, and performing discrete wavelet change on a high-frequency component to obtain a decomposed component;
s3, carrying out neural network identification and prediction by using each decomposed component, dynamically adjusting the scale, the number of hidden nodes and the positions of the neural network, and respectively predicting each component through the variable structure neural network;
s4, carrying out information recombination on the forecasting result of each component to obtain a final forecasting model.
Further, the ship motion measured data comprises ship six-degree-of-freedom attitude information and ship speed information, and the ship six-degree-of-freedom attitude information comprises the longitudinal position of the shipXLateral positionYVertical positionZAngle of rollPitch angle->Head rocking angle->The ship speed information includes longitudinal speed +.>Lateral speed->Angular velocity of head shaker。
Further, S2 specifically includes the following steps:
s21, performing empirical mode decomposition on ship motion actual measurement data to obtain a plurality of subsequences, wherein the subsequences comprise connotation mode components and residual error components, and the formula is as follows:
wherein,nthe decomposition order automatically determined for the EMD algorithm,rthe initial angular velocity;
s22, performing discrete wavelet decomposition on the high-frequency components in the subsequences to obtain decomposed data, wherein the formula is as follows:
wherein,D iI1 is thatIMF 1 Is the first of (2)iThe component of the individual details is,Kfor the number of detail components,A I1 is thatIMF 1 Is a component of the approximation of (a).
Further, the decomposed components are:
。
further, S3 specifically includes the following steps:
s31, 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 receiving a new input-output data, a 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;
s32, after receiving new data samples in each step, updating a 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;
s33, 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 nodesSelectingk 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):
s34, after each step of hidden node determination, updating the connection weight value from the hidden layer to the output layer;
response matrix of the obtained radial basis function neural networkTo the output layerQThe connection weight of the (C) is obtained by a method of solving partial least square;
in response matrixAnd output matrixYPerforming partial least square regression operation; extracting principal component matrixTAfter that, the response matrix is->And output matrixYRespectively projected to the principal component matrixTThe radial basis function neural network based on partial least squares regression is obtained as follows:
wherein,Tis thatIs a matrix of principal components of (a);Wis->Is a conversion matrix of (a);Ris a regression coefficient matrix;Fis a residual matrix.
Further, in S4, the prediction model is as follows:
for motions with shorter periods such as roll, pitch and heave motions, and for roll anglesPitch angle->Head rocking angle->Carrying out direct forecasting; while longitudinal position, transverse position and yaw angle change relatively slowly, the heave velocity is first predicteduSpeed of swayingvAnd yaw raterAnd further based on heave velocityuSpeed of swayingvAnd yaw raterObtaining the longitudinal directionPredictive values for position, lateral position and yaw angle.
The invention also provides a storage medium comprising a stored program, wherein when the program runs, the ship motion forecasting method based on the hierarchical multi-scale decomposition is executed.
The invention also provides an electronic device, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor runs and executes any of the ship motion forecasting methods based on the hierarchical multi-scale decomposition through the computer program.
Compared with the prior art, the invention has the following advantages:
the method for decomposing the hierarchical multi-scale information fully excavates the dynamic information of the actually measured ship motion state time sequence. The model established by the method reflects the time-varying dynamics of the ship motion, and the time sequence in the component has stronger homogeneity through multi-layer information decomposition of the information through empirical mode decomposition and discrete wavelet deformation, so that the difficulty of identifying and predicting the time sequence is reduced; by utilizing the nonlinear dynamic fitting capability of the variable structure radial basis function neural network based on the sliding window and the dynamic reflecting capability of the variable structure radial basis function neural network to the system state, the motion forecast of the marine vessel can be obtained more accurately.
The method has strong adaptability. The decomposition order is automatically determined by using EMD decomposition, and the DWT method only needs to decompose the component with the highest frequency, so that the decomposition order is not needed to be manually determined, and the self-adaption capability of an algorithm is improved; in the selection of the forecasting model, the Lipschitz quotient method is utilized for automatic determination, and uncertainty and time consumption caused by manual determination are avoided as well; in the network structure and parameter determination of the variable structure neural network, the number and the positions of hidden nodes are determined by using the standard error descent rate parameter, and the connection weight from the hidden layer to the output layer of the network is determined by using the least squares method, so that the self-adaptability of the algorithm is effectively improved.
