Flood risk warning method and device, computer program product and electronic equipment

Abstract

This disclosure relates to the field of flood alarm technology, providing a flood hazard alarm method and apparatus, computer program product, and electronic device. The method includes: acquiring computing task information corresponding to computing tasks generated from flood risk situation awareness data collected by monitoring stations in sub-basins of a target watershed, and computing resource information of edge nodes deployed within the target watershed, wherein each sub-basin has an edge server deployed; determining a target scheduling strategy based on the flood risk situation awareness data collected by the monitoring stations; allocating edge servers to computing tasks based on the target scheduling strategy, according to the computing task information and computing resource information; controlling each edge server to determine the execution priority of each computing task based on the flood risk situation awareness data, and processing the flood risk situation awareness data according to the execution priority to generate flood hazard alarm information for the corresponding sub-basin to which the monitoring station belongs. This solution can improve alarm efficiency.

Description

Technical Field

[0001] This disclosure relates to the field of flood alarm technology, and more specifically, to a flood hazard alarm method, a flood hazard alarm device, a computer program product, and an electronic device. Background Technology

[0002] Timely and accurate flood risk forecasting and early warning are of great significance for reducing casualties and economic losses.

[0003] Among related technologies, edge computing can be used to improve the timeliness of flood warnings. For example, the watershed to be forecasted can be divided into several sub-watersheds, and an integrated early warning device can be built in each sub-watershed as an edge node. Each edge node only processes flood monitoring data of its own sub-watershed.

[0004] However, in the above approach, each edge node has a static binding relationship with its corresponding sub-basin, and the computing resources of each edge node only serve a fixed sub-basin. When a sub-basin generates a large number of computing tasks due to extreme rainfall, its bound edge nodes may experience task backlog due to insufficient computing resources, while edge nodes in other sub-basins may be idle. This leads to an uneven utilization of overall computing resources, restricts the data processing efficiency of edge nodes, and consequently affects the timeliness of flood warnings.

[0005] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this disclosure is to provide a flood hazard warning method and device, computer program product and electronic device, thereby improving the timeliness of flood warnings to at least some extent.

[0007] Other features and advantages of this disclosure will become apparent from the following detailed description, or may be learned in part from practice of this disclosure.

[0008] According to a first aspect of this disclosure, a flood hazard alarm method is provided, comprising: controlling an alarm server to acquire computing task information corresponding to computing tasks generated from flood risk situation awareness data collected by monitoring stations in a target watershed, and computing resource information of edge nodes deployed in the target watershed, wherein the target watershed includes multiple sub-watersheds, each sub-watershed having at least one monitoring station and at least one edge node, and each edge node having an edge server deployed therein; controlling the alarm server to determine a target scheduling strategy for the computing task from candidate scheduling strategies based on the flood risk situation awareness data collected by the monitoring stations; controlling the alarm server to allocate edge servers to the computing task based on the target scheduling strategy, according to the computing task information and the computing resource information; controlling each edge server to determine the execution priority of each computing task allocated to it based on the flood risk situation awareness data, and processing the flood risk situation awareness data according to the execution priority to generate flood hazard alarm information for the sub-watershed to which the corresponding monitoring station belongs.

[0009] According to a second aspect of this disclosure, a flood hazard alarm device is provided, comprising an information acquisition module configured to control an alarm server to acquire computing task information corresponding to computing tasks generated from flood risk situation awareness data collected by monitoring stations in a target watershed, and computing resource information of edge nodes deployed within the target watershed. The target watershed includes multiple sub-watersheds, each sub-watershed having at least one monitoring station and at least one edge node, with edge servers deployed in each edge node; a scheduling strategy determination module configured to control the alarm server to determine a target scheduling strategy for the computing task from candidate scheduling strategies based on the flood risk situation awareness data collected by the monitoring stations; a computing task allocation module configured to control the alarm server to allocate edge servers to the computing task based on the target scheduling strategy, the computing task information, and the computing resource information; and an alarm module configured to control each edge server to determine the execution priority of each computing task allocated to it based on the flood risk situation awareness data, and process the flood risk situation awareness data according to the execution priority to generate flood hazard alarm information for the corresponding sub-watershed to which the monitoring station belongs.

[0010] According to a third aspect of this disclosure, a computer program product comprising instructions is provided, which, when run on a computer, causes the computer to perform the steps of the flood hazard alarm method as described in the first aspect.

[0011] According to a fourth aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the flood hazard alarm method as described in the first aspect of the above embodiments.

[0012] According to a fifth aspect of this disclosure, an electronic device is provided, comprising: a processor; and a storage device for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement the flood hazard alarm method as described in the first aspect of the above embodiments.

[0013] As can be seen from the above technical solutions, the flood hazard alarm method, flood hazard alarm device, and computer program product and electronic device implementing the flood hazard alarm method in the exemplary embodiments of this disclosure have at least the following advantages and positive effects: In some embodiments of this disclosure, the technical solutions provided, on the one hand, utilize an alarm server to schedule computational tasks within the target watershed, achieving dynamic allocation of computational tasks among multiple edge servers. When a monitoring station generates a large number of computational tasks due to extreme rainstorms, these tasks can be flexibly allocated to edge servers with lighter loads based on the real-time computational resource status of each edge server. This avoids task backlog on a single node, improves the overall balance of computational resource utilization and system throughput, and thus ensures the processing efficiency of early warning tasks in extreme scenarios. On the other hand, when an edge node fails or experiences computational resource shortages, the alarm server can allocate the computational tasks that the edge node originally needed to process to other available edge nodes, avoiding single-point failures. The failure prevented timely processing of flood monitoring data in the sub-basin, thus improving the system's fault tolerance and operational reliability. Furthermore, after receiving assigned computing tasks, each edge server determines its execution order based on the flood risk situation awareness data corresponding to each task. When multiple computing tasks arrive at the same edge server simultaneously, the task corresponding to the more urgent flood risk situation awareness data is automatically processed, reducing information delays caused by queuing for early warning computing tasks in urgent basins and improving the accuracy of early warning information. Additionally, scheduling strategies can be flexibly selected based on the basin scenario represented by the flood risk situation awareness data, simultaneously achieving rapid early warning of extreme rainstorms and efficient utilization of computing power for routine monitoring.

[0014] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0015] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure. It is obvious that the drawings described below are merely some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0016] Figure 1 A schematic diagram of an exemplary system architecture for a flood hazard alarm method applicable to embodiments of the present disclosure is shown; Figure 2 A flowchart illustrating a flood hazard alarm method according to an exemplary embodiment of this disclosure is shown. Figure 3 This diagram illustrates a flowchart of a method for determining a target scheduling strategy according to an exemplary embodiment of the present disclosure. Figure 4 A flowchart illustrating the implementation of a scheduling strategy based on a simulated annealing method in an exemplary embodiment of this disclosure is shown. Figure 5 This diagram illustrates a flowchart of a computational task allocation method based on simulated annealing, according to an exemplary embodiment of this disclosure. Figure 6 A flowchart illustrating a method for generating a second flood hazard alarm signal according to an exemplary embodiment of this disclosure is shown. Figure 7 This diagram illustrates the composition of a flood hazard alarm device according to an exemplary embodiment of the present disclosure; Figure 8 A schematic diagram of the structure of an electronic device according to an exemplary embodiment of the present disclosure is shown. Detailed Implementation

[0017] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this disclosure more comprehensive and complete, and to fully convey the concept of the example embodiments to those skilled in the art. The described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced with one or more of the specific details omitted, or other methods, components, apparatus, steps, etc., can be employed. In other instances, well-known technical solutions are not shown or described in detail to avoid obscuring various aspects of this disclosure.

[0018] The terms “a,” “an,” “the,” and “the” are used in this specification to indicate the presence of one or more elements / components / etc.; the terms “including” and “having” are used to indicate an open-ended inclusion and to mean that there may be other elements / components / etc. in addition to the listed elements / components / etc.; the terms “first” and “second” are used only as markings and are not a limitation on the number of objects.

[0019] Furthermore, the accompanying drawings are merely illustrative of this disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0020] Currently, extreme rainstorm and flood disaster forecasting is based on short-term rainfall data from meteorological bureaus. The data is shared with a disaster monitoring and early warning platform, which uses hydrological models to calculate forecast and early warning information. Finally, the wireless disaster reduction release system sends out forecast text messages, thereby reducing casualties and economic losses.

[0021] In related technologies, local data collection, calculation, and early warning are achieved through edge computing flash flood monitoring terminals. However, there is a static binding relationship between the edge nodes and the watershed, which leads to uneven utilization of computing resources and thus restricts the data processing efficiency of the edge nodes.

[0022] To address the aforementioned problems, this disclosure provides a flood emergency alarm method and apparatus, which can be applied to... Figure 1 In the system architecture of the exemplary application environment shown.

[0023] like Figure 1 As shown, the system architecture 100 may include a data acquisition layer, an edge layer, a data transmission layer, a service support layer, and a system application layer.

[0024] The data acquisition layer is the data acquisition front-end of the entire system architecture, and it includes two implementation methods: sensor collection in the flood-affected area and user upload. The former consists of various sensor devices such as hydraulic gauges, flow meters, rain gauges, and cameras, which automatically upload data at regular intervals; the latter requires system users to manually enter data in the software for uploading. Both methods work together to collect the data information needed for the system's computing tasks.

