A rapid prediction method for the environmental response of ships sailing in ice areas

Abstract

The present invention provides a rapid prediction method for the environmental response of ships sailing in ice areas, comprising the following steps: S1. Establish a mathematical model for the motion response of a ship under environmental disturbances of ice, wind and waves at sea; S2. Establish a prediction model for ice conditions and sea conditions in the ship's sailing area; S3. Enrich the database of the environmental motion and structural response of the ship sailing in ice areas or perform data assimilation according to measured information; S4. According to the planned route of the ship and the prediction results of the regional ice area environmental elements, conduct refined real-time prediction of the ice area environmental elements in the ship's route area hour by hour within 24 hours. In the prediction process, multi-level decomposition and prediction are adopted, the prediction is decomposed into different time scales for prediction, and different predictions are summed up to obtain the final prediction result. The present invention can quickly and accurately obtain the motion and structural response of a ship under the influence of a complex ice area environment.

Description

Technical Field

[0001] The present invention relates to the technical field of ice navigation, and particularly to a rapid prediction method for the environmental response of ships navigating in ice areas. Background Art

[0002] The real-time prediction of the sea attitude of ships plays an important role in ship steering in high winds and waves, the takeoff and landing of aircraft on the deck, the retraction and deployment of underwater vehicles, offshore transshipment, cable retraction and deployment, and other offshore construction operations. However, when a ship is sailing at sea, affected by environmental factors such as wind, waves, and currents, the rolling motion of the ship has complex characteristics such as non-linearity and dynamic time-variation, which increases the difficulty of its prediction.

[0003] Currently, the numerical simulation method is adopted for the motion response of ships during ice navigation. Due to the large amount of calculation, this method is difficult to apply in real time and cannot be applied to scenarios such as the online assessment of ice area ship risks and the online adjustment of shipping routes. In addition, the current prediction methods have problems such as a long general environmental prediction period and difficulty in real-time application. There are also problems such as strong uncertainty in environmental element prediction, resulting in poor prediction accuracy of environmental elements. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to propose a rapid prediction method for the environmental response of ships navigating in ice areas to solve the technical problem of poor prediction accuracy of the environmental response of existing ships navigating in ice areas.

[0005] The technical means adopted by the present invention are as follows:

[0006] A rapid prediction method for the environmental response of ships navigating in ice areas includes the following steps:

[0007] S1. Establish a mathematical model of the motion response and ship structure response of a ship under environmental disturbances of ice, wind, and waves at sea;

[0008] S2. Establish an ice condition and sea state prediction model for the ship's navigation area, and use a spatio-temporal generative adversarial neural network model with an attention mechanism to realize the assimilation and fusion of observed data and high-resolution model prediction data, and perform dynamic update prediction of the external ice, wind, wave, and current fields to obtain the initial marine environmental prediction result of the sea area passed by the ship;

[0009] S3. Enrich the environmental motion and structure response database for the ship's ice area navigation according to the real-time measured ice conditions, wind, wave, and current information, as well as the ship's motion and structure response information, or update the ice condition and sea state prediction model database for the ship's navigation area according to the measured information of wind, wave, and current elements and the ice load inversion information of the ship's structure response to obtain the real-time marine environmental prediction result;

[0010] S4. According to the planned route of the ship and the real-time ocean environmental forecast results, conduct refined real-time forecasting of ice zone environmental elements in the ship's route area hour by hour within 24 hours. During the forecasting process, adopt multi-level decomposition and prediction, decompose the prediction into different time scales for forecasting, and sum up different forecasts to obtain the final forecast result;

[0011] S5. According to the sea area passed by the ship's planned route, calculate the motion response and structural response of the ship passing through the ice zone, wind field and wave field in this sea area through the motion response and structural response mathematical models; the inputs of the motion response and structural response mathematical models include ship element information, environmental element information and dynamic element information, and the outputs of the motion response and structural response mathematical models are the ship's speed, motion attitude and structural response under the influence of the set ice, wind and waves;

[0012] S6. Based on the above-mentioned motion responses and structural responses of the ship such as speed and motion attitude under the influence of ice, wind and waves, and combined with the final forecast result, comprehensively conduct response forecasting for the ship sailing in the ice zone.

[0013] Furthermore, S1 specifically includes the following steps:

[0014] S11. Set the input of the training model and the output of the training model; the input of the training model includes ship element information, environmental element information and dynamic element information; the output of the training model includes the ship's speed, motion attitude and structural response under the influence of ice, wind and waves;

[0015] S12. Based on the model input and model output, establish a model input-output training sample through the method of numerical simulation of the ship's motion and structural response in ice, and calculate the response of a specific ship to environmental elements in terms of ship speed, motion attitude and structure; assume that there are m samples in the sample set, the input vector x has n independent variables, the input X is an m×n-dimensional matrix, the output vector y has l dependent variables, and the output Y is an m×l-dimensional matrix;

[0016] S13. According to the input and output data of the samples obtained in S12, offline train the neural network to make the neural network accurately approximate the input-output mapping, and establish a neural network mapping model of the ship's navigation motion and structural response model.

[0017] Furthermore, the ship element information includes ship main dimensions, structural parameters, hull parameters such as bow column inclination angle, superstructure layout, parameters of the ship's ice class, ship draft and draft difference, and propeller scale; the environmental element information includes hydrometeorological environmental elements of ice, wind and waves; the dynamic element information includes the relative position and motion relationship between the ship and environmental elements, propeller speed, and rudder angle.

[0018] Furthermore, S13 specifically includes the following steps:

[0019] S131. Generate random numbers for sampling from the input and output sample sets X and Y according to the uniform distribution sampling method. In the second step, extract the first set of data x1 and y1, in the second step, extract the second set of data x2 and y2, and so on. In the k-th step, extract the data x k and y k , and construct the training sample sets U ∈ R k×n and V ∈ R k×l ;

[0020] S132. Use the input and output training sample sets U and V to construct a radial basis function neural network, and use partial least squares method to perform regression calculations on the hidden layer response matrix and the output:

[0021] In the RBF network, use the Gaussian function as the radial basis function, and the vector u j (j = 1,…,k) becomes the center of the hidden node; after calculation by the radial basis function, the hidden layer response matrix A ∈ R k×k is obtained, where the element in the i-th row and j-th column is:

[0022]

[0023] where, u i is the i-th sample, here ||·|| represents the Euclidean distance, c j and σ j are the center and width of the j-th basis function respectively;

[0024] S133. Perform PLS principal component extraction and regression operations between the hidden layer response matrix A and the output matrix V;

[0025] After extracting the principal component T, project A and V onto the principal component matrix respectively, and obtain:

[0026] V = TR + F = AWR + F

[0027] where, T is the principal component matrix of A, which is a k×n T dimensional matrix; W is the transformation matrix of A, which is a k×n T dimensional matrix; R is an n T ×l dimensional regression coefficient matrix; F is a k×l dimensional residual matrix;

[0028] Obtain the generalization model:

[0029]

[0030] where A p1 and W1 are k×k dimensional and k×n T dimensional matrices respectively, and R1 is an n TAn \(l\)-dimensional matrix is used to generalize the samples in the set \(U\) by the obtained network. A precision threshold is set to determine the number of times \(n\) of principal component extraction. T Stop extracting principal components when the change in the root mean square error of generalization is less than the set precision change threshold \(e1\):

[0031]

[0032] S134. Generalize the obtained network to the training data in the entire data set \(X\), and calculate the generalization error for the entire data set:

[0033]

[0034] Where:

[0035]

[0036] S135. If RMSE O is greater than the set precision threshold \(e2\), then repeat steps S231 - S234;

[0037] S136. If RMSE O is less than the set precision threshold \(e2\), then end;

[0038] S137. The number of principal components \(n\) T and the number of loop steps \(k\) are the number of times of principal component extraction and the number of hidden nodes of the finally obtained radial basis function network. The vectors in the set \(U\) are the hidden nodes of the obtained radial basis function network.