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 process of the present invention.
FIG. 2 is a schematic view of six degrees of freedom motion of the marine vessel according to the present invention.
FIG. 3 is a flow chart of the model training process of the present invention.
Fig. 4 is a diagram of simulation results of a variable structure neural network prediction algorithm based on multi-scale decomposition.
FIG. 5 is a scatter plot of the prediction results of the variable structure neural network prediction algorithm based on multi-scale decomposition.
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.
The invention relates to a real-time forecasting method for the attitude of a marine ship. The real-time prediction of the ship in the marine attitude plays an important role in ship steering in high storms, take-off and landing of an aircraft on a deck, retraction and extension of an underwater vehicle, offshore refuting, cable retraction and other offshore construction operations. However, when the ship sails on the sea, the ship is influenced by environmental factors such as wind, waves, currents and the like, and the swinging motion of the ship has the complex characteristics of nonlinearity, dynamic time variation and the like, so that the forecasting difficulty is increased.
The method comprises the steps of firstly, carrying out information multi-scale decomposition on ship motion actual measurement data, decomposing the ship motion data by using an empirical mode decomposition method, and carrying out further Discrete Wavelet Transform (DWT) on high-frequency components in the frequency aliasing phenomenon of the empirical mode decomposition method (EMD); identifying and forecasting the data in the sliding data window, dynamically adjusting the scale, the number of hidden nodes and the positions of the neural network, and forecasting each component through the variable structure neural network; and (5) carrying out information recombination on the forecasting result of each component to obtain a final forecasting model.
As shown in fig. 1-3, the invention provides a ship motion prediction method based on hierarchical multi-scale decomposition, which comprises the following steps:
s1, collecting ship motion actual measurement data;
the six-degree-of-freedom motion element data of the ship motion is collected as standby information, and comprises (1) six-degree-of-freedom attitude information of the ship, including the longitudinal position of the shipXLateral positionYVertical positionZAngle of rollPitch angle->Head rocking angle->The method comprises the steps of carrying out a first treatment on the surface of the (2) Marine vessel speed information, including longitudinalDirection speed->Lateral speed->Angular velocity of head shakerIs a data of (a) a data of (b).
S2, performing information multi-scale decomposition on ship motion actual measurement data by using an empirical mode decomposition method, and performing discrete wavelet change on a high-frequency component to obtain a decomposed component;
and carrying out multi-scale decomposition on the ship motion data by using an empirical mode decomposition method.
The motions of the vessel at sea are affected by a number of factors, which are superimposed on each other, so that the vessel motions exhibit complex motion characteristics. The original ship motion time sequence can be automatically divided into a plurality of subsequences by empirical mode decomposition, wherein the subsequences comprise inclusion mode componentsIMF 1 ,IMF 2 , …,IMF n ) Sum residual component [ ]r) This facilitates the input of the subsequence as a model. In the scheme, taking ship movement roll as an example, the roll angle of the ship movement is measuredAnd performing empirical mode decomposition to obtain an connotation mode component and a residual component.
Wherein,nthe decomposition order is automatically determined for the EMD algorithm.
Aiming at the frequency aliasing phenomenon in the component obtained by an Empirical Mode Decomposition (EMD), discrete Wavelet Transform (DWT) is carried out on the obtained high-frequency component, so that a data structure is optimized to be convenient for identification and forecast. As for thereinIMF 1 And (3) carrying out DWT decomposition on the components to obtain:
wherein,D iI1 is thatIMF 1 Is the first of (2)iThe component of the individual details is,Kis the number of detail components.A I1 Is thatIMF 1 Is a component of the approximation of (a). In the usual case, pairIMF 1 The DWT decomposition is performed to effectively optimize the component.
