A standardized water conservancy professional model packaging and operation method based on Docker container technology

Abstract

The present invention discloses a standardized water conservancy professional model packaging and operation method based on Docker container technology, which belongs to the field of water conservancy projects. The specific steps of the packaging and operation method are as follows: Q1. Collecting each group of data of the target water conservancy model, and constructing a visual model knowledge graph, analyzing key computing modules and data dependencies; the present invention can enhance cross-organizational collaboration capabilities, break the limitations caused by data silos in traditional models, avoid efficiency reduction due to deviation or one-sidedness, reduce the development cost of model packaging and operation, and provide technical support for large-scale water conservancy project deployment; it can ensure the integrity and consistency of data, reduce errors and uncertainties caused by data dispersion, improve the quality and accuracy of model output, achieve more intelligent result optimization, and enhance the adaptability and scalability of the model.

Description

Technical Field

[0001] The present invention relates to the field of water conservancy projects, and in particular to a standardized water conservancy professional model encapsulation and operation method based on Docker container technology. Background Art

[0002] Water conservancy projects must cope with increasingly complex and variable hydrological conditions, placing higher demands on the accuracy, computational efficiency, and adaptability of specialized water conservancy models. Traditional water conservancy models typically rely on single-machine computing, with fixed computing resources and limited scalability. This makes them unable to meet the demands of the growing volume of hydrological data and the increasing computational complexity. Furthermore, the development and application of water conservancy models often rely on expert experience and static rules, resulting in a lack of flexibility in the model building process and difficulty adapting to diverse scenarios in dynamic environments. In recent years, the rapid development of cloud computing and cloud-native technologies has provided new solutions for the construction, operation, and management of water conservancy models. Docker container technology, by unifying applications and their runtime environments, addresses the complex environment dependencies and poor portability inherent in traditional software. Kubernetes, a container orchestration tool, enables dynamic allocation and automated management of computing resources, providing powerful support for parallel computing and load balancing of distributed tasks. The integration of these technologies can significantly improve the development efficiency and computational performance of water conservancy models, while enabling cross-platform deployment and multi-user collaboration. Therefore, it is crucial to develop a standardized method for packaging and operating water conservancy models using Docker container technology.

[0003] The existing standardized water conservancy professional model packaging and operation methods have poor cross-organizational collaboration capabilities and are easily reduced in efficiency due to bias or one-sidedness, which increases the efficiency reduction caused by bias or one-sidedness. In addition, the existing standardized water conservancy professional model packaging and operation methods cannot ensure the integrity and consistency of data, and the errors and uncertainties caused by data dispersion increase, which reduces the quality and accuracy of model output. To this end, we propose a standardized water conservancy professional model packaging and operation method based on Docker container technology. Summary of the Invention

[0004] The purpose of the present invention is to solve the defects in the prior art and propose a standardized water conservancy professional model packaging and operation method based on Docker container technology.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A standardized water conservancy professional model packaging and operation method using Docker container technology, the specific steps of the packaging and operation method are as follows:

[0007] Q1. Collect data from each set of target water conservancy models, build a visual model knowledge graph, and analyze key calculation modules and data dependencies.

[0008] Q2: Encapsulate the computing units of each water conservancy model into independent Docker images and divide the computing units into different priorities to form an orderly computing task queue;

[0009] Q3. Dynamically create container instances, allocate computing resources required for model execution, monitor resource usage of each container in real time, and dynamically optimize the scheduling process of container instances.

[0010] Q4. Based on the topological relationship of the task queue, the computing tasks are assigned to different computing nodes for parallel execution. During the parallel process, the task execution order is dynamically adjusted.

[0011] Q5. Set up automatic container restart and fault recovery mechanisms to handle abnormal or failed tasks, and collect and optimize the calculation results after the distributed tasks are completed;

[0012] Q6. Verify the rationality of the model calculation results, and continuously optimize the Docker image packaging, task queue division, and scheduling strategies based on the model operation results and user feedback.

[0013] As a further solution of the present invention, the specific steps of analyzing the key computing modules and data dependencies described in Q1 are as follows:

[0014] S1.1: Collect the basic data sets for the target hydraulic model from monitoring equipment, remote sensing data, historical records, and expert knowledge, including topographic data, hydrological data, hydraulic parameters, and engineering rules. Clean, remove duplicates, and standardize the collected basic data sets.

