A reservoir scheduling scheme intelligent generation method and system fusing an AI agent

Abstract

The application is suitable for the field of reservoir operation and management, and provides a reservoir scheduling scheme intelligent generation method and system fusing AI Agent, the method comprises the following steps: continuously acquiring network public multimedia data, and processing to obtain recognition results and corresponding position information by using a visual recognition model; fusing and analyzing the recognition results and real-time operation data of the reservoir to generate a cleaning task list; based on the cleaning task list, assigning cleaning tasks according to known ship state data, and performing operation path optimization calculation to generate an optimal cleaning path planning scheme; integrating the task assignment result and the cleaning path planning scheme to generate a visual cleaning scheduling scheme containing a ship, an operation route and a timing arrangement. The application changes reservoir operation and scheduling from passive to active perception, significantly improves cleaning efficiency, and greatly reduces labor cost and safety risk.

Description

Technical Field

[0001] This invention relates to the field of reservoir operation and maintenance management, specifically to a method and system for intelligently generating reservoir scheduling schemes that integrates AI agents. Background Technology

[0002] Managing floating debris (such as plastic waste, dead wood, and weeds) on the reservoir surface is an important part of the reservoir's daily operation and maintenance. Large amounts of floating debris not only affect the reservoir's landscape and water quality, but can also block water intakes, disrupt the normal operation of generator units, and even pose a potential threat to the safety of the dam.

[0003] Currently, the inspection and detection of floating debris in reservoirs mainly relies on manual inspection methods, including regular boat patrols and shoreline patrols. This method has obvious drawbacks: First, the inspection cycle is long and costly, and it cannot achieve all-weather, all-reservoir coverage, making it prone to monitoring blind spots; second, the response process from the discovery of floating debris to reporting and then to organizing cleanup is slow, resulting in delays in cleanup work and an inability to deal with the debris in its early stages of accumulation.

[0004] With the development of information technology, some reservoirs have attempted to use fixed cameras for monitoring. However, due to their limited field of view and reliance on manual review for video analysis, efficiency improvements are limited. Although some existing reservoir scheduling systems integrate artificial intelligence technology, they are mainly applied to hydrological forecasting and flood control and power generation scheduling optimization, focusing on structured data such as water level, flow rate, and rainfall. Existing technologies are lacking in handling unstructured, dynamically changing visual information such as reservoir surface conditions. This prevents reservoir management from obtaining a comprehensive picture of floating debris distribution in the reservoir area in a timely manner, resulting in a lack of accurate data support for cleanup operations, leaving them in a passive response and blind inspection mode. Therefore, this paper proposes an intelligent generation method and system for reservoir scheduling schemes that integrates AI agents, aiming to solve the above problems. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a method and system for intelligent generation of reservoir scheduling schemes that integrates AI Agent, so as to solve the problems existing in the above-mentioned background technology.

[0006] This invention is implemented as follows: a method for intelligently generating reservoir scheduling schemes by integrating AI agents, the method comprising the following steps: The system continuously acquires publicly available multimedia data related to the reservoir and processes the data using a visual recognition model to obtain recognition results and corresponding location information. The visual recognition model is driven by an AI Agent and used to identify floating objects on the water surface. The identification results are fused and analyzed with the real-time operation data of the reservoir to generate a list of cleanup tasks that includes target areas and priorities. The identification results include the type of floating debris, location information, and estimated area of ​​the floating debris accumulation area. Based on the cleanup task list, cleanup tasks are assigned according to known vessel status data, and the optimal cleanup path planning scheme is generated by performing operation path optimization calculations. The task allocation results and the cleanup path planning scheme are integrated to generate a visual cleanup scheduling scheme that includes ships, operation routes, and timing arrangements, and then pushed to the mobile terminal.

[0007] As a further aspect of the present invention: the step of continuously acquiring publicly available online multimedia data related to the reservoir, and processing the publicly available online multimedia data using a visual recognition model to obtain recognition results and corresponding location information, specifically includes: A distributed web crawler architecture is used to continuously acquire publicly available multimedia data from multiple social media platforms, which are associated with the reservoir and contain geotags or text descriptions. The visual recognition model is used to identify water scene keyframes in publicly available multimedia data on the network to obtain location information. For the identified floating objects, the texture, color and outline features are analyzed by visual recognition model to obtain the floating object type. At the same time, image georeferencing technology is used to locate and estimate the area of ​​the floating object cluster. The recognition results of the integrated visual recognition model are linked with the original multimedia data referenced.