[0039] Furthermore, S2 specifically includes:

[0040] Select the coverage ranges of ice areas, wind fields, wave fields, and current fields according to the sea areas passed by the ship's planned route. Select high - resolution ice condition and hydro - meteorological environment forecasting models. Combine the ice condition information obtained by identifying the ship's measured video and image information and the real - time observed hydro - meteorological environment data information for data assimilation. Use a spatio - temporal generative adversarial neural network with an attention mechanism to achieve the assimilation and fusion of observed data and high - resolution model forecast data, and improve the accuracy of hydro - meteorological forecasting in the navigation area;

[0041] Use the principal component analysis method to perform dimensionality reduction and denoising processing on the meteorological field data, remove redundant noise signals, retain the main component feature vectors of the meteorological field, reconstruct the main feature vectors after dimensionality reduction and denoising processing into meteorological field data, and construct an ocean environment forecasting model based on depth - separable convolution; among them, the depth - separable convolution module is used to improve the prediction efficiency and performance, and at the same time reduce the computational amount and hardware memory requirements of the model. The principal component analysis method filters the evolution characteristics of external meteorological elements efficiently by reducing the data dimension and redundant noise effect, and improves the forecasting accuracy.

[0042] Using the principal component analysis method, data cleaning and reconstruction processing are carried out on the data of external wind waves, currents and sea ice coverage areas through the empirical orthogonal decomposition theory. The principal component analysis method efficiently filters the evolution characteristics of meteorological elements by reducing the data dimension and redundant noise effects, thereby improving the prediction accuracy;

[0043] The calculation formula of the principal component analysis method is as follows:

[0044]

[0045] where Meteroloty m,t represents the corresponding wind waves, currents and ice cover tensor fields at time t. PC is the principal component variable, and EOF is the corresponding spatial eigenvector. t represents the information of the external ice, wind and wave prediction fields in the sea area passed by the ship at time t. n = 1,..., k represents the number of principal components PC, and m represents the number of tensors;

[0046] Adopt a depthwise separable 2D CNN layer to build a deep encoder-decoder. The 2D convolutional kernel in the model is used to mine the coupling interaction characteristics between the tensor factors of the external meteorological environment field; among them, the decoder module exports high-dimensional semantic information, and the encoder part extracts the high-dimensional feature maps of the tensor sequence with low-order non-linear information; the comprehensive decoding encoder forms a generator of a spatio-temporal generative adversarial network model based on the architecture of depthwise separable convolution;

[0047] Create a mathematical prediction model of a deep generative adversarial neural network based on the depthwise separable module for the ice field, wind field, wave field and current field in the sea area passed by the ship. The model uses depthwise separable convolution as the backbone and adopts a spatio-temporal generative adversarial neural network model with an attention mechanism to realize the assimilation and fusion of observation data and high-resolution model prediction data, and perform dynamic update prediction of the external ice conditions, wind waves and current fields to obtain the ocean environment prediction results of the sea area passed by the ship;

[0048] Among them, the spatio-temporal generative adversarial network model includes two modules: a generator G(Z) and a discriminator D(x). Both the generator and discriminator modules are built with a depthwise separable convolution backbone. The generator module G(Z) assimilates and fuses the observation data and the multi-modal data of the model prediction field reconstructed after dimensionality reduction and noise removal by the principal component analysis method, where Z represents the observation point information and the meteorological field; the meteorological field mapping generated by the generator module G and the real mapping decoded from the reanalysis data are fed back and input into the discriminator D. The input of the discriminator D is the generated meteorological field data and the reanalysis meteorological field data reconstructed after dimensionality reduction and noise removal; the output of D is a scalar used to judge the probability that the output mapping comes from the real input mapping; the stop criterion for training the spatio-temporal generative adversarial network model is that the discriminator cannot distinguish whether the generated meteorological field prediction mapping conforms to the real mapping;

[0049] The loss function of the spatio-temporal generative adversarial network model is as follows:

[0050]

[0051] Among them, L D is the discriminator loss, and L G is the generator loss function.

[0052] Furthermore, the parameters of the external ocean meteorological environment field prediction model are trained and updated based on the following minimization and maximization formulas:

[0053]

[0054] Among them, P data (x) is the reanalysis meteorological field mapping data distribution, P z (z) is the data distribution of the model input observations and the numerically simulated meteorological field data reconstructed after dimensionality reduction and denoising by the principal component analysis method, and E represents the expectation operator;

[0055] The number of features of the time series sequential tensor reconstructed after dimensionality reduction and denoising of the multi-modal meteorological factors input into the spatio-temporal generative adversarial neural network model is C, H is the height of the model input tensor, and W is the width of the model input tensor. represents the input tensor of the model. The attention mechanism module further calculates the three matrices K, Q, and V of the input tensor using dot-product to obtain three corresponding mapped weight feature matrix values;

[0056] Among them, the dimensions of the K and Q feature matrices should be kept consistent. The similarity between the i-th query eigenvalue and the j-th key eigenvalue in the iterative operation is calculated by the regularization function ;

[0057] Suppose the actually measured external ice conditions, wind, current, and wave meteorological elements of the shipborne instrument are S ice , S wind , S current , S wave , respectively. The two-dimensional field elements of the ice conditions, wind, current, and waves predicted by the ERA5 high-resolution numerical model reconstructed after dimensionality reduction and denoising by the principal component analysis method are PCA-ERA ice , PCA-ERA wind , PCA-ERA current , PCA-ERA wave , and the sea condition prediction model for the ship navigation area updated dynamically:

[0058] P ice,t = STGAN(Sice,t , PCAERA ice,t )

[0059] P wind,t = STGAN(S wind,t , PCA - ERA wind,t )

[0060] P current,t = STGAN(S curren,t , PCA - ERA current,t )

[0061] P wave,t = STGAN(S wave,t , PCA - ERA wave,t )

[0062] Wherein, STGAN is a spatio - temporal generative adversarial neural network model, and P*,t represents the information of the external ice condition, wind, wave and current prediction field in the sea area passed by the ship at time t.