[0025] The edge layer consists of three parts: data acquisition nodes directly connected to river sensors, alarm servers deployed around the river, and edge computing servers. The data acquisition nodes are responsible for packaging and encapsulating the raw data collected by the sensors at the monitoring stations; the alarm servers are responsible for planning the execution of computing tasks; and the edge computing servers are responsible for executing the computing tasks offloaded to them.

[0026] The data transmission layer mainly transmits the encapsulated data information from the nodes to the central terminal via 4G / 5G networks and NB-IoT (Narrowband Internet of Things) networks.

[0027] The service support layer primarily supports data persistence and spatial information processing for the entire model, achieved through a database server and a map server. The database server is responsible for classifying and storing various information data, such as flood attribute information and regional sensor data, providing data support for subsequent work. The map server provides map service interfaces to the application layer, enabling functions such as loading and displaying layers.

[0028] The system application layer is the user-facing interface of the system, providing users with services such as flood information management and hazard monitoring.

[0029] In one exemplary implementation, considering the different application scenarios of users with different roles, alarm information can be sent through both desktop web applications and mobile applications. The desktop web application is primarily implemented using the SSM framework (composed of three high-performance frameworks: Spring, Spring MVC, and MyBatis), while the mobile application is developed based on Android. Spring serves as the container management framework (dependency injection, transaction control), Spring MVC is the web controller framework that handles front-end requests, and MyBatis is the database persistence framework.

[0030] In one exemplary implementation, different service content can be provided to different types of users. For example, government water resources officials can regularly collect and verify information related to extreme rainstorms and transmit flood information to mobile devices. When a flood hazard is detected in a certain area, a warning message will be pushed to the user's mobile phone. System administrators can manage flood information and system user information. In this scenario, not only is the data volume enormous, but the various operations are also quite cumbersome. Therefore, using a desktop webpage to manage system data is more convenient and efficient. For ordinary users, only limited access to webpage functions is granted, such as viewing flood warning information.

[0031] In another exemplary implementation, multiple map servers can be compared, and the map server with the best performance can be selected to build the flood warning map server. For example, a map server with strong map editing capabilities, multiple network auxiliary platforms, and support for multiple data formats can be selected.

[0032] This disclosure establishes a map server, which can intuitively display the location and environmental information of rainfall and flood monitoring stations. At the same time, it can avoid errors in the recording of rainfall and flood monitoring station information to a certain extent. In addition, 3D map modeling technology can be used later to simulate and restore extreme rainstorm and flood scenarios, providing experience for the integrated development of water conservancy and informatization.

[0033] For example, Figure 2 This illustration shows a flood hazard alarm method according to an exemplary embodiment of the present disclosure. (Reference) Figure 2 The method includes: Step S210: Control the alarm server to obtain the computing task information corresponding to the computing task generated by the flood risk situation awareness data collected by the monitoring stations in the target watershed, as well as the computing resource information of the edge nodes deployed in the target watershed. The target watershed includes multiple sub-watersheds, each of which has at least one monitoring station and at least one edge node, and the edge node has an edge server deployed in it. Step S220: Control the alarm server to determine the target scheduling strategy for the computing task from the candidate scheduling strategies based on the flood risk situation awareness data collected by the monitoring station. Step S230: Control the alarm server to allocate an edge server to the computing task based on the target scheduling strategy, the computing task information, and the computing resource information; Step S240: Control each edge server to determine the execution priority of each computing task assigned to the edge server based on the flood risk situation awareness data, and process the flood risk situation awareness data according to the execution priority to generate flood hazard alarm information for the sub-basin to which the corresponding monitoring station belongs.

[0034] exist Figure 2In the technical solution provided by the illustrated embodiment, on the one hand, the alarm server schedules the computing tasks in the target watershed, realizing the dynamic allocation of computing tasks among multiple edge servers. When a monitoring station generates a large number of computing tasks due to extreme rainstorms, its computing tasks can be flexibly allocated to edge servers with lighter loads based on the real-time computing resource status of each edge server, avoiding task backlog on a single node, improving the overall balance of computing resource utilization and system throughput, thereby ensuring the processing efficiency of early warning tasks in extreme scenarios. On the other hand, when an edge node fails or computing resources are scarce, the alarm server can allocate the computing tasks that the edge node originally needed to process to other available edge nodes, avoiding single-point failures. This resulted in the inability to process flood monitoring data in the sub-basin in a timely manner, thus improving the system's fault tolerance and operational reliability. On the other hand, after receiving the assigned computing tasks, each edge server further determines its execution order on the edge server based on the flood risk situation awareness data corresponding to each computing task. When multiple computing tasks arrive at the same edge server at the same time, the computing tasks corresponding to the more urgent flood risk situation awareness data will be automatically processed, reducing the information lag caused by the queuing of early warning computing tasks in the emergency basin and improving the accuracy of early warning information. Furthermore, the scheduling strategy can be flexibly selected according to the basin scenario represented by the flood risk situation awareness data, so as to simultaneously achieve rapid early warning of extreme rainstorms and efficient utilization of computing power for routine monitoring.

[0035] Next, a detailed description will be given of the specific implementation method of "step S210, controlling the alarm server to obtain the computing task information corresponding to the computing task generated by the flood risk situation awareness data collected by the monitoring stations in the target watershed, and the computing resource information of the edge nodes deployed in the target watershed".

[0036] In one exemplary embodiment, the target watershed may include any watershed requiring flood risk warning monitoring. The target watershed may be divided into multiple sub-watersheds, each sub-watershed deploying at least one monitoring station and at least one edge node, with an edge server deployed in each edge node. The number of sub-watersheds into which the target watershed is divided can be customized according to requirements; for example, it may be divided into three sub-watersheds: an upstream sub-watershed, a midstream sub-watershed, and a downstream sub-watershed. This exemplary embodiment does not impose any special limitation on this.

[0037] For example, AI (Artificial Intelligence) models can be deployed at river monitoring stations. Edge computing can analyze data such as water levels and rainfall in real time, and promptly issue early warning information, buying valuable time for flood control decisions. For instance, a target watershed can be divided into multiple sub-watersheds, and at least one integrated flood risk monitoring and forecasting device can be built in each sub-watershed. Each integrated flood risk monitoring and forecasting device acts as an edge node, acquiring flood risk situation awareness data for the corresponding watershed sub-unit through these devices. Edge servers can be deployed on these edge nodes, and these servers contain hydrological models. Based on the flood risk situation awareness data, these hydrological models can be used to predict flood risks.

[0038] For example, at least one alarm server can be deployed in the target watershed. Each alarm server can be pre-associated with at least one monitoring station. Each alarm server is used to schedule and allocate computing tasks generated by the monitoring station associated with it, and to determine the edge server corresponding to each computing task.

[0039] For example, monitoring stations within the same sub-basin can be associated with the same alarm server. Configuration can also be based on the communication network topology, such as associating monitoring stations connected to the same base station or local area network with the same alarm server. Alternatively, configuration can be based on administrative regions or flood control management units, such as associating monitoring stations within the same administrative region with the same alarm server. Those skilled in the art can flexibly set the association method according to actual deployment needs; this embodiment does not impose any special limitations on this.

[0040] In one exemplary implementation, each monitoring station can periodically collect flood risk situation awareness data and package and encapsulate the collected data, for example, by encapsulating it into a message containing the monitoring station identifier, collection time, data type, and data value according to a preset data format. This generates a calculation task, which is then sent to an alarm server pre-associated with the monitoring station. The calculation task can be understood as a task of calculating flood risk warning based on the flood risk situation awareness data.

[0041] The monitoring stations use different data collection frequencies during the non-flood season and the flood season. For example, during the non-flood season, flood risk situation awareness data is collected at a first collection frequency, while during the flood season, a second collection frequency is used. The first collection frequency is lower than the second collection frequency; for example, the first collection frequency is once every 10 minutes, and the second collection frequency is once every 5 minutes. Through dynamic adjustment of the collection frequency, the system operates at a lower collection frequency during the non-flood season, reducing unnecessary computing tasks and data transmission, and lowering system energy consumption and network bandwidth usage. Under extreme rainfall conditions, the system automatically increases the collection frequency to capture rapidly changing rainfall and water conditions, providing high temporal resolution input data for accurate flood forecasting.

[0042] In one exemplary implementation, the data acquisition frequencies for different sub-basins of the target watershed can be the same or different, and this exemplary implementation does not impose any particular limitation on this. For example, when extreme rainfall is detected in one sub-basin while the weather is sunny in other sub-basins, the monitoring stations in the sub-basin experiencing extreme rainfall perform high-frequency data acquisition at a second acquisition frequency, while the monitoring stations in the sub-basins where no extreme rainfall occurred continue to perform regular data acquisition at the first acquisition frequency. This can further optimize the system resource utilization efficiency and avoid unnecessary computational tasks for monitoring stations in non-emergency areas.

[0043] In one exemplary embodiment, the flood risk situation awareness data includes one or more of the following: rainfall intensity, water level rise rate, distance between the current water level at the monitoring station and a preset warning water level, and soil moisture content. Of course, the flood risk situation awareness data may also include other content, and this exemplary embodiment does not impose any special limitations on it.