[0015] S1.2: Establish a structured database based on the types and relationships of each set of processed basic data, extract the hydraulic engineering rules of the target hydraulic model, and establish the dependency relationships between data nodes based on the rules;

[0016] S1.3: Convert the organized data, rules, and formulas into model knowledge graph nodes and edges, where nodes represent data, computational units, or rules, and edges represent upstream and downstream topological relationships or data dependencies.

[0017] S1.4: Convert the model knowledge graph into an interactive graph using the Neo4j graph database, analyze the key computing modules in the graph, remove redundant data or irrelevant nodes in the graph based on the analysis results, and determine the computational priority of each computing module.

[0018] As a further solution of the present invention, the specific steps of encapsulating the computing units of each water conservancy professional model into an independent Docker image as described in Q2 are as follows:

[0019] S2.1: Analyze the nodes and edges in the model knowledge graph, extract the nodes representing the key calculations of the water conservancy model, decompose the algorithms of the calculation units corresponding to each node, clarify the input, processing logic, and output, and, based on the analysis results, write the core algorithm programs for each calculation unit of the water conservancy model in various development languages;

[0020] S2.2: Determine the runtime environment required for algorithm implementation, including programming language versions, third-party libraries, and operating systems. Use Docker tools to build a Docker image, and clearly define all dependencies in the container environment configuration file.

[0021] S2.3: Encapsulate the algorithm code and its dependent environment, then test the image's functionality and performance locally. Once the local test passes, push the completed image to the container registry and add metadata tags to the image, marking the version number and functional module information.

[0022] As a further solution of the present invention, the specific steps of dividing the computing units into different priorities and forming an ordered computing task queue as described in Q2 are as follows:

[0023] S3.1: Based on the node and edge information in the model knowledge graph, construct a directed acyclic graph (DAG) and use an adjacency matrix or adjacency list to represent the DAG structure. Calculate the in-degree of each node, that is, the number of nodes that depend on its output. Then initialize a set of nodes with in-degree 0 as the starting point for topological sorting.

[0024] S3.2: Use the queue structure to record the nodes with the current in-degree of 0. Take the nodes out of the queue in order, add them to the topological sort list, remove their outgoing edges, update the in-degree of the downstream nodes, and if there is a downstream node whose in-degree becomes 0, add it to the queue. Repeat the queue extraction until the queue is empty, and finally generate a hierarchical topological sort.

[0025] S3.3: Output the node set of each layer after topological sorting to represent the task groupings of different priorities. After the grouping is completed, collect the information of each participant and the DAG structure and task hierarchy information they manage. Then each participant generates the DAG topology locally, including computing unit nodes and dependency edges.

[0026] S3.4: Each participant calculates the importance of each dependency edge and locally updates its DAG topology and task priority information. It then uses the gradient descent algorithm to optimize and adjust the parameters of its corresponding priority model. After the adjustment is complete, each participant uploads its locally optimized DAG model parameters to the central federated server.

[0027] S3.5: The server performs weighted average aggregation on the uploaded parameters and updates the global model parameters to generate a more comprehensive DAG topology and priority information. Each participant downloads the aggregated global model parameters and re-updates its local DAG topology.

[0028] S3.6: Based on the globally optimized priority information, the hierarchy is re-divided according to the node priority, and the nodes with the highest priority are processed first. At the same time, the local update and global aggregation steps are repeated, and the DAG topology relationship and task layering strategy are gradually iterated.

[0029] As a further solution of the present invention, the specific steps of the dynamic optimization container instance scheduling process described in Q3 are as follows:

[0030] S4.1: Collect resource usage data of Docker containers through monitoring tools or integrated monitoring platforms. Correlate the real-time monitored resource usage data with the current task allocation to form a task load matrix and collect the current load of each node.

[0031] S4.2: With the goal of minimizing the load difference between nodes, set the objective function of the task allocation scheme, set the initial temperature and cooling rate, and randomly perturb the task allocation scheme at the current temperature to generate a new scheme. Calculate the target value of the new scheme using the objective function, and compare the target value of the new scheme with the target value of the current scheme.

[0032] S4.3: If the target value of the new solution is lower than the current task allocation solution, the new solution is accepted and replaced with the original task allocation solution. Otherwise, the acceptance probability of the new solution is calculated and, based on the acceptance probability, a decision is made on whether to accept the new solution. The task allocation solution is updated repeatedly, and the temperature is gradually reduced based on the preset cooling rate.

[0033] S4.4: When the temperature reaches the set minimum value or the objective function converges to the preset range, the final task allocation plan is output. Based on the optimized task allocation plan, the Kubernetes API is called to dynamically adjust the Pod's allocation node.

