Ocean weather navigation system integration method based on cloud native architecture

Abstract

The application relates to the field of ocean navigation management, and discloses an ocean weather navigation system integration method based on a cloud native architecture, which comprises the following steps: building a cloud native base, collecting and preprocessing multi-source heterogeneous data such as weather, ships, oceans and geography; adopting micro-service splitting and containerized encapsulation of core algorithms to realize elastic scheduling of computing power; building a data-model hybrid driven prediction engine to accurately predict the speed loss and sailing speed of a ship; carrying out multi-dimensional risk coupling assessment and dynamic grading of weather-geography-ship to output a risk grade; and finally, completing route-sailing speed joint optimization based on the risk grade to generate an applicable route. The application has the advantages of strong architecture elasticity, high algorithm fusion degree, accurate risk assessment and intelligent route optimization, and improves the safety and economy of ocean navigation.

Description

Technical Field

[0001] This invention relates to the field of ocean navigation management, and in particular to an integration method for ocean meteorological navigation systems based on a cloud-native architecture. Background Technology

[0002] Ocean-going meteorological navigation is a core technology for ensuring ship navigation safety and reducing operating costs. Current mainstream solutions mostly adopt traditional monolithic architectures, rigidly coupling modules such as weather forecasting, risk assessment, and route optimization, relying on localized server deployments, resulting in fixed computing resources and poor scalability. At the data level, existing technologies only access meteorological data and basic ship information from single sources, lacking unified governance and fusion of multi-source heterogeneous data (global reanalysis data, AIS dynamics, geospatial data). At the algorithm level, they mostly use single physical models or basic machine learning models, making it difficult to adapt to the needs of ship stall prediction and accurate typhoon path forecasting in complex wind and wave scenarios. Furthermore, in scenarios with uncertainty in typhoon forecasts, route optimization is mostly static planning, unable to be dynamically adjusted in real time. At the service level, ship-shore data transmission suffers from high latency and weak collaboration, making it difficult to achieve integrated services throughout the entire process—pre-voyage, during-voyage, and post-voyage. At the same time, long algorithm iteration cycles and the susceptibility to failures leading to system-wide paralysis severely restrict the digital and intelligent development of ocean shipping.

[0003] Existing ocean-going meteorological navigation technologies suffer from three major pain points: First, insufficient architectural adaptability. The single-unit architecture has poor computing power elasticity and high module coupling, making it unable to support the processing of massive multi-source data and the parallel computing of complex algorithms, and difficult to cope with the high-concurrency computing requirements of extreme weather scenarios such as typhoons and severe convection. Second, low algorithm integration. The physical model has poor generalization ability, and the data-driven model depends on the sample size. The lack of an effective integration mechanism between the two leads to insufficient accuracy in ship stall prediction and typhoon path forecast, and risk assessment has a single dimension and inaccurate classification. Third, weak collaborative response capability. Ship-shore data is not synchronized, and route optimization is static, which cannot adapt to the dynamic risks brought about by the uncertainty of typhoon forecasts. At the same time, the system operation and maintenance is cumbersome and the iteration efficiency is low, making it difficult to quickly adapt to IMO regulatory updates and the personalized needs of shipping companies.

[0004] Therefore, this patent proposes an integration method for ocean-going meteorological navigation systems based on cloud-native architecture to solve the above problems. Summary of the Invention

[0005] This invention overcomes the shortcomings of existing technologies and provides an integration method for ocean-going meteorological navigation systems based on cloud-native architecture.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides an integration method for a cloud-native architecture-based oceanographic navigation system, comprising the following steps: Based on cloud-native algorithms, a cloud-native foundation architecture is built to collect and preprocess the associated core data of ocean meteorological navigation, and obtain target multi-source heterogeneous data. By introducing microservice decomposition and containerizing the core meteorological navigation algorithm, the configuration processing of target multi-source heterogeneous data is carried out. By combining standardized container images with target multi-source heterogeneous data, a data-model hybrid-driven prediction engine is constructed. Based on the output of the data-model hybrid driving prediction engine, a multi-dimensional risk coupling assessment and dynamic classification of meteorology, geography and ships is carried out. In the application layer of the cloud-native data management platform, the route-speed joint optimization of the target vessel is performed based on the risk level of each candidate indicator data to obtain the applicable route for the target vessel.

