A multi-task emergency shared resource budget arbitration scheduling method and system

Abstract

The present application belongs to the technical field of computer data processing, intelligent resource scheduling system, Internet of Things data processing, GIS spatial data processing, risk early warning system and facility operation and maintenance system, and specifically discloses a multi-task emergency shared resource budget arbitration scheduling method and system, which comprises the following steps: S1, standardized scheduling data is generated, and the size of a shared resource pool is determined according to available resource data; S2, parameterized scheduling constraints are generated; S3, a shared resource task set is determined, and the task pressure, marginal contribution proxy and regional yield rate distribution of each task in the shared resource task set are calculated; S4, the task budget proportion of each task is output; S5, the regional weight in the corresponding task is output; S6, a regional-level resource allocation scheme is obtained; S7, the regional-level resource allocation scheme is converted into a resource scheduling instruction, and the execution feedback of the scheduling instruction is received and used for state updating in the next scheduling period.

Description

Technical Field

[0001] This invention belongs to the technical fields of computer data processing, intelligent resource scheduling system, Internet of Things data processing, GIS spatial data processing, risk early warning system and facility operation and maintenance system, and specifically relates to a multi-task emergency shared resource budget arbitration scheduling method and system. Background Technology

[0002] In regional emergency support and facility operation and maintenance scenarios, there are usually multiple tasks within the same scheduling cycle, such as risk suppression, service restoration, access maintenance, equipment repair and material replenishment. These multiple tasks often share the same resource pool, such as emergency personnel, maintenance equipment, support materials, repair vehicles and mobile operation units.

[0003] Existing scheduling methods often allocate resources using fixed ratios, manual rules, or single-task sorting, making it difficult to dynamically adjust the resource ratios between tasks based on facility status, regional topology, service interruption level, task stage, and resource availability time window, which can easily lead to long-term supply shortages for certain tasks.

[0004] The existing scheduling methods also suffer from problems such as multi-source heterogeneous data fusion and decentralized constraint processing: the data provided by IoT interfaces, GIS databases, facility operation and maintenance systems, work order management platforms and resource positioning terminals have different formats, time granularities and spatial codes. If they are only used as a basis for manual judgment, it is difficult to directly generate personnel dispatch, equipment dispatch, material delivery or work unit movement instructions that can be executed by the scheduling terminal.

[0005] In addition, most existing scheduling methods merge multiple tasks and multiple regions into a joint action vector, and the unified model directly outputs the resource amount of all regions and tasks. The action dimension of this method increases synchronously with the number of tasks and the number of regional nodes, resulting in high training sample requirements and inference computation. Furthermore, it is difficult to distinguish whether resource changes are caused by changes in task-level pressure or by changes in regional priority within a task.

[0006] Therefore, there is a need for a unified processing method that integrates multi-source data acquisition, standardized processing, task-level budget arbitration, intra-task region allocation, parameterized constraint projection, resource scheduling instruction output, and execution feedback updates, in order to reduce the dimensionality of joint actions and generate executable resource configuration schemes that satisfy constraints. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention provides a multi-task emergency shared resource budget arbitration scheduling method and system.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solution:

[0009] The first aspect of this invention provides a multi-task emergency shared resource budget arbitration scheduling method, comprising the following steps:

[0010] S1. Obtain regional node data, facility status data, task target data, available resource data, and scheduling constraint data of the target area within the current scheduling cycle, generate standardized scheduling data, and determine the size of the shared resource pool based on the available resource data; the target area is at least one of the following scenarios: regional emergency support, facility operation and maintenance, or service recovery; the available resource data includes at least one of the following: emergency personnel, maintenance equipment, support materials, repair vehicles, or mobile operation units.

[0011] S2. Generate regional scheduling status features based on standardized scheduling data, and generate parameterized scheduling constraints; the parameterized scheduling constraints include minimum service ratio of tasks, effective regional node mask, total resource constraints, and regional resource upper limit constraints; the regional scheduling status features include task stage features and shared resource pool features.

[0012] S3. Based on the regional scheduling status characteristics and parameterized scheduling constraints, determine the set of shared resource tasks, and calculate the task pressure, marginal contribution proxy quantity, and regional rate of return distribution for each task in the set of shared resource tasks.

[0013] S4. Input the task pressure, marginal contribution proxy volume, task stage characteristics and shared resource pool size of each task into the task layer budget arbitrator, and output the task budget ratio of each task.

[0014] S5. Input the regional return distribution and effective regional node mask of each task into the corresponding task-based structured regional allocator, and output the regional weights within the corresponding task.

[0015] S6. Generate regional-level resource input quantities with task tags based on the size of the shared resource pool, the task budget ratio, and the regional weights within the task. Perform constraint projection on the regional-level resource input quantities to obtain the regional-level resource allocation scheme.

[0016] S7. Convert the regional resource allocation scheme into resource scheduling instructions and send them to the scheduling terminal, management platform or device control terminal. Receive the execution feedback of the scheduling instructions and use it for status updates in the next scheduling cycle.

[0017] Therefore, the joint resource actions of multiple tasks and regions are decomposed into three levels: task-level budget arbitration, intra-task regional allocation, and regional command output, in order to reduce the dimensionality of joint actions and reduce ineffective regional allocation and long-term supply shortages for tasks.

[0018] According to the above-mentioned multi-task emergency shared resource budget arbitration scheduling method, preferably, step S1 includes the following steps:

[0019] S1.1 Obtain raw scheduling data through IoT interfaces, GIS databases, facility operation and maintenance systems, work order management platforms, or resource location terminals.

[0020] S1.2 Perform field mapping, format conversion, outlier identification, missing value filling, time alignment, spatial matching and normalization on the original scheduling data to obtain standardized scheduling data.

[0021] S1.3. Update the scheduling constraint data for the current scheduling period based on the resource configuration results, task completion status, facility status changes, and resource consumption data of the previous scheduling period.

[0022] The size of the shared resource pool is obtained from available resource data according to a preset resource metering mapping:

[0023] ;

[0024] in, This indicates the current size of the shared resource pool. This represents the available resource data for the current scheduling period. This represents a mapping that converts different types of resources into a unified resource meter.

[0025] Specifically, when available resource data is recorded according to schedulable resource units, including available time windows and capacity coefficients, the size of the shared resource pool is expressed as:

[0026]

[0027] in, Indicates schedulable resource units Capacity conversion factor Indicates schedulable resource units During the scheduling period Available time window coefficient within, Indicates schedulable resource units The resource status or operational capacity reduction factor; the above factor is determined by historical scheduling data or preset by the management platform; all available resources are converted into a unified resource unit before entering, and no further category-based splitting is performed.

[0028] According to the above-mentioned multi-task emergency shared resource budget arbitration scheduling method, preferably, in step S2, the generation step of the regional scheduling status features is as follows: generate a regional node feature matrix based on regional node data and facility status data, generate a regional adjacency matrix based on spatial topology, road accessibility or service dependency, and generate regional coding features through a preset graph feature extraction model.

