A Ship Transportation Efficiency Intelligent Control System and Method Optimized Based on Large Models

Abstract

The present invention relates to the field of intelligent control of ship operation efficiency, and particularly to an intelligent control system and method for ship operation efficiency optimized based on a large model. It includes: collecting real-time data of ship operation and environment, eliminating dimension differences through normalization processing to obtain normalized data, forming a normalized data vector, and using a multi-dimensional data feature mapping algorithm to extract features from the normalized data vector to obtain a feature vector; calculating the optimal solution through a dynamic adaptive perturbation optimization algorithm; generating a dimensionless control vector based on the gap between the optimal solution and the initial solution, and converting the dimensionless control vector into an actual control vector through an inverse normalization operation. It solves the technical problems that existing methods cannot comprehensively consider the mutual influence between ships and environmental factors, especially the efficient optimization in complex dynamic environments; traditional optimization methods are prone to falling into local optimal solutions; and there is a lack of real-time and accurate adjustment in traditional control methods.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control of ship transportation efficiency, and particularly to a ship transportation efficiency intelligent control system and method optimized based on a large model. Background Art

[0002] In the field of modern ship transportation, with the advancement of global economic integration and the rapid growth of international trade, the shipping industry is facing more and more challenges. How to improve the operating efficiency of ships, reduce operating costs, optimize energy consumption during navigation, and at the same time reduce environmental pollution has become an important issue that the industry urgently needs to solve. Traditional ship transportation efficiency optimization methods mostly rely on experience and rules. Although they can improve efficiency to a certain extent, they often show great limitations when facing variable factors such as complex sea conditions, route changes, and diverse ship types.

[0003] Against this background, the ship transportation efficiency intelligent control method optimized based on a large model has emerged. The large model optimization technology realizes the refined management and real-time dynamic adjustment of the operating efficiency within the entire life cycle of the ship by leveraging advanced artificial intelligence, big data analysis, deep learning and other technologies, combined with real-time ship operation data and environmental data, thereby improving shipping efficiency, reducing energy consumption, optimizing the navigation path, and ultimately realizing the intelligent control of ship transportation efficiency.

[0004] However, the existing ship transportation efficiency intelligent control system and method optimized based on a large model still have technical problems such as the inability to fully optimize fuel consumption and energy utilization during the ship operation process, resulting in energy waste, high carbon emissions, and high operating costs; and the inability to fully consider the interaction effects of dynamic environmental factors, resulting in the failure to maximize the operation efficiency of the ship. Summary of the Invention

[0005] The present invention provides a ship transportation efficiency intelligent control system and method optimized based on a large model to solve the technical problems that the existing methods cannot comprehensively consider the mutual influence between the ship and environmental factors, especially the efficient optimization in a complex dynamic environment; the traditional optimization methods are prone to falling into local optimal solutions; and the lack of real-time and accurate adjustment in the traditional control methods.

[0006] The ship transportation efficiency intelligent control system and method optimized based on a large model of the present invention specifically include the following technical solutions:

[0007] A ship transportation efficiency intelligent control method optimized based on a large model includes the following steps:

[0008] S1. Collect real-time data of ship operation and environment, and eliminate the dimension difference through normalization processing to obtain the normalized data, which constitutes a normalized data vector. Use a multi-dimensional data feature mapping algorithm to extract features from the normalized data vector to obtain a feature vector;

[0009] S2. Perform perturbation search and adaptive adjustment through the dynamic adaptive perturbation optimization algorithm, and calculate the optimal solution; based on the gap between the optimal solution and the initial solution, generate a dimensionless control vector, and convert the dimensionless control vector into an actual control vector through an anti-normalization operation.

[0010] Preferably, the S1 specifically includes:

[0011] The multi-dimensional data feature mapping algorithm integrates the currently normalized data vector and the historically normalized data vector as the basic part of the input enhancement matrix, and introduces a pseudo-periodic perturbation vector and an environmental coupling factor vector to form the input enhancement matrix.

[0012] Preferably, the S1 specifically includes:

[0013] In the implementation process of the multi-dimensional data feature mapping algorithm, the input enhancement matrix is transformed into an intermediate feature matrix through non-linear feature mapping, and the calculation formula is:

[0014] ,

[0015] where, represents the intermediate feature matrix at time ; represents the transpose of the input enhancement matrix; represents the transpose; represents the input enhancement matrix at time ; represents the square of the Frobenius norm, which is part of the normalization factor; represents the environmental attenuation matrix, which reflects the attenuation effect of environmental factors at time .