[0007] Furthermore, in a preferred embodiment of the present invention, the step of building a cloud-native infrastructure based on cloud-native algorithms to collect and preprocess the associated core data of ocean-going meteorological navigation to obtain target multi-source heterogeneous data specifically includes: The central control unit used for the integration of the ocean-going meteorological navigation system is obtained, and it is designated as the target central control unit. Within the target central control unit, a Kubernetes network is introduced to build a cloud-native data management foundation. At the same time, the ship where the target central control unit is located is designated as the target ship. The cloud-native data management platform is divided into four independent architectures: data layer, algorithm layer, service layer and application layer. The independent architectures of different layers are isolated and have communication control. In the data layer of the cloud-native data management platform, multi-source heterogeneous data is accessed. The source of the access to the multi-source heterogeneous data is the cloud data network. The multi-source heterogeneous data includes global meteorological data, ship dynamic data, marine environmental data, and geospatial data. Within the data layer of the cloud-native data management platform, standardized cleaning processing is performed on the accessed multi-source heterogeneous data. The standardized cleaning processing involves using the 3σ outlier detection method to remove outliers from the accessed multi-source heterogeneous data, using a time series smoothing algorithm to eliminate time series fluctuations in the accessed multi-source heterogeneous data, and finally performing time series alignment on the accessed multi-source heterogeneous data to obtain preliminary pre-processed multi-source heterogeneous data. By introducing a unified description rule based on ontology concepts, the data source differences of the preliminary preprocessed multi-source heterogeneous data are eliminated, and a unified data field mapping rule for the preliminary preprocessing of multi-source heterogeneous data is established to obtain the target multi-source heterogeneous data.

[0008] Furthermore, in a preferred embodiment of the present invention, the step of configuring and processing target multi-source heterogeneous data by introducing microservice decomposition and containerized encapsulation of the core meteorological navigation algorithm specifically includes: Within the target control unit, the associated core algorithm for ocean meteorological navigation is retrieved through the cloud data network and identified as the core algorithm for ocean meteorological navigation to be analyzed. In the data layer of the cloud-native data management platform, combined with the target multi-source heterogeneous data, and based on the functional decomposition principle of high cohesion and low coupling, all the core algorithms of ocean meteorological navigation to be analyzed are decomposed into independent microservice modules. Each independent microservice module corresponds to a core ocean meteorological navigation algorithm to be analyzed. Each independent microservice module is configured with an independent operating environment. The method for configuring the independent operating environment is to retrieve the historical operating environments of different core ocean meteorological navigation algorithms to be analyzed from the cloud data network and output them as the independent operating environment of the independent microservice module. After configuring and processing the independent operating environment of each independent microservice module, standardized container images are generated. In the cloud-native data management platform, all standardized container images are orchestrated and managed. The orchestration and management rules are as follows: calculate the load requirements of each independent microservice module, sort the standardized container images in reverse order according to the load, and then dynamically orchestrate and manage them. In the cloud-native data management platform, communication links are built for all independent microservice modules.

[0009] Furthermore, in a preferred embodiment of the present invention, the step of combining standardized container images and target multi-source heterogeneous data to construct a data-model hybrid-driven prediction engine specifically includes: Within the algorithm layer of the cloud-native data management platform, a standardized container image corresponding to the core ocean meteorological navigation algorithm to be analyzed, used for stall prediction and speed prediction of target ships, is selected and loaded into the corresponding independent microservice module. For the target multi-source heterogeneous marine environmental data, wind speed, significant wave height and wave period correlation data are clustered into wind and wave scenarios and divided into general wind and wave scenarios and large wind and wave scenarios. The scenarios corresponding to different wind speed, significant wave height and wave period correlation data are retrieved from the cloud data network. For general wind and wave scenarios, the standardized container image corresponding to the core algorithm of the ocean meteorological navigation to be analyzed for predicting the stall of the target ship is called, and a serial hybrid drive mode is adopted to build the ship stall prediction sub-engine. For high wind and wave scenarios, the standardized container image corresponding to the core ocean meteorological navigation algorithm to be analyzed for predicting the target ship's speed is invoked, and a parallel hybrid drive mode is adopted to build the ship speed prediction sub-engine. By combining the ship stall prediction sub-engine and the ship speed prediction sub-engine, a data-model hybrid driven prediction engine is constructed to output the stall and speed prediction results of the target ship.

[0010] Furthermore, in a preferred embodiment of the present invention, the step of performing meteorological-geographical-ship multidimensional risk coupling assessment and dynamic classification based on the output results of the data-model hybrid-driven prediction engine specifically involves: In the service layer of the cloud-native data management platform, the stall and speed prediction results of the target ship output by the data-model hybrid driving prediction engine are retrieved and associated with meteorological elements, geospatial information and target ship parameters in the multi-source heterogeneous data of the target. Retrieve the knowledge graph of ocean-going vessel navigation risk assessment in the cloud data network. The knowledge graph of ocean-going vessel navigation risk assessment records the risk threshold of the indicator data related to the navigation risk of the target multi-source heterogeneous data, and marks the indicator data related to the navigation risk of the target multi-source heterogeneous data as candidate indicator data. In the service layer of the cloud-native data management platform, the risk matrix method is introduced. The risk threshold of the candidate indicator data is combined to perform the rating process of the candidate indicator data to obtain the risk output probability level and the risk output impact level. The risk output probability level and risk output impact level of all candidate indicator data are coupled and matched to determine the risk level of each candidate indicator data, and then output to the application layer of the cloud-native data management platform.

