CN-122022354-A - Multi-service scene-oriented transport capacity resource integrated intelligent scheduling method and system

CN122022354ACN 122022354 ACN122022354 ACN 122022354ACN-122022354-A

Abstract

The invention discloses an integrated intelligent scheduling method and system for capacity resources oriented to a multi-service scene, which relate to the field of intelligent scheduling and are used for constructing a layered collaborative decision-making framework comprising a macroscopic strategy layer intelligent body and a microscopic tactical layer optimizer, inputting a structural feature vector into the layered collaborative decision-making framework, dynamically distributing proper solving algorithms and parameters for the microscopic tactical layer optimizer according to a real-time scheduling situation by utilizing an online element learning optimizer, and outputting a pre-scheduling scheme. According to the invention, through the integration of the multi-service scene data, the comprehensive utilization of static basic information, real-time capacity data and prediction environment data is realized, and the data supporting capability of scheduling decision is improved. The feature interaction network of the multi-layer perceptron structure can accurately extract core features and provide effective input for scheduling decisions. The layered collaborative decision architecture is combined with an online element learning optimizer, so that a solution algorithm and parameters can be dynamically matched, and a better prescheduling scheme is output.

Inventors

WANG YINBO
YU YANG
Yu Suoshu
FENG BAOYUN
JIANG JIWEI
HU LINGHUI
CHEN JIANWEN
YANG RUI
ZHANG JIANJUAN

Assignees

云南合远科技有限公司
大连理工大学滇西产业发展研究院

Dates

Publication Date: 20260512
Application Date: 20260205

Claims (10)