[0029] ;

[0030] in, Indicates regional coding features, Represents the feature matrix of the region nodes. Represents the region adjacency matrix. Indicates the characteristics of the task phase. This indicates the execution feedback characteristics of the previous cycle. It represents the characteristics of the shared resource pool; when there are no available adjacency relationships, it generates regional scheduling status characteristics based on regional node characteristics and global statistical characteristics.

[0031] The graph feature extraction model is a graph structure feature calculation model executed by a processor, including one of graph convolutional networks, graph attention networks, message passing networks, graph embedding models based on adjacency matrices, or feature extraction models based on neighborhood aggregation rules.

[0032] The first of the regional node feature matrix The row is obtained by concatenating the following node features:

[0033]

[0034] in, Indicates the value of regional nodes. Indicates the importance of the pathway. Indicates the importance of the service. Indicates the mobility of terrain or road. Indicates the intensity of risk disturbance. Indicates the degree of damage or interruption. Indicates facility availability. This indicates the estimated arrival time of resources or the cost of passage.

[0035] The effective regional node mask is determined based on regional reachability, task applicability, data validity, resource deployment conditions, and task completion status; the total resource constraint is used to limit the total amount of uniform resource units that can be deployed in the current scheduling cycle; the regional resource upper limit constraint is used to limit the maximum amount of resources that a single regional node can receive in the current scheduling cycle.

[0036] The effective region node mask It is a binary mask; when the region node is unreachable, the task is inapplicable, the data is invalid, the resource cannot be deployed, or the corresponding task has been completed. Take zero; otherwise Choose one; when the part that needs to be expressed is available. It can be expanded to a soft mask of zero to one, the value of which represents the availability of a region node for the corresponding task.

[0037] According to the above-mentioned multi-task emergency shared resource budget arbitration scheduling method, preferably, in step S3, the shared resource task set includes at least two shared resource tasks; when the task set includes risk suppression tasks and service restoration tasks, the risk suppression tasks include inspection and reinforcement, anomaly isolation, risk reduction, path hazard elimination, and temporary protection operations, and the task pressure of the risk suppression tasks is generated based on at least some of the characteristics among risk disturbance intensity, path importance, service importance, anomaly degree, and regional neighborhood status; the service restoration tasks include emergency repair and restoration, equipment replacement, material replenishment, service restart, and operation review, and the task pressure of the service restoration tasks is generated based on at least some of the characteristics among service interruption degree, facility availability gap, service importance, repair time limit, and expected resource arrival time.

[0038] The marginal contribution proxy quantity is generated based on the resource input and state changes of the previous cycle. It is used to characterize the mitigating effect of risk suppression input on the pressure of subsequent service recovery, as well as the supporting effect of service recovery input on subsequent risk suppression conditions.

[0039] Specifically, the pressure of risk suppression and service recovery tasks are represented by the following weighted normalization methods:

[0040]

[0041]

[0042] in, Indicates the degree of abnormality. This represents a summary of the state of the region's neighborhood; each weight coefficient is a non-negative coefficient, determined based on historical scheduling samples, expert rules, or model training results.

[0043] This represents the normalization function, which normalizes the input value to the range of zero to one according to the preset upper and lower limits, the historical sample range, or the maximum value of the current scheduling period. The specific normalization method is determined by the system configuration parameters before the start of the scheduling period and remains unchanged within the same scheduling period. When the normalization denominator is less than the preset threshold, the zero value, the historical average, or the preset backoff value is used.

[0044] The marginal contribution proxy quantity is generated by the state difference before and after the previous cycle's investment, satisfying one of the following implementation methods:

[0045]

[0046]

[0047] in, This represents the amount of agency contribution that reduces residual losses due to risk mitigation investment. The amount of agency volume that represents the improvement in facility availability due to service restoration efforts. Represents a region node During the scheduling period The residual loss or risk residual amount, as opposed to an amount representing the expected arrival time of resources or the cost of passage. When there is a lack of valid feedback from the previous cycle, zero agent quantity or the historical average is used as the backoff value.

[0048] The task pressure and marginal contribution agent quantity are satisfied:

[0049]

[0050] in, Indicates the first Each task in the scheduling cycle Task pressure Indicates the first Each task in the scheduling cycle The marginal contribution of the agent. and These represent the pressure calculation function and the marginal contribution calculation function for the corresponding task, respectively.

[0051] The regional yield distribution is used to represent the expected contribution of allocating unit shared resources to corresponding tasks and regional nodes to improving facility availability, reducing service downtime, or reducing risk disturbances. It is represented as a set or matrix of yields organized according to shared resource tasks and regional nodes; the regional yields satisfy:

[0052]

[0053] in, Indicates the first The task is located at the regional node. and scheduling cycle Regional yield, Represents a region node During the scheduling period The region coding features, Indicates the first The function for calculating the regional return on investment for each task.

[0054] To avoid treating the regional return distribution merely as an abstract function, we first calculate the regional return and regional cost terms corresponding to the task, and then generate the non-negative regional return:

[0055]

[0056] in, This represents the regional yield of the corresponding task at the corresponding regional node. This represents the regional revenue item for the corresponding task in the corresponding regional node. This represents the cost term for the corresponding region.

[0057] The regional benefit of the risk suppression task is obtained by weighting the risk disturbance intensity, the importance of the route, the importance of the service, and the degree of anomaly; the regional benefit of the service restoration task is obtained by weighting the degree of service interruption, the facility availability gap, the importance of the service, and the repair time limit; the regional cost is determined by the estimated arrival time of resources, road accessibility, operational difficulty, and resource deployment conditions.

[0058] The above weighted normalization method is and One specific implementation; the above-mentioned regional revenue items and regional cost items are generated in the following way. One specific implementation is that, in actual deployment, lookup table rules, regression models, or neural network models can also be used.

[0059] According to the above-mentioned multi-task emergency shared resource budget arbitration scheduling method, preferably, in step S4, the task-level budget arbitrator allocates the remaining shared resources based on the minimum service ratio of the tasks, so that the budget ratios of each task satisfy the non-negativity constraint, the sum of all constraints, and the minimum service ratio constraint; the task budget ratio is expressed as:

[0060]

[0061] in, Indicates the first Each task in the scheduling cycle Task budget ratio Indicates the first Minimum service ratio for each task Indicates the first The original output value of the task layer for each task. This represents a set of shared resource tasks.

[0062] The minimum service ratio for the task must meet the following requirements:

[0063]

[0064] in, This indicates the minimum service ratio for the corresponding task. When the sum of the minimum service ratios exceeds one, the system generates a constraint exception flag according to the preset task priority or management platform configuration, and resets the minimum service ratio using a rollback ratio.

