Arctic ship navigation risk prediction method and system based on GA-BP neural network model

Abstract

The application discloses a kind of GA-BP neural network model-based arctic ship navigation risk prediction method and system, belong to risk prediction technical field.Step 1: identify the key risk factors affecting arctic ship navigation, determine the input and output of GA-BP neural network model, constitute data set;Step 2: the data set obtained in step 1 is normalized;Step 3: establish BP neural network model, determine model structure;Step 4: the weight and threshold of BP neural network are optimized using genetic algorithm GA, and GA-BP prediction model is obtained;Step 5: the data set normalized in step 2 is trained and tested GA-BP prediction model obtained in step 4, and the ice trouble and ship ice collision risk are predicted.The present application is used to solve the problem that the navigation risk assessment and analysis steps of the prior art for arctic ship navigation accident are complicated and the efficiency is low.

Description

Technical Field

[0001] This invention belongs to the field of risk prediction technology, specifically relating to a method and system for predicting Arctic ship navigation risks based on a GA-BP neural network model. Background Technology

[0002] In recent years, global warming has led to a continuous reduction in Arctic sea ice area and a gradual decrease in sea ice concentration. Current reports indicate that by the summer of 2050, the Arctic region may experience a period of ice-free operation. This trend suggests that navigation conditions for ships in the Arctic will improve, navigation cycles will lengthen, and Arctic shipping and resource extraction may become increasingly standardized. However, the harsh Arctic environment also presents risks and challenges to safe navigation, making ensuring safe passage for ships in Arctic waters a critical issue. Ice entrapment and ship-ice collisions are two common accidents involving ships navigating the Arctic, resulting in structural damage, property loss, environmental pollution, and personal injury. Current research on Arctic ship navigation accidents largely focuses on risk assessment and analysis. While these methods can calculate navigation risks, they are cumbersome and inefficient. Therefore, it is necessary to combine artificial neural network technology with the prediction of ship navigation risks in Arctic waters, developing and applying real-time, accurate, and rapid methods for predicting Arctic ship navigation risks to provide information support and intelligent decision-making basis for Arctic ship navigation. Summary of the Invention

[0003] This invention provides a method for predicting navigation risks of Arctic vessels based on a GA-BP neural network model, which solves the problem that the existing technology for assessing and analyzing navigation risks of Arctic vessels is cumbersome and inefficient.

[0004] This invention provides an Arctic vessel navigation risk prediction system based on a GA-BP neural network model, which is used to implement a method for predicting Arctic vessel navigation risks based on a GA-BP neural network model.

[0005] This invention is achieved through the following technical solution: A method for predicting navigation risks of ships in the Arctic based on a GA-BP neural network model, the method comprising the following steps: Step 1: Identify the key risk factors affecting Arctic ship navigation, determine the input and output of the GA-BP neural network model, and construct the dataset; Step 2: Normalize the dataset obtained in Step 1; Step 3: Establish a BP neural network model and determine the model structure; Step 4: Use the genetic algorithm (GA) to optimize the weights and thresholds of the BP neural network to obtain the GA-BP prediction model; Step 5: Train and test the GA-BP prediction model obtained in Step 4 on the dataset that has been normalized in Step 2 to predict the risk of ice trapping and ship-ice collision.

[0006] Furthermore, step 1 specifically involves conducting a sensitivity analysis on the existing Arctic vessel navigation risk assessment model to identify the key factors affecting vessel navigation risks in the Arctic Ocean; using the identified key risk factors as input to the GA-BP prediction model, and using the ice trap / ship-ice collision risk data calculated based on the risk assessment model as output to form a dataset.

[0007] Furthermore, step 2 specifically involves processing the input data before training the GA-BP neural network.

[0008] Furthermore, step 3 specifically involves determining the number of nodes in the input layer, hidden layer, and output layer of the BP model structure, initializing the weights and thresholds, and setting the maximum number of model iterations, error threshold, and learning rate. The number of input layer nodes is the number of key factors affecting the navigation risks of ships in the Arctic Ocean, and the number of output layer nodes is 1. The optimal number of hidden layer nodes for the model is selected based on the minimum mean squared error. The maximum number of iterations, error threshold, and learning rate are set.

