Test task scheduling and management system based on 5G electric power virtual private network

Abstract

The invention discloses a test task scheduling and management system based on a 5G power virtual private network, and relates to the technical field of 5G communication and power system automation. A two-layer distributed architecture based on core and region edge nodes is adopted, and task domain division management and resource synchronization are realized through a lightweight Raft protocol; calculating a task key index TCI by using a four-dimensional label and a key dimension index amplification method, and dynamically scheduling resources in combination with a Lyapunov optimization model; and according to the health degree index of the access equipment, adaptively correcting resource reservation through the prediction model. According to the method, the problems of high time delay, extensive resource allocation and the like of a traditional centralized architecture are solved, distributed collaborative scheduling and resource dynamic optimization of the power test tasks are realized, and the system reliability is improved.

Description

Technical Field

[0001] This invention relates to the field of 5G communication and power system automation technology, specifically to a test task scheduling and management system based on a 5G power virtual private network. Background Technology

[0002] With the application of 5G technology in power private networks, the scheduling and management of test tasks based on 5G power virtual private networks face many challenges.

[0003] In existing technologies, the traditional centralized dispatch architecture has significant drawbacks in power wide area networks: First, it has a high risk of single-point failure, and the failure of the central node can easily lead to the paralysis of the entire dispatch system; Second, it has a prominent communication latency bottleneck, with large latency in the transmission of data from equipment in remote areas back to the central node, making it difficult to meet the low latency requirements of real-time testing tasks such as relay protection; Third, it has poor protocol compatibility, and traditional fixed protocol parsing methods cannot dynamically adapt to equipment upgrades and the access of new protocols. In terms of resource allocation, existing technologies lack precise means to quantify the criticality of tasks, making it difficult to distinguish between tasks with different priorities and real-time requirements. The average allocation often leads to insufficient resource guarantees for critical tasks. In addition, existing scheduling models do not fully consider the impact of the health status of power equipment on test tasks and cannot dynamically adjust resource reservations according to the aging of equipment, which may lead to decreased test accuracy or waste of resources for aging equipment.

[0004] Therefore, there is an urgent need for a test task scheduling and management solution that can adapt to the characteristics of 5G power virtual private networks. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a test task scheduling and management system based on a 5G power virtual private network, which solves the problems of centralized architecture defects, overly even resource allocation, and inability to adjust resource reservation based on device health.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a test task scheduling and management system based on a 5G power virtual private network, comprising: The edge node module is deployed in the power grid coverage area and adopts a two-layer distributed architecture, containing multiple edge node MECs. The TCI calculation module automatically classifies test tasks based on four-dimensional labels to obtain a four-dimensional judgment vector, and calculates the task criticality index (TCI) using the key dimension exponential amplification method. The four-dimensional labels specifically include priority, latency requirements, bandwidth requirements, and security level. The dynamic scheduling module incorporates a Lyapunov optimization model, which dynamically adjusts task priority and resource allocation based on the Task Criticality Index (TCI) and combined with the real-time resource status of edge nodes and network parameters. The feedback adjustment module receives power equipment status monitoring data to generate an equipment health index, and dynamically adjusts the resource reservation for tasks based on the health index.

[0007] As a further improvement of the present invention, the two-layer distributed architecture specifically consists of core edge nodes and regional edge nodes. The core edge nodes are equipped with high-performance processors and large-capacity storage, and are responsible for global resource catalog maintenance, cross-regional task collaboration, and strategy optimization. The regional edge nodes adopt a lightweight design and focus on local task preprocessing and real-time scheduling.

[0008] As a further improvement of the present invention, the regional edge node needs to be configured with a pluggable protocol parsing module.

[0009] As a further improvement of the present invention, edge nodes are networked using the lightweight Raft protocol to achieve task domain management and global resource directory synchronization.

[0010] As a further improvement of the present invention, the specific steps for obtaining the four-dimensional judgment vector are as follows: When a test task accesses an edge node through a 5G slicing channel, the system reads the identity information carried by the task to obtain the initial priority, latency requirements, bandwidth requirements, and security level. Match the priority and security level with the weight level mapping table to obtain the weights corresponding to the two dimensions; The resulting four-dimensional judgment vector can ultimately be represented as (priority weight, latency requirement, bandwidth requirement, security weight).