[0011] Furthermore, in a preferred embodiment of the present invention, the step of performing joint route-speed optimization of the target vessel based on the risk level of each candidate indicator data in the application layer of the cloud-native data management platform to obtain a suitable route for the target vessel specifically involves: In the application layer of the cloud-native data management platform, a multi-objective route-speed joint optimization model is constructed by combining the risk level of each candidate indicator data and the geospatial data in the target multi-source heterogeneous data. The construction method is as follows: the training target is that the sailing time, fuel consumption and sailing risk are all less than the corresponding preset range. The geospatial constraints and ship maneuvering restrictions associated with the target ship are retrieved in the cloud data network. At the same time, the LSTM model is used as the model architecture for model training. A two-layer mapping swarm intelligent optimization algorithm is introduced to solve the optimization model, generate alternative routes and corresponding hourly speed schemes for the target vessel, and conduct a reasonableness and compliance check on all alternative routes for the target vessel to eliminate invalid routes. The alternative routes and corresponding hourly speed plans for the remaining target vessels are analyzed, and typhoon avoidance planning is carried out in the application layer of the cloud-native data management platform.

[0012] Furthermore, in a preferred embodiment of the present invention, the analysis of the remaining alternative routes and corresponding hourly speed plans for the target vessels, and the typhoon avoidance planning process performed in the application layer of the cloud-native data management platform, specifically involves: Analyze the data related to typhoons in the multi-source heterogeneous data of the target, identify the typhoon impact range, movement path and intensity change trend when the target ship is sailing, and label them as typhoon impact factors; Typhoon impact factors are imported into the cloud data network for historical data retrieval. The probability distribution map of the trajectory of typhoons when the similarity with the typhoon impact factors in history is greater than a preset value is retrieved and used to delineate typhoon danger zones. Determine whether the remaining alternative routes for the target vessel do not pass through typhoon-prone areas. If so, the corresponding alternative routes shall be adopted as the applicable routes for the target vessel. If not, the standardized container image of the independent microservice module associated with typhoon avoidance is called, and combined with sailing time, fuel consumption and sailing risk, the LSTM model is used as the model architecture to continue to build the typhoon scenario optimization model through the two-layer mapping swarm intelligent optimization algorithm, and iteratively generate new alternative routes for the target ship. The new alternative routes for the target vessel do not pass through typhoon-prone areas and are applicable to the target vessel. The applicable routes for the target vessel are output in the application layer of the cloud-native data management platform.

[0013] This invention addresses the technical deficiencies in the background technology and offers the following advantages: It establishes a cloud-native infrastructure to collect and preprocess heterogeneous data from multiple sources, including meteorology, shipping, oceanography, and geography; it employs microservice decomposition and containerized encapsulation of core algorithms to achieve elastic scheduling of computing power; it constructs a data-model hybrid-driven prediction engine to accurately predict ship stall and speed; it conducts multi-dimensional risk coupling assessment and dynamic classification of meteorology, geography, and shipping, outputting risk levels; and finally, based on the risk levels, it performs joint optimization of routes and speeds to generate applicable routes. This invention possesses advantages such as strong architectural flexibility, high algorithm integration, accurate risk assessment, and intelligent route optimization, thereby improving the safety and economy of ocean navigation. Attached Figure Description

[0014] Figure 1 A flowchart illustrating the integration method of a cloud-native architecture-based oceanographic navigation system is shown.

[0015] Figure 2 A flowchart illustrating the method for outputting the applicable routes for the target vessel is shown. Detailed Implementation

[0016] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0017] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0018] Figure 1 A flowchart illustrating the integration method for a cloud-native architecture-based oceanographic navigation system is shown, including the following steps: S102: Based on cloud-native algorithms, a cloud-native foundation architecture is built to collect and preprocess the associated core data of ocean meteorological navigation to obtain target multi-source heterogeneous data; S104: By introducing microservice decomposition and containerizing the core meteorological navigation algorithm, configuration processing of target multi-source heterogeneous data is carried out; S106: Combining standardized container images with target multi-source heterogeneous data to build a data-model hybrid-driven prediction engine; S108: Based on the output of the data-model hybrid driving prediction engine, conduct multi-dimensional risk coupling assessment and dynamic classification of meteorology, geography and ships; S110: In the application layer of the cloud-native data management platform, the route-speed joint optimization of the target ship is performed based on the risk level of each candidate indicator data to obtain the applicable route for the target ship.