1. The intelligent scheduling method for integrating the transport capacity resources for the multi-service scene is characterized by comprising the following steps of: Step 1, accessing static basic information of each service scene, acquiring a capacity dynamic state through a real-time data perception model, and integrating external environment data predicted based on a long-period memory network to form multi-service scene data comprising real-time data and predicted data; step 2, inputting multi-service scene data into a feature interaction network with a multi-layer perceptron structure, synchronously extracting core demand features, capacity types, endurance, preference matching features and resource competition and cooperative relationship features among scenes of the service, and outputting a structured feature vector; Step 3, constructing a layered collaborative decision-making architecture comprising a macroscopic strategy layer agent and a microscopic strategy layer optimizer, inputting a structural feature vector into the layered collaborative decision-making architecture, dynamically distributing a proper solving algorithm and parameters for the microscopic strategy layer optimizer according to a real-time scheduling situation by utilizing an online element learning optimizer, and outputting a prescheduling scheme; Step 4, placing the generated prescheduling scheme in an active digital twin simulation environment for multi-round pressure test and advanced intervention, and outputting an optimized prescheduling scheme with situation adaptation labels and anti-risk grades; And 5, comparing the optimized pre-scheduling scheme with the real-time data stream, calling a quick response sub-module of the micro tactical layer optimizer to carry out local conflict resolution when the capacity conflict or the time window violation is detected, synchronously presenting the adjustment suggestion and the digital twin pre-modeling result through a visual interface, and receiving and fusing a manual intervention instruction of a scheduler to form a final execution scheme.
2. The intelligent scheduling method for the integrated capacity resource of the multi-service scene according to the claim 1 is characterized by further comprising the step of 6 converting a final execution scheme into a structured instruction according to a predefined service scene template, wherein the structured instruction comprises a task ID, a capacity ID, a space path point sequence, a time window and an operation standard structured instruction, and carrying out elastic combined pushing through a driver APP, a fleet management platform and a short message channel according to the instruction priority and the state of a receiver, and attaching a unique traceability code of associated full-flow data to each instruction.
3. The intelligent scheduling method for integrating capacity resources for multi-service scenes according to claim 2 is characterized in that in step 1, a real-time data perception model is deployed at a vehicle-mounted internet of things terminal and a station edge computing node and is used for collecting vehicle position, speed, load, driver fatigue and station equipment occupancy rate data, and external environment data predicted based on a long-period memory network at least comprises prediction results of traffic flow states and weather disaster influence ranges for 2-6 hours in the future.
4. The intelligent scheduling method for integrating capacity resources facing multi-service scenes according to claim 3, wherein in step 2, the feature interaction network calculates correlation weights between demand features and capacity features of different service scenes through an attention mechanism, and the resource competition and coordination relationship features are used for quantifying conflict strength or engagement coordination potential of different scene tasks on the same capacity resource in the same time period.
5. The intelligent scheduling method for integrating capacity resources for multi-service scenarios according to claim 4, wherein in step 3, the macroscopic strategy layer agent is trained by adopting a deep reinforcement learning algorithm, and a reward function is set to be the weighted comprehensive optimum of the long-term capacity resource utilization rate, the global carbon emission reduction rate and the multi-service scenario task completion rate; The micro tactical layer optimizer is a mixed integer planning multi-objective optimization solver, and objective function weights are dynamically issued by a macro tactical layer agent according to real-time service priorities; The online element learning optimizer identifies the current traffic jam mode, order density level and capacity supply and demand gap situation in real time by constructing a scheduling situation and algorithm performance mapping database, dynamically selects a genetic algorithm, a tabu search algorithm or a particle swarm optimization algorithm for the micro tactical layer optimizer as a core solver based on the algorithm adaptation effect of the historical scheduling case, and simultaneously outputs a corresponding parameter configuration scheme.
6. The intelligent scheduling method for integrating capacity resources for multi-service scenarios according to claim 5, wherein when the micro tactical layer optimizer generates a pre-scheduling scheme in step 3, a core objective function is: ; wherein U is the comprehensive utilization rate of transport capacity resources, and the value range is 0, 1; Wherein Indicating the effective job time for the ith capacity to perform the kth task, Representing the maximum operable time of the corresponding capacity in a scheduling period, wherein N is the total capacity of the capacity participating in scheduling, C is the carbon emission of a global unit task, D is the average delay rate of the multi-service scene task, and the value range is 0, 1; 、、 weight coefficients of the capacity utilization rate, the carbon emission and the task delay rate respectively meet the following conditions And is dynamically issued by the macroscopic strategy layer agent according to the real-time business priority.
7. The intelligent scheduling method of integrated capacity and resource for multi-service scenario according to claim 6, wherein in step 4, an active digital twin simulation environment is constructed based on a physical engine and multi-agent simulation technology, a dynamic pressure test event is automatically generated according to the data distribution of historical abnormal events, and the execution process of a preset scheduling scheme under different events is simulated; when the simulation detects that the task delay rate of the pre-scheduling scheme under the pressure event exceeds a preset threshold value, automatically triggering capacity redundancy allocation simulation and path re-planning, calculating the performance retention rate after scheme adjustment, and generating low, medium and high three-level anti-risk grade labels and corresponding adaptation strategy descriptions by combining the situation characteristics; deep reinforcement learning training is adopted by the macroscopic strategy layer agent, and a reward function is comprehensively optimal for the long-term capacity resource utilization rate and the global carbon emission; The micro tactical layer optimizer is a multi-objective optimization solver, and the objective function weight receives real-time adjustment instructions of the macro tactical layer intelligent agent; the online element learning optimizer maintains a scheduling context and algorithm performance mapping table, and dynamically selects an algorithm core for the micro tactical layer optimizer according to the traffic mode and order density context identified in real time.
8. The intelligent scheduling method of integrating capacity resources for multi-service scenarios according to claim 7, wherein the core anti-risk evaluation formula for optimizing the prescheduling scheme in step 4 is: ; Wherein R is an anti-risk level quantized value, the value range [0,1], R is more than or equal to 0.8 and is high-grade, R is more than or equal to 0.5 and is less than or equal to 0.8 and is medium-grade, R is less than or equal to 0.5 and is low-grade, and P a is the comprehensive performance index of the prescheduling scheme under a normal scene; The comprehensive performance index of the scheme after adjustment under the pressure test event is obtained; S is the scheme to adjust the response speed score, and the value range [0,1] is calculated by the ratio of the time required for adjustment to the preset standard response time.
9. The intelligent scheduling method for integrating capacity and resources for multi-service scenarios according to claim 8, characterized in that in step 5, a fast response sub-module adopts a rolling time domain optimization strategy, sets a sliding optimization window for 30-60 minutes, only performs local decoupling and reassignment on capacity tasks with conflicts in the window, and invokes an adaptation strategy corresponding to the risk-resisting grade label output in step 4 in the solving process, so as to ensure conflict resolution efficiency and scheme stability.
10. The intelligent scheduling system for integrating the capacity resources facing the multi-service scene is characterized by being used for realizing the intelligent scheduling method for integrating the capacity resources facing the multi-service scene, and comprises a data integration module, a feature extraction module, a hierarchical decision module, a digital twin optimization module, a conflict resolution module and an instruction issuing module; The data integration module is used for accessing static basic information of each service scene, collecting operation capacity dynamic state through a real-time data perception model, integrating external environment data based on long-short-term memory network prediction, outputting multi-service scene data, the feature extraction module is used for inputting the multi-service scene data and extracting service core demand features, operation capacity features and inter-scene resource competition and cooperative relation features by adopting a feature interaction network with a multi-layer perceptron structure, outputting structural feature vectors, the hierarchical decision module comprises a macroscopic strategy layer agent, a microscopic strategy layer optimizer and an online element learning optimizer, the macroscopic strategy layer agent outputs target function weights through deep reinforcement learning, the online element learning optimizer dynamically matches solving algorithm and parameters, the microscopic strategy layer optimizer outputs a pre-scheduling scheme based on the core target function, the digital twin optimization module performs multi-round pressure test on the pre-scheduling scheme, outputs an optimization pre-scheduling scheme with a situation adaptation label and an anti-risk level through a core anti-risk evaluation formula, the conflict evaluation module compares the optimization pre-scheduling scheme with the real-time data stream, invokes a rapid response sub-module to perform local conflict resolution, a fusion instruction is generated, and the execution instruction is converted into an elastic instruction pushing instruction.