[0044] In one exemplary implementation, the computation task information includes one or more of the following: the total number of computation tasks, the computational load of the computation tasks, and the data volume of the computation tasks.

[0045] The total number of computational tasks refers to the total number of computational tasks generated by all monitoring stations, such as 100 flood risk situation awareness data processing tasks and 50 water accumulation image recognition tasks. The computational load of a task can be understood as the computational resources required by the task (represented by floating-point operations in FLOPs), such as one water level trend analysis task requiring 1 × 10⁻⁶ FLOPs. 9 FLOPs, one image recognition task requires 5 × 10 9 FLOPs. The data volume of a computation task refers to the amount of data that needs to be uploaded from the monitoring site to the edge server. For example, raw sensor data is 100KB / piece, and image data is 5MB / image.

[0046] In one exemplary embodiment, the computing resource information includes edge server information for the edge node, such as one or more of the edge server's computing power, communication bandwidth, and number of servers. Of course, the computing resource information may also include other information, and this exemplary embodiment does not impose any special limitations on it.

[0047] Next, a detailed description will be given of the specific implementation method of "step S220, controlling the alarm server to determine the target scheduling strategy of the computing task from the candidate scheduling strategies based on the flood risk situation awareness data collected by the monitoring station".

[0048] In one exemplary implementation, the candidate scheduling strategy includes a first candidate scheduling strategy and a second candidate scheduling strategy. The first candidate scheduling strategy may include a Min-Min scheduling strategy improved based on a priority queue, and the second candidate scheduling strategy may include a computational task scheduling strategy based on simulated annealing.

[0049] Simulated Annealing (SA) is inspired by the principle of solid-state annealing, and its idea is based on the Monte Carlo iterative algorithm. Before the algorithm starts, the initial temperature, the number of iterations at each temperature, the temperature decay coefficient, and the termination condition need to be initialized. The temperature will decrease as the number of iterations increases until the termination condition is reached.

[0050] The Min-Min algorithm is a classic algorithm that allocates server computing resources and plans the location of unloading computing tasks in the shortest possible time, focusing on reducing latency. In this disclosure, the Min-Min algorithm is improved by establishing a task priority model, thereby enhancing the load balancing performance of the original algorithm while ensuring that higher-risk computing tasks are allocated computing resources first.

[0051] For example, Figure 3 This diagram illustrates a flowchart of a method for determining a target scheduling strategy according to an exemplary embodiment of this disclosure. (Refer to...) Figure 3 As shown, the method may include steps S310 to S330. Specifically: in step S310, the rainfall level is determined to be either Level 1 or Level 2 based on flood risk situation awareness data. If it is Level 1, the process proceeds to step S320; if it is Level 2, the process proceeds to step S330. In step S320, a first candidate scheduling strategy is determined as the target scheduling strategy. In step S330, a second candidate scheduling strategy is determined as the target scheduling strategy.

[0052] In one exemplary implementation, the rainfall intensity of the first level is greater than that of the second level.

[0053] For example, rainfall levels can be divided into Level 1 and Level 2. Level 1 represents higher rainfall intensity. For instance, the rainfall level can be determined based on rainfall intensity from direct flood risk situation awareness data. If the rainfall intensity exceeds a preset value (e.g., real-time rainfall intensity exceeds 15 mm / h), it is classified as Level 1; otherwise, it is classified as Level 2. Alternatively, the rainfall level can be determined by comprehensively considering various dimensions of data from the flood risk situation awareness data. For example, after normalizing the flood risk situation awareness data, the normalized values ​​are weighted and fused. If the weighted fusion result is greater than a preset value, the rainfall level is determined as Level 1; otherwise, it is determined as Level 2. During normalization, positive normalization can be used for rainfall intensity, water level rise rate, and soil moisture content; that is, the larger the original value, the larger the corresponding normalized value. Reverse normalization can be used for the distance between the current water level at the monitoring station and the preset warning water level; that is, the larger the original value, the smaller the corresponding normalized value.

[0054] If the flood risk situation awareness data in any computing task at a monitoring station indicates a rainfall level of Level 1, it means that the sub-basin corresponding to that monitoring station has entered a rainstorm state. The resulting computing tasks will increase significantly, and timely alarm calculations are required. Therefore, the first candidate scheduling strategy can be adopted as the target scheduling strategy, assigning higher priority to computing tasks generated by monitoring stations with higher rainfall levels. This ensures that flood calculation tasks in high-risk areas can obtain edge server resources first and complete calculations in a timely manner. If the flood risk situation awareness data in the computing tasks of all monitoring stations indicates a rainfall level of Level 2, it indicates that the system is in a normal monitoring state outside the flood season. In this case, the second candidate scheduling strategy can be adopted, aiming to minimize the maximum completion time of all edge servers. The optimal task allocation scheme is solved globally to improve the overall resource utilization efficiency of the system.

[0055] In one exemplary implementation, the first candidate scheduling strategy includes: calculating the minimum completion time of each computing task on each edge server based on computing task information and computing resource information, and a pre-established task completion time model; determining the allocation priority of each computing task based on the minimum completion time and the rainfall in the flood situation awareness data of the computing task; and allocating the computing tasks according to the allocation priority to allocate the computing tasks to the edge server corresponding to the current minimum completion time of the computing task.

[0056] For example, given three edge servers E1, E2, and E3, the completion time of each computing task on each of the three edge servers can be determined based on a pre-established task completion time model. The minimum completion time among the three is defined as the minimum completion time. The minimum completion time and rainfall for each computing task are normalized to obtain the normalized minimum completion time and normalized rainfall. A first allocation score is obtained based on the negative correlation between the normalized minimum completion time and the allocation score, and a second priority score is obtained based on the positive correlation between the normalized rainfall and the allocation score. The target allocation score is obtained by weighting the first and second allocation scores; the higher the target allocation score, the higher the allocation priority. Then, according to the descending order of allocation priority, edge servers corresponding to the current minimum completion time are assigned to each computing task sequentially. That is, computing task A has a higher priority than computing task B. After assigning computing task A to edge server E1, when assigning an edge server to computing task B, the computing resources of server E1 occupied by computing task A need to be considered to determine the current minimum completion time of computing task B.

[0057] It should be noted that when determining the allocation scores, the minimum completion time for each computing task is determined independently. However, after the priority is determined, when the edge servers are actually allocated according to the priority, the completion time of the later computing tasks on a certain edge server needs to be added to the queuing waiting time of the earlier allocated computing tasks.

[0058] In one exemplary implementation, the second candidate strategy includes: determining the task completion time of each computing task on each edge server based on the computing task information and computing resource information and a pre-established task completion time model; optimizing the initial allocation scheme by using simulated annealing method with the goal of minimizing the maximum completion time of all edge servers until the iteration termination condition is met, stopping the iteration optimization, and allocating edge servers to each computing task based on the iteration optimization results.

[0059] For example, iteratively optimizing the initial allocation scheme based on simulated annealing until the iteration termination condition is met, stopping the iteration optimization, and allocating edge servers to each computing task according to the iteration optimization results includes: randomly generating an initial allocation scheme at a preset initial temperature; determining the task completion time of each computing task on the edge servers in the allocation scheme based on the computing task information and computing resource information, and a pre-established task completion time model; obtaining the total task completion time of all computing tasks allocated to each edge server based on the sum of the task completion times of all computing tasks allocated to each edge server under the allocation scheme, and taking the maximum value of the total task completion times of each edge server as the initial system completion time. The process involves: randomly adjusting the edge servers assigned to at least one computing task in the allocation scheme to update the allocation scheme, and calculating the system completion time corresponding to the updated allocation scheme; calculating the time difference between the system completion time corresponding to the updated allocation scheme and the system completion time corresponding to the previous allocation scheme; accepting the updated allocation scheme if the time difference is less than zero, and accepting the updated allocation scheme according to a preset probability if the time difference is greater than or equal to zero, wherein the preset probability is related to the time difference and the current temperature of the simulated annealing method; lowering the temperature and repeating the above process until a preset termination temperature is reached, stopping the iteration, and assigning edge servers to each computing task according to the allocation scheme at the end of the iteration.

[0060] For example, the task completion time model is pre-built in the following way: the data transmission duration corresponding to the computing task on the edge server is determined based on the ratio between the data size required to be transmitted for the computing task and the data transmission rate between the monitoring station and the base station connected to the edge server; the computing task execution duration corresponding to the computing task on the edge server is determined based on the ratio between the computing amount of the computing task and the computing power of the edge server; and the task completion time model is obtained based on the data transmission duration and the computing task execution duration.

[0061] For example, a task completion time model can be pre-established, and by substituting computation task information and server resource information, the completion time of any computation task on a certain edge server can be calculated. The runtime of the edge server is the time to complete all computation tasks unloaded there. The task completion time model is a core tool connecting task requirements and resource capabilities. It needs to comprehensively consider communication latency, queuing latency, computation latency, and result return latency to accurately calculate the "total completion time of a task unloaded to a certain edge server" and the "total runtime of the server" (i.e., the longest completion time for the server to process all unloaded tasks, which determines the overall efficiency of the system).