[0009] Furthermore, step 4 specifically includes the following steps: Step 4.1: Set the initial parameters of the GA-BP model; Step 4.2: Based on the initial parameters in Step 4.1, encode the chromosomes and generate the initial population. Use real number encoding to encode the weights and thresholds of the BP model, and randomly generate N data points as the initial population. Step 4.3: Use the reciprocal of the mean square error of the dataset as the fitness function to determine the quality of individuals in the population; Step 4.4: Calculate the fitness value of an individual using the standard geometric selection method; Step 4.5: Genetic operations, including crossover and mutation, are used to generate the next generation of individuals; the crossover operation can generate new individuals by exchanging the weights and thresholds between individuals; the mutation operation can introduce random perturbations to increase the diversity of the population. Step 4.6: Repeat steps 4.3 and 4.4 until the stopping condition is met, and select the individual with the highest fitness as the optimized initial weights and threshold.

[0010] Furthermore, step 5 specifically involves the GA-BP neural network model using optimized initial weights and thresholds to predict the navigation risks of ships in the Arctic.

[0011] A risk prediction system for Arctic vessel navigation based on a GA-BP neural network model, the system employing the aforementioned method for predicting Arctic vessel navigation risks based on a GA-BP neural network model, the system comprising: Dataset Modules: Identify key risk factors affecting Arctic ship navigation, determine the input and output of the GA-BP neural network model, and construct the dataset; Dataset preprocessing module: Normalizes the dataset obtained in step 1; GA-BP prediction model building module: Establishes the BP neural network model and determines the model structure; The weights and thresholds of the BP neural network are optimized using the genetic algorithm (GA) to obtain the GA-BP prediction model. Training and testing the GA-BP prediction model module: The GA-BP prediction model is trained and tested on the normalized dataset to predict the risk of ice trapping and ship-ice collision.

[0012] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method described above.

[0013] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described above.

[0014] The beneficial effects of this invention are: This invention can predict the risks of ice entrapment and ship-ice collisions for Arctic ships in real time, accurately and quickly, providing information support and intelligent decision-making basis for Arctic ship navigation. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method of the present invention.

[0016] Figure 2 This is a flowchart of the Arctic vessel navigation risk GA-BP prediction model of the present invention.

[0017] Figure 3 This is a schematic diagram of the Arctic shipping risk assessment model of the present invention.

[0018] Figure 4 This is a training result diagram of the ice-trapping risk based on the GA-BP neural network model of the present invention, where (a) is the ice-trapping risk training set, (b) is the ice-trapping risk validation set, (c) is the ice-trapping risk test set, and (d) is the complete ice-trapping risk dataset.

[0019] Figure 5This is a training result diagram of the ship-ice collision risk based on the GA-BP neural network model of the present invention, where (a) is the training set of ship-ice collision risk, (b) is the validation set of ship-ice collision risk, (c) is the test set of ship-ice collision risk, and (d) is the complete dataset of ship-ice collision risk.

[0020] Figure 6 This is a comparison chart of the predicted and calculated values ​​of the ice-induced risk using the GA-BP neural network model of this invention.

[0021] Figure 7 This is a comparison chart of the predicted and calculated values ​​of ship-ice collision risk using the GA-BP neural network model of this invention. Detailed Implementation

[0022] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0023] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0024] It should also be understood that the terminology used in this application specification is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this application specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0025] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0026] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0027] Implementation Method 1 This embodiment provides a method for predicting Arctic vessel navigation risks based on a GA-BP neural network model. This invention introduces a GA-BP neural network into the existing Bayesian network risk assessment model for Arctic vessel navigation. Using risk factors as input and risk data calculated from the risk assessment model as output, the GA-BP neural network is used to train the navigation risk prediction model. The algorithm is continuously optimized to achieve optimal parameters and training accuracy, enabling rapid and accurate prediction of the risk values ​​for ice entrapment and ship-ice collisions, thereby supporting vessel operational decisions.

[0028] The method includes the following steps: Step 1: Identify the key risk factors affecting Arctic ship navigation, determine the input and output of the GA-BP neural network model, and construct the dataset; Step 2: Normalize the dataset obtained in Step 1; Step 3: Establish a BP neural network model and determine the model structure; Step 4: Use the genetic algorithm (GA) to optimize the weights and thresholds of the BP neural network to obtain the GA-BP prediction model; Step 5: Train and test the GA-BP prediction model obtained in Step 4 on the dataset that has been normalized in Step 2 to predict the risk of ice trapping and ship-ice collision.