[0011] As a further improvement to the present invention, the weight level mapping table specifically includes: Priority is divided into three levels: high, medium, and low, with corresponding weights of 10, 5, and 1, respectively. The security level is divided into three levels: highest, medium, and basic, with corresponding weights of 1, 0.5, and 0, respectively.

[0012] As a further improvement to the present invention, the steps for calculating TCI using the key dimension exponential amplification method are as follows: Based on the four-dimensional judgment vector, the latency requirement D and bandwidth requirement B are normalized to obtain Dnorm and Bnorm. According to the formula The Task Criticality Index (TCI) is calculated, where P is the priority weight and S is the safety weight. , , As a regulating factor and , , .

[0013] As a further improvement to the present invention, the specific steps of Lyapunov optimization of resource allocation based on TCI are as follows: Establish a task queue Q={T1,T2,...,Tn}, and calculate TCIi for each task Ti using the key dimension exponential amplification method; Initialize the queue state Q(0), record the task arrival time, delay constraint Di, bandwidth requirement Bi and TCIi, and use them as input parameters for the Lyapunov optimization model; Edge nodes monitor the remaining slice bandwidth Bavail(t), edge computing power Cavail(t), and network latency fluctuation ΔD(t) in real time, forming a resource state vector R(t)=[Bavail(t),Cavail(t),ΔD(t)]; Establish the following constraints: Bi <= Bavail(t), Ci <= Cavail(t), ΔD(t) <= Di; Constructing Lyapunov functions ; A drift penalty mechanism is introduced to minimize the combined cost of queue latency and resource consumption. The specific formula is as follows: min(△L(Q(t))+V×C(t)), where △L is the queue stability index, C(t) is the resource consumption function, and V is the weight parameter for adjusting resource allocation and queue stability. Priority is given to allocating basic resources to high TCI tasks that meet latency constraints, and the remaining resources are dynamically allocated according to the TCIi ratio. ; Based on the optimization results, resource allocation instructions are generated, including slice bandwidth bi(t) and computing power quota ci(t). High TCI tasks are triggered for immediate execution, while low TCI tasks enter the waiting queue. According to the formula , Update the state of the resource layer; For the task layer, remove executed tasks from the queue.

[0014] As a further improvement of the present invention, the specific steps for dynamically adjusting the resource reservation of tasks based on the health index are as follows: By accessing power equipment status monitoring data and fusing multi-dimensional indicators through DS evidence theory, an equipment health index H∈[0,1] is generated; An LSTM prediction model is established for the equipment health index H. Based on historical data, the health level H(t+T) at time T is predicted, and the predicted H(t+T) is substituted into the correction factor formula K. H =1+q×(1-H(t+T)), where q is the adjustment coefficient; According to the formula The corrected resource reservation is calculated, where f(TCIi) is the basic resource reservation and ΔH is the health rate of change. This is the sensitivity coefficient to change.

[0015] As a further improvement to the present invention, according to the formula V=V0×K H The parameter V in the Lyapunov drift penalty is dynamically adjusted, where V0 is the initial parameter.

[0016] This invention provides a test task scheduling and management system based on a 5G power virtual private network, which has the following advantages compared with the prior art: (1) This invention adopts a two-layer distributed edge node architecture and a lightweight Raft protocol to realize task domain management and global resource directory synchronization, solve the single point of failure problem of centralized architecture, and improve system reliability; (2) This invention calculates TCI based on four-dimensional labels and key dimension exponential amplification method, accurately quantifies the criticality of tasks, and combines Lyapunov optimization to dynamically adjust resource allocation, giving priority to high-priority real-time tasks, avoiding resource waste caused by average allocation, and improving scheduling accuracy. (3) The health index of access devices in this invention realizes dynamic correction of resource reservation, and adjusts the test resources of aging devices in advance through the prediction model, thereby improving the integrity of test data and the identification rate of equipment defects and optimizing resource utilization efficiency. Attached Figure Description