[0063] Furthermore, S4 specifically includes the following steps:

[0064] Use the empirical mode decomposition method to decompose the time series of sea ice thickness variables:

[0065] I = IMF1 + IMF2 + … + IMF j + … + IMF n + r

[0066] For each component IMF j Use a neural network to identify and predict it;

[0067] Specifically, for IMF j Perform q - step - ahead prediction, and the input order of the prediction model is determined to be p by the Lipschitz quotient method. Then the structure of the q - step - ahead prediction is p - input - 1 - output, and the input and output of the neural network identification process are:

[0068]

[0069] Wherein, p j Is the number of neural network inputs for the j - th IMF component, which is determined according to the data by the Lipschitz quotient method to increase the adaptability of the method;

[0070] The inputs and outputs in the prediction process are respectively:

[0071]

[0072] Determine the order of time series prediction for each component by calculating the Lipschitz coefficient method. After determining the input-output order, use a variable-structure radial basis function neural network for time series prediction of each component.

[0073] Furthermore, the steps to establish a variable-structure neural network prediction model are as follows:

[0074] S41. Establish a sliding data window

[0075] To reflect the latest ship motion state, establish a sliding data window to observe the ship motion state, and use the real-time updated input-output data to dynamically adjust the fitting model based on the radial basis function neural network;

[0076] The sliding window is a first-in-first-out data sample sequence with a fixed width. When a new set of input-output data is received, the new data set is added to the sliding window, and the earliest set of data is removed from the sliding window. Denote the sliding window W at time t SD as:

[0077] W SD = [(p t-L+1 , q t-L+1 ), L, (p t , q t )]

[0078] where L is the width of the sliding window;

[0079] Adjust the sliding window in real time to reflect the latest ship motion and the response of the structure. Adopt a sliding window adjustment method based on time series segmentation, that is, segment the time series input of the sliding window with width L, and form an adaptive-width sliding data window from the earliest data point to the mutation point, and keep the window width not less than the set minimum width Lmin;

[0080] For the time series x(t) within the sliding window composed of L data inputs, calculate the average value μ1(i) and standard deviation s1(i) of the data before each data point, and the average value μ2(i) and standard deviation s2(i) of the data after it respectively. Then the combined deviation S D (i) is:

[0081]

[0082] where L1 and L2 are the number of data points in the data part before and after point i respectively. Use the t-test value T(i) to represent the difference in the mean values of the two parts of data before and after point i:

[0083]

[0084] Repeat the above calculations for each data in the sequence x(t) to obtain a sequence of statistical test values T(t) corresponding to the data values of x(t), and calculate the maximum value T in T(t). max The statistical significance P(T max ):

[0085] P(T max ) = Prob(T ≤ T max )

[0086] where P(T max ) represents the probability of obtaining a T value less than or equal to T max in the random process. Generally, P(T max ) can be approximately expressed as:

[0087]

[0088] The values of η and δ are obtained by Monte Carlo simulation. L is the sliding window width, i.e., the length of x(t), v = L - 2, I x (a, b) is the incomplete beta function. Set the P0 threshold. If P(T max ) ≥ P0, the two sub - sequences into which x(t) is divided are different sub - sequences, and this point is the mutation point;

[0089] A new sliding data window La is formed by the mutation point and the earliest data point in the sliding window. If the width La of the sliding data window is less than the set minimum width Lmin (La < Lmin), then the sliding data window is Lmin;

[0090] Use the input - output data group in the sliding window, that is, represent the real - time dynamics of the mapping relationship with the input matrix P and the corresponding output vector Q respectively:

[0091]

[0092] Q = [q t-L+1 L q t

[0093] In the formula, n p is the dimension of the input matrix;

[0094] Use the input matrix P and the corresponding output Q as the input and output of the radial basis function neural network respectively, and train and dynamically adjust the neural network;

[0095] S42. After receiving a new data sample at each step, update the sliding data window, add the latest sample to the window, and delete the earliest sample from the window. Directly add the new data sample to the hidden layer as a new hidden node;

[0096] ​S43. Calculate the response matrix Φ of the hidden layer, where:

[0097]

[0098] where c j is the center of the j-th hidden node, p i is the i-th sample, ||·|| represents the Euclidean distance, σ is the width of the basis function; M is the number of hidden nodes. Orthogonally decompose the vectors in Φ using the Gram-Schmidt rule Φ = WA to obtain W = [w1, L, w M ;

[0099] Calculate the error descent rate:

[0100]

[0101] Normalize the error descent rate:

[0102]

[0103] S44. Select those hidden nodes whose sum of contributions to the output is less than the set value until the sum of the normalized error descent rates of the selected hidden nodes Select k1,..., k S to form a set of candidate hidden nodes to be deleted

[0104] Take the intersection I of the sets of hidden nodes selected in the past consecutive M S steps and delete the hidden nodes in I:

[0105]

[0106] S45. After determining the hidden nodes at each step, update the connection weights from the hidden layer to the output layer using the least squares method:

[0107] Θ = Φ + Y = (Φ T Φ) -1 Φ T Y

[0108] For the neural network prediction value of the remainder It is recombined from the predicted values of each component:

[0109]

[0110] Thus, a neural network capable of online prediction is obtained.

[0111] The present invention also provides a storage medium, the storage medium includes a stored program, wherein, when the program runs, it executes the rapid prediction method for the environmental response of ice area navigation ships described in any one of the above.

[0112] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor runs through the computer program to execute the rapid prediction method for the environmental response of an ice navigation ship as described in any one of the above.

[0113] Compared with the prior art, the present invention has the following advantages:

[0114] The present invention uses a neural network fitting model established offline to calculate the motion response and structural response of the ship to the environment to increase the rapidity of ship response calculation; comprehensively considers the motion response and structural response of ship motion when the ship is navigating in ice areas; in the establishment of the neural network, the partial least squares method is used to solve the problem of too high input dimension, and a simple network structure is selected by gradually increasing the network nodes, improving the generalization ability of the obtained neural network. Differentiated processing is carried out on the wind field, wave field, and current field.

[0115] During the ice navigation of a ship, ice loads not only have a great impact on the motion attitude of the ship, but also pose a great threat to the ship's structure. Therefore, this patent divides the environmental response into motion response and structural response. In view of the fact that there are many sudden situations during the ice navigation of a ship, such as frequent ship steering and speed changes, and frequent interactions with ice blocks, etc., the ship's motion state and structural loads will change violently accordingly. In this case, it is necessary to adjust the sliding window in real time to reflect the latest motion and structural response of the ship. The present invention adopts a sliding window adjustment method based on time series segmentation.

[0116] The structural response and motion response of a ship to ice have their unique difficulties. The motion and structural responses of a ship to wind, waves, and ice are caused by wind and wave loads and ice loads. However, the ice load is different from the wave and wind loads. Wave and wind elements can be directly measured to calculate the wind and wave loads, but the ice load cannot be directly measured temporarily. Therefore, the present invention inversely calculates the ice load according to the measured structural response measured data of the ship.

[0117] The present invention uses environmental prediction information to perform response predictions on multiple time scales. On the basis of dynamically updating and predicting the ice area environmental information such as ice, wind, waves, and currents outside every 24 hours, within 24 hours, multi-scale decomposition and prediction are carried out using a neural network based on a sliding window, which can identify and extract the long-term, medium-term, and short-term change information of environmental elements, and accurately predict the environmental elements.