[0019] Furthermore, in a preferred embodiment of the present invention, the step of building a cloud-native infrastructure based on cloud-native algorithms to collect and preprocess the associated core data of ocean-going meteorological navigation to obtain target multi-source heterogeneous data specifically includes: The central control unit used for the integration of the ocean-going meteorological navigation system is obtained, and it is designated as the target central control unit. Within the target central control unit, a Kubernetes network is introduced to build a cloud-native data management foundation. At the same time, the ship where the target central control unit is located is designated as the target ship. The cloud-native data management platform is divided into four independent architectures: data layer, algorithm layer, service layer and application layer. The independent architectures of different layers are isolated and have communication control. In the data layer of the cloud-native data management platform, multi-source heterogeneous data is accessed. The source of the access to the multi-source heterogeneous data is the cloud data network. The multi-source heterogeneous data includes global meteorological data, ship dynamic data, marine environmental data, and geospatial data. Within the data layer of the cloud-native data management platform, standardized cleaning processing is performed on the accessed multi-source heterogeneous data. The standardized cleaning processing involves using the 3σ outlier detection method to remove outliers from the accessed multi-source heterogeneous data, using a time series smoothing algorithm to eliminate time series fluctuations in the accessed multi-source heterogeneous data, and finally performing time series alignment on the accessed multi-source heterogeneous data to obtain preliminary pre-processed multi-source heterogeneous data. By introducing a unified description rule based on ontology concepts, the data source differences of the preliminary preprocessed multi-source heterogeneous data are eliminated, and a unified data field mapping rule for the preliminary preprocessing of multi-source heterogeneous data is established to obtain the target multi-source heterogeneous data.

[0020] It's important to note that establishing a ship-cloud integrated basic control and computing environment on the ship involves first building a cloud-native foundation. This foundation's architecture comprises a data layer, algorithm layer, service layer, and application layer, using a Kubernetes network as the basic network architecture. This provides advantages such as high availability, strong elasticity, scalability, support for cross-ship / shore collaboration, high stability, and simple maintenance. Data from four major categories—meteorology, shipping, oceanography, and geography—is uniformly accessed from the cloud, covering all elements of ocean navigation. Compared to traditional navigation systems that rely solely on single meteorological or simple shipping data, this solution achieves full-element modeling, significantly improving prediction accuracy and the comprehensiveness of risk assessment. Subsequent preprocessing of the multi-source heterogeneous data includes 3σ outlier detection to remove jumps, errors, and extreme anomalies; time-series smoothing to remove random noise and fluctuations; and time-series alignment to ensure consistency in the time dimension of the multi-source data. Finally, ontology concepts are used to uniformly describe and eliminate semantic differences between data from different sources, establishing a unified field mapping to output high-quality, standardized, and directly modelable target multi-source heterogeneous data.

[0021] Furthermore, in a preferred embodiment of the present invention, the step of configuring and processing target multi-source heterogeneous data by introducing microservice decomposition and containerized encapsulation of the core meteorological navigation algorithm specifically includes: Within the target control unit, the associated core algorithm for ocean meteorological navigation is retrieved through the cloud data network and identified as the core algorithm for ocean meteorological navigation to be analyzed. In the data layer of the cloud-native data management platform, combined with the target multi-source heterogeneous data, and based on the functional decomposition principle of high cohesion and low coupling, all the core algorithms of ocean meteorological navigation to be analyzed are decomposed into independent microservice modules. Each independent microservice module corresponds to a core ocean meteorological navigation algorithm to be analyzed. Each independent microservice module is configured with an independent operating environment. The method for configuring the independent operating environment is to retrieve the historical operating environments of different core ocean meteorological navigation algorithms to be analyzed from the cloud data network and output them as the independent operating environment of the independent microservice module. After configuring and processing the independent operating environment of each independent microservice module, standardized container images are generated. In the cloud-native data management platform, all standardized container images are orchestrated and managed. The orchestration and management rules are as follows: calculate the load requirements of each independent microservice module, sort the standardized container images in reverse order according to the load, and then dynamically orchestrate and manage them. In the cloud-native data management platform, communication links are built for all independent microservice modules.

[0022] It should be noted that all core algorithms required for ocean-going meteorological navigation are retrieved uniformly from the cloud, including but not limited to algorithms for objective forecasting of high-impact factors for ships at sea globally, algorithms for predicting ship stall, algorithms for identifying meteorological risks on typical climate routes for ocean-going vessels, algorithms for dynamic risk assessment of navigation under complex weather and sea conditions, algorithms for joint optimization of routes and speeds, algorithms for planning typhoon avoidance routes for ships in uncertain typhoon forecast scenarios, and algorithms for assessing the seaworthiness of meteorological routes by integrating multi-source geospatial data. The aim is to uniformly collect algorithm resources across the entire chain, avoid algorithm fragmentation and omissions, provide a complete algorithm list for subsequent systematic decomposition and encapsulation, and ensure full algorithm coverage throughout the entire navigation process. Based on the preprocessed target multi-source heterogeneous data, and adhering to the principles of single-function independence and weak inter-module correlation, each core algorithm is broken down into independent microservices. One algorithm corresponds to one module, each module has independent functions, focuses on the performance of one algorithm, and each module has its own dedicated runtime environment. After the environment is configured, the algorithm code, dependencies, and configurations are packaged into a standardized container image. This standardized container image can be combined with load priority scheduling to solve the problems of computing power elasticity, load balancing, and high availability, adapt to the complex operating environment of ocean-going vessels, and realize the clustering and intelligent management of microservices.