[0017] Figure 1 This is a system framework diagram of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] like Figure 1 This invention provides a test task scheduling and management system based on a 5G power virtual private network, comprising: The edge node module is deployed in the power grid coverage area and adopts a two-layer distributed architecture, containing multiple edge node MECs. The two-layer distributed architecture consists of core edge nodes and regional edge nodes. Core edge nodes are deployed in key locations such as substations and transmission line convergence points covered by the power private network, while regional edge nodes are deployed in distribution substations and areas with dense terminal equipment. The core edge nodes are equipped with high-performance processors and large-capacity storage, and are responsible for global resource catalog maintenance, cross-regional task collaboration, and strategy optimization. The regional edge nodes adopt a lightweight design, focusing on local task preprocessing and real-time scheduling, and support plug-and-play deployment; Regional edge nodes need to be configured with pluggable protocol parsing modules to support dynamic loading of power-specific protocols such as IEC 61850 and Modbus. The specific reasons include: Power equipment protocols are highly regional and equipment-dependent; for example, protection devices in different substations may use protocols from different manufacturers. Regional nodes need to parse local device data in real time, and dynamically loading protocol modules can avoid compatibility issues caused by device upgrades; The core edge node, as the central hub for global resource scheduling, is mainly responsible for generating cross-regional task collaboration strategies, maintaining the global resource catalog, and interacting with the cloud management platform to implement global scheduling strategies. The information it processes has already been parsed by the regional edge nodes, so it does not need to directly deal with the underlying device protocols. The edge nodes are networked using the lightweight Raft protocol to achieve task domain management and global resource directory synchronization; The layered architecture above addresses the issue of division of labor and collaboration, where core edge nodes are responsible for global strategy formulation, while regional edge nodes are responsible for local task execution. The lightweight Raft protocol addresses the information synchronization problem and serves as a communication language for distributed systems. The specific reasons for choosing it are as follows: The global scheduling strategy formulated by the core edge nodes needs to be synchronized to all regional edge nodes in real time to ensure cross-layer information synchronization. Regional edge nodes need to share local resource status to avoid resource misjudgment when scheduling cross-regional tasks, so as to ensure consistency of status within the same layer. For example, a core edge node formulates a strategy of "prioritizing relay protection testing during peak load periods of the provincial power grid" and synchronizes it to all regional edge nodes via the Raft protocol. If a regional edge node does not receive this strategy and still allocates resources for routine testing tasks, it may cause delays in relay protection testing and violate power grid safety regulations. The master-slave replication mechanism of the Raft protocol ensures that when an edge node in the core edge node cluster fails, other edge nodes can quickly take over through election, and when communication between regional edge nodes and core edge nodes is interrupted, scheduling is maintained based on the latest strategy cached locally. For example, a lightning strike caused a communication interruption between the core edge node and 30 regional edge nodes in a certain city's power grid. The local caching mechanism of the Raft protocol enabled the regional edge nodes to maintain scheduling for 40 minutes, during which no high-priority tasks were lost, thus improving reliability compared to a system without a consistency protocol (where an interruption means paralysis).

[0020] The TCI calculation module automatically classifies test tasks based on four-dimensional labels to obtain a four-dimensional judgment vector. The four-dimensional labels include priority, latency requirements, bandwidth requirements, and security level. When a test task accesses an edge node via a 5G slicing channel, the system reads the identity information carried by the task, including: Task types, such as relay protection testing, drone inspection testing, and cybersecurity attack simulation; Technical specifications, such as maximum latency, minimum bandwidth, and data encryption level; Business attributes, such as whether it belongs to the core business testing of the power grid and whether it affects real-time power supply; The four-dimensional judgment vector for this task is obtained based on the read identity information; Match the priority and security level labels with the weight level mapping table to obtain the weights corresponding to the two dimensions; The specific contents of the weight level mapping table are as follows: Priority is divided into three levels: high, medium, and low. Tests with high priority that directly affect the safe operation of the power grid, such as relay protection verification and real-time load control testing, need to be dispatched immediately, with a priority weight of 10. Medium priority indicates routine performance tests, such as network bandwidth stability tests and device compatibility tests, which can be queued and have a priority weight of 5. Low priority indicates non-real-time tasks, such as historical data statistics and system log backups, which can be executed when resources are idle, with a priority weight of 1; Security levels are specifically divided into three levels: highest, medium, and basic. The highest level of security involves testing core power grid control commands and user privacy data, with a security weight of 1. Medium security involves testing common business data, with a security weight of 0.5. Basic security involves testing with publicly available data, with a security weight of 0. The resulting four-dimensional judgment vector can ultimately be represented as (priority weight, latency requirement, bandwidth requirement, security weight); Based on the four-dimensional judgment vector, the Task Criticality Index (TCI) is calculated using the key dimension exponential amplification method. The specific operation is as follows: In power testing, priority (whether it affects grid security) and latency requirements (whether it is sensitive to real-time performance) are the core dimensions that determine the criticality of the task, while bandwidth and security level serve as auxiliary correction factors, according to the formula... Calculate the task criticality index, where P is the priority weight and D is the priority weight. norm B is the normalized value of the time delay. norm Here, S is the bandwidth normalization value, and S is the security weight. , , As a regulating factor, , , All three adjustment factors need to be set manually according to the actual situation; In the first part of the formula, that is The main approach is to amplify the impact between the two using an exponential function, which means that only tasks with both high priority and high latency requirements will see a significant increase in the total contribution of the core dimension, thus avoiding misjudgment of tasks with high priority in one dimension but low priority in another. In the bandwidth correction section, i.e. It decreases as bandwidth demand increases, reflecting that high bandwidth usage weakens scheduling priority; In the section on security level correction, namely The priority decreases as the security level increases, because high-security-level tasks require additional resource guarantees, which may reduce the actual scheduling priority.