[0008] As a further aspect of the present invention: before obtaining publicly available multimedia data input to the visual recognition model, multi-level screening is performed, specifically including the following steps: The geographic tags embedded in each piece of publicly available multimedia data on the network are analyzed. When a data point falls within the geographic fence area associated with a reservoir, it is determined to be primary associated data. For data without geotags, content tags, titles, and descriptions are extracted and matched using a keyword database. When a match is successful, the data is identified as secondary related data. The keyword database includes reservoir names, reservoir aliases, surrounding place names, and scenic spot names. For data that fails to be identified, its visual features are extracted and matched with a pre-built benchmark image library that covers the typical features of reservoirs from multiple perspectives and in multiple seasons. If the match is successful, it is identified as visually related data. The primary association data, secondary association data, and visual association data are integrated, and the association confidence score is calculated for each data point. Data with a correlation confidence level exceeding a set threshold are integrated to obtain a publicly available multimedia data set on the internet.

[0009] As a further aspect of the present invention: the reference image library is a dynamically updated image library, and its update source includes publicly available online multimedia data that has been determined to be associated with the reservoir after multi-level screening and whose association confidence exceeds a set threshold; the visual recognition model performs standardization processing on such publicly available online multimedia data, extracts its key frames or clear image frames, and automatically labels the typical visual elements and environmental features of the reservoir contained therein, and integrates them into the reference image library as new reference samples.

[0010] As a further aspect of the present invention: the step of fusing and analyzing the identification results with the real-time operational data of the reservoir to generate a list of cleanup tasks including target areas and priorities specifically includes: The floating object type is mapped to a predefined category code, the estimated area is used as a continuous numerical feature, and the Euclidean distance from the floating object to a preset key point is calculated based on the location information as a distance feature. The preset key point is the starting point of the reservoir operation and maintenance work. The system analyzes real-time reservoir operation data to determine the current reservoir operation mode and selects feature weight vectors from a pre-defined weight configuration library based on the reservoir's operation model. The priority coefficients are obtained by performing a dot product of the category code, numerical features, and distance features based on the feature weight vector, and the floating object aggregation areas are sorted according to the priority coefficients. The system integrates and sorts floating debris clusters, debris types, location information, estimated area, and priority coefficients to generate a structured cleanup task list.

[0011] As a further aspect of the present invention: the step of allocating cleanup tasks based on the cleanup task list and known vessel status data, and performing operation path optimization calculations to generate the optimal cleanup path planning scheme, specifically includes: A directed edge-node network is generated based on the reservoir's electronic map and the location information of the floating debris accumulation area. The nodes are the ship mooring points, the locations of various floating debris, and the garbage unloading points. The directed edges are the waterways between the nodes, and weights are added according to the sailing distance. Based on minimizing the total operation time, tasks are allocated in the directed edge-node network using ship load and task time window as constraints. Based on the allocation results, an improved genetic algorithm is used to calculate the optimal path sequence for each ship; The optimal path sequence of all ships is integrated to generate a clear path planning scheme that includes the navigation sequence.

[0012] Another objective of this invention is to provide an intelligent generation system for reservoir scheduling schemes that integrates AI agents, the system comprising: The data acquisition and identification module is used to continuously acquire publicly available online multimedia data related to the reservoir, and to process the publicly available online multimedia data using a visual recognition model to obtain the identification results and corresponding location information. The visual recognition model is driven by an AI Agent and is used to identify floating objects on the water surface in the water area. The fusion analysis module is used to fuse the identification results with the real-time operation data of the reservoir to generate a list of cleanup tasks that includes target areas and priorities. The identification results include the type of floating debris, location information, and estimated area of ​​the floating debris accumulation area. The cleanup planning module is used to allocate cleanup tasks based on the cleanup task list and known vessel status data, and to perform operation path optimization calculations to generate the optimal cleanup path planning scheme. The scheme generation and push module is used to integrate the task allocation results and the cleanup path planning scheme to generate a visual cleanup scheduling scheme that includes ships, operation routes and time sequence arrangements, and push it to the mobile terminal.

[0013] As a further aspect of the present invention: the data acquisition and identification module includes: The data acquisition unit is used to continuously acquire publicly available multimedia data with geotags or text descriptions from multiple social media platforms through a distributed web crawler architecture, wherein the geotags and text descriptions are associated with the reservoir. The data recognition unit is used to use the visual recognition model to identify water scene keyframes in publicly available multimedia data on the network to obtain location information. The feature analysis unit is used to analyze the texture, color and outline features of the identified floating objects through a visual recognition model to obtain the floating object type, and at the same time to locate and estimate the area of ​​the floating object cluster area using image georeferencing technology. The result integration unit is used to integrate the recognition results of the output visual recognition model and associate them with the referenced original multimedia data.