[0118] In the ice area hydrometeorological prediction method, compared with the traditional generative adversarial model, the present invention introduces a dual attention mechanism module, which maps the high-dimensional feature information containing multiple meteorological factors to the multi-dimensional feature matrix of the encoder-decoder. And the basic framework of the generative adversarial model is replaced with a two-dimensional convolutional neural network module for directly inputting the two-dimensional ice area environmental factor field information. A 2D deep CNN layer is used to build the deep encoder-decoder of the generative adversarial model. The 2D convolution kernel is used to mine the deep coupling interaction influence features between various meteorological factors; this method's deep convolution block can provide a method to improve the prediction efficiency and performance, while reducing its computing and memory requirements, and can better perform feature extraction and aggregation. The encoder part extracts the high-dimensional feature map of the tensor sequence with low-order non-linear information, while the decoder module derives the high-dimensional semantic information; improving the prediction accuracy of the data fusion field elements. BRIEF DESCRIPTION OF THE DRAWINGS

[0119] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0120] Figure 1 It is the flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0121] In order to enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0122] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0123] As Figure 1 shown, the present invention provides a rapid prediction method for the environmental response of ships sailing in ice areas. This method fully utilizes the existing prediction information of ice area environmental elements, the measured ice area environmental elements of the ship itself, and the ship motion response. The numerical simulation of the ship motion and structural response during ice area navigation, which involves large computational amounts, is carried out offline. Then, a neural network is used to fit this database. In the prediction link of ice area environmental elements, a hierarchical prediction is adopted to improve the real-time prediction accuracy while ensuring prediction stability. This method can quickly and accurately obtain the ship motion and structural response under the influence of complex ice area environments.

[0124] (1) Establish a mathematical model for the motion response of the ship and the structural response of the ship under environmental disturbances such as ice, wind, and waves at sea.

[0125] First, set the input and output of the training model. Set the training model input as: ① information on ship elements such as the main dimensions of the ship, structural parameters, bow column inclination and other hull parameters, superstructure layout, ship ice class and other parameters, ship draft and draft difference, propeller scale, etc.; ② parameters of hydro-meteorological environmental elements such as ice, wind, and waves; ③ dynamic information such as the relative position and motion relationship between the ship and environmental elements, propeller speed, rudder angle, etc. Set the training model output as the ship speed, motion attitude, and structural response under the influence of ice, wind, and waves. Secondly, establish the input-output training samples of the model through the numerical simulation of the ship's motion in ice, and calculate the response of a specific ship to environmental elements in terms of ship speed, motion attitude, and structure. Finally, according to the input and output data of the obtained samples, train the neural network offline so that the neural network accurately approximates the input-output mapping, and establish a neural network mapping model for the motion and structural response of ship navigation. Specifically,

[0126] Suppose there are m samples in the sample set, the input vector x has n independent variables, the input X is an m×n-dimensional matrix, the output vector y has l dependent variables, and the output Y is an m×l-dimensional matrix.

[0127] The algorithm steps for establishing the neural network calculation model of ship motion response at sea are as follows:

[0128] I. Starting from step 1, at step k, use the uniform distribution sampling method to generate random numbers for sampling from the input and output sample sets X and Y, and extract a set of data x k and y k , and construct the training sample sets U∈R k×n and V∈R k×l respectively.

[0129] II. Use the input and output training sample sets U and V to construct an RBF neural network, and use the partial least squares method to perform regression calculations on the hidden layer response matrix and the output:

[0130] In the RBF network, the Gaussian function is used as the radial basis function, and the vector u j (j = 1, …, k) becomes the center of the hidden nodes. After calculation by the radial basis function, the hidden layer response matrix A ∈ R k×k is obtained. The element in the i-th row and j-th column is:

[0131]

[0132] where u i is the i-th sample, and here ||·|| represents the Euclidean distance, c j and σ j are respectively the center and width of the j-th basis function.

[0133] III. Perform PLS principal component extraction and regression operation between the response matrix A and the output matrix V.

[0134] After extracting the principal component T, project A and V onto the principal component matrix respectively, and obtain:

[0135] V = TR + F = AWR + F

[0136] where T is the principal component matrix of A, a k×n T dimensional matrix; W is the transformation matrix of A, a k×n T dimensional matrix; R is an n T ×l dimensional regression coefficient matrix; F is a k×l dimensional residual matrix.

[0137] Obtain the generalization model:

[0138]

[0139] where A p1 and W1 are k×k dimensional and k×n T dimensional matrices respectively. R1 is an n T ×l dimensional matrix. Generalize the samples in the set U by the obtained network, and set a precision threshold to determine the number of times n T of principal component extraction. Stop extracting principal components when the change in the root mean square error (RMSE) of generalization is less than the set precision change threshold e1:

[0140]

[0141] IV. Generalize the training data in the entire data set X by the obtained network, and calculate the generalization error for the entire data set:

[0142]

[0143] where

[0144]

[0145] V. If the RMSE O is greater than the precision threshold e2, repeat steps I-IV.

[0146] VI. The algorithm ends when the RMSE O is less than the precision threshold e2.

[0147] The number of principal components n T and the number of loop steps k are the number of times of principal component extraction and the number of hidden nodes of the final RBF network, respectively. The vectors in the set U are the hidden nodes of the obtained RBF network.

[0148] Introduce the principal component analysis method to construct an ocean environmental prediction model based on depthwise separable convolution; among them, the depthwise separable convolution module can improve the prediction efficiency and performance, while reducing the computational amount and hardware memory requirements of the model. Compared with traditional convolution, the depthwise separable convolution greatly reduces the number of model parameters and computational requirements. This has strong practicability and feasibility for realizing on-board weather forecasting when ships are sailing at sea. Among them, the principal component analysis method can efficiently filter the evolution characteristics of external meteorological elements and improve the forecasting accuracy by reducing the data dimension and redundant noise effect.

[0149] The principal component analysis method is used in the data preprocessing module of the ocean environmental element forecasting module to perform preprocessing of dimensionality reduction and noise reduction on the collected meteorological data. After the meteorological data is processed by the principal component analysis, it is used as the input of the external ocean environmental field forecasting model.

[0150] Introduce a method based on the principal component analysis method to perform data cleaning and reconstruction processing on the data of external wind waves, currents and sea ice coverage areas. The principal component analysis method can efficiently filter the evolution characteristics of meteorological elements and improve the forecasting accuracy by reducing the data dimension and redundant noise effect.

[0151] The reconstruction calculation formula of the meteorological element tensor based on empirical orthogonal decomposition is as follows:

[0152]

[0153] where PC is the principal component variable, EOF is the corresponding spatial eigenvector, t represents the information of the external ice, wind and wave forecasting fields in the sea area passed by the ship at time t, n = 1,..., k, representing the number of principal components PC, and m represents the number of tensors.

[0154] A depth - separable 2D CNN layer is adopted to build a depth - encoding decoder, and a depth - separable convolutional neural network model is constructed. The 2D convolutional kernels in the model are used to mine the coupling interaction features between extreme wind field tensor factors. Among them, the decoder module exports high - dimensional semantic information, and the encoder part extracts high - dimensional feature maps of tensor sequences with low - order non - linear information. The depth - separable convolutional block introduced by this method can significantly reduce the model's computational and memory requirements. Compared with traditional convolutions, it greatly reduces the number of required parameters and calculations. This makes them more efficient and faster to train and run on devices with limited computational resources.