[0016] Preferably, the S1 specifically includes:

[0017] In the implementation process of the multi-dimensional data feature mapping algorithm, through the reverse mapping of the intermediate feature matrix, the intermediate feature matrix is reverse mapped back to the input space to obtain the pseudo-inverse matrix of the intermediate feature matrix, and a dynamic projection matrix is designed. The pseudo-inverse matrix of the intermediate feature matrix is combined with the dynamic projection matrix to finally obtain the feature vector, including fuel efficiency and energy utilization rate.

[0018] Preferably, the S2 specifically includes:

[0019] The dynamic adaptive perturbation optimization algorithm introduces random perturbation and adaptive step size, and dynamically adjusts the search direction and step size by using perturbation feedback and historical experience; in the initialization stage, the optimization objective function is defined and the initial solution is set; the initial solution consists of the fuel efficiency and energy utilization rate at the current moment.

[0020] Preferably, S2 specifically includes:

[0021] In the implementation process of the dynamic adaptive perturbation optimization algorithm, iterative search is carried out. In each iteration, a random perturbation vector is generated to perturb the current solution, thereby obtaining a candidate solution; the random perturbation vector is generated through a standard normal distribution and is normalized after each generation to make it a unit vector, that is, the random perturbation direction.

[0022] Preferably, S2 specifically includes:

[0023] In the implementation process of the dynamic adaptive perturbation optimization algorithm, an adaptive step size mechanism is introduced. According to the difference between the current optimization objective function value and the historical maximum optimization objective function value, the perturbation step size is dynamically adjusted. The calculation formula of the perturbation step size is:

[0024] ,

[0025] where, represents the perturbation step size of the th iteration; represents the perturbation step size of the th iteration; represents the step size decay factor; represents the function used to adjust the perturbation step size; represents the optimization objective function value of the th iteration; represents the historical maximum optimization objective function value.

[0026] Preferably, S2 specifically includes:

[0027] In the implementation process of the dynamic adaptive perturbation optimization algorithm, based on the perturbation step size, candidate solutions are generated by means of weighted sum; after multiple rounds of iteration, the optimal solution is finally output, that is, the fuel efficiency and energy utilization rate that maximize the optimization objective function value during the entire optimization process.

[0028] A ship operation efficiency intelligent control system based on large model optimization includes the following parts:

[0029] Real-time data acquisition and normalization processing module, feature extraction module, adaptive optimization processing module, control generation module, denormalization module;

[0030] Real-time data acquisition and normalization module: It acquires the real-time data of ship operation and environment, eliminates the dimension difference through normalization processing to obtain the normalized data, forms the normalized data vector from the normalized data, and outputs the normalized data vector to the feature extraction module;

[0031] Feature extraction module: It uses the multi-dimensional data feature mapping algorithm to extract features from the normalized data vector from the real-time data acquisition and normalization module to obtain the feature vector, which includes fuel efficiency and energy utilization rate. The fuel efficiency and energy utilization rate form the initial solution, and the feature vector is output to the adaptive optimization processing module and the feature vector is output as the initial solution to the control generation module;

[0032] Adaptive optimization processing module: Based on the feature vector of the feature extraction module, it performs perturbation search and adaptive adjustment through the dynamic adaptive perturbation optimization algorithm, calculates the fuel efficiency and energy utilization rate that maximize the optimization objective function value to obtain the optimal solution, and outputs the optimal solution to the control generation module;

[0033] Control generation module: Based on the gap between the optimal solution of the adaptive optimization processing module and the initial solution of the feature extraction module, it generates a dimensionless control vector through the proportional control method and outputs the dimensionless control vector to the inverse normalization module;

[0034] Inverse normalization module: It converts the dimensionless control vector from the control generation module into an actual control vector through inverse normalization operation.

[0035] The beneficial effects of the technical solution of the present invention are:

[0036] 1. By acquiring the real-time data of ship operation and environment and performing normalization processing, the dimension difference is eliminated, ensuring the consistency and comparability of the data, enabling the ship operation data to be accurately analyzed and processed, laying a foundation for subsequent feature extraction and optimization decision-making, and ensuring the quality and reliability of the data.