[0029] Furthermore, step 1 specifically involves conducting a sensitivity analysis on the existing Arctic vessel navigation risk assessment model to identify the key factors affecting vessel navigation risks in the Arctic Ocean; using the identified key risk factors as input to the GA-BP prediction model, and using the ice trap / ship-ice collision risk data calculated based on the risk assessment model as output to form a dataset. Sensitivity analysis can be used to analyze the degree of influence of changes in model input parameters on the target output, thereby determining the parameter sensitivity of the model and effectively identifying the key input parameters that affect the output.

[0030] The principle of sensitivity analysis is to reflect the sensitivity of the basic nodes relative to the target node by the maximum absolute value change of the target node's value. Assume... A For a Bayesian network node undergoing sensitivity analysis, the node state can be represented as follows: For a given piece of evidence Node A is The conditional probability in a state can be expressed as ;assumed B For nodes A The parent node, whose node state can be represented as , For nodes B In state Conditional probability parameters at time; nodes A exist The conditional probability in a state can be expressed as Linear functions: (1) in: and These are real constants, representing the slope and intercept of the linear function, respectively; At the node B of Under the influence of state, nodes A exist The sensitivity value under a given state can be calculated using the following formula: (2) Equation (2) only considers the sensitivity value of a child node in a certain state under the influence of a single state of the parent node. Extending the above equation, we can obtain the sensitivity value of the child node in a certain state under the influence of a single state of the parent node. B Under the influence of various states ( ),node A Average sensitivity value: (3) in, n 0 represents the total number of states of parent node B. i This is the sample number.

[0031] Furthermore, step 2 specifically involves considering that when the input data differs significantly in numerical range, it will reduce the learning speed of the GA-BP neural network and may cause the network to fail to converge. Therefore, it is necessary to process the input data before training the GA-BP neural network to facilitate the uniform processing of data samples of different magnitudes, thereby enabling rapid analysis of the relationships between the data.

[0032] The data was normalized using the mapminmax function in Matlab. (4) Where X is the original sample, , These are the minimum and maximum values ​​in the original sample input. , These are the local and local minima after mapping. It is usually 0. It is usually 1. This is the data after normalization. This process can be applied to scale the original input sample proportionally.

[0033] Furthermore, step 3 specifically involves determining the BP model structure, including determining the number of nodes in the input layer, hidden layer, and output layer, initializing weights and thresholds, and setting the maximum number of model iterations, error threshold, and learning rate. The number of input layer nodes is the number of key factors affecting the navigation risks of ships in the Arctic Ocean, and the number of output layer nodes is 1. The number of hidden layer nodes has a significant impact on the prediction performance of the entire neural network model. It can be initially determined and appropriately adjusted according to Equation (5). The number of hidden layer nodes with the smallest mean square error of the model is selected as the optimal number of hidden layer nodes of the model. The mean square error (MSE) is the performance function of the network. This statistical parameter is the mean of the sum of squares of the errors of corresponding points in the predicted data and the original data. Its expression is Equation (5). Set the maximum number of iterations, error threshold and learning rate.

[0034] The maximum number of iterations is set to 1000, and the error threshold is set to 10. -6 The learning rate was set to 0.01. (5) Where N is the number of hidden layer nodes. k The number of nodes in the input layer. l This represents the number of nodes in the output layer. It is an integer between 0 and 10.

[0035] (6) in, Mean square error, The risk predicted by the GA-BP model, The risk calculated by the risk assessment model, n 1 represents the number of training data points.