[0014] As a further aspect of the present invention: the fusion analysis module includes: The data vectorization unit is used to map the floating object type to a predefined category code, use the estimated area as a continuous numerical feature, and calculate the Euclidean distance from the floating object to a preset key point as a distance feature based on the location information. The preset key point is the starting point of the reservoir operation and maintenance work. The feature matching unit is used to parse the real-time operation data of the reservoir, determine the current operation mode of the reservoir, and select feature weight vectors from the preset weight configuration library according to the operation model of the reservoir. The priority calculation unit is used to calculate the priority coefficient by performing a dot product of the category code, numerical features and distance features according to the feature weight vector, and to sort the floating object aggregation areas according to the priority coefficient; The list generation unit is used to integrate sorted floating debris accumulation areas, floating debris types, location information, estimated areas, and priority coefficients to generate a structured list of cleanup tasks.

[0015] As a further aspect of the present invention: the cleanup planning module includes: The network model building unit is used to generate a directed edge-node network based on the reservoir electronic map and the location information of the floating debris accumulation area. The nodes are the ship mooring points, the locations of each floating debris and the garbage unloading points, and the directed edges are the waterways between the nodes, with weights added according to the sailing distance. The task allocation unit is used to allocate tasks in the directed edge-node network based on minimizing the total operation time and using the ship's load and the task point time window as constraints. The optimal path sequence generation unit is used to calculate the optimal path sequence for each ship based on the allocation results using an improved genetic algorithm. The scheme output unit is used to integrate the optimal path sequences of all ships to generate a clear path planning scheme containing the navigation sequence.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention constructs a complete reservoir debris management system based on "Internet data perception - intelligent decision-making - automatic scheduling," effectively solving the problems of low efficiency in traditional manual inspections and monitoring. Leveraging the widespread use of short videos, this invention can proactively and accurately capture information about the presence of floating debris in the reservoir from massive amounts of online data. Then, through intelligent algorithms, it combines visual intelligence with real-time reservoir conditions to automatically formulate a clear and prioritized cleanup strategy. The resulting work plan not only precisely plans the route and timing of each vessel but can also be directly distributed to mobile terminals for execution. In summary, this invention transforms reservoir operation and maintenance scheduling from a passive to a proactive approach, significantly improving cleanup efficiency while greatly reducing labor costs and safety risks, thus promoting more refined reservoir management. Attached Figure Description

[0017] Figure 1 This is a flowchart of an intelligent generation method for reservoir scheduling schemes that integrates AI Agents.

[0018] Figure 2 This is a flowchart illustrating the process of obtaining publicly available multimedia data and recognition results in an intelligent generation method for reservoir scheduling schemes that integrates AI Agents.

[0019] Figure 3 This is a flowchart illustrating the multi-level filtering of publicly available multimedia data in a reservoir scheduling scheme intelligent generation method that integrates AI Agent.

[0020] Figure 4 This is a flowchart illustrating the fusion analysis of identification results and real-time operational data of a reservoir in an intelligent generation method that integrates AI agents for reservoir scheduling.

[0021] Figure 5 This is a flowchart illustrating a method for intelligently generating reservoir scheduling schemes that integrates AI agents. It describes how to allocate cleaning tasks based on known vessel status data and perform operation path optimization calculations to generate the optimal cleaning path planning scheme.

[0022] Figure 6 This is a schematic diagram of the structure of an intelligent generation system for reservoir scheduling schemes that integrates AI agents.

[0023] Figure 7 This is a schematic diagram of the data acquisition and recognition module in an intelligent generation system for reservoir scheduling schemes that integrates AI Agents.

[0024] Figure 8 This is a schematic diagram of the fusion analysis module in an intelligent generation system for reservoir scheduling schemes that integrates AI agents.

[0025] Figure 9 This is a schematic diagram of the cleaning and planning module in an intelligent generation system for reservoir scheduling schemes that integrates AI Agents. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0027] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0028] like Figure 1 As shown in the figure, this embodiment of the invention provides an intelligent generation method for reservoir scheduling schemes that integrates AI agents. The method includes the following steps: S100, continuously acquire publicly available multimedia data related to the reservoir, and use a visual recognition model to process the publicly available multimedia data to obtain recognition results and corresponding location information. The visual recognition model is driven by AIAgent and used to identify floating objects on the water surface in the water area. S200, integrates the identification results with the real-time operation data of the reservoir to generate a list of cleanup tasks that includes target areas and priorities. The identification results include the type of floating debris, location information, and estimated area of ​​the floating debris accumulation area. S300, based on the cleanup task list, allocates cleanup tasks according to known vessel status data, and performs operation path optimization calculations to generate the optimal cleanup path planning scheme. S400, integrate the task allocation results and the cleanup path planning scheme to generate a visual cleanup scheduling scheme that includes ships, operation routes and time sequence arrangements, and push it to the mobile terminal.