[0034] As a further solution of the present invention, the specific steps of collecting and optimizing the calculation results after the distributed tasks are completed in Q5 are as follows:

[0035] S5.1: Collect the output data from each task calculation and integrate the results into a unified model output format. Perform basic data cleanup on the results according to the model's predefined rules. Then, represent the overall model output as a point in the state space and construct an objective function based on the economic, stability, or accuracy indicators of the water conservancy model.

[0036] S5.2: Set up the path search model in state space and pass Generate the corresponding decision path and identify the control points corresponding to the key state points based on the calculation results. Each control point corresponds to an important decision point on the optimization path, where P(t) represents the control curve of the decision path, t represents the progress of the path, and n represents the total number of decision points. represents the kth order Bellstein basis function, C k represents the control point, which indicates the key decision variable configuration in the path;

[0037] S5.3: Based on the current state and objective function value, the path is optimized by adjusting the position of the control points. Through multiple iterations, the decision path is sampled multiple times, and the objective function value of each path point is calculated. The optimal path point is selected and the final output result is generated. The optimized result is then post-processed to generate a water conservancy project scheduling plan.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. The present invention constructs a directed acyclic graph based on the information of nodes and edges in the model knowledge graph, and uses an adjacency matrix or adjacency list to represent the DAG structure. The queue structure is used to record the nodes with the current in-degree of 0, and the nodes are taken out of the queue in order, added to the topological sorting list, and their out-edges are removed. The in-degree of the downstream node is updated. If the in-degree of the downstream node becomes 0, it is added to the queue, and the queue extraction is repeated until the queue is empty. Finally, a hierarchical topological sorting is generated, and the node set of each layer after the topological sorting is output to represent the task grouping of different priorities. After the grouping is completed, the information of each participant and the DAG structure and task hierarchical information managed by each participant are collected. After that, each participant generates a DAG topology locally, including computing unit nodes and dependent edges. Each participant calculates the importance of each dependent edge and updates its DAG topological relationship and task priority information locally. Gradient descent is used The algorithm optimizes and adjusts the parameters of its corresponding priority model. After the adjustment is completed, each participant uploads its locally optimized DAG model parameters to the central federal server. The server performs weighted average aggregation on the uploaded parameters and updates the global model parameters to generate a more comprehensive DAG topology and priority information. Each participant downloads the aggregated global model parameters and re-updates its local DAG topology. Based on the globally optimized priority information and the node priority, the hierarchy is re-divided, and the node with the highest priority is processed first. At the same time, the local update and global aggregation steps are repeated, and the DAG topology and task layering strategy are gradually iterated to improve cross-organizational collaboration capabilities, break the limitations of traditional models caused by data silos, avoid efficiency reduction due to deviation or one-sidedness, reduce the development cost of model packaging and operation, and provide technical support for the deployment of large-scale water conservancy projects.

[0040] 2. The present invention collects each group of output data of the task calculation and integrates the calculation results into a unified model output format. According to the predefined rules of the model, the basic data of the results are cleaned, and then the overall output results of the model are represented as points in the state space. The objective function is constructed according to the various indicators of the economy, stability or accuracy of the water conservancy model. A path search model is set in the state space, and a corresponding decision path is generated. The control points corresponding to the key state points are identified based on the calculation results. Each control point corresponds to an important decision point on the optimization path. According to the current state and the objective function value, the position of the control point is adjusted to optimize the path. Through multiple iterations, the decision path is sampled multiple times, and the objective function value of each path point is calculated at the same time. The optimal path point is selected to generate the final output result. The optimized result is then post-processed to generate a water conservancy project scheduling plan. This can ensure the integrity and consistency of the data, reduce the errors and uncertainties caused by data dispersion, improve the quality and accuracy of the model output, achieve more intelligent result optimization, and enhance the adaptability and scalability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0042] Figure 1 This is a flowchart of a standardized water conservancy professional model packaging and operation method based on Docker container technology proposed in the present invention. DETAILED DESCRIPTION

[0043] Example 1

[0044] Reference Figure 1 A standardized water conservancy professional model packaging and operation method based on Docker container technology. The specific steps of the packaging and operation method are as follows:

[0045] Collect data from each set of target water conservancy models, build a visual model knowledge graph, and analyze key calculation modules and data dependencies.