[0023] Furthermore, in a preferred embodiment of the present invention, the step of combining standardized container images and target multi-source heterogeneous data to construct a data-model hybrid-driven prediction engine specifically includes: Within the algorithm layer of the cloud-native data management platform, a standardized container image corresponding to the core ocean meteorological navigation algorithm to be analyzed, used for stall prediction and speed prediction of target ships, is selected and loaded into the corresponding independent microservice module. For the target multi-source heterogeneous marine environmental data, wind speed, significant wave height and wave period correlation data are clustered into wind and wave scenarios and divided into general wind and wave scenarios and large wind and wave scenarios. The scenarios corresponding to different wind speed, significant wave height and wave period correlation data are retrieved from the cloud data network. For general wind and wave scenarios, the standardized container image corresponding to the core algorithm of the ocean meteorological navigation to be analyzed for predicting the stall of the target ship is called, and a serial hybrid drive mode is adopted to build the ship stall prediction sub-engine. For high wind and wave scenarios, the standardized container image corresponding to the core ocean meteorological navigation algorithm to be analyzed for predicting the target ship's speed is invoked, and a parallel hybrid drive mode is adopted to build the ship speed prediction sub-engine. By combining the ship stall prediction sub-engine and the ship speed prediction sub-engine, a data-model hybrid driven prediction engine is constructed to output the stall and speed prediction results of the target ship.

[0024] It should be noted that, from the standardized image library at the algorithm layer, dedicated images for ship stall and speed prediction are selected and loaded into the corresponding independent microservices to complete the containerized deployment of the prediction algorithm, providing basic services for subsequent scenario calls. Core parameters such as wind speed, significant wave height, and wave period are extracted from marine environmental data. Two scenarios are divided using clustering algorithms. Due to the significant environmental impact of wind and waves, in general wind and wave scenarios, the stall prediction image is called in a serial mode: the physical model output is used as the data model input, and the data model is fused and fitted to construct a sub-engine. This serial fusion provides constraints to the physical model and optimizes the accuracy of the data model, balancing interpretability and accuracy. In contrast, in high wind and wave scenarios, the speed prediction image is called in a parallel mode: the physical model outputs a baseline value, the data model compensates for errors, and the data model is fused in parallel to construct a sub-engine. This method is suitable for scenarios with scarce samples and complex operating conditions, solving the problem of model failure under extreme wind and waves and ensuring the stability of speed prediction. Finally, the two types of sub-engines are merged to form a unified prediction engine, outputting the core results of ship stall and speed values ​​to support subsequent risk assessment and route optimization.

[0025] Furthermore, in a preferred embodiment of the present invention, the step of performing meteorological-geographical-ship multidimensional risk coupling assessment and dynamic classification based on the output results of the data-model hybrid-driven prediction engine specifically involves: In the service layer of the cloud-native data management platform, the stall and speed prediction results of the target ship output by the data-model hybrid driving prediction engine are retrieved and associated with meteorological elements, geospatial information and target ship parameters in the multi-source heterogeneous data of the target. Retrieve the knowledge graph of ocean-going vessel navigation risk assessment in the cloud data network. The knowledge graph of ocean-going vessel navigation risk assessment records the risk threshold of the indicator data related to the navigation risk of the target multi-source heterogeneous data, and marks the indicator data related to the navigation risk of the target multi-source heterogeneous data as candidate indicator data. In the service layer of the cloud-native data management platform, the risk matrix method is introduced. The risk threshold of the candidate indicator data is combined to perform the rating process of the candidate indicator data to obtain the risk output probability level and the risk output impact level. The risk output probability level and risk output impact level of all candidate indicator data are coupled and matched to determine the risk level of each candidate indicator data, and then output to the application layer of the cloud-native data management platform.

[0026] It should be noted that within the service layer of the four-layer architecture, predicted values ​​of the ship's actual navigation status are obtained. Simultaneously, pre-processed meteorological and environmental data, marine geographic data, and the ship's own performance parameters are integrated to form a complete navigation assessment data source for subsequent database construction. A mature professional knowledge map of ship navigation risks is retrieved from the cloud, and pre-defined critical values ​​for various navigation risks are extracted. Assessment indicators strongly correlated with navigation safety are selected from massive amounts of heterogeneous data to empower data-driven risk assessment. Risk thresholds are solidified based on massive amounts of historical navigation data, replacing the traditional, crude method of manually determining thresholds, resulting in more professional and unified assessment standards. Using a risk matrix assessment method, and comparing against predetermined risk thresholds, the probability of a navigation hazard occurring for each assessment indicator is determined, as well as the degree of harm to the ship's navigation if the hazard occurs. Two independent rating results are formed, including two core dimensions: probability of occurrence and severity of harm. The probability and impact levels of all assessment indicators are integrated and cross-coupled to obtain the final navigation risk level for each indicator, which is then uniformly summarized and transmitted to the system application layer. The innovation lies in the fact that it achieves comprehensive analysis of risk indicators across three dimensions: meteorology, geography, and shipping. This solves the technical pain points of traditional navigation risk assessment, which is fragmented and unable to be comprehensively coordinated. Its advantages are that the risk classification results are intuitive and consistent, and the data flow is smooth.