[0029] In this embodiment of the invention, efficient management of floating debris in reservoirs is achieved by constructing a complete intelligent decision-making closed loop. By automatically collecting publicly available image data of the reservoir area from the internet, when the visual recognition model processes videos of the reservoir area uploaded by users on social media platforms (such as Douyin, Weibo, etc.), it can intelligently identify specific types of floating debris (such as accumulated fallen leaves or plastic waste) gathered on the water surface and automatically associate them with their approximate locations. Subsequently, the recognition results are comprehensively analyzed in conjunction with the current operating conditions of the reservoir. For example, during periods of guaranteed water supply, the priority of cleaning floating debris near the water intake is automatically increased, or during peak tourist seasons, visible floating debris in scenic areas is prioritized, thereby generating a scientifically reasonable cleaning task list. Based on this task list, the invention comprehensively considers the current location and operational capabilities of available cleaning vessels, and automatically plans the optimal operational route that balances efficiency and safety through intelligent algorithms. Finally, a visualized scheduling scheme containing operational vessels, navigation paths, and work sequences is formed and directly pushed to the mobile devices of on-site personnel, realizing fully automated management from environmental perception to intelligent scheduling.

[0030] like Figure 2 As shown in the preferred embodiment of the present invention, the step of continuously acquiring publicly available online multimedia data related to the reservoir and processing the publicly available online multimedia data using a visual recognition model to obtain recognition results and corresponding location information specifically includes: S101, continuously obtain publicly available multimedia data with geotags or text descriptions from multiple social media platforms through a distributed web crawler architecture, wherein the geotags and text descriptions are associated with the reservoir; S102, using the visual recognition model to identify water scene keyframes in publicly available multimedia data on the network to obtain location information; S103, For the identified floating objects, the texture, color and outline features are analyzed by a visual recognition model to obtain the floating object type. At the same time, image georeferenced technology is used to locate and estimate the area of ​​the floating object cluster. S104 integrates the recognition results of the visual recognition model and links them with the original multimedia data referenced.

[0031] In this embodiment of the invention, intelligent identification of floating objects in a reservoir is achieved by constructing a multi-source information sensing network. A distributed crawler architecture continuously acquires publicly available multimedia data from various social media platforms and content-sharing websites. After initial screening using geolocation information or relevant text descriptions related to reservoir features, this data enters an intelligent analysis process driven by an AI Agent. The visual recognition model first performs scene understanding on the image data, identifying areas with reservoir water features. Then, it performs multi-dimensional feature analysis on floating targets in the water area image, using deep learning algorithms to distinguish different types of floating objects, such as organic plant debris from artificial plastic products. Simultaneously, combined with identifiable landmarks in the image, computer vision technology is used to estimate the relative position and range of the floating object distribution (i.e., location and area estimation). Finally, the structured data, including the identified floating object type, distribution area, and estimated size, is linked to the original image data to form a complete identification report, providing data support for subsequent scheduling decisions.

[0032] like Figure 3 As shown, in a preferred embodiment of the present invention, before obtaining publicly available multimedia data input to the visual recognition model, multi-level screening is performed. Specific steps include: S1011, parse the geographic tags embedded in each piece of publicly available multimedia data on the network, and when it falls within the geographic fence area associated with the reservoir, it is determined to be primary associated data; S1012, For data without geographic tags, extract its content tags, title and description text, and match it through a keyword library. When the match is successful, it is determined to be secondary related data. The keyword library includes reservoir name, reservoir alias, surrounding place names and scenic spot names. S1013. For data that fails to be determined, extract its visual features and perform similarity matching with a pre-built benchmark image library that covers the typical features of reservoirs from multiple perspectives and in multiple seasons. If the match is successful, it is determined to be visually related data. S1014, integrate the primary association data, secondary association data and visual association data, and calculate the association confidence for each data point; S1015, integrate the data whose association confidence exceeds the set threshold to obtain a publicly available multimedia data set on the network.

[0033] In this embodiment of the invention, for the acquired publicly available multimedia data from the internet, the geotagged information contained in each data entry is first parsed. When the data coordinates are detected to fall within a geofence area preset according to the reservoir management scope, it can be determined as primary related data. For example, when a photo taken along the reservoir shoreline with precise location is obtained, it will pass this first screening step.

[0034] For data lacking geotags, a text analysis mechanism is activated to automatically extract content tags, titles, and descriptive text, and then intelligently match them against a pre-built keyword database. This keyword database not only includes the standard name of the reservoir but also commonly used local nicknames, names of surrounding landform features, and names of well-known scenic spots, ensuring the identification of various relevant text descriptions. For example, when a user's video title uses local dialect to refer to a reservoir or mentions a little-known viewpoint, the system can still accurately identify it through fuzzy keyword matching.