[0062] Each monitoring station periodically generates a computational task with varying computational requirements. Several edge servers with different computing capabilities are connected to the monitoring station. The computational task is offloaded from the monitoring station to the connected edge servers for calculation. The task completion time is divided into three parts: task data transmission time, task execution time on the edge servers, and the time for the computation results to be uploaded from the edge servers to the central server.

[0063] For task data transmission time The wireless data transmission rate between the monitoring station and the edge server connecting the base station B j and the size of data transmitted by the computing task d i The decision, for example, is determined by the following formula (1): (1) For calculating task execution time The amount of computation required for the task c i and the computing power of edge computing servers The decision can be made, for example, by formula (2): (2) Regarding the time for uploading the calculation results to the central terminal, since the amount of data uploaded by the monitoring sites connected to the edge server is relatively small, this disclosure chooses to ignore this part of the time.

[0064] The task completion time is determined by the task data transmission time. Computational task execution time Together, it can be represented as the following formula (3): (3) in Represents computational task Task i Uninstall location , That is, to represent the computational task Task i In the number M If the edge server performs the calculation, then the total time for any edge server to complete the unloading task (i.e., the running time of the edge server) is as follows (4): (4) Therefore, the system task completion time is the longest running time among all edge servers, as shown in the following formula (5): (5) in, Indicates that in the number j All computing tasks run on the edge servers.

[0065] In other words, taking a total of 3 edge servers as an example, such as E1, E2, and E3, if computing tasks 1, 2, and 3 are assigned to E1, computing tasks 4, 7, and 8 are assigned to E2, and computing tasks 5, 6, 9, and 10 are assigned to E3, then the running time of E1 is the sum of the completion times of computing tasks 1, 2, and 3; the running time of E2 is the sum of the completion times of computing tasks 4, 7, and 8; and the running time of E3 is the sum of the completion times of computing tasks 5, 6, 9, and 10. The system task completion time is the maximum value among the running times of E1, E2, and E3.

[0066] In the scheduling strategy based on simulated annealing, in order to minimize the task completion time, the scheduling problem is transformed into the objective function solution problem shown in the following formula (6), that is, minimizing the maximum task completion time among all edge servers: (6) For example, Figure 4 This diagram illustrates a scheduling strategy implementation flowchart based on a simulated annealing method according to an exemplary embodiment of this disclosure. (See reference...) Figure 4 The method may include steps S410 to S430, wherein: in step S410, computing task information and computing resource information are obtained; in step S420, a task completion time model is established; and in step S430, the unloading location of the computing task is planned based on the simulated annealing method.

[0067] The computational task information and computational resource information in step S410 can be referred to the foregoing explanation, and will not be repeated here. The establishment of the task completion time model in step S420 can also be referred to the relevant foregoing explanation, and will not be repeated here.

[0068] For simulated annealing (SA) algorithm to plan unloading locations, it can be understood as global optimization and local exploration. SA is a probabilistic optimization algorithm that excels at solving combinatorial optimization problems (such as task-server allocation) and effectively avoids getting trapped in local optima (such as situations where "nearby servers are resource-constrained, but distant servers have surplus resources"). Its core logic balances "exploring new solutions" and "converging to the optimal solution" through a "random perturbation-probabilistic acceptance" mechanism, adapting to the needs of scenarios with "surge in task volume and dynamic resource changes" under extreme rainstorms.

[0069] For example, computational tasks are randomly assigned to edge servers. The runtime of each edge server is calculated based on a task completion time model, and the maximum runtime across all servers is taken as the system task completion time. The task allocation locations are then randomly adjusted, and the task completion time under the new allocation method is calculated. The new solution is accepted according to the Metropolis acceptance criterion, and the algorithm temperature is reduced. This iteration is repeated until the termination condition is met, and the optimal unloading location for the task is output. The Metropolis criterion is an important sampling method that accepts new solutions probabilistically, rather than based on completely deterministic rules.

[0070] Assume the following: End nodes: 100 water level sensors (each generates 1 task, computational load 1×10^6) 9 FLOPs (100KB of data transferred), 50 water-filled cameras (each generating one task, computational load 5×10⁻⁶). 9 FLOPs (transmitted data volume 5MB); Edge servers: 5 (each server has a peak computing capacity of 20 TFLOPs, and the current remaining computing capacities are 15, 12, 10, 8, and 5 TFLOPs respectively; base station uplink bandwidth is 100 Mbps, and downlink bandwidth is 200 Mbps); The simulated annealing algorithm first performs initial allocation during computation: after random allocation, the system time Tsystem = 120s (a server has a load rate of 0.95 due to being assigned too many camera tasks, resulting in large queuing delays); Next, simulated annealing optimization is performed: the initial temperature T0 = 100, and the cooling coefficient = 0.9; after 100 iterations, the temperature drops to The final optimal solution was obtained: some camera tasks were transferred from the overloaded server to the server with sufficient remaining computing power, and the system time was reduced to Tsystem=45S (meeting the "minute-level early warning" requirement).

[0071] For example, Figure 5 This diagram illustrates a flowchart of a computational task allocation method based on simulated annealing, according to an exemplary embodiment of this disclosure. (See reference...) Figure 5The method may include steps S501 to S510. Specifically: In step S501, an initial allocation scheme is randomly generated; in step S502, the first system completion time in the initial allocation scheme is determined based on a task completion time model; in step S503, a perturbation is generated to update the allocation scheme; in step S504, the second system completion time of the updated allocation scheme is calculated; in step S505, the difference between the second system completion time and the first system completion time is calculated; in step S506, it is determined whether the difference is less than 0. If it is less than 0, proceed to step S507; otherwise, proceed to step S508; in step S507, the updated allocation scheme is accepted; in step S508, the updated allocation scheme is accepted according to the Metropolils criterion; in step S509, it is determined whether the iteration number has been reached at a certain temperature. If not, proceed to step S503; then proceed to step S510, lower the algorithm temperature, and determine whether the minimum temperature has been reached. If yes, the process ends; otherwise, proceed to step S503.

[0072] For example, the process of optimizing the task unloading position based on the SA algorithm in this disclosure, i.e., finding the optimal solution of the above objective function, is implemented as follows: Initialize the algorithm: Set the initial temperature of the algorithm. Termination temperature Temperature decay coefficient and the number of iterations at each temperature And assume that there exists N Multiple rainfall and flood monitoring stations simultaneously generate a computational task with different computational information, producing a task computational workload matrix. With the matrix of transmitted data volume The formulas (7) and (8) are shown below, respectively.

[0073] (7) (8) Generate a data transmission time model matrix by combining the above content. Task execution time model matrix The results are shown in equations (9) and (10) respectively.

[0074] (9) (10) A random function is used to generate one. N M The 0-1 matrix is ​​denoted as the initial solution vector matrix. The solution vector matrix is ​​defined therein. ,in, , Its form is shown in equation (11), vector matrix This indicates the calculation method and task unloading location for all devices. Solution vector. =1, 2, 3…, M Indicates whether the computing task is selected to be performed on edge computing servers 1, 2... M Execute above. Represents computational task This method will not be selected for execution. Represents computational task Select this calculation method to perform the calculation.

[0075] (11)

[0076] Equation (11) above indicates that task 1 runs on edge server number 1, task 2 runs on edge server number 2, task 3 runs on edge server number 5, and tasks 4 and 5 run on edge server number 4.

[0077] By combining relevant linear algebra concepts and formulas, the completion time span matrix can be obtained. As shown in equation (12).

[0078] (12)

[0079] in Indicates the first i The runtime of each server completing the unloading of all tasks to this point is: Maximum value in the matrix System task completion time Simultaneously, randomly swap the original solution vector matrix. The positions of the three elements are used to generate a new solution vector matrix. And calculate the task completion time under the solution matrix. And calculate the difference in completion time, as shown in equation (13): (13) like Then the new interpretation must be accepted. ,like Accepting new solutions according to the Metropooils criterion That is, based on probability Receive new solutions (T is the current temperature). Determine whether the current iteration has reached the upper limit of the number of iterations at this temperature. If the upper limit of the number of iterations has not been reached, repeat the position swap and calculation of the completion time. If the upper limit of the number of iterations at this temperature has been reached, determine whether the specified minimum temperature has been reached. If the minimum temperature has not been reached, reduce the current temperature according to the following formula (14), reset the number of iterations, and repeat the above iteration process. If the specified minimum temperature has been reached, output the current solution vector matrix and the algorithm ends.

[0080] (14)

[0081] Next, a detailed description will be given of the specific implementation method of "step S230, controlling the alarm server to allocate an edge server to the computing task based on the target scheduling strategy, according to the computing task information and the computing resource information".

[0082] For example, the specific implementation of step S230 can refer to the relevant content of step S220 above (i.e. how to allocate the content of the edge server when the first candidate scheduling policy is the target scheduling policy and the second candidate scheduling policy is the target policy), and will not be repeated here.

[0083] Next, a detailed description will be given of the specific implementation method of "step S240, controlling each edge server to determine the execution priority of each computing task assigned to the edge server based on the flood risk situation perception data, and processing the flood risk situation perception data according to the execution priority to generate flood hazard alarm information for the sub-basin to which the corresponding monitoring station belongs".