For direct connectionq-step prediction, recognition and prediction using neural networks. In the selection of the forecasting model, the input order of each component is determined by using Lipschitz quotient method. For example, forIMFIs the first of (2)jIndividual componentsIMF j Determining the input order asp j Then forqStep prediction, wherein the prediction model structure is as followsp-input-1-output. Then for the followingIs a training model of (a):
after the network nodes and the connection weights of the neural network are determined through training, the network nodes and the connection weights are used for prediction, and the input and output of a prediction model are as follows:
for DWT-decomposed IMF1, e.g.Di I1 The training process comprises the following steps:
wherein the method comprises the steps ofp Dj Is the firstjThe input order corresponding to each detail component is determined by Lipschitz quotient method.
The inputs and outputs of the prediction process are:
as for thereinIMF 1 In the case of DWT decomposition of components, the final prediction results are obtained by reorganizing the prediction results for each of the following components:
s3, carrying out neural network identification and prediction by using each decomposed component, dynamically adjusting the scale, the number of hidden nodes and the positions of the neural network, and respectively predicting each component through the variable structure neural network;
(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:
。
select pairsThose hidden nodes that contribute less than the set point are output. 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, updating the connection weight value from the hidden layer to the output layer.
Response matrix of the obtained radial basis function neural networkTo the output layerQThe connection weight of the (C) is obtained by a method of solving partial least square.
In response matrixAnd output matrixYPerforming partial least square regression operation; extracting principal component matrixTAfter that, the response matrix is->And output matrixYRespectively projected to the principal component matrixTThe radial basis function neural network based on partial least squares regression is obtained as follows:
wherein,Tis thatIs a matrix of principal components of (a);Wis->Is a conversion matrix of (a);Ris a regression coefficient matrix;Fis a residual matrix.
S4, carrying out information recombination on the forecasting result of each component to obtain a final forecasting model.
For motions with shorter periods such as rolling, pitching and heave motions, the scheme aims at the roll anglePitch angle->Head rocking angle->Carrying out direct forecasting; the longitudinal position, the transverse position and the initial rocking angle change relatively slowly, and the scheme firstly predicts the pitching speeduSpeed of swayingvAnd yaw raterAnd further based on heave velocityuSpeed of swayingvAnd yaw raterObtaining forecast values of longitudinal position, transverse position and head rocking angle:
。
examples
And carrying out a prediction simulation experiment on the actually measured roll data of a certain ship by using the neural network prediction algorithm based on decomposition multi-scale decomposition. The experiment was performed for 600 steps at 1s sampling intervals. The experimental results are shown in fig. 4 and 5.
In order to verify the effectiveness of the algorithm, a comparison experiment was performed between the algorithm and other predictive algorithms. The experimental comparison indexes are Root Mean Square Error (RMSE), correlation Coefficient (CC) and single step operation time. The experimental results are shown in table 1. The result shows that the algorithm has higher prediction precision and high operation speed, and is convenient for practical application.
TABLE 1 results of simulation tests for ship roll prediction
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 (8)
1. The ship motion prediction method based on the hierarchical multi-scale decomposition is characterized by comprising the following steps of:
s1, collecting ship motion actual measurement data;
s2, performing information multi-scale decomposition on ship motion actual measurement data by using an empirical mode decomposition method, and performing discrete wavelet change on a high-frequency component to obtain a decomposed component;
s3, carrying out neural network identification and prediction by using each decomposed component, dynamically adjusting the scale, the number of hidden nodes and the positions of the neural network, and respectively predicting each component through the variable structure neural network;
s4, carrying out information recombination on the forecasting result of each component to obtain a final forecasting model.