[0035] Of course, if the first two levels of screening fail, a third-level visual feature matching process will be initiated. This step extracts the visual features of the data and compares them with a pre-built benchmark image library for similarity. This benchmark image library specifically considers the landscape changes of the reservoir under different seasons and water levels, including typical landscape features of each season. For example, the system can accurately identify the image features of the same shoreline during the high-water and low-water seasons, avoiding misjudgments caused by seasonal changes. After completing the three-level screening, a correlation confidence score is calculated for each data point. This calculation process comprehensively considers multiple factors such as the reliability of the data source and the accuracy of the matching. For example, data with precise location information and text matching results will receive a higher confidence score. Finally, data with correlation confidence scores exceeding a set threshold will be integrated to form a high-quality publicly available online multimedia dataset, providing a reliable data foundation for subsequent visual recognition analysis. This multi-level linkage screening mechanism ensures both the comprehensiveness of data collection and the quality and relevance of the data. It should be noted that the benchmark image library is a dynamically updated image library. Its update source includes publicly available online multimedia data that has been determined to be associated with the reservoir after multi-level screening and whose association confidence exceeds a set threshold. The visual recognition model performs standardization processing on such publicly available online multimedia data, extracts its key frames or clear image frames, and automatically labels the typical visual elements and environmental features of the reservoir contained therein, integrating them into the benchmark image library as new benchmark samples.

[0036] like Figure 4 As shown, in a preferred embodiment of the present invention, the step of fusing and analyzing the identification results with the real-time operational data of the reservoir to generate a cleanup task list containing target areas and priorities specifically includes: S201, map the floating object type to a predefined category code, use the estimated area as a continuous numerical feature, and calculate the Euclidean distance from the floating object to a preset key point based on the location information as a distance feature. The preset key point is the starting point of the reservoir operation and maintenance work. S202, parse the real-time operation data of the reservoir to determine the current operation mode of the reservoir, and select feature weight vectors from the preset weight configuration library according to the operation model of the reservoir. S203, calculate the priority coefficient by performing a dot product of the category code, numerical features and distance features according to the feature weight vector, and sort the floating object aggregation areas according to the priority coefficient; S204 integrates and sorts the floating debris accumulation areas, floating debris types, location information, estimated area, and priority coefficients to generate a structured cleanup task list.

[0037] In this embodiment of the invention, the identified floating debris features are first digitally encoded: the floating debris type is converted into a corresponding category code according to a predefined classification system, such as rotten plant debris, plastic products, and wood, each corresponding to different code values; simultaneously, the estimated floating debris coverage area is used as a continuous numerical feature; and based on the location information of the floating debris accumulation area, the Euclidean distance from it to a preset key operation and maintenance starting point (such as water intake, lock, dam front area, etc.) is calculated as a distance feature. Simultaneously, real-time reservoir operation data is analyzed to intelligently determine the current dominant operation mode of the reservoir. For example, during peak water supply periods, the system automatically identifies and prioritizes water supply security as the main operation mode; during flood season, flood control scheduling is the main mode; and during peak tourist season, landscape maintenance is emphasized. Based on the identified operation mode, the system retrieves the corresponding feature weight vector from a preset weight configuration library. This weight vector clearly defines the relative importance of various floating debris types, distribution areas, and distances from key facilities in priority assessment under different operation modes. The system performs a dot product operation between various feature values ​​and the corresponding weight vector to calculate the comprehensive priority coefficient for each floating debris accumulation area. This calculation process ensures that the evaluation criteria can be automatically adjusted under different operating conditions, so that the floating debris accumulation areas that have the greatest impact on the current operational objectives receive higher priority scores. Finally, this invention sorts all accumulation areas according to priority coefficients and integrates the sorting results with detailed data such as floating debris type, location information, and estimated area to generate a structured cleanup task list. This cleanup task list not only clearly marks the priority order of each cleanup target area, but also provides complete decision-making basis information.

[0038] like Figure 5As shown, in a preferred embodiment of the present invention, the step of allocating cleanup tasks based on the cleanup task list and known vessel status data, and performing operation path optimization calculations to generate the optimal cleanup path planning scheme, specifically includes: S301. A directed edge-node network is generated based on the reservoir electronic map and the location information of the floating debris accumulation area. The nodes are the ship mooring points, the locations of each floating debris and the garbage unloading points. The directed edges are the waterways between the nodes, and weights are added according to the sailing distance. S302, based on minimizing the total operation time, uses the ship's load and the task point time window as constraints to allocate tasks in the directed edge-node network; S303, based on the allocation results, uses an improved genetic algorithm to calculate the optimal path sequence for each ship; S304, integrates the optimal path sequences of all ships to generate a clear path planning scheme that includes the navigation sequence.