[0046] Specifically, various sets of basic data of the target water conservancy model, including terrain data, hydrological data, hydraulic parameters and engineering rules, are collected from monitoring equipment, remote sensing data, historical records and expert knowledge. The collected basic data are cleaned, deduplicated and standardized. A structured database is established according to the types and relationships of the processed basic data sets, and the water conservancy engineering rules of the target water conservancy model are extracted. At the same time, the dependency relationship between data nodes is established according to the rules, and the sorted data, rules and formulas are converted into the form of model knowledge graph nodes and edges, where nodes represent data, computing units or rules, and edges represent upstream and downstream topological relationships or data dependency relationships. The model knowledge graph is converted into an interactive graph through the Neo4j graph database, and the key computing modules in the graph are analyzed. Based on the analysis results, redundant data or irrelevant nodes in the graph are removed, and the calculation priority of each computing module is determined.

[0047] The computing units of each water conservancy professional model are encapsulated into independent Docker images, and the computing units are divided into different priorities to form an orderly computing task queue.

[0048] Specifically, we analyze the nodes and edges in the model knowledge graph, extract the nodes representing the key calculations of the water conservancy model, decompose the algorithms of the calculation units corresponding to each node, clarify the input, processing logic and output, and based on the analysis results, determine the operating environment required for the algorithm implementation based on the core algorithm programs of each calculation unit of the water conservancy professional model written in various development languages, including the programming language version, third-party libraries and operating system, and use the Docker tool to build a Docker image. Clearly define all dependencies in the container environment configuration file, encapsulate the algorithm code and its dependent environment, and then test the function and performance of the image locally. After the local test passes, push the built image to the container registry, and add metadata tags to the image to mark the version number and functional module information.

[0049] Specifically, according to the information of nodes and edges in the model knowledge graph, a directed acyclic graph is constructed, and the adjacency matrix or adjacency list is used to represent the DAG structure. The in-degree of each node is calculated, that is, the number of nodes that depend on its output. Then, a set of nodes with an in-degree of 0 is initialized as the starting point of the topological sorting. The queue structure is used to record the nodes with the current in-degree of 0. The nodes are taken out from the queue in order, added to the topological sorting list, and their out-edges are removed. The in-degree of the downstream nodes is updated. If the in-degree of the downstream node becomes 0, it is added to the queue. The queue extraction is repeated until the queue is empty. Finally, a hierarchical topological sorting is generated. The node set of each layer after the topological sorting is output to represent the task grouping of different priorities. After the grouping is completed, the information of each participant and the DAG structure and task hierarchical information managed by each participant are collected. Then, each participant generates a DAG locally. Topology, including computing unit nodes and dependent edges. Each participant calculates the importance of each dependent edge and updates its DAG topology and task priority information locally. It uses the gradient descent algorithm to optimize and adjust the parameters of its corresponding priority model. After the adjustment is completed, each participant uploads its locally optimized DAG model parameters to the central federated server. The server performs weighted average aggregation on the uploaded parameters and updates the global model parameters to generate a more comprehensive DAG topology and priority information. Each participant downloads the aggregated global model parameters and re-updates its local DAG topology structure. Based on the globally optimized priority information, the hierarchy is re-divided according to the priority of the nodes, and the nodes with the highest priority are processed first. The local update and global aggregation steps are repeated at the same time, and the DAG topology and task layering strategy are gradually iterated and improved.

[0050] Example 2

[0051] Reference Figure 1 A standardized water conservancy professional model packaging and operation method based on Docker container technology. The specific steps of the packaging and operation method are as follows:

[0052] Dynamically create container instances, allocate computing resources required for model operation, monitor resource usage of each container in real time, and dynamically optimize the scheduling process of container instances.

[0053] Specifically, resource usage data of Docker containers is collected through monitoring tools or integrated monitoring platforms, and the real-time monitored resource usage data is associated with the current task allocation situation to form a task load matrix. The current load of each node is collected, with the goal of minimizing the load difference between nodes. The objective function of the task allocation scheme is set, the initial temperature and cooling rate are set, and the task allocation scheme is randomly perturbed at the current temperature to generate a new scheme. The target value of the new scheme is calculated through the objective function, and the target value of the new scheme is compared with the target value of the current scheme. If the target value of the new scheme is lower than the current task allocation scheme, the new scheme is accepted and the original task allocation scheme is replaced. Otherwise, the acceptance probability of the new scheme is calculated, and based on the acceptance probability, it is decided whether to accept the new scheme. The task allocation scheme is updated repeatedly, and the temperature is gradually lowered based on the preset cooling rate. When the temperature reaches the set minimum value or the objective function converges to the preset range, the final task allocation scheme is output, and according to the optimized task allocation scheme, the Kubernetes API is called to dynamically adjust the Pod's allocation node.