2. The method for predicting the motion of a vessel based on the hierarchical multi-scale decomposition according to claim 1, wherein the measured data of the motion of the vessel comprises six degrees of freedom attitude information of the vessel including a longitudinal position of the vessel and vessel speed informationXLateral positionYVertical positionZAngle of rollPitch angle->Head rocking angle->The ship speed information includes longitudinal speed +.>Lateral speed->Angular velocity of head shaker。
3. The ship motion prediction method based on the hierarchical multi-scale decomposition according to claim 1, wherein S2 specifically comprises the following steps:
s21, performing empirical mode decomposition on ship motion actual measurement data to obtain a plurality of subsequences, wherein the subsequences comprise connotation mode components and residual error components, and the formula is as follows:
wherein,nthe decomposition order automatically determined for the EMD algorithm,rthe initial angular velocity;
s22, performing discrete wavelet decomposition on the high-frequency components in the subsequences to obtain decomposed data, wherein the formula is as follows:
wherein,D iI1 is thatIMF 1 Is the first of (2)iThe component of the individual details is,Kfor the number of detail components,A I1 is thatIMF 1 Is a component of the approximation of (a).
4. A method of marine vessel motion prediction based on a hierarchical multi-scale decomposition according to claim 3, wherein said decomposed components are:
。
5. the ship motion prediction method based on the hierarchical multi-scale decomposition according to claim 1, wherein S3 specifically comprises the following steps:
s31, 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, 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;
s32, after receiving new data samples in each step, updating a 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;
s33, 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=WAObtain->；
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 nodesSelectingk 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):
s34, after each step of hidden node determination, updating the connection weight value from the hidden layer to the output layer;
response matrix of the obtained radial basis function neural networkTo the output layerQThe connection weight of the (C) is obtained by a method of solving partial least square;
in response matrixAnd output matrixYPerforming partial least square regression operation; extracting principal component matrixTAfter that, the response matrix is->And output matrixYRespectively projected to the principal component matrixTThe radial basis function neural network based on partial least squares regression is obtained as follows:
wherein,Tis thatIs a matrix of principal components of (a);Wis->Is a conversion matrix of (a);Ris a regression coefficient matrix;Fis a residual matrix.
6. The ship motion prediction method based on the hierarchical multi-scale decomposition according to claim 1, wherein in S4, the prediction model is as follows:
for motions with shorter periods such as roll, pitch and heave motions, and for roll anglesPitch angle->Head rocking angle->Carrying out direct forecasting; while longitudinal position, transverse position and yaw angle change relatively slowly, the heave velocity is first predicteduSpeed of swayingvAnd yaw raterAnd further based on heave velocityuSpeed of swayingvAnd yaw raterAnd obtaining forecast values of the longitudinal position, the transverse position and the head rocking angle.
7. A storage medium comprising a stored program, wherein the program, when run, performs the method of marine motion prediction based on a hierarchical multi-scale decomposition of any one of claims 1 to 6.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor is operative to perform the hierarchical multiscale decomposition-based vessel motion prediction method according to any one of claims 1 to 6 by means of the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311846304.5A CN117494057A (en) | 2023-12-29 | 2023-12-29 | Ship motion prediction method based on hierarchical multi-scale decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311846304.5A CN117494057A (en) | 2023-12-29 | 2023-12-29 | Ship motion prediction method based on hierarchical multi-scale decomposition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117494057A true CN117494057A (en) | 2024-02-02 |
Family
ID=89681408
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311846304.5A Pending CN117494057A (en) | 2023-12-29 | 2023-12-29 | Ship motion prediction method based on hierarchical multi-scale decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117494057A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117893575A (en) * | 2024-03-15 | 2024-04-16 | 青岛哈尔滨工程大学创新发展中心 | Ship motion prediction method and system with self-attention mechanism integrated by graph neural network |