[0036] Furthermore, step 4 specifically includes the following steps: Step 4.1: Set the initial parameters of the GA-BP model; Step 4.2: Encode chromosomes and generate an initial population based on the initial parameters from Step 4.1. Since the genetic algorithm cannot directly process the parameters of the problem space, it is necessary to encode the problem to be solved as chromosomes or individuals in the genetic space. Real number encoding is used to encode the weights and thresholds of the BP model. The real number space is [1e-6, 1]. N data points are randomly generated as the initial population. Step 4.3: The reciprocal of the mean squared error of the dataset is used as the fitness function to determine the quality of individuals in the population; the formula for the fitness function in Step 4.3 is as follows: (7); Step 4.4: Use the standard geometric selection method (normGeomSelect function in Matlab Genetic Algorithm Toolbox) to calculate the fitness value of individuals, and select parent individuals based on the fitness value. Individuals with higher fitness are more likely to be selected. Step 4.5: Genetic operations, including crossover and mutation, are used to generate the next generation of individuals. Crossover generates new individuals by exchanging weights and thresholds between them. Mutation introduces random perturbations to increase population diversity. The crossover operation uses arithmetic crossover (the `arithXover` function in the Matlab Genetic Algorithm Toolbox), simulating gene exchange in biological evolution by combining the genetic information of two individuals to generate two new individuals. The mutation operation uses non-uniform mutation (the `nonUnifMutation` function in the Matlab Genetic Algorithm Toolbox), simulating gene mutation in biological evolution by introducing random perturbations into the genetic information of individuals to increase the exploration capability of the search space. The probability of mutation gradually decreases with increasing iteration count. Step 4.6: Repeat steps 4.3 and 4.4 until the stopping condition is met (e.g., reaching the maximum number of iterations or obtaining a satisfactory fitness level). Finally, select the individual with the highest fitness as the optimized initial weights and thresholds.

[0037] Furthermore, step 5 specifically involves the GA-BP neural network model using optimized initial weights and thresholds to predict the navigation risks of ships in the Arctic.

[0038] A method for predicting navigation risks in the Arctic based on a GA-BP neural network model, taking a ship navigating in the Arctic shipping route as an example, is implemented in the following steps: Step 1: Identify the key risk factors affecting Arctic ship navigation, determine the input and output of the GA-BP neural network model, and construct the dataset; In step 1, a sensitivity analysis is performed on the existing Arctic vessel navigation risk assessment model to identify the key factors affecting vessel navigation risks in the Arctic Ocean. The identified key risk factors are used as input to the GA-BP prediction model, and the ice entanglement / ship-ice collision risk data calculated based on the risk assessment model are used as output to form a dataset.

[0039] Sensitivity analysis can be used to analyze the degree of influence of changes in model input parameters on the target output, thereby determining the parameter sensitivity of the model and effectively identifying the key input parameters that affect the output.

[0040] The principle of sensitivity analysis is to reflect the sensitivity of the basic nodes relative to the target node by the maximum absolute value change of the target node's value. Assume... AFor a Bayesian network node undergoing sensitivity analysis, the node state can be represented as follows: For a given piece of evidence ,node A exist The conditional probability in a state can be expressed as .assumed B For nodes A The parent node, whose node state can be represented as , For nodes B In state The conditional probability parameter at that time. Node A exist The conditional probability in a state can be expressed as Linear functions: (1) in: and These are real constants, representing the slope and intercept of the linear function, respectively.

[0041] At the node B of Under the influence of state, nodes A exist The sensitivity value under a given state can be calculated using the following formula: (2) Equation (2) only considers the sensitivity value of a child node in a certain state under the influence of a single state of the parent node. Extending the above equation, we can obtain the sensitivity value of the child node in a certain state under the influence of a single state of the parent node. B Under the influence of various states ( ),node A Average sensitivity value: (3) in, n 0 represents the total number of states of parent node B. i For the first i Network nodes.

[0042] Step 1 is specifically implemented by revising existing Arctic vessel navigation risk assessment models (such as...). Figure 3 Sensitivity analysis was conducted to identify key factors affecting the navigation risks of ships in the Arctic Ocean, as shown in Tables 1 and 2. Seven of the most important risk factors, namely "sea ice concentration", "sea ice thickness", "ship ice class", "ship speed", "ice navigation experience", "professional skills and qualifications", and "visibility", were selected as inputs to the GA-BP prediction model. The ice trap / ship-ice collision risk data calculated based on the risk assessment model were used as outputs, forming 120 datasets.

[0043] Table 1 Sensitivity Analysis of Ice-Related Risks in the Arctic Shipping Routes

[0044] Table 2 Sensitivity Analysis of Ship Ice Collision Risk in the Arctic Shipping Route

[0045] Step 2: Normalize the obtained dataset; The `mapminmax` function in Matlab is used to normalize the data to facilitate and uniformly process data samples of different magnitudes. The specific formula is as follows: (4) Where X is the original sample. , These are the minimum and maximum values ​​in the original sample input. , These are the local and local minima after mapping. It is usually 0. It is usually 1. This is the data after normalization.

[0046] Step 3: Establish a BP neural network model and determine the model structure; Step 3 specifically involves determining the BP model structure, including determining the number of nodes in the input layer, hidden layer, and output layer, initializing weights and thresholds, and setting the maximum number of model iterations, error threshold, and learning rate.