[0039] In this embodiment of the invention, a directed edge-node network is first constructed based on the reservoir electronic map and the location information of the floating debris accumulation area. This network uses vessel mooring points, the locations of each floating debris, and garbage unloading points as nodes, and the navigable waterways between nodes as directed edges. Each directed edge is assigned a corresponding weight value based on the actual navigation distance, forming a complete waterway network model. During the task allocation phase, minimizing the total operation time is the primary objective, while considering constraints such as vessel carrying capacity and task time windows. Optimization calculations are performed within the directed edge-node network. Based on the allocation results, the system uses an improved genetic algorithm to calculate the optimal path sequence for each vessel. This algorithm continuously optimizes path selection by simulating a natural evolutionary process, ultimately generating the optimal operation path for each vessel from its starting point to all allocated task points and then to the unloading point. Finally, the system integrates the optimal path sequences of all vessels to generate a cleaning path planning scheme containing detailed navigation sequences. Throughout the entire process, the system transforms the complex scheduling problem into a computable optimization problem by establishing a mathematical model. Based on a directed edge-node network, the system comprehensively considers multiple constraints such as ship status, task characteristics, and environmental factors, and utilizes intelligent algorithms to achieve collaborative optimization of task allocation and path planning. The final generated cleaning path planning scheme not only ensures operational efficiency but also fully considers various constraints in actual operations.

[0040] like Figure 6 As shown in the figure, this embodiment of the invention also provides an intelligent generation system for reservoir scheduling schemes that integrates AI agents, the system comprising: The data acquisition and recognition module 100 is used to continuously acquire publicly available online multimedia data related to the reservoir, and to process the publicly available online multimedia data using a visual recognition model to obtain recognition results and corresponding location information. The visual recognition model is driven by an AI Agent and is used to identify floating objects on the water surface in the water area. The fusion analysis module 200 is used to fuse and analyze the identification results with the real-time operation data of the reservoir to generate a list of cleanup tasks that includes target areas and priorities. The identification results include the type of floating debris, location information, and estimated area of ​​the floating debris accumulation area. The cleaning planning module 300 is used to allocate cleaning tasks based on the cleaning task list and known vessel status data, and to perform operation path optimization calculations to generate the optimal cleaning path planning scheme. The scheme generation and push module 400 is used to integrate the task allocation results and the cleanup path planning scheme to generate a visual cleanup scheduling scheme that includes ships, operation routes and time sequence arrangements, and push it to the mobile terminal.

[0041] In this embodiment of the invention, the system achieves intelligent scheduling of floating debris removal in reservoirs through four core modules. The data acquisition and identification module 100 collects relevant image data of the reservoir from the internet and automatically detects and locates floating debris using AI visual recognition technology. The fusion analysis module 200 combines the identification results with real-time reservoir operation data to generate a priority-ordered list of cleaning tasks. The cleaning planning module 300 allocates tasks and optimizes routes based on the task list and vessel status, formulating the optimal operation route. The scheme generation and push module 400 finally integrates and generates a visualized scheduling scheme containing vessel information, routes, and timing, and directly pushes it to mobile terminals for execution. The entire system achieves fully automated management from environmental perception to intelligent scheduling.

[0042] like Figure 7 As shown, in a preferred embodiment of the present invention, the data acquisition and identification module 100 includes: Data acquisition unit 101 is used to continuously acquire publicly available multimedia data with geotags or text descriptions from multiple social media platforms through a distributed web crawler architecture, wherein the geotags and text descriptions are associated with the reservoir; Data recognition unit 102 is used to use the visual recognition model to identify water scene keyframes in publicly available multimedia data on the network to obtain location information; The feature analysis unit 103 is used to analyze the texture, color and outline features of the identified floating objects through a visual recognition model to obtain the floating object type, and at the same time to locate and estimate the area of ​​the floating object aggregation area using image georeferencing technology. The result integration unit 104 is used to integrate the recognition results of the output visual recognition model and associate them with the referenced original multimedia data.

[0043] like Figure 8 As shown, in a preferred embodiment of the present invention, the fusion analysis module 200 includes: The data vectorization unit 201 is used to map the floating object type to a predefined category code, use the estimated area as a continuous numerical feature, and calculate the Euclidean distance from the floating object to a preset key point as a distance feature based on the location information. The preset key point is the starting point of the reservoir operation and maintenance work. The feature matching unit 202 is used to parse the real-time operation data of the reservoir, determine the current operation mode of the reservoir, and select feature weight vectors from the preset weight configuration library according to the operation model of the reservoir. The priority calculation unit 203 is used to calculate the priority coefficient by performing a dot product of the category code, numerical features and distance features according to the feature weight vector, and sort the floating object aggregation areas according to the priority coefficient; The list generation unit 204 is used to integrate the sorted floating debris gathering areas, floating debris types, location information, estimated area and priority coefficients to generate a structured cleanup task list.