[0054] According to the topological relationship of the task queue, the computing tasks are assigned to different computing nodes for parallel execution. During the parallel process, the task execution order is dynamically adjusted.

[0055] Set up container automatic restart and fault recovery mechanisms to handle abnormal or failed tasks, and collect and optimize the calculation results after the distributed tasks are completed.

[0056] Specifically, each set of output data of the task calculation is collected, and the calculation results are integrated into a unified model output format. According to the predefined rules of the model, the basic data of the results are cleaned, and then the overall output results of the model are represented as points in the state space. The objective function is constructed according to the various indicators of the economy, stability or accuracy of the water conservancy model. A path search model is set in the state space, and the corresponding decision path is generated. Based on the calculation results, the control points corresponding to the key state points are identified. Each control point corresponds to an important decision point on the optimization path. According to the current state and the objective function value, the position of the control point is adjusted to optimize the path. Through multiple iterations, the decision path is sampled multiple times, and the objective function value of each path point is calculated at the same time. The optimal path point is selected to generate the final output result. The optimized result is then post-processed to generate a water conservancy project scheduling plan.

[0057] In this embodiment, the specific calculation formula of the decision path is as follows:

[0058]

[0059] Where P(t) represents the control curve of the decision path, t represents the progress of the path, and n represents the total number of decision points. represents the kth order Bellstein basis function, Ck Represents the control point, which indicates the key decision variable configuration in the path.

[0060] Verify the rationality of the model calculation results, and continuously optimize the Docker image packaging, task queue division and scheduling strategy based on the model operation effect and user feedback.

Claims

1. A standardized water conservancy professional model packaging and operation method based on Docker container technology, characterized by: The specific steps of the package operation method are as follows: Q1. Collect data from each set of target water conservancy models, build a visual model knowledge graph, and analyze key calculation modules and data dependencies. Q2: Encapsulate the computing units of each water conservancy model into independent Docker images and divide the computing units into different priorities to form an orderly computing task queue; Q3. Dynamically create container instances, allocate computing resources required for model execution, monitor resource usage of each container in real time, and dynamically optimize the scheduling process of container instances. Q4. Based on the topological relationship of the task queue, the computing tasks are assigned to different computing nodes for parallel execution. During the parallel process, the task execution order is dynamically adjusted. Q5. Set up automatic container restart and fault recovery mechanisms to handle abnormal or failed tasks, and collect and optimize the calculation results after the distributed tasks are completed; Q6: Verify the rationality of model calculation results and continuously optimize Docker image packaging, task queue division, and scheduling strategies based on model operation results and user feedback. The specific steps for dividing computing units into different priorities and forming an ordered computing task queue as described in Q2 are as follows: S3.1: Based on the node and edge information in the model knowledge graph, construct a directed acyclic graph (DAG) and use an adjacency matrix or adjacency list to represent the DAG structure. Calculate the in-degree of each node, that is, the number of nodes that depend on its output. Then initialize a set of nodes with in-degree 0 as the starting point for topological sorting. S3.2: Use the queue structure to record the nodes with the current in-degree of 0. Take the nodes out of the queue in order, add them to the topological sort list, remove their outgoing edges, update the in-degree of the downstream nodes, and if there is a downstream node whose in-degree becomes 0, add it to the queue. Repeat the queue extraction until the queue is empty, and finally generate a hierarchical topological sort. S3.3: Output the node set of each layer after topological sorting to represent the task groupings of different priorities. After the grouping is completed, collect the information of each participant and the DAG structure and task hierarchy information they manage. Then each participant generates the DAG topology locally, including computing unit nodes and dependency edges. S3.4: Each participant calculates the importance of each dependency edge and locally updates its DAG topology and task priority information. It then uses the gradient descent algorithm to optimize and adjust the parameters of its corresponding priority model. After the adjustment is complete, each participant uploads its locally optimized DAG model parameters to the central federated server. S3.5: The server performs weighted average aggregation on the uploaded parameters and updates the global model parameters to generate a more comprehensive DAG topology and priority information. Each participant downloads the aggregated global model parameters and re-updates its local DAG topology. S3.6: Based on the globally optimized priority information, the hierarchy is re-divided according to the node priority, and the nodes with the highest priority are processed first. At the same time, the local update and global aggregation steps are repeated, and the DAG topology relationship and task layering strategy are gradually iterated.