[0047] The number of input layer nodes is the number of key factors affecting the navigation risks of ships in the Arctic Ocean, and the number of output layer nodes is 1.

[0048] The number of hidden layer nodes has a significant impact on the prediction performance of the entire neural network model. It can be initially determined and appropriately adjusted according to equation (5), selecting the number of hidden layer nodes with the smallest mean square error as the optimal number of hidden layer nodes for the model. Mean square error (MSE) is the performance function of the network; this statistical parameter is the mean of the sum of squares of the errors at corresponding points in the predicted data and the original data, and its expression is given by equation (5). The maximum number of iterations is set to 1000, and the error threshold is set to 10. -6 The learning rate was set to 0.01. (5) Where N is the number of hidden layer nodes. k The number of nodes in the input layer. l This represents the number of nodes in the output layer, where a is an integer between 0 and 10.

[0049] (6) in, Mean square error, The risk predicted by the GA-BP model, The risk calculated by the risk assessment model, n 1 represents the number of training data points.

[0050] Step 3 specifically involves designing a BP neural network model and experimenting with structures having different numbers of hidden layer nodes. As shown in Table 3, the structure with the minimum mean square error was selected as the optimal model structure. The BP neural network structure parameters were set as follows: the number of input layer nodes was 7 (the number of key factors affecting the navigation risk of ships in the Arctic Ocean), the number of output layer nodes was 1, the number of hidden layer nodes was 5, the maximum number of iterations was set to 1000, and the error threshold was set to 10. -6 The learning rate was set to 0.01.

[0051] Table 3. Mean Squared Error of Training Data for Different Numbers of Hidden Layer Nodes

[0052] Step 4: Optimize the weights and thresholds of the BP neural network using a genetic algorithm (GA); Step 4.1: The initial parameters of the GA-BP model are set as follows: the initial population size is set to 20, the number of generations is set to 100, the crossover probability is set to 0.9, and the mutation probability is set to 0.1. Step 4.2: Chromosome encoding and initial population generation; Since genetic algorithms cannot directly process parameters in the problem space, it is necessary to represent the problem to be solved as chromosomes or individuals in the genetic space through encoding. Real-number encoding is used to encode the weights and thresholds of the BP model, and the real-number space is

[10] . -6 [1], randomly generate N data points as the initial population.

[0053] Step 4.3: Calculate the fitness value; The reciprocal of the mean squared error of the dataset is used as the fitness function to determine the quality of individuals in the population. The fitness function formula is as follows: (7) Step 4.4: Select an operation; The standard geometric selection method (normGeomSelect function in Matlab's genetic algorithm toolbox) is used to calculate the fitness value of individuals, and parent individuals are selected based on the fitness value. Individuals with higher fitness are more likely to be selected.

[0054] Step 4.5: Genetic manipulation; Genetic operations, including crossover and mutation, are used to generate the next generation of individuals. Crossover generates new individuals by exchanging the weights and thresholds between them. Mutation introduces random perturbations to increase population diversity. The crossover operation uses arithmetic crossover (the `arithXover` function in the Matlab Genetic Algorithm Toolbox), simulating gene exchange in biological evolution by combining the genetic information of two individuals to generate two new individuals. The mutation operation uses non-uniform mutation (the `nonUnifMutation` function in the Matlab Genetic Algorithm Toolbox), simulating gene mutation in biological evolution by introducing random perturbations into the genetic information of individuals to increase the exploration capability of the search space. The probability of mutation gradually decreases with increasing iterations.

[0055] Step 4.6: Repeat steps 4.3 and 4.4 until the stopping condition is met (e.g., reaching the maximum number of iterations or obtaining a satisfactory fitness level). Finally, select the individual with the highest fitness as the optimized initial weights and thresholds.

[0056] Step 5: Train and test the GA-BP prediction model to predict the risk of ice trapping and ship-ice collision; The dataset in step 1 consists of 120 groups, randomly divided into a training set (70%), a validation set (15%), and a test set (15%). The training set is used to adjust network weights and optimize the model structure; the test set is used to check the generalization ability of the neural network; and the validation set is used to validate the model. The GA-BP neural network model of this invention is trained using the initial weights and thresholds optimized in step 4. The training regression results are as follows: Figure 4 As shown.