[0021] The dynamic scheduling module incorporates a Lyapunov optimization model and constructs a Task Criticality Index (TCI) based on a four-dimensional judgment vector. It then dynamically adjusts task priority and resource allocation by combining the real-time resource status of edge nodes with network parameters. The specific reasons for choosing the Lyapunov optimization model in the built-in algorithm are as follows: Power testing tasks and resource status are highly dynamic, such as sudden access to relay protection testing and real-time fluctuations in 5G slicing bandwidth. Lyapunov optimization can handle the priority processing of sudden tasks by real-time queue status monitoring and decision-making without predicting the future arrival mode of tasks. When a power grid fault triggers an emergency test task with high TCI, the model can immediately adjust the queue priority and allocate resources first, avoiding the response delay of traditional static scheduling. In power testing scenarios, the arrival patterns of tasks and resource availability are often unpredictable. Lyapunov's optimized online scheduling mechanism avoids the dependence of traditional models (such as MPC) on future information. When new types of test tasks are introduced, the model only needs to calculate its TCI in real time to dynamically adjust resource allocation without retraining or parameter tuning. The dedicated power grid adopts a layered architecture of core + regional edge nodes, which can achieve Lyapunov-optimized distributed consistency characteristics; The specific steps for implementing Lyapunov-based optimized resource allocation using TCI are as follows: Establish a task queue Q={T1,T2,...,Tn}, and calculate TCIi for each task Ti using the key dimension exponential amplification method to form a weighted priority queue; Initialize the queue state Q(0), record the task arrival time, delay constraint Di, bandwidth requirement Bi and TCIi, and use them as input parameters for the Lyapunov optimization model; Edge nodes monitor available resources in real time, including remaining slice bandwidth Bavail(t), edge computing power Cavail(t), and network latency fluctuation ΔD(t), forming a resource state vector R(t)=[Bavail(t),Cavail(t),ΔD(t)]; Resource requirements for power testing tasks may change suddenly. For example, when drone inspections switch from standard definition to high definition, bandwidth requirements may increase from 100Mbps to 500Mbps. Real-time data collection ensures that allocation strategies are based on the latest resource availability. Establish the following constraints: Bi <= Bavail(t), Ci <= Cavail(t), ΔD(t) <= Di, to ensure that the allocation strategy meets the basic requirements of the task; By using constraints, we can prevent multiple high-TCI tasks from simultaneously preempting resources and causing system crashes. Constructing Lyapunov functions By introducing TCIi, the queue fluctuations of highly critical tasks are amplified, forcing the model to prioritize these tasks. A drift penalty mechanism is introduced to minimize the combined cost of queue latency and resource consumption. The specific formula is as follows: min(△L(Q(t))+V×C(t)), where △L is the queue stability index, C(t) is the resource consumption function, and V is the weight parameter for adjusting resource allocation and queue stability. Penalties prevent resources from being excessively allocated to high-TCI tasks, allowing low-TCI tasks to execute when resources are plentiful, thus improving overall utilization. Priority is given to allocating basic resources to high TCI tasks that meet latency constraints, and the remaining resources are dynamically allocated according to the TCIi ratio. ; For example, an edge node simultaneously handles a relay protection task (TCI=24, requiring 200Mbps) and two drone inspection tasks (TCI=5, each requiring 150Mbps), but the total bandwidth requirement is 500Mbps, with a remaining bandwidth of 400Mbps. The Lyapunov optimization model will prioritize satisfying the 200Mbps of the relay protection task, and allocate the remaining 200Mbps to the inspection task according to the TCI ratio (5:5), with each receiving 100Mbps. This ensures both the latency of critical tasks and controls the latency of inspection tasks. Based on the optimization results, resource allocation instructions are generated, including slice bandwidth bi(t) and computing power quota ci(t). High TCI tasks are triggered for immediate execution, while low TCI tasks enter the waiting queue. According to the formula , Update the state of the resource layer; For the task layer, remove executed tasks from the queue; By synchronizing the state of edge nodes through the lightweight Raft protocol, resource allocation conflicts caused by state asynchrony in distributed systems can be prevented. For example, when the core node issues the strategy of "prioritizing real-time control tasks during peak grid load periods", the regional node immediately adjusts the allocation: 200Mbps bandwidth is fixed for relay protection tasks (accounting for 40% of the slice), 150Mbps (30%) is allocated for equipment status monitoring tasks, and 150Mbps (30%) is allocated for non-real-time data backup.