[0037] 2. By extracting the key features in the ship operation data, including fuel efficiency and energy utilization rate, the operation state of the ship can be accurately described. By introducing the pseudo-periodic perturbation vector and the environmental coupling factor vector, the time dependence and environmental interaction effects can be captured, reflecting the dynamic changes of the ship under different operating conditions, providing accurate data support for ship operation efficiency optimization, and effectively improving the accuracy of optimization decision-making.

[0038] 3. Through non - linear feature mapping, the complex patterns hidden in the data are further revealed. At the same time, by integrating the environmental impact through the environmental attenuation matrix and considering factors such as the normalized wave and wind speed on ship efficiency, it can accurately capture the inhibitory effect of environmental factors on ship operation efficiency, thus providing a more comprehensive and refined feature description for ship operation efficiency optimization and making the optimization results more in line with actual operating conditions.

[0039] 4. The introduction of the dynamic projection matrix enhances the directionality of the eigenvectors. Especially by introducing the influence of the normalized heading angle, it further enhances the ability to represent the ship's operating state, making the calculation of the final fuel efficiency and energy utilization rate more accurate and providing a reliable basis for ship optimization decisions.

[0040] 5. By introducing random perturbations and adaptive step sizes, the dynamic adaptive perturbation optimization algorithm can efficiently explore the non - linear space, avoiding the limitations of traditional methods. It can find a balance between the global optimal solution and the local optimal solution. By continuously adjusting the perturbation step size and direction, it can quickly approach the maximum value of the optimization objective function, effectively improving the accuracy and efficiency of ship optimization decisions and ensuring that the ship can achieve the best fuel efficiency and energy utilization rate in a complex environment.

[0041] 6. By designing a time - smoothness penalty term, the optimization objective function can ensure that the ship remains stable during the optimization process, without significant performance fluctuations, improving the stability and reliability of the ship operation efficiency intelligent control system, avoiding the instability caused by over - adjustment, and enhancing the operability and adaptability of the ship in practical applications.

[0042] 7. Through the anti - normalization operation, the dimensionless control vector is converted into an actual control vector, enabling the adjustment of the ship's speed and heading to be reasonably executed in actual operation, ensuring that the ship can accurately adjust its operating state according to the optimization results, thereby improving fuel efficiency and energy utilization rate, reducing operating costs, and enhancing the ship's operating performance.

[0043] 8. Introducing a constraint boundary mechanism ensures that each optimized solution meets physical feasibility, guarantees that the ship operates within the actual operable range, improves the safety and practicality of the optimization process, and avoids the emergence of solutions that do not meet actual operating conditions. Brief Description of the Drawings

[0044] Figure 1 It is a structural diagram of a ship operation efficiency intelligent control system based on large - model optimization according to the present invention;

[0045] Figure 2 It is a flowchart of a ship operation efficiency intelligent control method based on large - model optimization according to the present invention. Detailed Embodiments

[0046] To further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0047] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0048] The following specifically describes the specific solutions of a ship operation efficiency intelligent control system and method based on large model optimization provided by the present invention in conjunction with the accompanying drawings.

[0049] Referring to the attached Figure 1 , which shows the structure diagram of a ship operation efficiency intelligent control system based on large model optimization provided by an embodiment of the present invention. The system includes the following parts:

[0050] A real-time data acquisition and normalization processing module, a feature extraction module, an adaptive optimization processing module, a control generation module, and an inverse normalization module;

[0051] Real-time data acquisition and normalization processing module: Collect real-time data of ship operation and environment, and eliminate the dimension difference through normalization processing to obtain normalized data, including: normalized ship speed, engine power, fuel consumption rate, wind speed, wave height, and heading angle. The normalized data is composed of a normalized data vector, and the normalized data vector is output to the feature extraction module;

[0052] Feature extraction module: Use a multi-dimensional data feature mapping algorithm to extract features from the normalized data vector from the real-time data acquisition and normalization processing module to obtain a feature vector, including fuel efficiency and energy utilization rate. The fuel efficiency and energy utilization rate constitute an initial solution, and the feature vector is output to the adaptive optimization processing module and the feature vector is output as an initial solution to the control generation module;

[0053] Adaptive optimization processing module: Based on the feature vector of the feature extraction module, perform perturbation search and adaptive adjustment through a dynamic adaptive perturbation optimization algorithm, calculate the fuel efficiency and energy utilization rate that maximize the optimization objective function value, obtain the optimal solution, and output the optimal solution to the control generation module;

[0054] Control generation module: Based on the gap between the optimal solution of the adaptive optimization processing module and the initial solution of the feature extraction module, a dimensionless control vector is generated through a proportional control method, and the dimensionless control vector is output to the denormalization module;

[0055] Denormalization module: The dimensionless control vector from the control generation module is converted into an actual control vector through denormalization operations.