[0027] Figure 2 The flowchart of the method for outputting the applicable route for the target vessel is shown, including the following steps: S202: In the application layer of the cloud-native data management platform, the route-speed joint optimization of the target ship is carried out based on the risk level of each candidate indicator data to obtain the applicable route for the target ship; S204: Analyze the alternative routes and corresponding hourly speed plans for the remaining target vessels, and perform typhoon avoidance planning for the routes in the application layer of the cloud-native data management platform.

[0028] Furthermore, in a preferred embodiment of the present invention, the step of performing joint route-speed optimization of the target vessel based on the risk level of each candidate indicator data in the application layer of the cloud-native data management platform to obtain a suitable route for the target vessel specifically involves: In the application layer of the cloud-native data management platform, a multi-objective route-speed joint optimization model is constructed by combining the risk level of each candidate indicator data and the geospatial data in the target multi-source heterogeneous data. The construction method is as follows: the training target is that the sailing time, fuel consumption and sailing risk are all less than the corresponding preset range. The geospatial constraints and ship maneuvering restrictions associated with the target ship are retrieved in the cloud data network. At the same time, the LSTM model is used as the model architecture for model training. A two-layer mapping swarm intelligent optimization algorithm is introduced to solve the optimization model, generate alternative routes and corresponding hourly speed schemes for the target vessel, and conduct a reasonableness and compliance check on all alternative routes for the target vessel to eliminate invalid routes. The alternative routes and corresponding hourly speed plans for the remaining target vessels are analyzed, and typhoon avoidance planning is carried out in the application layer of the cloud-native data management platform.

[0029] It should be noted that within the application layer of the four-layer architecture, the results of various navigation risk levels assessed earlier are integrated, along with preprocessed geospatial data such as sea area topography, waterway distribution, and navigable areas, to build the main body of an optimization model for simultaneous route and speed planning—that is, a multi-objective route-speed joint optimization model. This model can incorporate quantified risk levels as core constraints into the modeling process, achieving navigation plan planning under the premise of safety priority. First, the three core training objectives are set as shortening navigation time, reducing fuel consumption, and strictly controlling navigation risks. Hard constraint rules such as waterway restricted areas, water depth restrictions, ship turning radius, and maximum speed are retrieved from the cloud, and a long short-term memory network is used to build the model framework for training. Among them, the LSTM model can solve the problem that traditional static planning cannot adapt to dynamic sea state changes. Its advantages are that the objectives are balanced and reasonable, the constraints are complete, the model has time-series prediction capabilities, and the accuracy of long-distance planning is higher. Subsequently, a two-layer mapping swarm intelligent optimization algorithm is used to solve the trained model, generating multiple navigation routes and matching time-segment speed schemes in batches. Each route is then verified to ensure compliance with navigation rules, water depth requirements, and air traffic control requirements, eliminating invalid routes that are illegal or impassable. The use of a two-layer mapping swarm intelligent algorithm instead of a traditional single pathfinding algorithm provides stronger global optimization capabilities, enabling simultaneous collaborative solution of route paths and segmented speeds.

[0030] Furthermore, in a preferred embodiment of the present invention, the analysis of the remaining alternative routes and corresponding hourly speed plans for the target vessels, and the typhoon avoidance planning process performed in the application layer of the cloud-native data management platform, specifically involves: Analyze the data related to typhoons in the multi-source heterogeneous data of the target, identify the typhoon impact range, movement path and intensity change trend when the target ship is sailing, and label them as typhoon impact factors; Typhoon impact factors are imported into the cloud data network for historical data retrieval. The probability distribution map of the trajectory of typhoons when the similarity with the typhoon impact factors in history is greater than a preset value is retrieved and used to delineate typhoon danger zones. Determine whether the remaining alternative routes for the target vessel do not pass through typhoon-prone areas. If so, the corresponding alternative routes shall be adopted as the applicable routes for the target vessel. If not, the standardized container image of the independent microservice module associated with typhoon avoidance is called, and combined with sailing time, fuel consumption and sailing risk, the LSTM model is used as the model architecture to continue to build the typhoon scenario optimization model through the two-layer mapping swarm intelligent optimization algorithm, and iteratively generate new alternative routes for the target ship. The new alternative routes for the target vessel do not pass through typhoon-prone areas and are applicable to the target vessel. The applicable routes for the target vessel are output in the application layer of the cloud-native data management platform.