[0065] The original output value of the task layer is generated by a rule model or a neural network model; when implemented using a neural network, the original output value of the task layer satisfies:

[0066]

[0067] in, and For the first The trainable parameters corresponding to each task; when using a rule-based model, the original output value of the task layer is obtained by weighting the task pressure, marginal contribution proxy quantity, task stage and the proportion of the previous cycle's task budget according to preset weights.

[0068] According to the above-mentioned multi-task emergency shared resource budget arbitration scheduling method, preferably, in step S5, the task-in-task structured region allocator generates the region weights within the task based on the region yield distribution, effective region node mask, threshold parameter, and concentration parameter; first, the yield threshold is determined:

[0069]

[0070] Then, perform low-yield truncation and concentration transformation based on the effective region node mask:

[0071]

[0072] Finally, safety normalization is performed to obtain the region weights within the task:

[0073]

[0074] in, Indicates the first The task is located at the regional node. and scheduling cycle Regional yield, Indicates the effective region node mask. Represents the set of currently valid region nodes. Indicates the yield threshold. Indicates the threshold parameter. Represents the concentration parameter. Indicates the region weights within the task. This indicates the zero-prevention threshold; when the candidate region is empty or the normalized denominator is less than the preset threshold, a task-specific coverage distribution or a uniform distribution within the effective region set is used as the backoff distribution.

[0075] The task-specific coverage distribution is generated based on the task objectives and regional basic characteristics. Specifically, the task-specific coverage distribution for risk suppression tasks is obtained by weighted normalization based on path importance, risk disturbance intensity, and anomaly degree. The task-specific coverage distribution for service recovery tasks is obtained by weighted normalization based on service importance, facility availability gap, and service interruption degree. The task-specific coverage distribution is only used as a fallback distribution when the candidate region is empty or the normalization denominator is less than a preset threshold.

[0076] According to the above-mentioned multi-task emergency shared resource budget arbitration scheduling method, preferably, in step S6, the regional-level resource input is expressed as:

[0077]

[0078] in, Indicates the first The task is located at the regional node. and scheduling cycle Regional-level resource input, This indicates the current size of the shared resource pool. Represents a set of shared resource tasks. This represents the set of currently valid region nodes.

[0079] According to the above-mentioned multi-task emergency shared resource budget arbitration scheduling method, preferably, in step S6, the constraint projection includes zeroing out invalid region nodes, non-negative truncation, region resource upper limit truncation, task minimum service supplementation, resource total normalization, and redistribution of remaining resources according to region weight. The constraint projection is used to correct the region-level resource input to a resource configuration scheme that satisfies parameterized scheduling constraints, which satisfies:

[0080]

[0081]

[0082] in, This represents the projected regional-level resource input. This represents the feasible region, determined by the minimum service ratio of the task, the effective regional node mask, the total resource constraint, and the regional resource upper limit constraint. Indicates the first The task is located at the regional node. and scheduling cycle The upper limit of regional resources.

[0083] Specifically, the constraint projection is performed in the following order: First, the resource input of regions with invalid effective region nodes is set to zero; second, the remaining regional resource input is truncated non-negatively; third, excess input is truncated according to the regional resource limit; fourth, the minimum service provision is performed when the task budget is lower than the resource amount corresponding to the minimum service ratio; fifth, the total resource input of all regional resources is normalized according to the total resource constraint; sixth, the remaining resources after truncation or normalization are redistributed according to the regional weights within the task.

[0084] According to the above-mentioned multi-task emergency shared resource budget arbitration scheduling method, preferably, in step S7, when the regional level resource input is converted into a discretized resource scheduling instruction, the continuous input is discretized and quantified, and the remaining resource units are supplemented and allocated according to the quantification loss or execution priority; the resource scheduling instruction includes regional node identifier, task tag, resource quantity, operation time window, execution priority, and resource shortage identifier; the discretized resource scheduling instruction is the terminal executable form of the resource scheduling instruction, used to map the continuous resource input into the specific quantity and operation time window of personnel dispatch, equipment dispatch, material delivery, or operation unit movement.

[0085] Under the constraint of total resource allocation, the total regional resource input for all tasks and all regions shall not exceed the current size of the shared resource pool. When there are still unallocated resources after constraint projection, the system generates a resource shortage flag or an unallocated resource flag in the resource scheduling instruction and writes it into the execution feedback.

[0086] The execution feedback includes execution status, arrival status, task completion status, facility status changes, service interruption level changes, and resource consumption data. The processor writes the execution feedback into the storage module and uses it as the basis for updating the execution feedback characteristics and scheduling constraints of the previous cycle in the next scheduling cycle.

[0087] The task-layer budget arbitrator, the in-task structured region allocator, or a combination thereof, are rule-based models, neural network models, or reinforcement learning strategy models. When trainable parameters are included, the trainable parameters are updated during the digital twin emergency drill or rolling prediction phase. During the actual deployment phase, reward calculation and strategy parameter updates are not performed; only resource scheduling instructions, task budget ratios, region weights, and anomaly handling flags are output.

[0088] The second aspect of the present invention provides a multi-task emergency shared resource budget arbitration scheduling system, including a data acquisition module, a data preprocessing module, a feature construction module, a parameterized constraint allocator module, a pressure-reward calculation module, a task-level budget arbitration module, an intra-task region allocation module, a resource configuration module, an output feedback module, a storage update module, a processor, a memory, and a communication interface.

[0089] The data acquisition module is used to acquire regional node data, facility status data, task target data, available resource data, and scheduling constraint data through IoT interfaces, GIS databases, facility operation and maintenance systems, work order management platforms, or resource positioning terminals.

[0090] The data preprocessing module is used to perform field mapping, format conversion, outlier identification, missing value filling, time alignment, spatial matching and normalization to generate standardized scheduling data.

[0091] The feature construction module is used to generate regional scheduling status features that include task phase features and shared resource pool features.

[0092] The parameterized constraint allocator module is used to generate the minimum service ratio for tasks, the effective regional node mask, the total resource constraint, and the regional resource upper limit constraint.

[0093] The pressure-reward calculation module is used to calculate the task pressure, marginal contribution proxy volume, and regional return distribution for each task.

[0094] The task-level budget arbitration module is used to output the task budget ratio for each task.

[0095] The task-in-region allocation module is used to output the region weights within the corresponding task.

[0096] The resource allocation module is used to generate regional-level resource input based on the size of the shared resource pool, the task budget ratio, and the regional weight within the task, and to generate a regional-level resource allocation scheme through constraint projection.

[0097] The output feedback module is used to convert regional resource allocation schemes into resource scheduling instructions and send them to scheduling terminals, management platforms or device control terminals, and to receive execution feedback.

[0098] The storage update module is used to save resource scheduling instructions and execution feedback, and update the status input for the next scheduling cycle according to the execution feedback; and executes the multi-task emergency shared resource budget arbitration scheduling method described in the first aspect through the cooperation of the processor, memory and communication interface.