[0057] in Figure 4 The image shows the training results of the ice-induced risk model based on the GA-BP neural network model, where (a) is the ice-induced risk training set, (b) is the ice-induced risk validation set, (c) is the ice-induced risk test set, and (d) is the complete ice-induced risk dataset. T=Y, where T is the true value and Y is the predicted value. Figure 5 The image shows the training results of the GA-BP neural network model for ice collision risk, where (a) is the training set, (b) is the validation set, (c) is the test set, and (d) is the complete dataset. T = Y, where T is the true value and Y is the predicted value. Twenty sets of untrained data are input into the constructed model for prediction. The prediction results are compared with the results calculated by the risk assessment model. Figure 6 As shown, from Figure 6 The results show that the predicted values ​​are close to the calculated values, indicating that the constructed GA-BP navigation risk prediction model has good predictive performance.

[0058] Therefore, the Arctic vessel navigation risk prediction method based on the GA-BP neural network model of this invention can quickly and accurately predict the risk values ​​of ice trapping and ship-ice collision, thereby guiding ship operations and supporting ship operation decisions.

[0059] Implementation Method 2 A risk prediction system for Arctic vessel navigation based on a GA-BP neural network model is provided. The system uses a risk prediction method for Arctic vessel navigation based on a GA-BP neural network model as described in Embodiment 1. The system includes: a dataset composition module: identifying key risk factors affecting Arctic vessel navigation, determining the input and output of the GA-BP neural network model, and constructing a dataset. Dataset preprocessing module: Normalizes the dataset obtained in step 1; GA-BP prediction model building module: Establishes the BP neural network model and determines the model structure; The weights and thresholds of the BP neural network are optimized using the genetic algorithm (GA) to obtain the GA-BP prediction model. Training and testing the GA-BP prediction model module: The GA-BP prediction model is trained and tested on the normalized dataset to predict the risk of ice trapping and ship-ice collision.

[0060] As can be seen from the above, the Arctic vessel navigation risk prediction method provided by the embodiments of the present invention uses a GA-BP neural network model based on a BP neural network model to predict the risks of ice entrapment and ship-ice collision. The algorithm is continuously optimized to achieve optimal parameters and training accuracy, enabling rapid and accurate prediction of the risk values ​​of ice entrapment and ship-ice collision, thereby supporting vessel operation decisions.

[0061] Implementation Method 3 A prediction method as described in Embodiment 1 is applied to the field of ship navigation risk prediction in the Arctic shipping route, which can quickly and accurately predict the risk values ​​of ice trapping and ship-ice collision, thereby supporting ship operation decisions.

[0062] Implementation Method 4 This invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor. The memory stores software programs and modules, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory and processor are connected via a bus. Specifically, the processor implements any step in Embodiment 1 by running the computer program stored in the memory.

[0063] It should be understood that, in the embodiments of this invention, the processor may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0064] Memory may include read-only memory, flash memory, and random access memory, and provides instructions and data to the processor. Some or all of the memory may also include non-volatile random access memory.

[0065] As can be seen from the above, the electronic device provided in this embodiment of the invention can implement the Arctic vessel navigation risk prediction method as described in Embodiment 1 by running a computer program. The GA-BP neural network model, based on the BP neural network model, is used to predict the risks of ice entrapment and ship-ice collision. The algorithm is continuously optimized to achieve optimal parameters and training accuracy, quickly and accurately predicting the risk values ​​of ice entrapment and ship-ice collision, thereby supporting ship operation decisions.

[0066] 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 skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for predicting Arctic ship navigation risks based on a GA-BP neural network model, characterized in that, The method includes the following steps: Step 1: Conduct sensitivity analysis on the existing Bayesian network risk assessment model for Arctic vessel navigation to identify key factors affecting the risk of vessel navigation in the Arctic. These key factors include: sea ice concentration, sea ice thickness, vessel ice class, vessel speed, experience in ice-covered navigation, professional skills and qualifications, and visibility. Use the quantified values ​​of these key factors as input to the GA-BP neural network model, and use the ice trap / ship-ice collision risk data calculated based on the Bayesian network risk assessment model as output to identify key risk factors affecting Arctic vessel navigation, determine the input and output of the GA-BP neural network model, and construct a dataset. Step 2: Normalize the dataset obtained in Step 1; Step 3: Establish a BP neural network model and determine the model structure; Step 4: Use the genetic algorithm (GA) to optimize the weights and thresholds of the BP neural network to obtain the GA-BP prediction model; Step 5: Train and test the GA-BP prediction model obtained in Step 4 on the dataset that has been normalized in Step 2 to predict the risk of ice trapping and ship-ice collision. Step 1 specifically involves conducting a sensitivity analysis on the existing Arctic vessel navigation risk assessment model to identify the key factors affecting vessel navigation risks in the Arctic Ocean; using the identified key risk factors as input to the GA-BP prediction model, and using the ice trap / ship-ice collision risk data calculated based on the risk assessment model as output to form a dataset.