[0155] A mathematical prediction model based on a deep generative adversarial neural network for the ice field, wind field, wave field, and current field in the sea area passed by the ship is created. The model takes depth - separable convolution as the backbone and uses a spatio - temporal generative adversarial neural network model with an attention mechanism to realize the assimilation and fusion of observed data and high - resolution model prediction data, and conducts dynamic update prediction of the external ice condition, wind, wave, and current fields to obtain the ocean environmental prediction results of the sea area passed by the ship.

[0156] The ocean environmental element prediction module of the present invention is a deep generative adversarial neural network model, and the convolutional backbone in the model adopts depth - separable convolution. The decoder and encoder are two major sub - modules of the generator in the generative adversarial network model.

[0157] Among them, the spatio - temporal generative adversarial network model includes two modules: a generator G(Z) and a discriminator D(x). Both the generator and discriminator modules are built with a depth - separable convolutional backbone. The generator module G(Z) assimilates and fuses observed data and multi - modal data of the model prediction field reconstructed after dimension reduction and denoising by the principal component analysis method, where Z represents observation point information and the meteorological field; the meteorological field mapping generated by the generator module G and the real mapping decoded from the re - analysis data are fed back and input into the discriminator D. The input of the discriminator D is the generated meteorological field data and the re - analyzed meteorological field data reconstructed after dimension reduction and denoising; the output of D is a scalar used to determine the probability that the output mapping comes from the real input mapping; the stopping criterion for training the spatio - temporal generative adversarial network model is that the discriminator cannot distinguish whether the generated meteorological field prediction mapping conforms to the real mapping.

[0158] The loss function of the spatio - temporal generative adversarial network model is as follows:

[0159]

[0160] Among them, L D is the discriminator loss, and L G is the generator loss function;

[0161] The generative adversarial network model is the external ocean environmental meteorological field prediction model in the present invention, where S2 is the module for creating the backbone and structure of the model, and S3 is the module for applying the prediction of the model.

[0162] The parameters of the external marine meteorological environmental field prediction model are trained and updated based on the following minimization and maximization formulas:

[0163]

[0164] where P data (x) is the reanalysis meteorological field mapping data distribution, and P z (z) is the data distribution of the model input observations and the numerically simulated meteorological field data reconstructed after dimensionality reduction and denoising by the principal component analysis method. E represents the expectation operator;

[0165] The number of features of the time series sequential tensor reconstructed after dimensionality reduction and denoising of the multi-modal meteorological factors by the principal component analysis method input into the spatio-temporal generative adversarial network model is C, H is the height of the model input tensor, and W is the width of the model input tensor. represents the model input tensor. The attention mechanism module further calculates the three matrices K, Q, and V of the input tensor using dot-product to obtain three corresponding mapped weight feature matrix values;

[0166] where the dimensions of the K and Q feature matrices should be kept consistent, and the similarity between the i-th query eigenvalue and the j-th key eigenvalue in the iterative operation is calculated by the regularization function ;

[0167] where further, assuming that the actually measured external ice conditions, wind, current, and wave and other meteorological elements by on-board instruments are S ice , S wind , S current , S wave , and the two-dimensional field elements of ice conditions, wind, current, and wave predicted by the ERA5 high-resolution numerical model reconstructed after dimensionality reduction and denoising by the principal component analysis method are PCA-ERA ice , PCA-ERA wind , PCA-ERA current , PCA-ERA wave , the sea condition prediction model for the ship navigation area updated dynamically:

[0168] P ice,t = STGAN(S ice,t , PCAERA ice,t )

[0169] P wind,t = STGAN(S wind,t , PCA-ERA wind,t )

[0170] Pcurrent,t = STGAN(S curren,t , PCA-ERA current,t )

[0171] P wave,t = STGAN(S wave,t , PCA-ERA wave,t )

[0172] Among them, STGAN is a spatio-temporal generative adversarial neural network model, and P*,t represents the information of the external ice condition, wind, wave and current forecast fields in the sea area where the ship sails at time t.

[0173] According to the above-created generative adversarial ocean environment prediction model, accurately predict the ice condition coverage, wind field, wave field, current field and other ocean environments in the sea area passed by the ship in real time and dynamically, and provide real-time guidance for ship route planning.

[0174] (2) Enrich the environmental motion and structural response database for the ship's ice area navigation according to the ice condition, wind, wave and current information measured in real time and the ship's motion and structural response information, or update the database according to the measured information of elements such as wind, wave and current and the ice load inversion information of the ship's structural response.

[0175] Obtain the ship's motion and structural response information through ship stress measurement sensors (fiber Bragg grating sensors). Since it is currently difficult to directly measure ice loads, ice loads are indirectly inverted through the measurement data of the local or overall response of the ship's structure such as ship structure strain and the six-degree-of-freedom motion parameters of the ship.

[0176] (3) Make a refined real-time prediction of the ice area environmental elements in the waters passed by the ship.

[0177] According to the ship's planned route and the prediction results of the regional ice area environmental elements, make a refined real-time prediction of the ice area environmental elements in the ship's route area hourly within 24 hours. In the prediction process, use multi-level decomposition and prediction, decompose the prediction into different time scales for prediction, and sum up different predictions to obtain the final prediction result. Make separate predictions for different environmental elements. Taking the sea ice thickness as an example, it is described as follows:

[0178] Use the empirical mode decomposition (EMD) method to decompose the time series of the sea ice thickness variable:

[0179] I = IMF1 + IMF2 + … + IMF j + … + IMF n + r

[0180] For each component IMF j Use a neural network to identify and predict it.

[0181] Specifically, for the IMF j perform q-step-ahead prediction. The input order of the prediction model is determined to be p using the Lipschitz quotient method. Then, the structure of the q-step-ahead prediction is p-input-1-output, and the input and output of the neural network identification process are as follows:

[0182]

[0183] where p j is the number of neural network inputs for the j-th IMF component, which is determined based on the data using the Lipschitz quotient method to increase the adaptability of the method.

[0184] The inputs and outputs during the prediction process are respectively:

[0185]

[0186] Determine the order of the time series prediction for each component by calculating methods such as the Lipschitz coefficient. After determining the input and output orders, use a variable-structure radial basis function neural network to perform the time series prediction for each component.