[0056] Refer to the appendix Figure 2 , which shows a flowchart of a ship operation efficiency intelligent control method based on large model optimization provided by an embodiment of the present invention. The method includes the following steps:

[0057] S1. Collect real-time data of ship operation and environment, eliminate dimensional differences through normalization processing to obtain normalized data, form a normalized data vector, and use a multi-dimensional data feature mapping algorithm to extract features from the normalized data vector to obtain a feature vector;

[0058] Collect real-time data of ship operation and environment, and eliminate dimensional differences through normalization processing to obtain normalized data, including: normalized ship speed, engine power, fuel consumption rate, wind speed, wave height, and heading angle. The normalized data is formed into a normalized data vector, and the expression form is as follows:

[0059] ,

[0060] where, represents at time normalized data vector; represents at time normalized ship speed; represents at time normalized engine power; represents at time normalized fuel consumption rate; represents at time normalized wind speed; represents at time normalized wave height; represents at time normalized heading angle; represents transpose;

[0061] In order to extract practical operation features from the real-time data of ship operation and provide data support for ship operation efficiency optimization decision-making, a multi-dimensional data feature mapping algorithm is used to extract features from the normalized data vector to obtain a feature vector, including fuel efficiency and energy utilization rate;

[0062] The multi-dimensional data feature mapping algorithm integrates the currently normalized data vector and the historically normalized data vector as the basic part of the input enhancement matrix. To capture the time dependence and environmental interaction effects, the multi-dimensional data feature mapping algorithm introduces two additional perturbation terms: one is the pseudo-periodic perturbation vector, which generates periodically varying features based on the normalized ship speed and the normalized wind speed through sine and cosine functions. Specifically, the sine value is calculated by multiplying the normalized ship speed by the time step, and the cosine value is calculated by multiplying the normalized wind speed by the time step, forming a two-dimensional vector that reflects the periodic dynamics of the normalized data; the other is the environmental coupling factor vector, which generates another two-dimensional vector through the product of the normalized wind speed and the normalized wave height, and the product of the normalized ship speed and the cosine value of the normalized course angle, and is used to characterize the interaction between environmental factors.

[0063] The input enhancement matrix is expressed as follows:

[0064] ,

[0065] where, represents the input enhancement matrix at time , which is used to capture the time dependence and environmental interaction effects of ship operation; represents the normalized data vector at time ; represents the normalized data vector at time ; represents the pseudo-periodic perturbation vector, which reflects the pseudo-periodic perturbation based on the normalized ship speed and the normalized wind speed at time . By introducing periodic perturbations, the time dynamic characteristics of the normalized data are enhanced. The pseudo-periodic perturbation vector is expressed as: , represents the sine transformation of the normalized ship speed; represents the cosine transformation of the normalized wind speed; represents the time step, which can be specifically set according to the specific implementation scenario and is not limited here; represents the environmental coupling factor vector, which reflects the coupling effect between environmental factors at time . The environmental coupling factor vector is expressed as: , represents the product of the normalized wind speed and the normalized wave height at time , which reflects the wind-wave interaction effect, represents the product of the normalized ship speed and the cosine value of the normalized course angle at time , which reflects the influence of the course on the ship speed;

[0066] The input enhancement matrix is ​​further transformed into an intermediate feature matrix through nonlinear feature mapping to explore hidden patterns in the normalized data while avoiding the limitations of traditional activation functions. Specifically, the input enhancement matrix is ​​multiplied by its own transpose to obtain a new matrix, which emphasizes the intrinsic correlation between the data. At the same time, in order to avoid calculation instability caused by too large or too small values, normalization processing is performed. The normalization processing method is to calculate the Frobenius norm of the input enhancement matrix, that is, the square root of the sum of squares of all elements, and then the square of the Frobenius norm plus 1 is used as the denominator, and the product of the input enhancement matrix and its own transpose is divided by the denominator to obtain a scaled matrix;

[0067] In order to further incorporate environmental influences, an environmental attenuation matrix is ​​designed. The first element of the environmental attenuation matrix is ​​the result of calculating the exponential function after taking the negative value of the normalized wave height, and the second element is the result of calculating the exponential function after taking the negative value of the normalized wind speed. The non-diagonal elements are zero, reflecting the inhibitory effect of environmental factors on the characteristics. For example, the higher the waves or the greater the wind speed, the lower the operating efficiency of the ship may be.