2. The method for predicting Arctic ship navigation risks based on the GA-BP neural network model according to claim 1, characterized in that, Specifically, step 2 involves processing the input data before training the GA-BP neural network. When the input data differs significantly in numerical range, it will reduce the learning speed of the GA-BP neural network and may cause the network to fail to converge.

3. The method for predicting Arctic ship navigation risks based on the GA-BP neural network model according to claim 1, characterized in that, Step 3 specifically involves determining the number of nodes in the input layer, hidden layer, and output layer of the BP model structure, initializing weights and thresholds, and setting the maximum number of model iterations, error threshold, and learning rate. The number of input layer nodes is the number of key factors affecting the navigation risks of ships in the Arctic Ocean, and the number of output layer nodes is 1. The number of hidden layer nodes that minimizes the mean square error of the model is selected as the optimal number of hidden layer nodes for the model. Set the maximum number of iterations, the error threshold, and the learning rate.

4. The method for predicting Arctic ship navigation risks based on the GA-BP neural network model according to claim 1, characterized in that, Step 4 specifically includes the following steps: Step 4.1: Set the initial parameters of the GA-BP model; Step 4.2: Based on the initial parameters in Step 4.1, encode the chromosomes and generate the initial population. Use real number encoding to encode the weights and thresholds of the BP model, and randomly generate N data points as the initial population. Step 4.3: Use the reciprocal of the mean square error of the dataset as the fitness function to determine the quality of individuals in the population; Step 4.4: Calculate the fitness value of an individual using the standard geometric selection method; Step 4.5: Genetic operations, including crossover and mutation, are used to generate the next generation of individuals; the crossover operation can generate new individuals by exchanging the weights and thresholds between individuals; the mutation operation can introduce random perturbations to increase the diversity of the population. Step 4.6: Repeat steps 4.3 and 4.4 until the stopping condition is met, and select the individual with the highest fitness as the optimized initial weights and threshold.

5. The method for predicting Arctic ship navigation risks based on the GA-BP neural network model according to claim 1, characterized in that, Step 5 specifically involves the GA-BP neural network model using optimized initial weights and thresholds to predict the navigation risks of ships in the Arctic.

6. A risk prediction system for Arctic ship navigation based on a GA-BP neural network model, characterized in that, The system uses an Arctic vessel navigation risk prediction method based on a GA-BP neural network model as described in any one of claims 1-5, and the system includes: Dataset Construction Module: Sensitivity analysis is performed on existing Bayesian network risk assessment models for Arctic vessel navigation to identify key factors affecting vessel navigation risks in the Arctic. These key factors include: sea ice concentration, sea ice thickness, vessel ice class, vessel speed, ice navigation experience, professional skills, and visibility. The quantified values ​​of these key factors are used as input to the GA-BP neural network model, and the ice entrapment / ship-ice collision risk data calculated based on the Bayesian network risk assessment model are used as output to identify key risk factors affecting Arctic vessel navigation. The input and output of the GA-BP neural network model are then determined, forming the dataset. Dataset preprocessing module: Normalizes the dataset obtained in step 1; GA-BP prediction model building module: Establishes the BP neural network model and determines the model structure; The weights and thresholds of the BP neural network are optimized using the genetic algorithm (GA) to obtain the GA-BP prediction model. Training and testing the GA-BP prediction model module: The GA-BP prediction model is trained and tested on the normalized dataset to predict the risk of ice trapping and ship-ice collision.

7. A prediction method according to any one of claims 1-5 is applied to the field of ship navigation risk prediction in the Arctic shipping route, which can quickly and accurately predict the risk values ​​of ice trapping and ship-ice collision, thereby supporting ship operation decisions.

8. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the method as described in any one of claims 1-5.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1-5.

Images