[0022] The feedback adjustment module receives power equipment status monitoring data to generate an equipment health index, and dynamically adjusts the resource reservation for tasks based on the health index. Existing strategies rely on historical data to adjust resource reservations on an average basis, without fully considering the impact of real-time equipment health status on task criticality. For example, the testing of aging relay protection devices actually requires higher resource guarantees, but traditional methods only allocate resources based on the historical average TCI, which may lead to insufficient test accuracy. The specific steps are as follows: Access power equipment status monitoring data (such as circuit breaker operation count, transformer oil temperature, and protection device action time error), and generate an equipment health index H∈[0,1] by integrating multi-dimensional indicators through DS evidence theory; Power equipment status data comes from diverse sources, and the DS evidence theory can handle multi-source information with different confidence levels, avoiding misjudgment based on a single indicator. An LSTM prediction model is established for the equipment health index H. Based on historical data, the health level H(t+T) at time T is predicted, and the predicted H(t+T) is substituted into the correction factor formula K. H =1+q×(1-H(t+T)), where q is the adjustment coefficient, which balances the aggressiveness of resource reservation; Power equipment aging is often accompanied by nonlinear degradation. Predictive models can capture trend changes and adjust resource reservations in advance. If adjustments are made only based on the current H, resources may be added only after equipment performance has deteriorated, leading to test failures and the need for retrying. According to the formula The corrected resource reservation is calculated, where f(TCIi) is the basic resource reservation and ΔH is the health rate of change. The coefficient of sensitivity to change; Traditional models only consider the TCI characteristics of the task itself and do not reflect the status of the equipment performing the task. By using KH and ΔH, the risk of equipment aging is transformed into a calculable increase in resource demand, which is in line with the power operation and maintenance logic that high-risk equipment needs more testing and protection. This can prevent frequent resource adjustments caused by small fluctuations in health and reduce system turbulence; The parameter V in the Lyapunov drift penalty is also related to K. H Relatedly, according to the formula V=V0×K H The system is dynamically adjusted to amplify the impact of task queue fluctuations of aging equipment on system stability and prioritize resource allocation for it. Here, V0 is the initial parameter.

[0023] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0024] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A test task scheduling and management system based on a 5G power virtual private network, characterized in that, include: The edge node module is deployed in the power grid coverage area and adopts a two-layer distributed architecture, containing multiple edge node MECs. The TCI calculation module automatically classifies test tasks based on four-dimensional labels to obtain a four-dimensional judgment vector, and calculates the task criticality index (TCI) using the key dimension exponential amplification method. The four-dimensional labels specifically include priority, latency requirements, bandwidth requirements, and security level. The dynamic scheduling module incorporates a Lyapunov optimization model, which dynamically adjusts task priority and resource allocation based on the Task Criticality Index (TCI) and combined with the real-time resource status of edge nodes and network parameters. The feedback adjustment module receives power equipment status monitoring data to generate an equipment health index, and dynamically adjusts the resource reservation for tasks based on the health index.

2. The test task scheduling and management system based on a 5G power virtual private network according to claim 1, characterized in that, The two-layer distributed architecture consists of core edge nodes and regional edge nodes. The core edge nodes are equipped with high-performance processors and large-capacity storage, and are responsible for global resource catalog maintenance, cross-regional task collaboration, and strategy optimization. The regional edge nodes adopt a lightweight design and focus on local task preprocessing and real-time scheduling.