[0187] Establish a variable-structure neural network prediction model through the following steps:

[0188] ① Establish a sliding data window

[0189] The movement of a ship at sea has the characteristics of dynamic time-variation. To reflect the latest ship movement state, establish a sliding data window to observe the ship movement state, and use the real-time updated input and output data to dynamically adjust the fitting model based on the radial basis function neural network;

[0190] The described sliding window is a first-in-first-out data sample sequence with a fixed width. When a new set of input-output data is received, the new data set is added to the sliding window, and the earliest set of data is removed from the sliding window. Represent the sliding window W at time t SD as:

[0191] W SD = [(p t-L+1 , q t-L+1 ), L, (p t , q t )]

[0192] where L is the width of the sliding window;

[0193] When a ship is sailing in ice areas, there are many sudden situations, such as frequent ship steering and speed changes, and many interactions with ice blocks. In such cases, the ship's motion state and structural loads will change significantly accordingly. In this situation, it is necessary to adjust the sliding window in real time to reflect the ship's latest motion and the response of the structure. The present invention adopts a sliding window adjustment method based on time series segmentation, that is, the input of the sliding window with a width of L is segmented by time series, and an adaptive-width sliding data window is formed from the earliest data point to the mutation point, and the window width is kept not less than the set minimum width Lmin.

[0194] For the time series x(t) within the sliding window composed of L data inputs, calculate the average value μ1(i) and the standard deviation s1(i) of the data before each data point, and the average value μ2(i) and the standard deviation s2(i) of the data after it respectively. Then the combined deviation S D (i) is

[0195]

[0196] where L1 and L2 are the number of data points in the data part before and after the i-th point respectively, and the t-test value T(i) represents the difference in the mean values of the two parts of data before and after the i-th point:

[0197]

[0198] Repeat the above calculations for each data in the sequence x(t) to obtain a statistical test value sequence T(t) corresponding to the data values of x(t), and calculate the statistical significance P(Tmax) of the maximum value Tmax in T(t):

[0199] P(T max ) = Prob(T ≤ T max )

[0200] where P(Tmax) represents the probability of obtaining a T value less than or equal to Tmax in the random process. Generally, P(Tmax) can be approximately expressed as

[0201]

[0202] The values of η and δ can be obtained by Monte Carlo simulation. L is the sliding window width, that is, the length of x(t), v = L - 2, and Ix(a, b) is the incomplete beta function. Set the P0 threshold. If P(Tmax) ≥ P0, the two subsequences of x(t) that are segmented and have differences are the subsequences with differences, and this point is the mutation point.

[0203] A new sliding data window La is formed by the mutation point and the earliest data point within the sliding window. If the width La of the sliding data window is less than the set minimum width Lmin (La < Lmin), the sliding data window is Lmin.

[0204] Using the input-output data groups within the sliding window, that is, the real-time dynamics of the mapping relationship are represented by the input matrix P and the corresponding output vector Q respectively:

[0205]

[0206] Q = [q t-L+1 L q t

[0207] In the formula, n p is the dimension of the input matrix;

[0208] Using the input matrix P and the corresponding output Q as the input and output of the radial basis function neural network respectively, the neural network is trained and dynamically adjusted.

[0209] ② After receiving a new data sample at each step, update the sliding data window, add the latest sample to the window, and delete the earliest sample from the window. Directly add the new data sample to the hidden layer as a new hidden node.

[0210] ③ Calculate the response matrix Φ of the hidden layer, where

[0211]

[0212] where c j is the center of the jth hidden node, p i is the ith sample, ||·|| represents the Euclidean distance, σ is the width of the basis function; M is the number of hidden nodes. Orthogonally decompose the vectors in Φ using the Gram-Schmidt rule Φ = WA to obtain W = [w1, L, w M .

[0213] Calculate the error descent rate:

[0214]

[0215] Normalize the error descent rate:

[0216]

[0217] ③ Select those hidden nodes whose sum of contributions to the output is less than the set value. Until the sum of the normalized error descent rates of the selected hidden nodes Select k1,..., k S to form the set of candidate hidden nodes to be deleted ​

[0218] Take the intersection I of the sets of hidden nodes selected in the past consecutive M S steps and delete the hidden nodes in I:

[0219]

[0220] ④ After determining the hidden nodes at each step, for convenience, use the least squares method to update the connection weights from the hidden layer to the output layer.

[0221] Θ = Φ + Y = (Φ T Φ) -1 Φ T Y.

[0222] For the neural network prediction value of the residual It is recombined from the predicted values of each component:

[0223]

[0224] Thus, a neural network that can perform online prediction is obtained.

[0225] (4) According to the sea areas passed by the planned route of the ship, calculate the motion response and structural response of the ship passing through the ice area, wind field and wave field in this sea area. The input of the calculation model is ① information on ship elements such as the main dimensions of the ship, the inclination angle of the bow column, etc., the layout of the superstructure, the ice class of the ship, etc., the current draft and draft difference of the ship, and the scale of the propeller; ② information on the parameters of hydro-meteorological environmental elements such as ice, wind, and waves at the location of the ship and the expected future location; ③ dynamic information such as the relative position and relative motion relationship between the ship and environmental elements such as ice, wind, and waves, the rotational speed of the propeller, and the rudder angle. Set the output of the training model to the ship's speed, motion attitude, and structural response under the influence of the set ice, wind, and waves.

[0226] Based on the above motion responses such as the ship's speed and motion attitude and the differential structure response under the influence of ice, wind, and waves, combined with the prediction results of the flow field, comprehensively conduct the response prediction of the ship sailing in the ice area. On this basis, applications such as real-time risk prediction and real-time route optimization for the ship sailing in the ice area can be carried out.

[0227] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A rapid prediction method for environmental response of ships sailing in ice areas, characterized in that: The steps include: S1. Establish a mathematical model of the ship's motion response and ship structure response under the environmental interference of ice, wind and waves at sea; S2. Establish a forecast model for ice and sea conditions in the sea area where ships sail, use a spatiotemporal generative adversarial neural network model with an attention mechanism to achieve assimilation and fusion of observation data and high-resolution model forecast data, dynamically update the forecast of external ice, wind, wave and flow fields, and obtain the initial marine environment forecast results for the sea area where the ship passes; S3. Based on the real-time measured ice conditions, wind, wave, current information, and ship motion and structural response information, the environmental motion and structural response database of ships sailing in ice areas is enriched, or based on the measured information of wind, wave and current elements and the ice load inversion information of ship structural response, the ice conditions and sea state forecast model database of ships sailing in the sea area is updated to obtain real-time marine environment forecast results; S4. According to the planned route of the ship and the real-time marine environment forecast results, a refined real-time forecast of ice environment elements is made for the ship route area on an hourly basis within 24 hours. In the forecast process, multi-level decomposition and prediction are adopted, the forecast is decomposed into different time scales for forecasting, and different forecasts are summed to obtain the final forecast result; S5. According to the sea area through which the ship's planned route passes, the motion response and structural response of the ship passing through the ice area, wind field and wave field of the sea area are calculated by using the motion response and structural response mathematical models; The input of the mathematical model of motion response and structural response includes ship element information, environmental element information and dynamic element information, and the output of the mathematical model of motion response and structural response is the speed, motion attitude and structural response of the ship under the influence of set ice, wind and waves; S6. Based on the above-mentioned ship's speed, motion attitude, motion response and structural response under the influence of ice, wind and waves, combined with the final forecast results, a comprehensive forecast of the ship's response to navigation in ice areas is carried out.