[0099] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0100] (1) This invention effectively reduces the dimension of joint actions by decomposing multi-task and multi-region joint resource actions into three levels: task-level budget arbitration, intra-task regional allocation, and regional instruction output; it reduces long-term task shortages, invalid regional allocations, and resource overruns by constraining minimum service ratios, effective regional node masks, total resources, and regional resource limits; it improves the executability of computer system output results by converting resource configuration schemes into executable resource scheduling instructions for scheduling terminals, management platforms, or device control terminals; and it enables resource scheduling to form a closed loop by using execution feedback to update the status input for the next scheduling cycle.

[0101] (2) The present invention can solve the problems of high joint action dimension, unclear task budget ratio, invalid area allocation, long-term task shortage and difficulty in converting resource allocation results into terminal instructions in multi-task emergency resource sharing scenarios.

[0102] (3) This invention is executed by a computer scheduling system including a processor, memory and communication interface; the data acquisition module obtains regional node data, facility status data, task target data, available resource data and scheduling constraint data within the current scheduling cycle through IoT interface, GIS database, facility operation and maintenance system, work order management platform and resource positioning terminal; the data preprocessing module performs format conversion, outlier identification, missing value filling, time alignment and normalization processing; the feature construction module generates regional node features, adjacency relationship features, task stage features and shared resource pool features; the parameterized constraint allocator module generates the minimum service ratio of tasks, effective regional node mask, total resource constraint, regional resource upper limit and fallback distribution; the pressure and benefit calculation module calculates the task pressure and marginal contribution of risk suppression tasks and service recovery tasks. The invention features a shared resource pool, hierarchical budget arbitration, in-task regional allocation, and a feedback loop. It reduces the dimensionality of joint scheduling actions, minimizes ineffective node allocation and task starvation, and improves the efficiency of scheduling instruction generation and service recovery stability when multiple emergency tasks share personnel, equipment, materials, and mobile work units. The resource allocation module generates regional-level resource input based on the shared resource pool size, task budget ratio, and regional weight, and executes constraint projection. The output feedback module sends the resource allocation scheme to the scheduling terminal, management platform, or equipment control terminal. The storage and update module saves the execution feedback and updates the status input for the next scheduling cycle. Attached Figure Description

[0103] Figure 1 This is a schematic diagram of the process of the present invention in Example 1.

[0104] Figure 2 This is a structural block diagram of the present invention in Example 2. Detailed Implementation

[0105] The present invention will be further illustrated by specific embodiments below, but this does not limit the scope of the invention.

[0106] Example 1

[0107] A multi-task emergency shared resource budget arbitration scheduling method, the process of which is as follows: Figure 1 As shown, it includes the following steps:

[0108] S1. Obtain regional node data, facility status data, task target data, available resource data, and scheduling constraint data of the target area within the current scheduling cycle, generate standardized scheduling data, and determine the size of the shared resource pool based on the available resource data; the target area is at least one of the following scenarios: regional emergency support, facility operation and maintenance, or service recovery; the available resource data includes at least one of the following: emergency personnel, maintenance equipment, support materials, repair vehicles, or mobile operation units.

[0109] Step S1 includes the following steps:

[0110] S1.1 Obtain raw scheduling data through IoT interfaces, GIS databases, facility operation and maintenance systems, work order management platforms, or resource location terminals.

[0111] S1.2 Perform field mapping, format conversion, outlier identification, missing value filling, time alignment, spatial matching and normalization on the original scheduling data to obtain standardized scheduling data.

[0112] S1.3. Update the scheduling constraint data for the current scheduling period based on the resource configuration results, task completion status, facility status changes, and resource consumption data of the previous scheduling period.

[0113] The size of the shared resource pool is obtained from available resource data according to a preset resource metering mapping:

[0114] ;

[0115] in, This indicates the current size of the shared resource pool. This represents the available resource data for the current scheduling period. This represents a mapping that converts different types of resources into a unified resource meter.

[0116] Specifically, when available resource data is recorded according to schedulable resource units, including available time windows and capacity coefficients, the size of the shared resource pool is expressed as:

[0117]

[0118] in, Indicates schedulable resource units Capacity conversion factor Indicates schedulable resource units During the scheduling period Available time window coefficient within, Indicates schedulable resource units The resource status or operational capacity reduction factor; the above factor is determined by historical scheduling data or preset by the management platform; all available resources are converted into a unified resource unit before entering, and no further category-based splitting is performed.

[0119] S2. Generate regional scheduling status features based on standardized scheduling data, and generate parameterized scheduling constraints; the parameterized scheduling constraints include minimum service ratio of tasks, effective regional node mask, total resource constraints, and regional resource upper limit constraints; the regional scheduling status features include task stage features and shared resource pool features.

[0120] The steps for generating the regional scheduling status features are as follows: generating a regional node feature matrix based on regional node data and facility status data; generating a regional adjacency matrix based on spatial topology, road accessibility, or service dependency; and generating regional coding features through a preset graph feature extraction model.

[0121] ;

[0122] in, Indicates regional coding features, Represents the feature matrix of the region nodes. Represents the region adjacency matrix. Indicates the characteristics of the task phase. This indicates the execution feedback characteristics of the previous cycle. This represents the characteristics of a shared resource pool; when no available adjacency relationships exist, regional scheduling status characteristics are generated based on regional node characteristics and global statistical characteristics. The graph feature extraction model is a graph structure feature calculation model executed by a processor, including one of the following: graph convolutional network, graph attention network, message passing network, graph embedding model based on adjacency matrix, or feature extraction model based on neighborhood aggregation rules.

[0123] The first of the regional node feature matrix The row is obtained by concatenating the following node features:

[0124]

[0125] in, Indicates the value of regional nodes. Indicates the importance of the pathway. Indicates the importance of the service. Indicates the mobility of terrain or road. Indicates the intensity of risk disturbance. Indicates the degree of damage or interruption. Indicates facility availability. This indicates the estimated arrival time of resources or the cost of passage.

[0126] The effective regional node mask is determined based on regional reachability, task applicability, data validity, resource deployment conditions, and task completion status; the total resource constraint is used to limit the total amount of uniform resource units that can be deployed in the current scheduling cycle; the regional resource upper limit constraint is used to limit the maximum amount of resources that a single regional node can receive in the current scheduling cycle.

[0127] The effective region node mask It is a binary mask; when the region node is unreachable, the task is inapplicable, the data is invalid, the resource cannot be deployed, or the corresponding task has been completed. Take zero; otherwise Choose one; when the part that needs to be expressed is available. It can be expanded to a soft mask of zero to one, the value of which represents the availability of a region node for the corresponding task.

[0128] S3. Based on the regional scheduling status characteristics and parameterized scheduling constraints, determine the set of shared resource tasks, and calculate the task pressure, marginal contribution proxy quantity, and regional rate of return distribution for each task in the set of shared resource tasks.