[0031] It should be noted that by retrieving typhoon-related information from pre-processed multi-source heterogeneous data, the coverage, path, and intensity evolution patterns of typhoons are accurately assessed, and these core characteristics are uniformly defined as typhoon impact factors. The calibrated typhoon impact factors are uploaded to a cloud database, and historical typhoon cases meeting the similarity criteria are compared and selected. Corresponding typhoon probability distribution maps are retrieved, and based on this, typhoon-prone sea areas where ships cannot pass are accurately delineated. The massive historical typhoon big data can be used to extrapolate the probable path of future typhoons, overcoming the limitations of limited range and high uncertainty in single real-time forecasts. Each compliant alternative route is compared with the previously screened and retained routes, prioritizing routes that do not cross typhoon-prone areas throughout their journey. These routes are directly determined as the final usable navigation plans, prioritizing the original optimal routes, eliminating the need for repetitive calculations, simplifying the decision-making process, and improving route output efficiency. If all existing routes pass through typhoon-prone areas, the containerized typhoon-avoidance microservice image (previously described) is retrieved. Using the established LSTM model architecture and two-layer mapping swarm intelligent optimization algorithm, and still with flight time, fuel consumption, and navigation risk as constraints, a typhoon-specific optimization model is built to iteratively generate new detour route solutions. Finally, ensuring that the iteratively generated new route completely avoids typhoon-prone areas, it is designated as the officially applicable flight route. The final result is output at the application layer of the cloud-native four-layer architecture, completing the entire navigation decision-making process.

[0032] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for integrating a cloud-native architecture-based oceanographic navigation system, characterized in that: Includes the following steps: Based on cloud-native algorithms, a cloud-native foundation architecture is built to collect and preprocess the associated core data of ocean meteorological navigation, and obtain target multi-source heterogeneous data. By introducing microservice decomposition and containerizing the core meteorological navigation algorithm, the configuration processing of target multi-source heterogeneous data is carried out. By combining standardized container images with target multi-source heterogeneous data, a data-model hybrid-driven prediction engine is constructed. Based on the output of the data-model hybrid driving prediction engine, a multi-dimensional risk coupling assessment and dynamic classification of meteorology, geography and ships is carried out. In the application layer of the cloud-native data management platform, the route-speed joint optimization of the target vessel is performed based on the risk level of each candidate indicator data to obtain the applicable route for the target vessel.

2. The method for integrating a cloud-native architecture-based oceanographic navigation system according to claim 1, characterized in that, The cloud-native algorithm-based cloud-native infrastructure is used to collect and preprocess the core data associated with ocean-going meteorological navigation, resulting in multi-source heterogeneous target data. Specifically: The central control unit used for the integration of the ocean-going meteorological navigation system is obtained, and it is designated as the target central control unit. Within the target central control unit, a Kubernetes network is introduced to build a cloud-native data management foundation. At the same time, the ship where the target central control unit is located is designated as the target ship. The cloud-native data management platform is divided into four independent architectures: data layer, algorithm layer, service layer and application layer. The independent architectures of different layers are isolated and have communication control. In the data layer of the cloud-native data management platform, multi-source heterogeneous data is accessed. The source of the access to the multi-source heterogeneous data is the cloud data network. The multi-source heterogeneous data includes global meteorological data, ship dynamic data, marine environmental data, and geospatial data. Within the data layer of the cloud-native data management platform, standardized cleaning processing is performed on the accessed multi-source heterogeneous data. The standardized cleaning processing involves using the 3σ outlier detection method to remove outliers from the accessed multi-source heterogeneous data, using a time series smoothing algorithm to eliminate time series fluctuations in the accessed multi-source heterogeneous data, and finally performing time series alignment on the accessed multi-source heterogeneous data to obtain preliminary pre-processed multi-source heterogeneous data. By introducing a unified description rule based on ontology concepts, the data source differences of the preliminary preprocessed multi-source heterogeneous data are eliminated, and a unified data field mapping rule for the preliminary preprocessing of multi-source heterogeneous data is established to obtain the target multi-source heterogeneous data.

3. The method for integrating a cloud-native architecture-based oceanographic navigation system according to claim 1, characterized in that, The process of configuring and processing target multi-source heterogeneous data involves introducing microservice decomposition and containerizing the core meteorological navigation algorithm. Within the target control unit, the associated core algorithm for ocean meteorological navigation is retrieved through the cloud data network and identified as the core algorithm for ocean meteorological navigation to be analyzed. In the data layer of the cloud-native data management platform, combined with the target multi-source heterogeneous data, and based on the functional decomposition principle of high cohesion and low coupling, all the core algorithms of ocean meteorological navigation to be analyzed are decomposed into independent microservice modules. Each independent microservice module corresponds to a core ocean meteorological navigation algorithm to be analyzed. Each independent microservice module is configured with an independent operating environment. The method for configuring the independent operating environment is to retrieve the historical operating environments of different core ocean meteorological navigation algorithms to be analyzed from the cloud data network and output them as the independent operating environment of the independent microservice module. After configuring and processing the independent operating environment of each independent microservice module, standardized container images are generated. In the cloud-native data management platform, all standardized container images are orchestrated and managed. The orchestration and management rules are as follows: calculate the load requirements of each independent microservice module, sort the standardized container images in reverse order according to the load, and then dynamically orchestrate and manage them. In the cloud-native data management platform, communication links are built for all independent microservice modules.