3. The test task scheduling and management system based on a 5G power virtual private network according to claim 2, characterized in that, The edge nodes of the region need to be configured with a pluggable protocol parsing module.

4. The test task scheduling and management system based on a 5G power virtual private network according to claim 1, characterized in that, Edge nodes are networked using the lightweight Raft protocol to achieve task domain management and global resource catalog synchronization.

5. The test task scheduling and management system based on a 5G power virtual private network according to claim 1, characterized in that, The specific steps to obtain the four-dimensional judgment vector are as follows: When a test task accesses an edge node through a 5G slicing channel, the system reads the identity information carried by the task to obtain the initial priority, latency requirements, bandwidth requirements, and security level. Match the priority and security level with the weight level mapping table to obtain the weights corresponding to the two dimensions; The resulting four-dimensional judgment vector can ultimately be represented as (priority weight, latency requirement, bandwidth requirement, security weight).

6. The test task scheduling and management system based on a 5G power virtual private network according to claim 5, characterized in that, The weight level mapping table specifically includes: Priority is divided into three levels: high, medium, and low, with corresponding weights of 10, 5, and 1, respectively. The security level is divided into three levels: highest, medium, and basic, with corresponding weights of 1, 0.5, and 0, respectively.

7. The test task scheduling and management system based on 5G power virtual private network according to claim 1, characterized in that, The steps for calculating TCI using the key dimension exponential amplification method are as follows: Based on the four-dimensional judgment vector, the latency requirement D and bandwidth requirement B are normalized to obtain Dnorm and Bnorm. According to the formula The Task Criticality Index (TCI) is calculated, where P is the priority weight and S is the safety weight. , , As a regulating factor and , , .

8. The test task scheduling and management system based on 5G power virtual private network according to claim 1, characterized in that, The specific steps of Lyapunov optimization of resource allocation based on TCI are as follows: Establish a task queue Q={T1,T2,...,Tn}, and calculate TCIi for each task Ti using the key dimension exponential amplification method; Initialize the queue state Q(0), record the task arrival time, delay constraint Di, bandwidth requirement Bi and TCIi, and use them as input parameters for the Lyapunov optimization model; Edge nodes monitor the remaining slice bandwidth Bavail(t), edge computing power Cavail(t), and network latency fluctuation ΔD(t) in real time, forming a resource state vector R(t)=[Bavail(t),Cavail(t),ΔD(t)]; Establish the following constraints: Bi <= Bavail(t), Ci <= Cavail(t), ΔD(t) <= Di; Constructing Lyapunov functions ; A drift penalty mechanism is introduced to minimize the combined cost of queue latency and resource consumption. The specific formula is as follows: min(△L(Q(t))+V×C(t)), where △L is the queue stability index, C(t) is the resource consumption function, and V is the weight parameter for adjusting resource allocation and queue stability. Priority is given to allocating basic resources to high TCI tasks that meet latency constraints, and the remaining resources are dynamically allocated according to the TCIi ratio. ; Based on the optimization results, resource allocation instructions are generated, including slice bandwidth bi(t) and computing power quota ci(t). High TCI tasks are triggered for immediate execution, while low TCI tasks enter the waiting queue. According to the formula , Update the state of the resource layer; For the task layer, remove executed tasks from the queue.

9. The test task scheduling and management system based on a 5G power virtual private network according to claim 1, characterized in that, The specific steps for dynamically adjusting task resource allocation based on the health index are as follows: By accessing power equipment status monitoring data and fusing multi-dimensional indicators through DS evidence theory, an equipment health index H∈[0,1] is generated; An LSTM prediction model is established for the equipment health index H. Based on historical data, the health level H(t+T) at time T is predicted, and the predicted H(t+T) is substituted into the correction factor formula K. H =1+q×(1-H(t+T)), where q is the adjustment coefficient; According to the formula The corrected resource reservation is calculated, where f(TCIi) is the basic resource reservation and ΔH is the health rate of change. This is the sensitivity coefficient to change.

10. The test task scheduling and management system based on a 5G power virtual private network according to claim 9, characterized in that, According to the formula V=V0×K H The parameter V in the Lyapunov drift penalty is dynamically adjusted, where V0 is the initial parameter.

Images