[0129] The shared resource task set includes at least two shared resource tasks. When the task set includes risk suppression tasks and service restoration tasks, the risk suppression tasks include inspection and reinforcement, anomaly isolation, risk reduction, pathway hazard mitigation, and temporary protection operations. The task pressure of the risk suppression tasks is generated based on at least some of the characteristics among risk disturbance intensity, pathway importance, service importance, anomaly degree, and regional neighborhood status. The service restoration tasks include emergency repair and restoration, equipment replacement, material replenishment, service restart, and operation review. The task pressure of the service restoration tasks is generated based on at least some of the characteristics among service interruption degree, facility availability gap, service importance, repair time limit, and estimated resource arrival time.

[0130] The marginal contribution proxy quantity is generated based on the resource input and state changes of the previous cycle. It is used to characterize the mitigating effect of risk suppression input on the pressure of subsequent service recovery, as well as the supporting effect of service recovery input on subsequent risk suppression conditions.

[0131] Specifically, the pressure of risk suppression and service recovery tasks are represented by the following weighted normalization methods:

[0132]

[0133]

[0134] in, Indicates the degree of abnormality. This represents a summary of the state of the region's neighborhood; each weight coefficient is a non-negative coefficient, determined based on historical scheduling samples, expert rules, or model training results.

[0135] This represents the normalization function, which normalizes the input value to the range of zero to one according to the preset upper and lower limits, the historical sample range, or the maximum value of the current scheduling period. The specific normalization method is determined by the system configuration parameters before the start of the scheduling period and remains unchanged within the same scheduling period. When the normalization denominator is less than the preset threshold, the zero value, the historical average, or the preset backoff value is used.

[0136] The marginal contribution proxy quantity is generated by the state difference before and after the previous cycle's investment, satisfying one of the following implementation methods:

[0137]

[0138]

[0139] in, This represents the amount of agency contribution that reduces residual losses due to risk mitigation investment. The amount of agency volume that represents the improvement in facility availability due to service restoration efforts. Represents a region node During the scheduling period The residual loss or risk residual amount, as opposed to an amount representing the expected arrival time of resources or the cost of passage. When there is a lack of valid feedback from the previous cycle, zero agent quantity or the historical average is used as the backoff value.

[0140] The task pressure and marginal contribution agent quantity are satisfied:

[0141]

[0142] in, Indicates the first Each task in the scheduling cycle Task pressure Indicates the first Each task in the scheduling cycle The marginal contribution of the agent. and These represent the pressure calculation function and the marginal contribution calculation function for the corresponding task, respectively.

[0143] The regional yield distribution is used to represent the expected contribution of allocating unit shared resources to corresponding tasks and regional nodes to improving facility availability, reducing service downtime, or reducing risk disturbances. It is represented as a set or matrix of yields organized according to shared resource tasks and regional nodes; the regional yields satisfy:

[0144]

[0145] in, Indicates the first The task is located at the regional node. and scheduling cycle Regional yield, Represents a region node During the scheduling period The region coding features, Indicates the first The function for calculating the regional return on investment for each task.

[0146] To avoid treating the regional return distribution merely as an abstract function, we first calculate the regional return and regional cost terms corresponding to the task, and then generate the non-negative regional return:

[0147]

[0148] in, This represents the regional yield of the corresponding task at the corresponding regional node. This represents the regional revenue item for the corresponding task in the corresponding regional node. This represents the cost term for the corresponding region.

[0149] The regional benefit of the risk suppression task is obtained by weighting the risk disturbance intensity, the importance of the route, the importance of the service, and the degree of anomaly; the regional benefit of the service restoration task is obtained by weighting the degree of service interruption, the facility availability gap, the importance of the service, and the repair time limit; the regional cost is determined by the estimated arrival time of resources, road accessibility, operational difficulty, and resource deployment conditions.

[0150] The above weighted normalization method is and One specific implementation; the above-mentioned regional revenue items and regional cost items are generated in the following way. One specific implementation is that, in actual deployment, lookup table rules, regression models, or neural network models can also be used.

[0151] S4. Input the task pressure, marginal contribution proxy volume, task stage characteristics and shared resource pool size of each task into the task layer budget arbitrator, and output the task budget ratio of each task.

[0152] The task-level budget arbitrator allocates remaining shared resources based on the minimum service ratio of each task, ensuring that the budget ratios of each task satisfy the non-negativity constraint, the sum of all constraints, and the minimum service ratio constraint; the task budget ratio is expressed as:

[0153]

[0154] in, Indicates the first Each task in the scheduling cycle Task budget ratio Indicates the first Minimum service ratio for each task Indicates the first The original output value of the task layer for each task. This represents a set of shared resource tasks.

[0155] The minimum service ratio for the task must meet the following requirements:

[0156]

[0157] in, This indicates the minimum service ratio for the corresponding task. When the sum of the minimum service ratios exceeds one, the system generates a constraint exception flag according to the preset task priority or management platform configuration, and resets the minimum service ratio using a rollback ratio.

[0158] The original output value of the task layer is generated by a rule model or a neural network model; when implemented using a neural network, the original output value of the task layer satisfies:

[0159]

[0160] in, and For the first The trainable parameters corresponding to each task; when using a rule-based model, the original output value of the task layer is obtained by weighting the task pressure, marginal contribution proxy quantity, task stage and the proportion of the previous cycle's task budget according to preset weights.

[0161] S5. Input the regional return distribution and effective regional node mask of each task into the corresponding task-based structured regional allocator, and output the regional weights within the corresponding task.

[0162] The task-based structured region allocator generates region weights within the task based on the region return distribution, effective region node mask, threshold parameter, and concentration parameter; first, the return threshold is determined:

[0163]

[0164] Then, perform low-yield truncation and concentration transformation based on the effective region node mask:

[0165]

[0166] Finally, safety normalization is performed to obtain the region weights within the task:

[0167]

[0168] in, Indicates the first The task is located at the regional node. and scheduling cycle Regional yield, Indicates the effective region node mask. Represents the set of currently valid region nodes. Indicates the yield threshold. Indicates the threshold parameter. Represents the concentration parameter. Indicates the region weights within the task. This indicates the zero-prevention threshold; when the candidate region is empty or the normalized denominator is less than the preset threshold, a task-specific coverage distribution or a uniform distribution within the effective region set is used as the backoff distribution.

[0169] The task-specific coverage distribution is generated based on the task objectives and regional basic characteristics. Specifically, the task-specific coverage distribution for risk suppression tasks is obtained by weighted normalization based on path importance, risk disturbance intensity, and anomaly degree. The task-specific coverage distribution for service recovery tasks is obtained by weighted normalization based on service importance, facility availability gap, and service interruption degree. The task-specific coverage distribution is only used as a fallback distribution when the candidate region is empty or the normalization denominator is less than a preset threshold.

[0170] S6. Generate regional-level resource input quantities with task tags based on the size of the shared resource pool, the task budget ratio, and the regional weights within the task. Perform constraint projection on the regional-level resource input quantities to obtain the regional-level resource allocation scheme.