[0044] like Figure 9 As shown, in a preferred embodiment of the present invention, the cleanup planning module 300 includes: The network model building unit 301 is used to generate a directed edge-node network based on the reservoir electronic map and the location information of the floating debris accumulation area. The nodes are the ship mooring points, the locations of each floating debris and the garbage unloading points. The directed edges are the waterways between the nodes, and weights are added according to the sailing distance. Task allocation unit 302 is used to allocate tasks in a directed edge-node network based on minimizing the total operation time and using the ship's load and task point time window as constraints. The optimal path sequence generation unit 303 is used to calculate the optimal path sequence for each ship based on the allocation results using an improved genetic algorithm. The scheme output unit 304 is used to integrate the optimal path sequences of all ships to generate a clear path planning scheme containing the navigation sequence.

[0045] The above description only details the preferred embodiments of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0046] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0047] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0048] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the disclosure in the specification and embodiments. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

Claims

1. A method for intelligently generating reservoir scheduling schemes integrating AI agents, characterized in that, The method includes the following steps: The system continuously acquires publicly available multimedia data related to the reservoir and processes the data using a visual recognition model to obtain recognition results and corresponding location information. The visual recognition model is driven by an AI Agent and used to identify floating objects on the water surface. The identification results are fused and analyzed with the real-time operation data of the reservoir to generate a list of cleanup tasks that includes target areas and priorities. The identification results include the type of floating debris, location information, and estimated area of ​​the floating debris accumulation area. Based on the cleanup task list, cleanup tasks are assigned according to known vessel status data, and the optimal cleanup path planning scheme is generated by performing operation path optimization calculations. The task allocation results and the cleanup path planning scheme are integrated to generate a visual cleanup scheduling scheme that includes ships, operation routes, and timing arrangements, and then pushed to the mobile terminal.

2. The intelligent generation method for reservoir scheduling schemes integrating AI Agent according to claim 1, characterized in that, The steps of continuously acquiring publicly available online multimedia data related to the reservoir, and processing the publicly available online multimedia data using a visual recognition model to obtain recognition results and corresponding location information, specifically include: A distributed web crawler architecture is used to continuously acquire publicly available multimedia data from multiple social media platforms, which are associated with the reservoir and contain geotags or text descriptions. The visual recognition model is used to identify water scene keyframes in publicly available multimedia data on the network to obtain location information. For the identified floating objects, the texture, color and outline features are analyzed by visual recognition model to obtain the floating object type. At the same time, image georeferencing technology is used to locate and estimate the area of ​​the floating object cluster. The recognition results of the integrated visual recognition model are linked with the original multimedia data referenced.

3. The intelligent generation method for reservoir scheduling schemes integrating AI Agents according to claim 2, characterized in that, Before obtaining publicly available multimedia data from the internet to input into the visual recognition model, a multi-level screening process is performed. Specific steps include: The geographic tags embedded in each piece of publicly available multimedia data on the network are analyzed. When a data point falls within the geographic fence area associated with a reservoir, it is determined to be primary associated data. For data without geotags, content tags, titles, and descriptions are extracted and matched using a keyword database. When a match is successful, the data is identified as secondary related data. The keyword database includes reservoir names, reservoir aliases, surrounding place names, and scenic spot names. For data that fails to be identified, its visual features are extracted and matched with a pre-built benchmark image library that covers the typical features of reservoirs from multiple perspectives and in multiple seasons. If the match is successful, it is identified as visually related data. The primary association data, secondary association data, and visual association data are integrated, and the association confidence score is calculated for each data point. Data with a correlation confidence level exceeding a set threshold are integrated to obtain a publicly available multimedia data set on the internet.

4. The intelligent generation method for reservoir scheduling schemes integrating AI Agents according to claim 3, characterized in that, The benchmark image library is a dynamically updated image library. Its update source includes publicly available online multimedia data that has been determined to be associated with the reservoir after multi-level screening and whose association confidence exceeds a set threshold. The visual recognition model performs standardization processing on such publicly available online multimedia data, extracts its key frames or clear image frames, and automatically labels the typical visual elements and environmental features of the reservoir contained therein, integrating them into the benchmark image library as new benchmark samples.

5. The intelligent generation method for reservoir scheduling schemes integrating AI Agent according to claim 1, characterized in that, The step of fusing and analyzing the identification results with the real-time operational data of the reservoir to generate a list of cleanup tasks including target areas and priorities specifically includes: The floating object type is mapped to a predefined category code, the estimated area is used as a continuous numerical feature, and the Euclidean distance from the floating object to a preset key point is calculated based on the location information as a distance feature. The preset key point is the starting point of the reservoir operation and maintenance work. The system analyzes real-time reservoir operation data to determine the current reservoir operation mode and selects feature weight vectors from a pre-defined weight configuration library based on the reservoir's operation model. The priority coefficients are obtained by performing a dot product of the category code, numerical features, and distance features based on the feature weight vector, and the floating object aggregation areas are sorted according to the priority coefficients. The system integrates and sorts floating debris clusters, debris types, location information, estimated area, and priority coefficients to generate a structured cleanup task list.