4. The method for integrating a cloud-native architecture-based oceanographic navigation system according to claim 1, characterized in that, The method of combining standardized container images and target multi-source heterogeneous data to construct a data-model hybrid-driven prediction engine is as follows: Within the algorithm layer of the cloud-native data management platform, a standardized container image corresponding to the core ocean meteorological navigation algorithm to be analyzed, used for stall prediction and speed prediction of target ships, is selected and loaded into the corresponding independent microservice module. For the target multi-source heterogeneous marine environmental data, wind speed, significant wave height and wave period correlation data are clustered into wind and wave scenarios and divided into general wind and wave scenarios and large wind and wave scenarios. The scenarios corresponding to different wind speed, significant wave height and wave period correlation data are retrieved from the cloud data network. For general wind and wave scenarios, the standardized container image corresponding to the core algorithm of the ocean meteorological navigation to be analyzed for predicting the stall of the target ship is called, and a serial hybrid drive mode is adopted to build the ship stall prediction sub-engine. For high wind and wave scenarios, the standardized container image corresponding to the core ocean meteorological navigation algorithm to be analyzed for predicting the target ship's speed is invoked, and a parallel hybrid drive mode is adopted to build the ship speed prediction sub-engine. By combining the ship stall prediction sub-engine and the ship speed prediction sub-engine, a data-model hybrid driven prediction engine is constructed to output the stall and speed prediction results of the target ship.

5. The method for integrating a cloud-native architecture-based oceanographic navigation system according to claim 1, characterized in that, The output of the data-model hybrid driving prediction engine is used to perform multi-dimensional risk coupling assessment and dynamic classification of meteorology, geography, and shipping, specifically as follows: In the service layer of the cloud-native data management platform, the stall and speed prediction results of the target ship output by the data-model hybrid driving prediction engine are retrieved and associated with meteorological elements, geospatial information and target ship parameters in the multi-source heterogeneous data of the target. Retrieve the knowledge graph of ocean-going vessel navigation risk assessment in the cloud data network. The knowledge graph of ocean-going vessel navigation risk assessment records the risk threshold of the indicator data related to the navigation risk of the target multi-source heterogeneous data, and marks the indicator data related to the navigation risk of the target multi-source heterogeneous data as candidate indicator data. In the service layer of the cloud-native data management platform, the risk matrix method is introduced. The risk threshold of the candidate indicator data is combined to perform the rating process of the candidate indicator data to obtain the risk output probability level and the risk output impact level. The risk output probability level and risk output impact level of all candidate indicator data are coupled and matched to determine the risk level of each candidate indicator data, and then output to the application layer of the cloud-native data management platform.

6. The method for integrating a cloud-native architecture-based oceanographic navigation system according to claim 1, characterized in that, In the application layer of the cloud-native data management platform, the route-speed joint optimization of the target vessel is performed based on the risk level of each candidate indicator data to obtain the applicable route for the target vessel, specifically as follows: In the application layer of the cloud-native data management platform, a multi-objective route-speed joint optimization model is constructed by combining the risk level of each candidate indicator data and the geospatial data in the target multi-source heterogeneous data. The construction method is as follows: the training target is that the sailing time, fuel consumption and sailing risk are all less than the corresponding preset range. The geospatial constraints and ship maneuvering restrictions associated with the target ship are retrieved in the cloud data network. At the same time, the LSTM model is used as the model architecture for model training. A two-layer mapping swarm intelligent optimization algorithm is introduced to solve the optimization model, generate alternative routes and corresponding hourly speed schemes for the target vessel, and conduct a reasonableness and compliance check on all alternative routes for the target vessel to eliminate invalid routes. The alternative routes and corresponding hourly speed plans for the remaining target vessels are analyzed, and typhoon avoidance planning is carried out in the application layer of the cloud-native data management platform.

7. The method for integrating a cloud-native architecture-based oceanographic navigation system according to claim 6, characterized in that, The analysis of the remaining alternative routes and corresponding hourly speed plans for the target vessels, and the typhoon avoidance planning of the routes within the application layer of the cloud-native data management platform, specifically involves: Analyze the data related to typhoons in the multi-source heterogeneous data of the target, identify the typhoon impact range, movement path and intensity change trend when the target ship is sailing, and label them as typhoon impact factors; Typhoon impact factors are imported into the cloud data network for historical data retrieval. The probability distribution map of the trajectory of typhoons when the similarity with the typhoon impact factors in history is greater than a preset value is retrieved and used to delineate typhoon danger zones. Determine whether the remaining alternative routes for the target vessel do not pass through typhoon-prone areas. If so, the corresponding alternative routes shall be adopted as the applicable routes for the target vessel. If not, the standardized container image of the independent microservice module associated with typhoon avoidance is called, and combined with sailing time, fuel consumption and sailing risk, the LSTM model is used as the model architecture to continue to build the typhoon scenario optimization model through the two-layer mapping swarm intelligent optimization algorithm, and iteratively generate new alternative routes for the target ship. The new alternative routes for the target vessel do not pass through typhoon-prone areas and are applicable to the target vessel. The applicable routes for the target vessel are output in the application layer of the cloud-native data management platform.

Images