[0084] For example, controlling each edge server to determine the execution priority of each computing task assigned to the edge server based on the flood risk situation awareness data includes: obtaining a target priority score by weighted combination of priority scores corresponding to each flood risk situation awareness data; determining the execution priority of each computing task assigned to the edge server according to the descending order of the target priority scores, wherein the higher the target priority score, the higher the execution priority; wherein the rainfall intensity is positively correlated with the priority score, the water level rise rate is positively correlated with the priority score, the distance between the current water level of the monitoring station and the preset warning water level is negatively correlated with the priority score, and the soil moisture content is positively correlated with the priority score.

[0085] For example, each data point in the flood risk situation awareness data can be normalized to obtain a corresponding normalized value. Then, based on the correlation between each flood risk situation awareness data point and its priority score, the priority score corresponding to each normalized value is determined. The priority scores corresponding to each normalized value are then weighted to obtain a target priority score. The weighting coefficients can be determined based on requirements or the degree of impact of each flood risk situation awareness data point on flood disasters; this exemplary implementation does not impose any special limitations on this. A higher target priority score indicates a higher computational priority for the task on the corresponding edge server, thus ensuring that urgent tasks are executed first on the edge server.

[0086] In one exemplary implementation, the flood risk situation awareness data may further include rainfall intensity and hydrological model environmental parameters, which include one or more of topographic parameters, runoff generation and runoff parameters, and receding parameters.

[0087] The terrain parameters include: vegetation index, soil moisture content, and average elevation. Runoff generation and runoff parameters include: water storage capacity, water storage capacity curve index, and the ratio of impervious area to the area of ​​the watershed to be predicted. Recession parameters include: slope and roughness.

[0088] Rainfall intensity can include multi-source rainfall intensity, which can include rainfall data sources based on rain radar, short-term rainfall data sources from meteorological bureaus, and gridded rainfall data sources. The multi-source rainfall data can be fused using the Bayesian averaging method to obtain a multi-source rainfall data fusion value, and flood risk prediction and warning can be made based on this fusion value.

[0089] For example, processing the flood risk situation awareness data according to the execution priority to generate flood hazard alarm information for the sub-basin to which the corresponding monitoring station belongs includes: determining the outlet section runoff process calculation unit of the sub-basin based on the rainfall intensity, the rainfall intensity based on the sub-basin, the terrain parameters, the runoff generation and confluence parameters, and the dissipation parameters; obtaining the alarm index value of the sub-basin based on the outlet section runoff process calculation unit, wherein the alarm index value includes at least one of the predicted water level value and the predicted flow rate value; determining that there is a flood risk in the sub-basin when the alarm index value of the sub-basin is greater than a first preset threshold, and generating a first flood hazard alarm signal for the sub-basin; and controlling the alarm device in the edge node to send the first flood hazard alarm signal to the target user.

[0090] For example, the runoff value of a watershed sub-unit can be calculated based on its topographic parameters and runoff generation parameters. The runoff direction of the sub-unit can be determined by comparing the average elevation of adjacent sub-units, and the corresponding recession coefficient can be obtained based on the recession parameters. The runoff process at the corresponding outlet section can be obtained from the multi-source stormwater fusion values, runoff values, runoff direction, and recession coefficients of each sub-unit, thus yielding the predicted water level and flow threshold for the sub-watershed. If either of these exceeds its corresponding preset threshold, a flood risk is identified in the sub-watershed, generating a single-point hazard alarm signal, i.e., the first flood hazard alarm signal. This first flood hazard alarm signal can be sent to the corresponding user's terminal via the integrated flood hazard monitoring and forecasting equipment at the sub-watershed's edge nodes, thereby issuing a flood hazard warning.

[0091] In one exemplary implementation, the early warning threshold can be automatically adjusted according to the flood season / non-flood season to adapt to extreme rainstorm scenarios in different river basins and seasons.

[0092] For example, Figure 6 This diagram illustrates a flowchart of a method for generating a second flood hazard alarm signal according to an exemplary embodiment of this disclosure. (See reference...) Figure 6 The method may include steps S610 to S650. Wherein: In step S610, if the alarm index value of the sub-basin is less than or equal to the first preset threshold, the sub-basin is identified as a candidate sub-basin. In step S620, based on the historical water level data and / or historical flow data of multiple sub-basins within the target basin, the water level correlation and / or flow correlation between each sub-basin are determined. In step S630, based on the water level correlation and / or flow correlation, combined with the current water level data and / or current flow data of each sub-basin, the predicted water level value of the candidate sub-basin is determined. In step S640, based on the difference between the predicted water level value and the measured current water level value of the candidate sub-basin, the collaborative risk score of the candidate sub-basin is determined. In step S650, if the collaborative risk score is greater than the second preset threshold and the duration of the collaborative risk score being greater than the second preset threshold exceeds a preset duration, a second flood hazard alarm signal for the candidate sub-basin is generated.

[0093] For example, hourly synchronous water level monitoring sequence data with a unified time base can be obtained in advance during typical rainstorm and flood events in the target watershed over the past 5 years. This data, sampled at hourly intervals, covers the entire stage of flood rise, peak, and recede, with all station collection timestamps corresponding one-to-one, used to calibrate the hydrological transmission relationship between sub-watersheds. Based on this historical data, the central server analyzes the correspondence between upstream and downstream sub-watershed water levels, calibrating the fixed lag time for upstream water level changes to be transmitted downstream, i.e., how long it takes for a change in water level upstream to be reflected downstream; secondly, it analyzes the numerical patterns of downstream water levels corresponding to different upstream water levels to verify the existence of a stable and strong correlation between the two.

[0094] Based on historical flood data, the water level response equations of the upstream and downstream sub-basins were obtained by linear regression fitting, with "water level value of the upstream sub-basin" as the independent variable and "water level value of the downstream sub-basin after a fixed lag time" as the dependent variable.

[0095] For example, for any sub-basin, determine the set of upstream sub-basins. Let the sub-basin be D, and its directly upstream sub-basins be A1, A2, ..., A... n For each upstream sub-basin A i Historical synchronous water level monitoring sequences (with unified time reference) of the upstream and downstream sub-basins D were obtained, and A was determined through cross-correlation analysis. i The fixed lag time Δt for water level changes to be transmitted to D i The central server uses the water level of D at time t as the dependent variable, and the water level of each upstream sub-basin at (t-Δt) as the dependent variable. i The water level at time ( ) is used as the independent variable to establish a multiple linear regression equation, thereby obtaining the water level response equation.

[0096] During the real-time prediction phase, the central server obtains the current water level values ​​of each upstream sub-basin, substitutes them into the water level response equation mentioned above, and calculates the predicted water level value of candidate sub-basin D.

[0097] After obtaining the predicted water level values ​​for candidate sub-basins, the difference between the predicted water level and the current actual water level can be calculated. This difference represents the additional rise in water level after upstream flooding is transmitted downstream, and is the core source of cross-basin collaborative risk. Based on this, a collaborative risk score is calculated by combining the remaining safe distance between the current water level and the warning level of the candidate sub-basin. That is, the collaborative risk score is the ratio of the aforementioned difference to the remaining safe distance between the current water level and the warning level. The collaborative risk score is used to measure the impact of upstream sub-basins on the remaining safety controls of candidate sub-basins. The remaining safe distance is the difference between the warning level and the current measured water level of the candidate sub-basin.

[0098] A higher collaborative risk score indicates a larger proportion of the safe space occupied by upstream water, and a higher risk of lag due to cross-basin transmission. A score greater than 1 indicates that upstream water will directly push the downstream water level above the warning level, with higher scores indicating a greater extent of exceeding the warning level. A score less than 1 indicates that the downstream water level has not yet exceeded the warning level after upstream water arrives, but the closer the score is to 1, the more imminent the risk. The second preset value can be a value less than 1 and greater than 0.5, such as 0.8, which means that when upstream water occupies more than 80% of the remaining safe space, the candidate sub-basin is judged to have potential collaborative risks. At the same time, a preset duration (such as 15 minutes) is set to filter false alarms caused by instantaneous data fluctuations.

[0099] For example, candidate sub-basins can be continuously monitored. When the collaborative risk score exceeds a second preset threshold for multiple consecutive monitoring periods, the sub-basin is determined to have an upstream-transmitted delayed flood risk, and a second flood hazard alarm signal can be generated. This alarm signal is marked as an upstream delayed risk warning and includes information such as the expected rise in water level, the expected peak arrival time, the predicted highest water level, and the risk level. It is pushed to basin operation and maintenance personnel and responsible units in riverside villages and towns, achieving an earlier warning than the local first flood hazard alarm signal, extending the preparation time for the evacuation and shelter of the masses. In other words, when the sub-basin itself has not yet triggered an alarm (water level is normal), the system can identify delayed risks from upstream in advance based on the upstream-downstream relationship, buying valuable emergency response time for the downstream.

[0100] In one exemplary implementation, the first flood hazard alarm signal is generated in real time by edge computing nodes deployed locally in each sub-basin based on local monitoring data. The second flood hazard alarm signal is generated by the central server of the basin management end based on real-time data and historical hydrological transmission relationships of multiple sub-basins in the whole basin, and sent to the edge computing nodes of the corresponding candidate sub-basins to execute alarm output.