6. The intelligent generation method for reservoir scheduling schemes integrating AI Agent according to claim 1, characterized in that, The steps of allocating cleanup tasks based on the cleanup task list and known vessel status data, and performing operation path optimization calculations to generate the optimal cleanup path planning scheme, specifically include: A directed edge-node network is generated based on the reservoir's electronic map and the location information of the floating debris accumulation area. The nodes are the ship mooring points, the locations of various floating debris, and the garbage unloading points. The directed edges are the waterways between the nodes, and weights are added according to the sailing distance. Based on minimizing the total operation time, tasks are allocated in the directed edge-node network using ship load and task time window as constraints. Based on the allocation results, an improved genetic algorithm is used to calculate the optimal path sequence for each ship; The optimal path sequence of all ships is integrated to generate a clear path planning scheme that includes the navigation sequence.

7. A smart generation system for reservoir scheduling schemes integrating AI Agent, characterized in that, The system includes: The data acquisition and identification module is used to continuously acquire publicly available online multimedia data related to the reservoir, and to process the publicly available online multimedia data using a visual recognition model to obtain the identification results and corresponding location information. The visual recognition model is driven by an AI Agent and is used to identify floating objects on the water surface in the water area. The fusion analysis module is used to fuse the identification results with the real-time operation data of the reservoir to generate a list of cleanup tasks that includes target areas and priorities. The identification results include the type of floating debris, location information, and estimated area of ​​the floating debris accumulation area. The cleanup planning module is used to allocate cleanup tasks based on the cleanup task list and known vessel status data, and to perform operation path optimization calculations to generate the optimal cleanup path planning scheme. The scheme generation and push module is used to integrate the task allocation results and the cleanup path planning scheme to generate a visual cleanup scheduling scheme that includes ships, operation routes and time sequence arrangements, and push it to the mobile terminal.

8. The intelligent generation system for reservoir scheduling schemes integrating AI Agent according to claim 7, characterized in that, The data acquisition and recognition module includes: The data acquisition unit is used to continuously acquire publicly available multimedia data with geotags or text descriptions from multiple social media platforms through a distributed web crawler architecture, wherein the geotags and text descriptions are associated with the reservoir. The data recognition unit is used to use the visual recognition model to identify water scene keyframes in publicly available multimedia data on the network to obtain location information. The feature analysis unit is used to analyze the texture, color and outline features of the identified floating objects through a visual recognition model to obtain the floating object type, and at the same time to locate and estimate the area of ​​the floating object cluster area using image georeferencing technology. The result integration unit is used to integrate the recognition results of the output visual recognition model and associate them with the referenced original multimedia data.

9. The intelligent generation system for reservoir scheduling schemes integrating AI Agent according to claim 7, characterized in that, The fusion analysis module includes: The data vectorization unit is used to map the floating object type to a predefined category code, use the estimated area as a continuous numerical feature, and calculate the Euclidean distance from the floating object to a preset key point as a distance feature based on the location information. The preset key point is the starting point of the reservoir operation and maintenance work. The feature matching unit is used to parse the real-time operation data of the reservoir, determine the current operation mode of the reservoir, and select feature weight vectors from the preset weight configuration library according to the operation model of the reservoir. The priority calculation unit is used to calculate the priority coefficient by performing a dot product of the category code, numerical features and distance features according to the feature weight vector, and to sort the floating object aggregation areas according to the priority coefficient; The list generation unit is used to integrate sorted floating debris accumulation areas, floating debris types, location information, estimated areas, and priority coefficients to generate a structured list of cleanup tasks.

10. The intelligent generation system for reservoir scheduling schemes integrating AI Agent according to claim 7, characterized in that, The cleanup planning module includes: The network model building unit is used to generate a directed edge-node network based on the reservoir electronic map and the location information of the floating debris accumulation area. The nodes are the ship mooring points, the locations of each floating debris and the garbage unloading points, and the directed edges are the waterways between the nodes, with weights added according to the sailing distance. The task allocation unit is used to allocate tasks in the directed edge-node network based on minimizing the total operation time and using the ship's load and task point time window as constraints. The optimal path sequence generation unit is used to calculate the optimal path sequence for each ship based on the allocation results using an improved genetic algorithm. The scheme output unit is used to integrate the optimal path sequences of all ships to generate a clear path planning scheme containing the navigation sequence.

Images