2. The standardized water conservancy professional model packaging and operation method based on Docker container technology according to claim 1 is characterized in that: The specific steps for analyzing key computing modules and data dependencies described in Q1 are as follows: S1.1: Collect the basic data sets for the target hydraulic model from monitoring equipment, remote sensing data, historical records, and expert knowledge, including topographic data, hydrological data, hydraulic parameters, and engineering rules. Clean, remove duplicates, and standardize the collected basic data sets. S1.2: Establish a structured database based on the types and relationships of each set of processed basic data, extract the hydraulic engineering rules of the target hydraulic model, and establish the dependency relationships between data nodes based on the rules; S1.3: Convert the organized data, rules, and formulas into model knowledge graph nodes and edges, where nodes represent data, computational units, or rules, and edges represent upstream and downstream topological relationships or data dependencies. S1.4: Convert the model knowledge graph into an interactive graph using the Neo4j graph database, analyze the key computing modules in the graph, remove redundant data or irrelevant nodes in the graph based on the analysis results, and determine the computational priority of each computing module.

3. The standardized water conservancy professional model packaging and operation method based on Docker container technology according to claim 2 is characterized in that: The specific steps for encapsulating the computing units of each water conservancy model into independent Docker images as described in Q2 are as follows: S2.1: Analyze the nodes and edges in the model knowledge graph, extract the nodes representing the key calculations of the water conservancy model, decompose the algorithms of the calculation units corresponding to each node, clarify the input, processing logic, and output, and, based on the analysis results, write the core algorithm programs for each calculation unit of the water conservancy model in various development languages; S2.2: Determine the runtime environment required for algorithm implementation, including programming language versions, third-party libraries, and operating systems. Use Docker tools to build a Docker image, and clearly define all dependencies in the container environment configuration file. S2.3: Encapsulate the algorithm code and its dependent environment, then test the image's functionality and performance locally. Once the local test passes, push the completed image to the container registry and add metadata tags to the image, marking the version number and functional module information.

4. The standardized water conservancy professional model packaging and operation method based on Docker container technology according to claim 1 is characterized in that: The specific steps of the dynamic optimization container instance scheduling process described in Q3 are as follows: S4.1: Collect resource usage data of Docker containers through monitoring tools or integrated monitoring platforms. Correlate the real-time monitored resource usage data with the current task allocation to form a task load matrix and collect the current load of each node. S4.2: With the goal of minimizing the load difference between nodes, set the objective function of the task allocation scheme, set the initial temperature and cooling rate, and randomly perturb the task allocation scheme at the current temperature to generate a new scheme. Calculate the target value of the new scheme using the objective function, and compare the target value of the new scheme with the target value of the current scheme. S4.3: If the target value of the new solution is lower than the current task allocation solution, the new solution is accepted and replaced with the original task allocation solution. Otherwise, the acceptance probability of the new solution is calculated and, based on the acceptance probability, a decision is made on whether to accept the new solution. The task allocation solution is updated repeatedly, and the temperature is gradually reduced based on the preset cooling rate. S4.4: When the temperature reaches the set minimum value or the objective function converges to the preset range, the final task allocation plan is output. Based on the optimized task allocation plan, the Kubernetes API is called to dynamically adjust the Pod's allocation node.

5. The standardized water conservancy professional model packaging and operation method based on Docker container technology according to claim 1 is characterized in that: The specific steps for collecting and optimizing the calculation results after the distributed tasks are completed as described in Q5 are as follows: S5.1: Collect the output data from each task calculation and integrate the results into a unified model output format. Perform basic data cleanup on the results according to the model's predefined rules. Then, represent the overall model output as a point in the state space and construct an objective function based on the economic, stability, or accuracy indicators of the water conservancy model. S5.2: Set up the path search model in state space and pass , generate the corresponding decision path, and identify the control points corresponding to the key state points based on the calculation results. Each control point corresponds to an important decision point on the optimization path, where The control curve representing the decision path, Represents the progress of the path, represents the total number of decision points, Representative Baelstein basis functions of order, represents the control point, which indicates the key decision variable configuration in the path; S5.3: Based on the current state and objective function value, the path is optimized by adjusting the position of the control points. Through multiple iterations, the decision path is sampled multiple times, and the objective function value of each path point is calculated. The optimal path point is selected and the final output result is generated. The optimized result is then post-processed to generate a water conservancy project scheduling plan.

Images