[0171] The regional-level resource input is expressed as follows:

[0172]

[0173] in, Indicates the first The task is located at the regional node. and scheduling cycle Regional-level resource input, This indicates the current size of the shared resource pool. Represents a set of shared resource tasks. This represents the set of currently valid region nodes.

[0174] The constraint projection includes zeroing out invalid region nodes, non-negative truncation, region resource upper limit truncation, minimum service completion for tasks, resource total normalization, and redistribution of remaining resources according to region weights. Constraint projection is used to correct the region-level resource allocation to a resource configuration scheme that satisfies parameterized scheduling constraints, which satisfy:

[0175]

[0176]

[0177] in, This represents the projected regional-level resource input. This represents the feasible region, determined by the minimum service ratio of the task, the effective regional node mask, the total resource constraint, and the regional resource upper limit constraint. Indicates the first The task is located at the regional node. and scheduling cycle The upper limit of regional resources.

[0178] Specifically, the constraint projection is performed in the following order: First, the resource input of regions with invalid effective region nodes is set to zero; second, the remaining regional resource input is truncated non-negatively; third, excess input is truncated according to the regional resource limit; fourth, the minimum service provision is performed when the task budget is lower than the resource amount corresponding to the minimum service ratio; fifth, the total resource input of all regional resources is normalized according to the total resource constraint; sixth, the remaining resources after truncation or normalization are redistributed according to the regional weights within the task.

[0179] S7. Convert the regional resource allocation scheme into resource scheduling instructions and send them to the scheduling terminal, management platform or device control terminal. Receive the execution feedback of the scheduling instructions and use it for status updates in the next scheduling cycle.

[0180] When regional-level resource input is converted into discretized resource scheduling instructions, the continuous input is discretely quantized, and the remaining resource units are supplemented and allocated according to the quantization loss or execution priority. The resource scheduling instructions include regional node identifiers, task tags, resource quantities, operation time windows, execution priorities, and resource shortage identifiers. The discretized resource scheduling instructions are the terminal executable form of resource scheduling instructions, used to map the continuous resource input to the specific quantity and operation time window of personnel dispatch, equipment dispatch, material delivery, or operation unit movement.

[0181] Under the constraint of total resource allocation, the total regional resource input for all tasks and all regions shall not exceed the current size of the shared resource pool. When there are still unallocated resources after constraint projection, the system generates a resource shortage flag or an unallocated resource flag in the resource scheduling instruction and writes it into the execution feedback.

[0182] The execution feedback includes execution status, arrival status, task completion status, facility status changes, service interruption level changes, and resource consumption data. The processor writes the execution feedback into the storage module and uses it as the basis for updating the execution feedback characteristics and scheduling constraints of the previous cycle in the next scheduling cycle.

[0183] The task-layer budget arbitrator, the in-task structured region allocator, or a combination thereof, are rule-based models, neural network models, or reinforcement learning strategy models. When trainable parameters are included, the trainable parameters are updated during the digital twin emergency drill or rolling prediction phase. During the actual deployment phase, reward calculation and strategy parameter updates are not performed; only resource scheduling instructions, task budget ratios, region weights, and anomaly handling flags are output.

[0184] Therefore, the joint resource actions of multiple tasks and regions are decomposed into three levels: task-level budget arbitration, intra-task regional allocation, and regional command output, in order to reduce the dimensionality of joint actions and reduce ineffective regional allocation and long-term supply shortages for tasks.

[0185] In this embodiment, specifically:

[0186] S1. Data collection and resource pool determination: The system is deployed in the regional emergency support management platform. It obtains facility availability, anomaly level and service interruption level through the Internet of Things interface, regional coordinates, adjacency relationship and road accessibility through the GIS database, asset type and maintenance status through the facility operation and maintenance system, task objectives and completion deadlines through the work order management platform, and location, quantity and available time window of emergency personnel, maintenance equipment, support materials, repair vehicles and mobile operation units through resource positioning terminals. Based on the available resource data, the current shared resource pool size is determined.

[0187] S2. Perform data preprocessing and feature construction; the data preprocessing module maps data fields from different interfaces, performs time alignment on data at different time granularities, performs regional node matching on spatial coordinates, handles outliers and missing values, and writes the processed standardized scheduling data into the storage module; the feature construction module generates regional node feature matrices, regional adjacency matrices, task stage features, previous cycle execution feedback features, and shared resource pool features based on the standardized scheduling data.

[0188] S3. Generate parameterized constraints; generate the lower bound of the task budget based on the minimum service ratio of the task; generate the effective regional node mask based on regional accessibility, task applicability, data validity, resource deployment conditions and task completion status; generate resource constraints based on the total amount of resources and the upper limit of regional resources; and set an abnormal rollback distribution for candidate regions that are empty or whose normalized denominator is too small.

[0189] S4. Calculate task pressure, marginal contribution proxy volume, and regional return distribution. The pressure-return calculation module uses the intensity of risk disturbance, path importance, service importance, anomaly degree, and regional neighborhood status to calculate the pressure of risk suppression tasks. It uses the degree of service interruption, facility availability gap, service importance, repair time limit, and expected resource arrival time to calculate the pressure of service recovery tasks. It also generates the marginal contribution proxy volume based on the changes in facility status before and after the resource investment in the previous cycle.

[0190] In this embodiment, if the residual loss decreases after the risk suppression investment in the previous cycle, the marginal contribution proxy of the risk suppression task increases; if the facility availability improves after the service restoration investment in the previous cycle, the marginal contribution proxy of the service restoration task increases; the pressure-reward calculation module inputs the marginal contribution proxy and the current task pressure into the task-level budget arbitration module, so that the task budget ratio not only reflects the current pressure, but also reflects the cross-cycle effect generated by the investment in the previous cycle.

[0191] S5. Perform task-level budget arbitration; The task-level budget arbitration module receives task pressure, marginal contribution proxy volume, task stage characteristics and shared resource pool size, allocates the remaining shared resources based on the minimum service ratio of the task, and outputs the task budget ratio of each task.

[0192] S6. Perform intra-task region allocation and constraint projection; the intra-task region allocation module generates intra-task region weights based on the regional yield distribution and effective region node masks; the resource allocation module generates regional-level resource input based on the shared resource pool size, task budget ratio, and region weights, and performs invalid region node zeroing, non-negative truncation, regional resource upper limit truncation, task minimum service supplementation, resource total normalization, and remaining resource reallocation.

[0193] S7. Generate resource scheduling instructions and update feedback; use the output feedback module to convert the regional resource allocation scheme into personnel dispatch, equipment dispatch, material distribution, work unit movement, work order adjustment or equipment control instructions, and send them to the scheduling terminal, management platform or equipment control terminal; use the storage update module to save resource scheduling instructions and execution feedback, and update the status input for the next scheduling cycle according to the execution feedback.