[0068] Add the scaled matrix to the environmental attenuation matrix to get the intermediate feature matrix, which is calculated as:

[0069] ,

[0070] in, Indicates at time The intermediate feature matrix is ​​used to capture the nonlinear patterns and environmental effects of the input data; represents the transpose of the input augmentation matrix, which is the same as the input augmentation matrix Multiply to generate the matrix self-product result, which is used to extract the intrinsic relationship between data; Represents the square of the Frobenius norm, which is used as part of the normalization factor to prevent the result of the matrix multiplication from being too large and maintain numerical stability; Represents the environmental attenuation matrix, reflecting the time The attenuation effect of environmental factors, the environmental attenuation matrix is ​​expressed as:

[0071] ,

[0072] in, represents the exponential decay term of the normalized wave height; represents the exponential decay term of the normalized wind speed;

[0073] Extract the final feature vector from the intermediate feature matrix, which includes fuel efficiency and energy utilization. Specifically, by processing the inverse mapping of the intermediate feature matrix, since the intermediate feature matrix is not necessarily a square matrix or invertible, the intermediate feature matrix is inversely mapped back to the input space to obtain the pseudo-inverse matrix of the intermediate feature matrix, which is used to ensure computational stability. To further introduce the influence of the normalized course angle and enhance the directivity of the features, a dynamic projection matrix is designed. The elements of the dynamic projection matrix are calculated based on the tangent value of the normalized course angle. The elements on the main diagonal are all 1, the upper right corner element is the tangent value of the normalized course angle, and the lower left corner element is the negative value of the tangent value of the normalized course angle.

[0074] The calculation formula for the feature vector is:

[0075] ,

[0076] where, represents the feature vector at time , which includes fuel efficiency and energy utilization and serves as the initial solution, reflecting the operating efficiency of the ship. The feature vector is expressed as: , represents the fuel efficiency at time , represents the energy utilization at time ; represents 's pseudo-inverse matrix, which inversely maps back to the input space to ensure computational stability, especially when is non-invertible and still solvable; represents the dynamic projection matrix, which introduces the influence of the normalized course angle, adjusts the directivity of the feature vector, and enhances the characterization ability of the ship's operating state. The dynamic projection matrix is expressed as:

[0077] ,

[0078] where, represents the tangent function, reflecting the influence of the course angle of the intermediate feature matrix; represents the back projection, which is used for directivity adjustment;

[0079] By extracting effective features from the ship's operating data, a more accurate assessment of the ship's fuel efficiency and energy utilization can be made, providing a reliable basis for subsequent optimization decisions;

[0080] S2. Conduct perturbation search and adaptive adjustment through the dynamic adaptive perturbation optimization algorithm, and calculate the optimal solution; based on the gap between the optimal solution and the initial solution, generate a dimensionless control vector, and convert the dimensionless control vector into an actual control vector through an anti-normalization operation;

[0081] The operation of a ship needs to find the optimal balance between economy and performance. By conducting perturbation search and adaptive adjustment through the dynamic adaptive perturbation optimization algorithm, it can efficiently explore the non-linear space, avoid the limitations of traditional methods, calculate the fuel efficiency and energy utilization rate that maximize the value of the optimization objective function, and provide the target value for subsequent control;

[0082] The dynamic adaptive perturbation optimization algorithm is used to explore the global optimal solution in non-linear and non-convex optimization problems by introducing random perturbations and adaptive step sizes. It dynamically adjusts the search direction and step size using perturbation feedback and historical experience, quickly approaching the maximum value of the optimization objective function in the search space while avoiding falling into local optimal solutions. The specific implementation is as follows:

[0083] In the initialization stage, define the optimization objective function and set the initial solution; the initial solution consists of the fuel efficiency and energy utilization rate at the current moment;