[0101] In this disclosure, two watershed scenarios were selected to test the method of this disclosure: a small watershed in a mountainous area and a watershed in a plain. The test data for the two watersheds were constructed by fusing data collected from field monitoring, public datasets of multi-source remote sensing, and historical measured flood data.

[0102] For collecting on-site monitoring data, integrated hydrological monitoring stations are deployed within the watershed, equipped with rain gauges, water level gauges, soil moisture sensors, and flow velocity monitoring equipment. These stations continuously collect measured data on extreme rainfall events during the flood season, specifically including: minute-by-minute rainfall intensity, hourly cumulative rainfall, real-time river water level, water level rise rate, surface soil moisture content, and cross-sectional flow velocity and flow rate. High-precision measured data within the watershed, such as rainfall data, continuous monitoring data on river and tributary water levels and flow rates, and monitoring data on vegetation cover in cultivated land / water areas, can be obtained through a comprehensive national hydrological station and automatic weather station monitoring network. Simultaneously, hardware resource data such as uplink bandwidth of edge base stations, real-time remaining computing power of edge servers, and data transmission latency are collected to calculate task latency.

[0103] To address the issues of sparse monitoring stations and data blind spots in mountainous areas, publicly available remote sensing datasets, such as the GPM (Global Precipitation Measurement) dataset and the MODIS (Moderate Resolution Imaging Spectroradiometer) vegetation remote sensing dataset, are used to supplement the watershed's overall data. This data is then used to calculate flood situation awareness data, such as vegetation indices, and to retrieve watershed soil moisture content, supporting experiments on multi-source rainfall fusion and underlying surface parameter calibration.

[0104] For historical flood data, the data compiled from the past 10 years of extreme rainstorms and flash floods published by the local water resources department is used. This data includes typical flood process lines, peak flow, spatiotemporal distribution of rainstorms, and records of watershed emergencies. These data serve as the basis for setting alarm thresholds, ensuring that the test results accurately reflect real flash flood disaster scenarios.

[0105] For example, based on the above two types of real monitoring data of the watershed, flood early warning calculation tasks with corresponding quantity, data volume and computational volume are generated. The traditional static binding scheduling scheme (unscheduled) and the dual-strategy dynamic scheduling scheme of this disclosure (simulated annealing algorithm and priority scheduling) are run in a real edge server cluster. The time consumption of each task, the maximum system latency and the alarm generation time are recorded throughout the process. The test results are shown in Table 1 below.

[0106] Table 1 Comparison of task completion times before and after scheduling.

[0107] As can be seen from Table 1, the method in this disclosure can optimize the CPU utilization of edge nodes by 60%-80%, avoid resource overload or idleness, and shorten the task completion time by 40%-60% compared with existing edge computing terminals, thus ensuring timely early warning.

[0108] This disclosure improves the timeliness of flood risk warnings based on an AI edge computing architecture. Data is processed locally through distributed sensor networks (such as rain gauges, water level gauges, cameras, and weather stations deployed at key locations in the watershed) and edge gateways and microservers (such as high-performance edge computing devices deployed at pumping stations, hydrological monitoring stations, or urban drainage nodes). Structures such as LSTM (Long Short-Term Memory) and Temporal Convolution Network (TCN) can be used to capture the long- and short-term dependencies of hydrological sequences. Graph Neural Networks (GNNs) are used to improve the spatial generalization ability of predictions, considering river network topology and spatial correlations. Simultaneously, a federated learning mechanism is used. Through collaborative training between the center and edge nodes, model parameters are continuously optimized while ensuring data privacy. When data distribution drift is detected (such as extreme rainfall), edge nodes can automatically trigger model retraining and adjust the prediction frequency based on real-time monitoring, ensuring efficient use of computing power and energy.

[0109] For example, this disclosure provides a method for determining dynamic early warning thresholds based on multi-dimensional analysis. By analyzing historical flood data, monitoring personnel can establish a database of characteristic values ​​for floods with different return periods using this system. Considering the impacts of watershed underlying surface changes and climate change, a dynamic correction factor is introduced to adapt the early warning thresholds to environmental changes. Based on this, a three-tiered early warning threshold system is established: Level 1 early warning is based on single-site monitoring data, setting early warning thresholds for single elements such as water level and flow; Level 2 early warning is based on multi-site collaborative analysis, considering the correlation between upstream and downstream hydrological stations, setting early warning thresholds at the watershed scale; Level 3 early warning is based on comprehensive multi-element analysis, incorporating factors such as rainfall, snowmelt, and temperature to establish a comprehensive early warning indicator system. The determination of the early warning thresholds is achieved through a fuzzy comprehensive evaluation method, combined with expert experience and data analysis, to improve the scientific validity and reliability of the thresholds.

[0110] In one exemplary implementation, a multi-layered, comprehensive data acquisition network is established for data collection. At the watershed level, data such as rainfall, water level, flow rate, wind speed, and wind direction are collected from hydrological and meteorological stations. At the engineering site, sensors such as strain gauges, piezometers, and displacement gauges are used to monitor dam stress, seepage, deformation, and other structural parameters in real time. Regarding equipment and facilities, vibration sensors and temperature sensors are used to collect data on the operational status of electromechanical equipment. For data transmission, IoT technology is used to automatically collect and upload sensor data. For remote areas or regions with poor communication conditions, satellite communication ensures effective data transmission. Edge computing technology is employed to perform preliminary processing of raw data at the data source, reducing the volume of data transmission, improving transmission efficiency, and ensuring that the data accurately, precisely, and rapidly reflects the engineering's operational status.

[0111] In the risk identification phase, risk identification models can be built using historical risk event data and real-time monitoring data, employing machine learning algorithms. For example, training can be conducted using a large amount of historical data to establish a mapping relationship between risk characteristics and operational data, enabling automatic diagnosis of potential hazards; support vector machines can classify and identify risk states by finding the optimal classification hyperplane. Regarding risk assessment, a dynamic risk assessment indicator system can be created, selecting assessment indicators from two dimensions: the probability of risk occurrence and the severity of its consequences.

[0112] This disclosure also utilizes AI algorithms to perform real-time analysis of video streams and sensor data, detecting abnormal behaviors or states and determining thresholds in real time. During the regular non-flood season, to improve the timeliness of risk warnings, it is necessary to reduce the completion time of computational tasks generated by nodes at specific times, so as to promptly identify potential hazards at rainfall and flood monitoring stations. A computational task completion time model can be established based on parameters such as computational task information and edge server computing capabilities. This model abstracts the scenario problem into a solvable objective function, and simulated annealing is used to allocate computational tasks, improving the overall performance of the alarm system. During the flood season, rainfall and flood monitoring stations in remote areas face the problem of inaccurate data. To further improve data quality and reduce the impact of flood accidents, computational tasks at higher-risk rainfall and flood monitoring stations need to be prioritized. Therefore, this disclosure improves the traditional Min-Min algorithm by incorporating priority. First, rainfall information is classified into levels, and the minimum task completion time is calculated. Second, the rainfall level and the minimum task completion time are normalized and weighted summed to form a new priority computational task queue. Finally, the priority computational task queue is scheduled according to the Min-Min algorithm's scheduling process.

[0113] For example, in this disclosure, the message push module of the rainfall and flood monitoring station and the early warning system can be used to send the emergency information to the relevant person in charge as soon as possible when an emergency occurs in the basin.

[0114] In this system, the person in charge is responsible for subscribing to messages from the central server, calculating the regional range ID, constructing subscription constraints, and sending them to the proxy server. The central server is responsible for constructing the hazard information and sending it to the proxy server. The proxy server is responsible for connecting the publisher and subscriber, receiving information from both ends, matching it, and then pushing the hazard information to system users.

[0115] Furthermore, it should be noted that the above figures are merely illustrative representations of the processes included in the method according to exemplary embodiments of the present invention, and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Additionally, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0116] Furthermore, an exemplary embodiment of this disclosure also provides a flood hazard alarm device. (See reference...) Figure 7 As shown, the flood hazard alarm device 700 includes the following program modules: an information acquisition module 710, configured to control the alarm server to acquire computing task information corresponding to the computing tasks generated by the flood risk situation awareness data collected by the monitoring stations in the target watershed, and computing resource information of edge nodes deployed in the target watershed. The target watershed includes multiple sub-watersheds, each of which has at least one monitoring station and at least one edge node, and the edge node has an edge server deployed in it; and a scheduling strategy determination module 720, configured to control the alarm server to determine the scheduling strategy based on the flood risk situation awareness data collected by the monitoring stations. According to the data, a target scheduling strategy for the computing task is determined from the candidate scheduling strategies; the computing task allocation module 730 is configured to control the alarm server to allocate edge servers to the computing task based on the target scheduling strategy, according to the computing task information and the computing resource information; the alarm module 740 is configured to control each edge server to determine the execution priority of each computing task allocated to the edge server according to the flood risk situation awareness data, and process the flood risk situation awareness data according to the execution priority to generate flood hazard alarm information for the sub-basin to which the corresponding monitoring station belongs.