[0194] Example 2

[0195] A multi-task emergency shared resource budget arbitration and scheduling system, such as Figure 2 As shown, it includes a data acquisition module, a data preprocessing module, a feature construction module, a parameterized constraint allocator module, a pressure-reward calculation module, a task-level budget arbitration module, a task-in-task region allocation module, a resource configuration module, an output feedback module, a storage update module, a processor, a memory, and a communication interface.

[0196] The data acquisition module is used to acquire regional node data, facility status data, task target data, available resource data, and scheduling constraint data through IoT interfaces, GIS databases, facility operation and maintenance systems, work order management platforms, or resource location terminals. The data preprocessing module performs field mapping, format conversion, outlier identification, missing value imputation, time alignment, spatial matching, and normalization to generate standardized scheduling data. The feature construction module generates regional scheduling status features that include task stage features and shared resource pool features.

[0197] The parameterized constraint allocator module generates the minimum service ratio for tasks, effective region node masks, total resource constraints, and region resource ceiling constraints. The pressure-reward calculation module calculates the task pressure, marginal contribution proxy volume, and region return distribution for each task. The task-layer budget arbitration module outputs the task budget ratio for each task. The intra-task region allocation module outputs the region weights within the corresponding task.

[0198] The resource allocation module generates regional-level resource input based on the shared resource pool size, task budget ratio, and regional weights within the task, and generates a regional-level resource allocation scheme through constraint projection. The output feedback module converts the regional-level resource allocation scheme into resource scheduling instructions and sends them to the scheduling terminal, management platform, or device control terminal, and receives execution feedback.

[0199] The storage update module is used to save resource scheduling instructions and execution feedback, and update the status input for the next scheduling cycle according to the execution feedback; the multi-task emergency shared resource budget arbitration scheduling method described in Embodiment 1 is executed through the cooperation of the processor, memory and communication interface.

[0200] This embodiment uses "multiple service interruptions and risk disturbances occur in a certain infrastructure area, requiring the simultaneous execution of risk suppression and service restoration tasks" as a specific example; the target area is divided into several regional nodes, each corresponding to a road area, communication area, energy security area, or public service area; shared resources include emergency personnel, maintenance equipment, support materials, repair vehicles, and mobile operation units.

[0201] At the start of the current scheduling cycle, the data acquisition module obtains the facility availability, anomaly level, service interruption level, road accessibility, task completion time limit, and estimated resource arrival time of each regional node; the data preprocessing module unifies the timestamp and regional number to obtain standardized scheduling data; the feature construction module forms regional scheduling status features based on the standardized scheduling data; and the parameterized constraint allocator module forms the minimum service ratio of tasks, effective regional node mask, total resource constraints, and regional resource limits.

[0202] The pressure-return calculation module first determines the risk suppression task and the service recovery task, and then calculates the task pressure, marginal contribution agency volume, and regional return distribution respectively. The task-level budget arbitration module outputs the task budget ratio based on the task pressure and marginal contribution agency volume, so that the risk suppression task and the service recovery task are not lower than the corresponding minimum service ratio. The in-task regional allocation module generates regional weights within each task based on the regional return distribution, and outputs zero weights for unreachable regions, regions where the task is not applicable, or regions with invalid data.

[0203] The resource allocation module multiplies the size of the shared resource pool, the task budget ratio, and the regional weight within the task to obtain the regional resource input amount with task tags. Then, the system uses constraint projection to meet the total resource amount and regional resource upper limit requirements. When the regional resource input amount needs to be converted into a discretized resource scheduling instruction, the system supplements the remaining resource units according to the quantization loss or execution priority to form an executable resource scheduling instruction.

[0204] The output feedback module sends resource scheduling instructions to the scheduling terminal, management platform, or device control terminal. During the execution process, the arrival status, task completion status, facility availability changes, service interruption degree changes, and resource consumption data returned are written to the storage module and used as the basis for updating the execution feedback characteristics and scheduling constraints of the previous cycle in the next scheduling cycle. Thus, the system completes the closed-loop scheduling from multi-source data collection to hierarchical budget arbitration, intra-task area allocation, constraint projection, instruction output, and feedback update.

[0205] In a small-scale numerical example, the task set includes risk suppression tasks and service recovery tasks. The regional nodes include Region 1, Region 2, and Region 3, and the shared resource pool size is ten resource units. The minimum service ratio for both tasks is 20%. The effective regional node masks for Region 1 to Region 3 for the risk suppression task are 1, 1, and 0, respectively, and for the service recovery task, they are 1, 0, and 1, respectively. Since the risk suppression task has a higher risk disturbance intensity in the current period and a larger marginal contribution proxy from the risk suppression investment in the previous period, the task-level budget arbitration module outputs a budget ratio of 60% for the risk suppression task and 40% for the service recovery task. Since Region 3 is ineffective for the risk suppression task and Region 2 is ineffective for the service recovery task, the corresponding regional weights are set to zero. The task-internal region allocation module outputs regional weights of 70%, 30%, and 0 for the risk suppression task and 50%, 0%, and 50% for the service recovery task.

[0206] Based on this, the resource allocation module obtained the regional resource input: the risk suppression task invested 4.2 resource units in region 1, 1.8 resource units in region 2, and zero resource units in region 3; the service restoration task invested 2 resource units in region 1, zero resource units in region 2, and 2 resource units in region 3.

[0207] After performing discrete quantization on the above continuous input, the system can generate scheduling instructions: dispatch four risk suppression operation units and two service recovery operation units to region 1, dispatch two risk suppression operation units to region 2, and dispatch two service recovery operation units to region 3, and record the remaining difference in the resource shortage identifier or quantization loss field for use in the status update and constraint correction of the next scheduling cycle.

[0208] In summary, this invention effectively overcomes the shortcomings of the prior art and has high industrial applicability. The above embodiments are intended to illustrate the substantive content of this invention, but are not intended to limit the scope of protection of this invention. Those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of this invention without departing from the essence and scope of protection of this invention.

Claims

1. A multi-task emergency shared resource budget arbitration scheduling method, characterized in that, Includes the following steps: S1. Obtain regional node data, facility status data, task target data, available resource data, and scheduling constraint data of the target area, generate standardized scheduling data, and determine the size of the shared resource pool based on the available resource data; S2. Generate regional scheduling status features based on standardized scheduling data, and generate parameterized scheduling constraints; the parameterized scheduling constraints include minimum service ratio of tasks, effective regional node mask, total resource constraints, and regional resource upper limit constraints; the regional scheduling status features include task stage features and shared resource pool features; S3. Based on the regional scheduling status characteristics and parameterized scheduling constraints, determine the set of shared resource tasks, and calculate the task pressure, marginal contribution proxy quantity, and regional rate of return distribution of each task in the set of shared resource tasks. S4. Input the task pressure, marginal contribution proxy volume, task stage characteristics and shared resource pool size of each task into the task layer budget arbitrator, and output the task budget ratio of each task. S5. Input the regional return distribution and effective regional node mask of each task into the corresponding task-in-task structured region allocator, and output the regional weights within the corresponding task. S6. Generate regional-level resource input with task tags based on the size of the shared resource pool, the task budget ratio, and the regional weight within the task. Perform constraint projection on the regional-level resource input to obtain the regional-level resource allocation scheme. S7. Convert the regional resource allocation scheme into resource scheduling instructions, receive the execution feedback of the scheduling instructions and use it for status updates in the next scheduling cycle.