[0084] The optimization objective function aims to balance the fuel efficiency and energy utilization rate of the ship, and at the same time consider time smoothness to ensure the smooth operation of the ship and prevent performance fluctuations caused by excessive adjustments. Specifically, by weighting the fuel efficiency and energy utilization rate, the optimization objective function can comprehensively consider the fuel consumption and energy utilization of the ship. Through the time smoothness penalty term, it ensures that there will be no drastic performance fluctuations during the optimization process of the ship, thereby improving the stability and reliability of the ship operation intelligent control system;

[0085] The optimization objective function is defined as follows:

[0086] ,

[0087] where represents the value of the optimization objective function, which is used to measure the quality of the current solution and serves as the evaluation criterion for the optimization process to determine whether to update the current solution; represents the weighting coefficient of fuel efficiency, which is used to control the importance of fuel efficiency in the optimization process and can be specifically set according to the specific implementation scenario and is not limited here; represents the weighting coefficient of energy utilization rate, which is used to control the importance of energy utilization rate in the optimization process and can be specifically set according to the specific implementation scenario and is not limited here; represents the time smoothness penalty term, which is used to avoid drastic changes in a short period of time, punish the initial solution with drastic changes, thereby reducing oscillations and increasing stability; denotes the smoothing coefficient, which is used to control the intensity of the smoothing penalty, avoid excessive rapid changes, and can be specifically set according to the specific implementation scenario, and is not limited here; denotes at time the fuel efficiency; denotes at time the energy utilization rate;

[0088] During the iterative search process, the dynamic adaptive perturbation optimization algorithm dynamically adjusts the search step size and direction based on the current solution and historical state. By introducing random perturbations, it ensures a wide exploration of potential optimal solutions. In each iteration, a random perturbation vector is generated to perturb the current solution to obtain a candidate solution; the random perturbation vector is generated through a standard normal distribution and is normalized after each generation to make it a unit vector (with a length of 1) to ensure the consistency of the random perturbation direction, enabling the dynamic adaptive perturbation optimization algorithm to conduct a comprehensive exploration in high-dimensional space and effectively avoid falling into local optimal solutions;

[0089] The formula for the random perturbation direction is:

[0090] ,

[0091] where, denotes the random perturbation direction of the th iteration. The candidate solution is generated through the perturbation direction to ensure the randomness of the search; denotes the random perturbation vector, which follows a two-dimensional standard normal distribution and is generated through the standard normal distribution; denotes the Euclidean norm of the random perturbation vector, which reflects the magnitude of the generated random perturbation vector;

[0092] To ensure the convergence and efficiency of the dynamic adaptive perturbation optimization algorithm, an adaptive step size mechanism is introduced. In each iteration, according to the difference between the current optimization objective function value and the historical maximum optimization objective function value, the perturbation step size is dynamically adjusted. If the current optimization objective function value is close to the historical maximum optimization objective function value, the perturbation step size will be reduced to improve the accuracy of local search. On the contrary, if the current optimization objective function value is far from the historical maximum optimization objective function value, the perturbation step size will be increased to accelerate the speed of global search; the perturbation step size is adjusted through the cosine function, thus realizing a smooth search process, which can conduct a relatively wide exploration in the initial stage and can finely adjust the solution in the later stage, thereby improving the convergence speed and accuracy of the dynamic adaptive perturbation optimization algorithm;

[0093] The calculation formula for the perturbation step size is:

[0094] ,

[0095] Among them, represents the perturbation step size of the -th iteration, which determines the size of each search step and is adaptively adjusted to improve the search accuracy; represents the perturbation step size of the -th iteration; represents the step size decay factor, which is used to adjust the change speed of the perturbation step size during each iteration and can be specifically set according to the specific implementation scenario and is not limited here; represents the function used to adjust the perturbation step size. By using the cosine function, it is ensured that when the value of the optimization objective function approaches the historical maximum optimization objective function value, the perturbation step size will gradually decrease, thereby finely searching the area near the optimal solution; represents the value of the optimization objective function at the -th iteration; represents the historical maximum optimization objective function value;

[0096] In each round of iteration, the generated perturbation step size will affect the update of the current solution. Based on the perturbation step size, a new candidate solution will be generated in the way of weighted sum, and the calculation formula is:

[0097] ,

[0098] Among them, represents the candidate solution at the -th iteration; represents the candidate solution at the -th iteration; represents the momentum weight, which is used to control the influence of historical momentum on the current perturbation and can be specifically set according to the specific implementation scenario and is not limited here; represents the momentum vector at the -th iteration, which is used to adjust the direction when updating the current solution. In each round of iteration, the momentum vector is obtained from the random perturbation direction and update history of the previous round, and the calculation formula is: ; represents the momentum vector at the -th iteration;

[0099] To ensure the physical feasibility of the candidate solution, a constraint boundary mechanism is introduced, so that each element of the candidate solution is subject to certain constraints to ensure that the optimal solution is within the effective range. The constraint boundary value is set as: ;

[0100] If the value of the current optimization objective function is greater than the value of the optimization objective function in the previous round of iteration, it means that the current solution is better, and the solution and momentum need to be updated. Otherwise, the solution and momentum remain unchanged;

[0101] The optimization process is judged whether to end by two termination conditions: the maximum number of iterations and the minimum step size. If the maximum number of iterations is reached, or the step size has been reduced to a sufficiently small level, the process is terminated and the final optimal solution is output, ensuring that the final result can be given after fully exploring the solution space, and at the same time avoiding meaningless over-iteration.

[0102] After multiple rounds of iteration, the optimal solution is finally output, that is, the fuel efficiency and energy utilization rate that maximize the value of the optimization objective function during the entire optimization process. The optimal solution is expressed as:

[0103] ,

[0104] where, represents the optimal fuel efficiency; represents the optimal energy utilization rate;

[0105] Based on the gap between the optimal solution and the initial solution, a dimensionless control vector is generated through a proportional control method, which is a well-known technical means to those skilled in the art and will not be elaborated here. The formula is as follows:

[0106] ,

[0107] where, represents the dimensionless control vector, which is used to adjust the speed and heading of the ship and is expressed as: , represents the control amount of the ship speed at time ; represents the control amount of the ship's heading angle at time ; represents the proportional control gain matrix, which is used to convert the difference between the optimal solution and the initial solution into the adjustment amount of the ship's speed and heading, and determines the adjustment strength of each control variable (speed and heading). It can be specifically set according to the specific implementation scenario and is not limited here. The expression form is: , represents the speed adjustment gain, represents the heading angle adjustment gain;

[0108] The dimensionless control vector is converted into the actual control vector through an anti-normalization operation to ensure that the ship can perform reasonable speed and heading adjustments. The anti-normalization formula is:

[0109] ,

[0110] where, represents the actual control vector, which performs anti-normalization processing on the control vector and converts it into a physical quantity, representing the actual ship speed and heading that need to be adjusted; Represents the maximum value of the control quantity; Represents the minimum value of the control quantity; and For the inverse normalization process to ensure that the obtained actual control vector is within a reasonable range in practical applications;

[0111] By adjusting the speed and heading of the ship in real time, it is possible to effectively improve the fuel efficiency and energy utilization rate of the ship, reduce energy waste, adapt to complex environmental changes, lower operating costs, and ultimately achieve the optimization of ship operation.

[0112] In summary, a ship operation efficiency intelligent control system and method based on large model optimization have been completed.

[0113] The sequence of the invention embodiments is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0114] Each embodiment in this specification is described in a progressive manner. The same or similar parts between each embodiment can be referred to each other, and the key point of each embodiment is to illustrate the differences from other embodiments.

[0115] 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 for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of each embodiment of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A ship operation efficiency intelligent control method optimized based on large models, characterized in that Including the following steps: S1. Collect the real-time data of ship operation and environment, eliminate the dimension difference through normalization processing to obtain the normalized data, form a normalized data vector, and use the multi-dimensional data feature mapping algorithm to extract features from the normalized data vector to obtain a feature vector; S2. Perform perturbation search and adaptive adjustment through the dynamic adaptive perturbation optimization algorithm to calculate the optimal solution; the dynamic adaptive perturbation optimization algorithm introduces random perturbation and adaptive step size, and dynamically adjusts the search direction and step size by using perturbation feedback and historical experience; in the initialization stage, define the optimization objective function and set the initial solution; the initial solution consists of the fuel efficiency and energy utilization rate at the current moment; In the iterative search process, in each round of iteration, a random perturbation vector is generated to perturb the current solution to obtain a candidate solution; the random perturbation vector is generated through a standard normal distribution and is normalized after each generation to make it a unit vector, that is, the random perturbation direction; Introduce an adaptive step size mechanism, and dynamically adjust the perturbation step size according to the difference between the current optimization objective function value and the historical maximum optimization objective function value. The calculation formula for the perturbation step size is: ， Among them, represents the perturbation step size of the -th iteration; represents the perturbation step size of the -th iteration; represents the step size decay factor; represents the function used to adjust the perturbation step size; represents the value of the optimization objective function at the -th iteration; represents the historical maximum value of the optimization objective function; Based on the gap between the optimal solution and the initial solution, generate a dimensionless control vector, and convert the dimensionless control vector into an actual control vector through an anti-normalization operation.