[0117] In one exemplary embodiment, the candidate scheduling strategy includes a first candidate scheduling strategy and a second candidate scheduling strategy. The control of the alarm server to determine the target scheduling strategy for the computation task from the candidate scheduling strategies based on flood risk situation awareness data collected by the monitoring station includes: determining the first candidate scheduling strategy as the target scheduling strategy when the flood risk situation awareness data indicates a rainfall level of level one; and determining the second candidate scheduling strategy as the target scheduling strategy when the flood risk situation awareness data indicates a rainfall level of level two; wherein the rainfall intensity of level one is greater than that of level two; wherein the first candidate scheduling strategy includes: calculating each... The first candidate strategy involves determining the minimum completion time of a computation task on each edge server, and then, based on the minimum completion time and the rainfall data in the flood situation awareness data of the computation task, determining the allocation priority of each computation task. The second candidate strategy includes: determining the completion time of each computation task on each edge server based on the computation task information and computation resource information, using a pre-established task completion time model; optimizing the initial allocation scheme by minimizing the maximum completion time of all edge servers, using simulated annealing until the iteration termination condition is met, stopping the iteration optimization, and allocating edge servers to each computation task based on the iteration optimization results.

[0118] In one exemplary implementation, the step of iteratively optimizing the initial allocation scheme based on simulated annealing until the iteration termination condition is met, stopping the iteration optimization, and allocating edge servers to each computing task according to the iteration optimization results includes: randomly generating an initial allocation scheme at a preset initial temperature; determining the task completion time of each computing task on the edge servers in the allocation scheme based on the computing task information and computing resource information, and a pre-established task completion time model; obtaining the total task completion time of all computing tasks allocated to each edge server based on the sum of the task completion times of all computing tasks allocated to each edge server under the allocation scheme, and using the maximum value of the total task completion times of each edge server as the initial system... The system completion time is calculated; the edge servers allocated to at least one computing task in the allocation scheme are randomly adjusted to update the allocation scheme, and the system completion time corresponding to the updated allocation scheme is calculated; the time difference between the system completion time corresponding to the updated allocation scheme and the system completion time corresponding to the previous allocation scheme is calculated; if the time difference is less than zero, the updated allocation scheme is accepted; if the time difference is greater than or equal to zero, the updated allocation scheme is accepted according to a preset probability, wherein the preset probability is related to the time difference and the current temperature of the simulated annealing method; the temperature is reduced, and the above process is repeated until a preset termination temperature is reached, the iteration is stopped, and edge servers are allocated to each computing task according to the allocation scheme at the end of the iteration.

[0119] In one exemplary implementation, the task completion time model is pre-constructed as follows: the data transmission duration corresponding to the computing task on the edge server is determined based on the ratio between the data size required to be transmitted for the computing task and the data transmission rate between the monitoring station and the base station connected to the edge server; the computing task execution duration corresponding to the computing task on the edge server is determined based on the ratio between the computing amount of the computing task and the computing power of the edge server; and the task completion time model is obtained based on the data transmission duration and the computing task execution duration.

[0120] In one exemplary implementation, the flood risk situation awareness data includes one or more of the following: rainfall intensity, water level rise rate, distance between the current water level of the monitoring station and the preset warning water level, and soil moisture content. The step of controlling each edge server to determine the execution priority of each computational task assigned to it based on the flood risk situation awareness data includes: obtaining a target priority score by weighted combination of priority scores corresponding to each flood risk situation awareness data point; determining the execution priority of each computational task assigned to the edge server according to the descending order of the target priority scores, wherein a higher target priority score corresponds to a higher execution priority; wherein the rainfall intensity is positively correlated with the priority score, the water level rise rate is positively correlated with the priority score, the distance between the current water level of the monitoring station and the preset warning water level is negatively correlated with the priority score, and soil moisture content is positively correlated with the priority score.

[0121] In one exemplary implementation, the flood risk situation awareness data includes rainfall intensity and hydrological model environmental parameters, the hydrological model environmental parameters including topographic parameters, runoff generation and runoff parameters, and recession parameters; the step of processing the flood risk situation awareness data according to the execution priority to generate flood hazard alarm information for the sub-basin to which the corresponding monitoring station belongs includes: determining the outlet section runoff process calculation unit of the sub-basin based on the rainfall intensity, topographic parameters, runoff generation and runoff parameters, and recession parameters of the sub-basin according to the execution priority; obtaining the alarm index value of the sub-basin according to the outlet section runoff process calculation unit, the alarm index value including at least one of water level prediction value and flow prediction value; if the alarm index value of the sub-basin is greater than a first preset threshold, determining that there is a flood risk in the sub-basin, and generating a first flood hazard alarm signal for the sub-basin; controlling the alarm device in the edge node to send the first flood hazard alarm signal to the target user.

[0122] In one exemplary embodiment, the device further includes a second alarm signal generation module, which can be configured to: determine the sub-basin as a candidate sub-basin when the alarm index value of the sub-basin is less than or equal to the first preset threshold; determine the water level correlation and / or flow correlation between each sub-basin based on historical water level data and / or historical flow data of multiple sub-basins within the target basin; determine the predicted water level value of the candidate sub-basin based on the water level correlation and / or flow correlation, combined with the current water level data and / or current flow data of each sub-basin; determine the collaborative risk score of the candidate sub-basin based on the difference between the predicted water level value and the measured current water level value of the candidate sub-basin; and generate a second flood hazard alarm signal for the candidate sub-basin when the collaborative risk score is greater than a second preset threshold and the duration of the collaborative risk score being greater than the second preset threshold exceeds a preset duration.

[0123] The specific details of each part of the above-mentioned device have been described in detail in the method section of the implementation plan. For any undisclosed details, please refer to the implementation plan of the method section, and therefore will not be repeated here.

[0124] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to exemplary embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0125] Furthermore, although the steps of the method in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0126] Exemplary embodiments of this disclosure also provide a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the aforementioned flood hazard alarm method.

[0127] In one implementation, the computer program product can be a tangible product containing a computer program, such as a computer-readable storage medium storing the computer program. The readable storage medium can be a storage medium based on electrical, magnetic, optical, electromagnetic, infrared, or other signals, including but not limited to: random access memory (RAM), read-only memory (ROM), magnetic tape, floppy disk, flash memory, hard disk drive (HDD), solid-state drive (SSD), etc. For example, the computer program product can be implemented as a non-volatile storage medium storing a computer program, such as read-only memory, NAND flash memory, etc.

[0128] In one implementation, the computer program product can be an intangible product containing a computer program. For example, the computer program product can be implemented as a virtual digital product, such as an executable file, installation package, or other digital file storing the computer program.

[0129] Computer program code can be written in one or more programming languages. Examples of programming languages ​​include C, Java, C++, and Python. Program code can execute entirely on the user's computing device, partially on the user's computing device, or as a standalone software package. It can also execute partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, such as a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via an internet connection provided by a mobile network operator).

[0130] Computer programs can be carried or transmitted via signals such as electricity, magnetism, light, electromagnetic fields, and infrared radiation. Electronic devices can convert signals carrying computer programs into digital signals, thereby running the computer programs. When a computer program runs on an electronic device, its code is used to cause the electronic device to execute (more specifically, to execute by the processor of the electronic device) the method steps of various exemplary embodiments of this disclosure, such as the aforementioned flood hazard alarm method.

[0131] Exemplary embodiments of this disclosure also provide an electronic device, which may include a processor and a memory. The memory stores executable instructions of the processor, such as a computer program. The processor executes the executable instructions to perform the method steps of various exemplary embodiments of this disclosure. Furthermore, the electronic device may also include a display for displaying a graphical user interface.

[0132] The following is for reference. Figure 8 The electronic device is illustrated by way of a general-purpose computing device. It should be understood that... Figure 8The electronic device 800 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0133] like Figure 8 As shown, the electronic device 800 may include: a processor 810, a memory 820, a bus 830, an I / O (input / output) interface 840, a network adapter 850, and a display 860.

[0134] The memory 820 may include volatile memory, such as RAM 821 and cache unit 822, and may also include non-volatile memory, such as ROM 823. The memory 820 may also include one or more program modules 824, including but not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. For example, program module 824 may include the modules described above.

[0135] The processor 810 may include one or more processing units, such as an AP (Application Processor), a modem processor, a GPU (Graphics Processing Unit), an ISP (Image Signal Processor), a controller, an encoder, a decoder, a DSP (Digital Signal Processor), a baseband processor, and / or an NPU (Neural-Network Processing Unit).

[0136] The processor 810 can be used to execute executable instructions stored in the memory 820, such as the flood hazard alarm method described above.

[0137] Bus 830 is used to connect different components of electronic device 800 and may include data bus, address bus and control bus.

[0138] Electronic device 800 can communicate with one or more external devices 900 (such as keyboard, mouse, external controller, etc.) through I / O interface 840.

[0139] Electronic device 800 can communicate with one or more networks via network adapter 850. For example, network adapter 850 can provide mobile communication solutions such as 3G / 4G / 5G, or wireless communication solutions such as wireless LAN, Bluetooth, and near-field communication. Network adapter 850 can communicate with other modules of electronic device 800 via bus 830.

[0140] Electronic device 800 can display a graphical user interface via display 860, such as a flood emergency alarm interface.

[0141] although Figure 8 As not shown in the diagram, other hardware and / or software modules may also be configured in the electronic device 800, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0142] Those skilled in the art will understand that various aspects of this disclosure can be implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which may be referred to as "circuit", "module" or "system" respectively.