Description

Multi-service scene-oriented transport capacity resource integrated intelligent scheduling method and system Technical Field The invention relates to the technical field of intelligent scheduling, in particular to a multi-service scene-oriented capacity resource integrated intelligent scheduling method and system. Background The current capacity scheduling is developed aiming at a single service scene, and resource cooperation is lacked among the service scenes, so that the overall utilization rate of capacity resources is low. Meanwhile, the integration and utilization of real-time capacity state data and external environment prediction data in the scheduling process are insufficient, and the scheduling situation is difficult to comprehensively reflect. Most of the existing dispatching decision-making architecture is in a single level, the coordination of macroscopic strategic guidance and microscopic tactical optimization is lacking, an effective dynamic pressure testing mechanism is not established, when the abnormal events such as vehicle faults, traffic interruption, order sharp increase and the like are faced, the anti-risk capability of a dispatching scheme is weak, task delay or transport capacity conflict easily occurs, and the efficient and stable dispatching requirements under a multi-service scene cannot be met. Disclosure of Invention In order to solve the technical problems, the invention provides a multi-service scene-oriented capacity resource integrated intelligent scheduling method and system. The following technical scheme is adopted: The intelligent scheduling method for the capacity resource integration facing the multi-service scene comprises the following steps: Step 1, accessing static basic information of each service scene, acquiring a capacity dynamic state through a real-time data perception model, and integrating external environment data predicted based on a long-period memory network to form multi-service scene data comprising real-time data and predicted data; step 2, inputting multi-service scene data into a feature interaction network with a multi-layer perceptron structure, synchronously extracting core demand features, capacity types, endurance, preference matching features and resource competition and cooperative relationship features among scenes of the service, and outputting a structured feature vector; Step 3, constructing a layered collaborative decision-making architecture comprising a macroscopic strategy layer agent and a microscopic strategy layer optimizer, inputting a structural feature vector into the layered collaborative decision-making architecture, dynamically distributing a proper solving algorithm and parameters for the microscopic strategy layer optimizer according to a real-time scheduling situation by utilizing an online element learning optimizer, and outputting a prescheduling scheme; Step 4, placing the generated prescheduling scheme in an active digital twin simulation environment for multi-round pressure test and advanced intervention, and outputting an optimized prescheduling scheme with situation adaptation labels and anti-risk grades; And 5, comparing the optimized pre-scheduling scheme with the real-time data stream, calling a quick response sub-module of the micro tactical layer optimizer to carry out local conflict resolution when the capacity conflict or the time window violation is detected, synchronously presenting the adjustment suggestion and the digital twin pre-modeling result through a visual interface, and receiving and fusing a manual intervention instruction of a scheduler to form a final execution scheme. Optionally, the method further comprises the step 6 of converting the final execution scheme into a structured instruction according to a predefined business scene template, wherein the structured instruction comprises a task ID, an operation capacity ID, a space path point sequence, a time window and a structured instruction with operation specifications, and carrying out elastic combined pushing through a driver APP, a fleet management platform and a short message channel according to the instruction priority and the state of a receiver, and attaching a unique traceability code of associated full-flow data to each instruction. Optionally, in step 1, the real-time data perception model is deployed at a vehicle-mounted internet of things terminal and a station edge computing node, and is used for collecting vehicle position, speed, load, driver fatigue and station equipment occupancy rate data, and the external environment data predicted based on the long-short-period memory network at least comprises prediction results of traffic flow states and weather disaster influence ranges for 2-6 hours in the future. Optionally, in step 2, the feature interaction network calculates the correlation weights between the demand features and the capacity features of different service scenarios through an attention mechanism, and the resourc