2. The rapid prediction method for environmental response of ships sailing in ice areas according to claim 1 is characterized in that: S1 specifically includes the following steps: S11, setting the input of the training model and the output of the training model; the input of the training model includes ship element information, environmental element information and dynamic element information; the output of the training model includes the speed, motion posture and structural response of the ship under the influence of ice, wind and waves; S12. Based on the model input and model output, a model input-output training sample is established by numerically simulating the motion and structural response of the ship in ice, and the response of a specific ship to environmental factors in terms of ship speed, motion posture and structure is calculated; assuming that there are m samples in the sample set, the input vector x has n independent variables, the input X is an m×n dimensional matrix, the output vector y has l dependent variables, and the output Y is an m×l dimensional matrix; S13. Based on the input and output data of the samples obtained in S12, offline training of the neural network makes the neural network accurately approximate the input-output mapping, and establishes a neural network mapping model of the motion and structural response model of the ship's navigation.

3. The rapid prediction method for environmental response of ships sailing in ice areas according to claim 2 is characterized in that: The ship element information includes the main dimensions of the ship, structural parameters and hull parameters of the bow inclination, superstructure layout, parameters of the ship's ice class, ship draft and draft difference, and propeller dimensions; the environmental element information includes hydrological and meteorological environmental elements of ice, wind, and waves; the dynamic element information includes the relative position and movement relationship between the ship and the environmental elements, propeller speed, and rudder angle.

4. The rapid prediction method for environmental response of ships sailing in ice areas according to claim 2 is characterized in that: S13 specifically includes the following steps: S131, according to the uniform distribution sampling method, generate random numbers from the input and output sample sets X and Y to perform sampling, extract the first group of data x1 and y1 in the second step, extract the second group of data x2 and y2 in the second step, and so on, extract data x in the kth step. k and k , construct a training sample set U∈R k×n and V∈R k×l ; S132, constructing a radial basis function neural network using the input and output training sample sets U and V, and performing regression calculation of the hidden layer response matrix and the output using the partial least squares method: In the RBF network, the Gaussian function is used as the radial basis function, and the vector u j Become the hidden node center, j = 1, ..., k; the hidden layer response matrix A∈R is obtained by radial basis function calculation k×k , where the element in row i and column j is: Among them, u i is the i-th sample, where ||·|| represents the Euclidean distance, c j and σ j are the center and width of the jth basis function respectively; S133, performing PLS principal component extraction and regression operation between the hidden layer response matrix A and the output matrix V; After extracting the principal component T, project A and V onto the principal component matrix respectively, and we get: V=TR+F=AWR+F Among them, T is the principal component matrix of A, which is k×n T dimensional matrix; W is the transformation matrix of A, which is k×n T dimensional matrix; R is n T ×l-dimensional regression coefficient matrix; F is the k×l-dimensional residual matrix; Get the generalized model: Among them, A p1 and W1 are k×k and k×n respectively. T dimensional matrix, R1 is n T ×l-dimensional matrix, generalize the obtained network to the samples in the set U, and set the accuracy threshold to determine the number of principal component extraction times n T , stop extracting the principal components when the change in the generalized root mean square error is less than the set accuracy change threshold e1: S134, generalize the obtained network to the training data in the entire data set X, and calculate the generalization error for the entire data set: in: S135, if RMSE O If it is greater than the set accuracy threshold e2, steps S231 to S234 are repeated; S136, if RMSE O If it is less than the set accuracy threshold e2, then the process ends; S137, the number of principal components obtained n T The number of loop steps k is the number of principal component extractions and the number of hidden nodes of the final radial basis function network, and the vectors in the set U are the hidden nodes of the obtained radial basis function network.

5. The rapid prediction method for environmental response of ships sailing in ice areas according to claim 1 is characterized in that: S2 specifically includes: According to the coverage of ice areas, wind fields, wave fields and flow fields selected according to the sea areas through which the ship's planned route passes, a high-resolution ice area ice condition and hydrological and meteorological environment forecast model is selected, and the ice condition information obtained by identifying the ship's measured video and image information and the real-time observed hydrological and meteorological environment data information are assimilated. The spatiotemporal generative adversarial neural network with an attention mechanism is used to realize the assimilation and fusion of the observed data and the high-resolution model forecast data, thereby improving the accuracy of the hydrological and meteorological forecast in the navigation area; The principal component analysis method is used to reduce the dimension of meteorological field data and remove redundant noise signals, retain the main component feature vectors of the meteorological field, reconstruct the main feature vectors after dimensionality reduction and denoising as meteorological field data, and build a marine environment forecast model based on deep separable convolution; the deep separable convolution module is used to improve the prediction efficiency and performance, while reducing the calculation amount and hardware memory requirements of the model; the principal component analysis method reduces the data dimension and redundant noise effect to efficiently filter the evolution characteristics of external meteorological elements and improve the forecast accuracy; The principal component analysis method is used to clean and reconstruct the external wind wave current and sea ice coverage area data through empirical orthogonal decomposition theory. The principal component analysis method reduces the data dimension and redundant noise effect, and effectively filters the evolution characteristics of meteorological elements to improve the forecast accuracy. The calculation formula of the principal component analysis method is as follows: Among them, Meterology m,t represents the corresponding wind, wave, current and ice cover tensor field at time t, PC is the principal component variable, EOF is the corresponding spatial feature vector, t represents the external ice, wind and wave forecast field information of the sea area where the ship passes at time t, n=1,…k represents the number of principal component PCs, and m represents the number of tensors; A deep separable 2D CNN layer is used to build a deep encoder-decoder. The 2D convolution kernel in the model is used to mine the coupling interaction characteristics between the tensor factors of the external meteorological environment field. The decoder module exports high-dimensional semantic information, and the encoder part extracts high-dimensional feature maps of tensor sequences with low-order nonlinear information. The integrated decoder-encoder constructs a spatiotemporal generative adversarial network model generator based on the deep separable convolution architecture. Create a mathematical forecasting model of ice field, wind field, wave field and flow field in the sea area where the ship passes through, which is based on the deep separable module architecture of deep generative adversarial neural network. The model uses deep separable convolution as the skeleton and adopts the spatiotemporal generative adversarial neural network model with the introduction of attention mechanism to realize the assimilation and fusion of observation data and high-resolution model forecast data, and dynamically update the forecast of external ice conditions, wind and wave fields to obtain the forecast results of the marine environment in the sea area where the ship passes through. The spatiotemporal generative adversarial network model includes two modules, a generator G(Z) and a discriminator D(x), wherein both the generator and the discriminator modules are constructed using a deep separable convolutional skeleton, the generator module G(Z) assimilates and fuses the observation data and the multimodal data of the model forecast field reconstructed after dimensionality reduction and denoising by the principal component analysis method, wherein Z represents the observation point information and the meteorological field; the meteorological field map generated by the generator module G and the real map decoded in the reanalysis data are fed back to the discriminator D, the input of the discriminator D is the generated meteorological field data and the reanalysis meteorological field data reconstructed after dimensionality reduction and denoising; the output of D is a scalar, which is used to judge the probability of whether the output map comes from the real input map; the stopping criterion for the training of the spatiotemporal generative adversarial network model is that the discriminator cannot distinguish whether the generated meteorological field forecast map conforms to the real map; The loss function of the spatiotemporal generative adversarial network model is as follows: Among them, L D is the discriminator loss, L G is the generator loss function.