[0143] It should be understood that this disclosure is not limited to the specific methods, steps, or structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. Those skilled in the art will readily conceive of other embodiments based on the specific implementations provided in this disclosure. Therefore, the specific implementations provided in this disclosure are merely exemplary, and the scope and spirit of this disclosure are indicated by the claims, and should cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary technical means in the art not disclosed in this disclosure.

Claims

1. A method for issuing a flood emergency warning, characterized in that, include: The control alarm server obtains the computing task information corresponding to the computing task generated by the flood risk situation awareness data collected by the monitoring stations in the target watershed, as well as the computing resource information of the edge nodes deployed in the target watershed. The target watershed includes multiple sub-watersheds, each of which has at least one monitoring station and at least one edge node, and the edge node has an edge server deployed in it. The alarm server is controlled to determine the target scheduling strategy for the computing task from the candidate scheduling strategies based on the flood risk situation awareness data collected by the monitoring stations. The alarm server is controlled to allocate edge servers to the computing task based on the target scheduling strategy, according to the computing task information and the computing resource information; The system controls each edge server to determine the execution priority of each computing task assigned to it based on the flood risk situation awareness data, and processes the flood risk situation awareness data according to the execution priority to generate flood hazard alarm information for the sub-basin to which the corresponding monitoring station belongs. The candidate scheduling strategies include a first candidate scheduling strategy and a second candidate scheduling strategy. The process of controlling the alarm server to determine the target scheduling strategy for the computation task from the candidate scheduling strategies based on the flood risk situation awareness data collected by the monitoring stations includes: When the flood risk situation awareness data indicates that the rainfall level is at the first level, the first candidate scheduling strategy is determined as the target scheduling strategy; If the flood risk situation awareness data indicates that the rainfall level is level two, the second candidate scheduling strategy will be determined as the target scheduling strategy. The rainfall intensity of the first level is greater than that of the second level. The first candidate scheduling strategy includes: calculating the minimum completion time of each computing task on each edge server based on computing task information and computing resource information and a pre-established task completion time model; determining the allocation priority of each computing task based on the minimum completion time and the rainfall in the flood situation awareness data of the computing task; and allocating the computing tasks according to the allocation priority to allocate the computing tasks to the edge server corresponding to the current minimum completion time of the computing task. The second candidate scheduling strategy includes: determining the task completion time of each computing task on each edge server based on the computing task information and computing resource information and a pre-established task completion time model; taking minimizing the maximum completion time of all edge servers as the optimization objective, iteratively optimizing the initial allocation scheme based on the simulated annealing method until the iteration termination condition is met, stopping the iterative optimization, and allocating edge servers to each computing task according to the iterative optimization results.

2. The method according to claim 1, characterized in that, The simulated annealing method iteratively optimizes the initial allocation scheme until the iteration termination condition is met, at which point the iteration optimization stops, and edge servers are allocated to each computing task based on the iteration optimization results. At a preset initial temperature, an initial allocation scheme is randomly generated; Based on the computing task information and computing resource information, and using a pre-established task completion time model, the task completion time of each computing task on the edge server in the allocation scheme is determined. The total completion time of all computing tasks allocated to each edge server under the allocation scheme is obtained by summing the task completion times of all computing tasks allocated to each edge server, and the maximum value of the total task completion times of each edge server is taken as the initial system completion time. Randomly adjust the edge servers assigned to at least one computing task in the allocation scheme to update the allocation scheme, and calculate the system completion time corresponding to the updated allocation scheme; Calculate the time difference between the system completion time corresponding to the updated allocation scheme and the system completion time corresponding to the original allocation scheme; If the time difference is less than zero, the updated allocation scheme is accepted; if the time difference is greater than or equal to zero, the updated allocation scheme is accepted according to a preset probability, wherein the preset probability is related to the time difference and the current temperature of the simulated annealing method. Lower the temperature and repeat the above process until the preset termination temperature is reached. Stop the iteration and allocate edge servers to each computing task according to the allocation scheme at the end of the iteration.

3. The method according to claim 1, characterized in that, The task completion time model is pre-built in the following way: The data transmission duration corresponding to the computing task on the edge server is determined based on the ratio between the data size required for the computing task and the data transmission rate between the monitoring station and the base station connected to the edge server. The execution time of the computing task on the edge server is determined based on the ratio between the computational workload of the computing task and the computing capacity of the edge server. The task completion time model is obtained based on the data transmission duration and the computation task execution duration.

4. The method according to claim 1, characterized in that, The flood risk situation awareness data includes one or more of the following: rainfall intensity, water level rise rate, distance between the current water level at the monitoring station and the preset warning water level, and soil moisture content; the control of each edge server to determine the execution priority of each computing task assigned to the edge server based on the flood risk situation awareness data includes: The target priority score is obtained by weighting and combining the priority scores corresponding to each flood risk situation awareness data. The execution priority of each computing task assigned to the edge server is determined based on the descending order of the target priority scores. The rainfall intensity is positively correlated with the priority score, the water level rise rate is positively correlated with the priority score, the distance between the current water level and the preset warning water level at the monitoring station is negatively correlated with the priority score, and the soil moisture content is positively correlated with the priority score.

5. The method according to claim 1, characterized in that, The flood risk situation awareness data includes rainfall intensity and hydrological model environmental parameters, which include topographic parameters, runoff generation and runoff parameters, and receding parameters. The step of processing the flood risk situation awareness data according to the execution priority to generate flood hazard alarm information for the sub-basin to which the corresponding monitoring station belongs includes: Based on the execution priority, the sub-basin outlet section runoff process calculation unit is determined based on the rainfall intensity of the sub-basin, the topographic parameters, the runoff generation and runoff parameters, and the dissipation parameters. The alarm index value of the sub-basin is obtained by the runoff process calculation unit at the outlet section of the sub-basin, and the alarm index value includes at least one of the water level prediction value and the flow prediction value. If the alarm index value of the sub-basin is greater than the first preset threshold, it is determined that there is a flood risk in the sub-basin, and a first flood hazard alarm signal for the sub-basin is generated. The alarm device in the edge node is controlled to send the first flood hazard alarm signal to the target user.

6. The method according to claim 5, characterized in that, The method further includes: If the alarm index value of the sub-basin is less than or equal to the first preset threshold, the sub-basin is determined as a candidate sub-basin. Based on historical water level data and / or historical flow data of multiple sub-basins within the target watershed, determine the water level correlation and / or flow correlation among the sub-basins; Based on the water level correlation and / or flow correlation, and combined with the current water level data and / or current flow data of each sub-basin, the predicted water level value of the candidate sub-basin is determined. The collaborative risk score of the candidate sub-basin is determined based on the difference between the predicted water level of the candidate sub-basin and the measured current water level of the candidate sub-basin. If the collaborative risk score is greater than a second preset threshold and the duration of the collaborative risk score being greater than the second preset threshold exceeds a preset duration, a second flood hazard alarm signal for the candidate sub-basin is generated.

7. A flood emergency alarm device, characterized in that, include: The information acquisition module is configured to control the alarm server to acquire the computing task information corresponding to the computing task generated by the flood risk situation awareness data collected by the monitoring stations in the target watershed, as well as the computing resource information of the edge nodes deployed in the target watershed. The target watershed includes multiple sub-watersheds, each of which has at least one monitoring station and at least one edge node, and the edge node has an edge server deployed in it. The scheduling strategy determination module is configured to control the alarm server to determine the target scheduling strategy for the computing task from candidate scheduling strategies based on the flood risk situation awareness data collected by the monitoring station. The computing task allocation module is configured to control the alarm server to allocate edge servers to the computing task based on the target scheduling strategy, the computing task information, and the computing resource information. The alarm module is configured to control each edge server to determine the execution priority of each computing task assigned to the edge server based on the flood risk situation perception data, and to process the flood risk situation perception data according to the execution priority to generate flood hazard alarm information for the sub-basin to which the corresponding monitoring station belongs. The candidate scheduling strategies include a first candidate scheduling strategy and a second candidate scheduling strategy. The process of controlling the alarm server to determine the target scheduling strategy for the computation task from the candidate scheduling strategies based on the flood risk situation awareness data collected by the monitoring stations includes: When the flood risk situation awareness data indicates that the rainfall level is at the first level, the first candidate scheduling strategy is determined as the target scheduling strategy; If the flood risk situation awareness data indicates that the rainfall level is level two, the second candidate scheduling strategy will be determined as the target scheduling strategy. The rainfall intensity of the first level is greater than that of the second level. The first candidate scheduling strategy includes: calculating the minimum completion time of each computing task on each edge server based on computing task information and computing resource information and a pre-established task completion time model; determining the allocation priority of each computing task based on the minimum completion time and the rainfall in the flood situation awareness data of the computing task; and allocating the computing tasks according to the allocation priority to allocate the computing tasks to the edge server corresponding to the current minimum completion time of the computing task. The second candidate scheduling strategy includes: determining the task completion time of each computing task on each edge server based on the computing task information and computing resource information and a pre-established task completion time model; taking minimizing the maximum completion time of all edge servers as the optimization objective, iteratively optimizing the initial allocation scheme based on the simulated annealing method until the iteration termination condition is met, stopping the iterative optimization, and allocating edge servers to each computing task according to the iterative optimization results.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1 to 6.

9. An electronic device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 6.

Images