2. The intelligent control method for ship operation efficiency optimization based on large model according to claim 1, wherein, The S1 specifically includes: The multi-dimensional data feature mapping algorithm integrates the current normalized data vector and the historical normalized data vector as the basic part of the input enhancement matrix, and introduces a pseudo-periodic perturbation vector and an environmental coupling factor vector to form the input enhancement matrix.

3. The intelligent control method for ship operation efficiency optimized based on a large model according to claim 2, wherein, The S1 specifically includes: In the implementation process of the multi-dimensional data feature mapping algorithm, the input enhancement matrix is transformed into an intermediate feature matrix through non-linear feature mapping. The calculation formula is: ， Among them, represents the intermediate feature matrix at time ; represents the transpose of the input enhancement matrix; represents the transpose; represents the input enhancement matrix at time ; represents the square of the Frobenius norm, which is part of the normalization factor; represents the environmental attenuation matrix, which reflects the attenuation effect of environmental factors at time .

4. A ship operation efficiency intelligent control method optimized based on a large model according to claim 3, characterized in that, The S1 specifically includes: In the implementation process of the multi-dimensional data feature mapping algorithm, through the reverse mapping processing of the intermediate feature matrix, the intermediate feature matrix is reverse mapped back to the input space to obtain the pseudo-inverse matrix of the intermediate feature matrix, and a dynamic projection matrix is designed. The pseudo-inverse matrix of the intermediate feature matrix is combined with the dynamic projection matrix to finally obtain a feature vector, including fuel efficiency and energy utilization rate.

5. A ship operation efficiency intelligent control method optimized based on a large model according to claim 1, characterized in that, The S2 specifically includes: In the implementation process of the dynamic adaptive perturbation optimization algorithm, based on the perturbation step size, generate candidate solutions in a weighted sum manner; after multiple rounds of iteration, finally output the optimal solution, that is, the fuel efficiency and energy utilization rate that maximize the optimization objective function value during the entire optimization process.

6. A ship operation efficiency intelligent control system optimized based on a large model, which is applied to a ship operation efficiency intelligent control method optimized based on a large model as described in claim 1, and is characterized in that, Including the following parts: Real-time data acquisition and normalization processing module, feature extraction module, adaptive optimization processing module, control generation module, anti-normalization module; Real-time data acquisition and normalization processing module: Collect the real-time data of ship operation and environment, eliminate the dimension difference through normalization processing to obtain the normalized data, form a normalized data vector from the normalized data, and output the normalized data vector to the feature extraction module; Feature extraction module: Using a multi-dimensional data feature mapping algorithm to extract features from the normalized data vectors from the real-time data acquisition and normalization processing module, obtaining feature vectors that include fuel efficiency and energy utilization rate. The fuel efficiency and energy utilization rate form the initial solution, and the feature vectors are output to the adaptive optimization processing module and used as the initial solution to be output to the control generation module; Adaptive optimization processing module: Based on the feature vectors of the feature extraction module, perform perturbation search and adaptive adjustment through a dynamic adaptive perturbation optimization algorithm, calculate the fuel efficiency and energy utilization rate that maximize the value of the optimization objective function, obtain the optimal solution, and output the optimal solution to the control generation module; Control generation module: Based on the gap between the optimal solution of the adaptive optimization processing module and the initial solution of the feature extraction module, generate a dimensionless control vector through a proportional control method and output the dimensionless control vector to the inverse normalization module; Inverse normalization module: Convert the dimensionless control vector from the control generation module into an actual control vector through inverse normalization operations.

Images