2. The multi-task emergency shared resource budget arbitration scheduling method according to claim 1, characterized in that, Step S1 includes the following steps: S1.1 Obtain raw scheduling data through IoT interfaces, GIS databases, facility operation and maintenance systems, work order management platforms, or resource location terminals; S1.2 Perform field mapping, format conversion, outlier identification, missing value filling, time alignment, spatial matching and normalization on the original scheduling data to obtain standardized scheduling data; S1.

3. Update the scheduling constraint data for the current scheduling period based on the resource configuration results, task completion status, facility status changes, and resource consumption data of the previous scheduling period.

3. The multi-task emergency shared resource budget arbitration scheduling method according to claim 1, characterized in that, In step S2, the steps for generating the regional scheduling status features are as follows: generating a regional node feature matrix based on regional node data and facility status data, generating a regional adjacency matrix based on spatial topology, road accessibility, or service dependency, and generating regional coding features through a preset graph feature extraction model. ； in, Indicates region coding features, Represents the feature matrix of the region nodes. Represents the region adjacency matrix. Indicates the characteristics of the task phase. This indicates the execution feedback characteristics of the previous cycle. This indicates the characteristics of a shared resource pool.

4. The multi-task emergency shared resource budget arbitration scheduling method according to claim 1, characterized in that, In step S3, the shared resource task set includes at least two shared resource tasks; when the task set includes risk suppression tasks and service restoration tasks, the task pressure of the risk suppression task is generated based on at least some of the characteristics among risk disturbance intensity, path importance, service importance, anomaly degree, and regional neighborhood status; the task pressure of the service restoration task is generated based on at least some of the characteristics among service interruption degree, facility availability gap, service importance, repair time limit, and estimated resource arrival time. The marginal contribution proxy quantity is generated based on the resource input and state changes of the previous cycle. It is used to characterize the mitigating effect of risk suppression input on the pressure of subsequent service recovery, as well as the supporting effect of service recovery input on subsequent risk suppression conditions.

5. The multi-task emergency shared resource budget arbitration scheduling method according to claim 1, characterized in that, In step S4, the task-level budget arbitrator allocates remaining shared resources based on the minimum service ratio of each task, ensuring that the budget ratios of each task satisfy the non-negativity constraint, the sum of all constraints, and the minimum service ratio constraint; the task budget ratio is expressed as: in, Indicates the first Each task in the scheduling cycle Task budget ratio Indicates the first Minimum service ratio for each task Indicates the first The original output value of the task layer for each task. This represents a set of shared resource tasks.

6. The multi-task emergency shared resource budget arbitration scheduling method according to claim 1, characterized in that, In step S5, the task-based structured region allocator generates task-specific region weights based on the region return distribution, effective region node mask, threshold parameter, and concentration parameter. in, Indicates the first The task is located at the regional node. and scheduling cycle Regional yield, Indicates the effective region node mask. Represents the set of currently valid region nodes. Indicates the yield threshold. Indicates the threshold parameter. Represents the concentration parameter. Indicates the region weights within the task. This indicates the threshold for preventing zero.

7. The multi-task emergency shared resource budget arbitration scheduling method according to claim 1, characterized in that, In step S6, the regional-level resource input is expressed as: in, Indicates the first The task is located at the regional node. and scheduling cycle Regional-level resource input, This indicates the current size of the shared resource pool. Represents a set of shared resource tasks. This represents the set of currently valid region nodes.

8. The multi-task emergency shared resource budget arbitration scheduling method according to claim 7, characterized in that, In step S6, the constraint projection includes zeroing out invalid region nodes, non-negative truncation, region resource upper limit truncation, minimum service completion for tasks, normalization of total resources, and redistribution of remaining resources according to region weights. The constraint projection is used to correct the region-level resource input to a resource configuration scheme that satisfies parameterized scheduling constraints, which satisfies: in, This represents the projected regional-level resource input. This represents the feasible region, determined by the minimum service ratio of the task, the effective regional node mask, the total resource constraint, and the regional resource upper limit constraint. Indicates the first The task is located at the regional node. and scheduling cycle The upper limit of regional resources.

9. The multi-task emergency shared resource budget arbitration scheduling method according to claim 1, characterized in that, In step S7, when the regional-level resource input is converted into a discretized resource scheduling instruction, the continuous input is discretized and quantized, and the remaining resource units are allocated according to the quantization loss or execution priority. The resource scheduling instruction includes a regional node identifier, task tag, resource quantity, operation time window, execution priority, and resource shortage identifier. The discretized resource scheduling instruction is the terminal executable form of the resource scheduling instruction, which is used to map the continuous resource input into the specific quantity and operation time window of personnel dispatch, equipment dispatch, material delivery, or operation unit movement.

10. A multi-task emergency shared resource budget arbitration scheduling system, characterized in that, It includes a data acquisition module, a data preprocessing module, a feature construction module, a parameterized constraint allocator module, a pressure-reward calculation module, a task-level budget arbitration module, an in-task region allocation module, a resource configuration module, an output feedback module, a storage update module, a processor, a memory, and a communication interface; The data acquisition module is used to acquire regional node data, facility status data, task target data, available resource data, and scheduling constraint data through IoT interfaces, GIS databases, facility operation and maintenance systems, work order management platforms, or resource positioning terminals. The data preprocessing module is used to perform field mapping, format conversion, outlier identification, missing value filling, time alignment, spatial matching and normalization to generate standardized scheduling data. The feature construction module is used to generate regional scheduling status features that include task phase features and shared resource pool features; The parameterized constraint allocator module is used to generate the minimum service ratio of the task, the effective regional node mask, the total resource constraint, and the regional resource upper limit constraint. The pressure-reward calculation module is used to calculate the task pressure, marginal contribution proxy volume, and regional return distribution of each task. The task-level budget arbitration module is used to output the task budget ratio for each task. The in-task region allocation module is used to output the region weights within the corresponding task. The resource allocation module is used to generate regional-level resource input based on the size of the shared resource pool, the task budget ratio, and the regional weight within the task, and to generate a regional-level resource allocation scheme through constraint projection. The output feedback module is used to convert the regional resource allocation scheme into resource scheduling instructions and send them to the scheduling terminal, management platform or equipment control terminal, and receive execution feedback. The storage update module is used to save resource scheduling instructions and execution feedback, and update the status input for the next scheduling cycle according to the execution feedback; the multi-task emergency shared resource budget arbitration scheduling method according to any one of claims 1-9 is executed through the cooperation of the processor, memory and communication interface.

Images