6. The rapid prediction method for environmental response of ships sailing in ice areas according to claim 5 is characterized in that: The parameters of the external marine meteorological environment forecast model are trained and updated based on the following minimization and maximization formulas: Among them, P data (x) is the distribution of reanalysis meteorological field mapping data, P z (z) is the model input observation and the numerical simulation meteorological field data distribution reconstructed after dimensionality reduction and denoising by principal component analysis, and E represents the expectation operator; The number of features of the time series sequential tensor reconstructed after dimensionality reduction and denoising of the multimodal meteorological factors input to the spatiotemporal generative adversarial neural network model by principal component analysis is C, H is the height of the model input tensor, W is the width of the model input tensor, Represents the input tensor of the model. The attention mechanism module uses dot-product to further calculate the three matrices K, Q and V of the input tensor to obtain three The corresponding mapping weight feature matrix value; The dimensions of the K and Q feature matrices should be consistent, and the i-th query eigenvalue in the iterative operation is and the jth key eigenvalue The similarity between them is determined by the regularization function Calculate acquisition; Assume that the shipboard instruments measure the external ice conditions, wind, current and wave meteorological elements respectively as S ice ,S wind ,S current ,S wave The ERA5 high-resolution numerical model forecasts ice conditions, wind, current and wave two-dimensional field elements after dimensionality reduction and denoising by principal component analysis, respectively: PCA-ERA ice ,PCA-ERA wind ,PCA-ERA current ,PCA-ERA wave ,Dynamically updated sea condition forecast model for ships sailing in the sea area: P ice,t =STGAN(S ice,t ,PCAERA ice,t ) P wind,t =STGAN(S wind,t ,PCA-ERA wind,t ) P current,t =STGAN(S curren,t ,PCA-ERA current,t ) P wave,t =STGAN(S wave,t ,PCA-ERA wave,t ) Among them, STGAN is a spatiotemporal generative adversarial neural network model, and P*,t represents the external ice condition, wind and wave forecast information of the sea area where the ship passes at time t.

7. The rapid prediction method for environmental response of ships sailing in ice areas according to claim 1 is characterized in that: S4 specifically includes the following steps: The empirical mode decomposition method is used to decompose the time series of sea ice thickness variables: I=IMF1+IMF2+…+IMF j +…+IMF n +r For each component IMF j Both use neural networks to identify and predict them; Specifically, for the IMF j A q-step-ahead forecast is performed, and the input order of the prediction model is determined to be p using the Lipschitz quotient method. Then the structure of the q-step-ahead forecast is p-input-1-output, and the input and output of the neural network identification process are: Among them, p j is the number of neural network inputs to the jth IMF component, which is determined by the Lipschitz quotient method based on the data to increase the adaptability of the method; The input and output in the prediction process are: The order of time series forecast for each component is determined by calculating the Lipschitz coefficient method. After the input and output orders are determined, the variable structure radial basis function neural network is used to predict the time series of each component.

8. The rapid prediction method for environmental response of ships sailing in ice areas according to claim 7 is characterized in that: The steps to establish a variable structure neural network prediction model are as follows: S41, establishing a sliding data window; In order to reflect the latest ship motion status, a sliding data window is established to observe the ship motion status, and the fitting model based on the radial basis function neural network is dynamically adjusted using the real-time updated input and output data; The sliding data window is a fixed-width first-in-first-out data sample sequence. When a new set of input-output data is received, the new data set is added to the sliding data window, and the earliest set of data is removed from the sliding data window. The sliding data window W at time t is SD It is expressed as: W SD =[(p t-L+1 ,q t-L+1 ),L,(p t ,q t )] Where L is the width of the sliding data window; The sliding data window is adjusted in real time to reflect the latest movement of the ship and the response of the structure. The sliding data window adjustment method based on time series segmentation is adopted, that is, the sliding data window input with a width of L is segmented into a time series, and a sliding data window with an adaptive width is formed from the earliest data point to the mutation point, and the window width is kept not less than the set minimum width Lmin; For the time series x(t) in the sliding data window consisting of L data inputs, the mean μ1(i) and standard deviation s1(i) of the data before each data point, as well as the mean μ2(i) and standard deviation s2(i) of the data after each data point are calculated, then the combined deviation S of point i is D (i) is: Among them, L1 and L2 are the number of points before and after point i, respectively. The t-test value T(i) represents the difference in the mean of the two parts of data before and after point i: Repeat the above calculation for each data in the sequence x(t) to obtain the statistical test value sequence T(t) corresponding to the data value of x(t), and calculate the maximum value T in T(t). max The statistical significance of P(T max ): P(T max )=Prob(T≤T max ) Among them, P(T max ) indicates that the value of T is less than or equal to T in the random process max The probability of max ) can be approximately expressed as: The values ​​of η and δ are obtained by Monte Carlo simulation, where L is the width of the sliding data window, i.e., the length of x(t), v = L-2, and I x (a, b) is an incomplete β function, set the P0 threshold, if P(T max )≥P0, if there is a subsequence that is different between the two subsequences into which x(t) is divided, then this point is a mutation point; A new sliding data window La is formed by the mutation point and the earliest data point within the sliding data window. If the width La of the sliding data window is less than the set minimum width Lmin (La < Lmin), the sliding data window is Lmin; The input-output data groups within the sliding data window are used, that is, the input matrix P and the corresponding output vector Q are respectively used to represent the real-time dynamics of the mapping relationship; Q=[q t-L+1 L q t ] Where n p is the dimension of the input matrix; The input matrix P and the corresponding output Q are respectively used as the input and output of the radial basis function neural network, and the neural network is trained and dynamically adjusted; S42. After receiving a new data sample at each step, update the sliding data window, add the latest sample to the window, and delete the earliest sample from the window. Directly add the new data sample to the hidden layer as a new hidden node; S43. Calculate the response matrix Φ of the hidden layer, where: Among them, c j is the center of the jth hidden node, p i is the i-th sample, ||·|| represents the Euclidean distance, σ is the basis function width; M is the number of hidden nodes, and the vectors in Φ are orthogonally decomposed using the Gram-Schmidt rule. Φ=WA, and we get W=[w1,L,w M ]; Calculate the error descent rate: Normalize the error descent rate: S44, select those hidden nodes whose sum of output contribution is less than the set value, until the sum of the standardized error reduction rates of the selected hidden nodes is Choose k1,...,k S Construct a set of hidden nodes to be deleted Take the past continuous M S The intersection I of the hidden node sets selected in the step and delete the hidden nodes in I: S45. After determining the hidden node at each step, use the least squares method to update the connection weights from the hidden layer to the output layer; I = F + Y=(Φ T F) -1 F T Y Neural network predictions for residual values It is obtained by reorganizing the predicted values ​​of each component: Thus, a neural network that can perform online prediction is obtained.

9. A storage medium, characterized in that: The storage medium includes a stored program, wherein when the program runs, it executes the rapid prediction method for the environmental response of an ice-going ship according to any one of claims 1 to 8.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: The processor runs through the computer program to execute the rapid prediction method for the environmental response of an ice-going ship according to any one of claims 1 to 8.

Images