| CN117893575B (en) * | 2024-03-15 | 2024-05-31 | 青岛哈尔滨工程大学创新发展中心 | Ship motion prediction method and system with self-attention mechanism integrated by graph neural network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109871977A (en) * | 2018-12-25 | 2019-06-11 | 广东电网有限责任公司 | Load forecasting method based on wavelet transform and the minimum support vector machines of optimization |
| CN113156815A (en) * | 2021-03-10 | 2021-07-23 | 广东海洋大学 | Data-driven marine ship motion attitude real-time forecasting method |
| CN116702095A (en) * | 2023-06-01 | 2023-09-05 | 大连海事大学 | Modularized marine ship motion attitude real-time forecasting method |
-
2023
- 2023-12-29 CN CN202311846304.5A patent/CN117494057A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109871977A (en) * | 2018-12-25 | 2019-06-11 | 广东电网有限责任公司 | Load forecasting method based on wavelet transform and the minimum support vector machines of optimization |
| CN113156815A (en) * | 2021-03-10 | 2021-07-23 | 广东海洋大学 | Data-driven marine ship motion attitude real-time forecasting method |
| CN116702095A (en) * | 2023-06-01 | 2023-09-05 | 大连海事大学 | Modularized marine ship motion attitude real-time forecasting method |
Non-Patent Citations (2)
| Title |
|---|
| JIANCHUAN YIN ETC.: ""An adaptive real-time modular tidal level prediction mechanism based on EMD and Lipschitz quotients method"", 《OCEAN ENGINEERING》, 11 November 2023 (2023-11-11), pages 3 * |
| 刘金培等: ""基于非结构数据和EMD-WTS二层分解的AQI组合预测方法"", 《重庆工商大学学报（自然科学版）》, vol. 38, no. 2, 30 April 2021 (2021-04-30), pages 1 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117893575A (en) * | 2024-03-15 | 2024-04-16 | 青岛哈尔滨工程大学创新发展中心 | Ship motion prediction method and system with self-attention mechanism integrated by graph neural network |
| CN117893575B (en) * | 2024-03-15 | 2024-05-31 | 青岛哈尔滨工程大学创新发展中心 | Ship motion prediction method and system with self-attention mechanism integrated by graph neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113156815B (en) | Data-driven marine ship motion attitude real-time forecasting method | |
| Corchado et al. | Hybrid artificial intelligence methods in oceanographic forecast models | |
| CN113837454B (en) | Ship three-degree-of-freedom hybrid neural network model prediction method and system | |
| Kawai et al. | Sea state estimation using monitoring data by convolutional neural network (CNN) | |
| CN112116162B (en) | Power transmission line icing thickness prediction method based on CEEMDAN-QFAOA-LSTM | |
| CN116702095B (en) | Modularized marine ship motion attitude real-time forecasting method | |
| Mittendorf et al. | Sea state identification using machine learning—A comparative study based on in-service data from a container vessel | |
| CN114445634A (en) | Sea wave height prediction method and system based on deep learning model | |
| CN112836893B (en) | Method for predicting ship oil consumption under severe sea conditions based on sea condition and ship navigation condition | |
| Meng et al. | Physics-guided generative adversarial networks for sea subsurface temperature prediction | |
| CN105836066A (en) | Method and system for predicting posture of ship moored at open type wharf | |
| CN112946784B (en) | SuperDARN radar convection diagram short-term forecasting method based on deep learning | |
| Cheng et al. | Spectralseanet: Spectrogram and convolutional network-based sea state estimation | |
| CN116021981A (en) | Method, device, equipment and storage medium for predicting ice coating faults of power distribution network line | |
| CN109389553A (en) | Meteorological causes isopleth interpolation method based on T batten | |
| Liu et al. | WaveNet: learning to predict wave height and period from accelerometer data using convolutional neural network | |
| CN116108995A (en) | Tidal river reach ship oil consumption prediction method and device and electronic equipment | |
| CN115859116A (en) | Marine environment field reconstruction method based on radial basis function regression interpolation method | |
| CN117104452A (en) | Method and system for inverting waves by ship-following swinging motion based on artificial neural network | |
| CN115470957A (en) | System, computer equipment and storage medium for predicting wave height of offshore waves during typhoon based on deep learning | |
| CN116306790B (en) | Offshore ship track real-time prediction method, system, equipment and medium based on CNN-GRU and attention mechanism | |
| Wang et al. | CNN-BiLSTM-attention model in forecasting wave height over South-East China Seas | |
| CN117494057A (en) | Ship motion prediction method based on hierarchical multi-scale decomposition | |
| CN116885711A (en) | Wind power prediction method, device, equipment and readable storage medium | |
| CN112085779A (en) | Wave parameter estimation method and device |
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 |