An adaptive intelligent scheduling method and system based on real-time state of power grid and multi-dimensional service label

Abstract

The application discloses a kind of based on power grid real-time state and multi-dimension service label adaptive intelligent scheduling method and system, the method includes: containerization division power grid secondary operation main station system, power grid secondary operation main station system is decoupled with containerization modeling to microservice;Based on the containerization division result of power grid secondary operation main station system, containerization modeling power grid service and defining multi-dimension service label system, define the standardization service label for each service;Real-time state input to intelligent dispatcher is perceived to power grid, and intelligent dispatcher calculates the dynamic weight of service label according to weight mapping rule, calculates each node score, generates optimal node, and corresponding deployable unit is scheduled to optimal node.The application establishes a set of strongly associated label system with power grid service, and designs a service sensing dispatcher capable of real-time sensing power grid operation state and dynamically adjusting scheduling strategy weight, to realize accurate, dynamic and safe matching of resource and service demand.

Description

Technical Field

[0001] This invention belongs to the field of power system automation and cloud computing technology. More specifically, it relates to a method and system for adaptive intelligent scheduling of microservices of the power grid secondary operation and maintenance master station system in a containerized environment, which can dynamically adjust the scheduling strategy according to the real-time operating status of the power grid. Background Technology

[0002] The service architecture is evolving, and containerization technologies (such as Docker) are being used for deployment to improve elasticity and efficiency. General-purpose container orchestration platforms (such as Kubernetes and Rancher) primarily rely on basic resources such as CPU and memory for balanced scheduling, and are unable to perceive the specific characteristics of power grid services.

[0003] The secondary system of the power grid has a wide variety of services with vastly different requirements for real-time performance, reliability, priority, and data consistency. The main problems include: (1) Real-time monitoring services require millisecond-level response. If deployed together with report generation services, monitoring delays may occur due to resource contention, leading to power grid risks; (2) Relay protection fault analysis is the highest priority service. When cluster resources are scarce, it is necessary to preempt resources from non-critical services such as historical data backup. The current general scheduling strategies cannot meet the above requirements, resulting in the system performance, reliability, and security after containerization deployment failing to meet the standards of the power grid production control area. This has become a bottleneck for the in-depth application of container technology in the core power grid system.

[0004] Specifically, in the practice of containerization of the power grid secondary operation and maintenance master station system, the existing and general container orchestration and scheduling technologies have the following inherent defects, making it difficult to directly apply them to power production control scenarios with extreme requirements for real-time performance, reliability and security.

[0005] 1. The scheduling strategy is out of sync with the semantics of power grid operations, resulting in unreasonable resource allocation. The default scheduler on platforms like Kubernetes is essentially a "resource scheduler," whose core decision-making is based on the availability and balance of resources such as CPU and memory on nodes. It cannot understand the inherent meaning of power grid services such as "relay protection," "real-time monitoring," and "power flow calculation," nor their differentiated requirements for the underlying infrastructure. This disconnect stems from its general-purpose design. General-purpose schedulers are designed to support stateless internet applications; their model abstraction level remains at the computing, network, and storage resource level, lacking the ability to perceive the attributes of higher-level services.

[0006] The consequences of this defect include: A. Insufficient guarantee of critical business resources: A high-priority "fault waveform analysis service" Pod may be waiting for scheduling together with a low-priority "report generation service" Pod due to insufficient resources, which fails to reflect the urgency of the business.

[0007] B. Performance requirements cannot be met: A "real-time data processing service" that requires millisecond-level response may be scheduled to a node with higher network latency simply because that node has more CPU resources, resulting in business processing timeouts and affecting real-time monitoring of the power grid.

[0008] C. Non-critical business instances preempting resources: When cluster resources are scarce, a large number of non-core business instances may occupy the resources of high-performance nodes, while the core businesses that truly need these resources cannot be scheduled.

[0009] 2. The lack of mandatory constraints for power grid security zoning poses a security risk. General-purpose schedulers typically implement soft constraints through labels and selectors, but they lack native support for mandatory, hard isolation requirements such as "security partitions." Misconfiguration can lead to services in the production control zone being scheduled to nodes in the management information zone. This risk stems from the design philosophy of general-purpose platforms, which prioritize maximizing resource integration and sharing. Their security model focuses on network policies rather than providing strong physical or logical isolation for workload placement at the scheduling level.

[0010] The consequences of this flaw include a direct violation of the core principles of "security zoning, dedicated networks, and lateral isolation" in power monitoring system security regulations. Once cross-regional dispatching occurs, it logically breaks down security zoning, making an attack path from the management information zone to the production control zone possible, thus posing a serious cybersecurity risk.

[0011] 3. The high availability mode is limited and cannot adapt to the complex redundancy mechanisms of the power grid. Kubernetes' Deployment and other workload controllers, with their multi-instance redundancy mode, are essentially "multi-active" load balancing modes. However, core power grid control services (such as SCADA and stability control) widely employ the traditional, proven "active-standby" mode, where the standby machine is in a hot standby state but does not process services. This mismatch stems from the fact that general-purpose orchestration platforms originate from internet applications, and their design paradigm assumes that all instances are completely peer-to-peer and substitutable, which is inconsistent with the state machine model of many stateful control services in the power grid.

[0012] The consequences of this defect include: A. Complex master-slave failover logic: Implementing the classic master-slave mode requires relying on external master election tools (such as through Kubernetes' Leader Election mechanism), which increases the complexity of the system and the potential for failure.

[0013] B. Resource waste: A simple "master-slave mode" means that the standby instance does not bear the load during normal times, resulting in idle resources. The general scheduler cannot intelligently and safely distribute low-priority batch processing tasks on the standby node.

[0014] C. Fault recovery is not intuitive: When the primary node fails, the scheduler will reschedule a new instance, but the startup, data synchronization and state recovery process of this new instance may not be as fast and reliable as a simple "standby to primary" process.

[0015] The fundamental reason for the aforementioned shortcomings of existing general-purpose container orchestration technologies lies in the essential misalignment between their original design intent and the inherent characteristics of power grid production control operations (high real-time performance, high reliability, strong security, and complex redundancy). Simply applying them to power grid secondary systems will only lead to compromises in performance, security, and reliability, failing to truly leverage the advantages of containerization. The most critical issue is that their scheduling strategies are statically preset and cannot perceive or respond to the real-time dynamic changes in the power grid's operating status. Therefore, there is an urgent need for an intelligent scheduling solution that can understand the semantics of power grid operations and dynamically adapt to changes in the power grid's operating status. Summary of the Invention

[0016] To address the shortcomings of existing technologies, this invention provides an adaptive intelligent scheduling method and system based on real-time power grid status and multi-dimensional business tags. The method and system establish a tagging system strongly correlated with power grid services and design a business-aware scheduler capable of real-time sensing of power grid operating status and dynamically adjusting scheduling strategy weights, thereby achieving accurate, dynamic, and secure matching of resources and business needs.

[0017] The present invention adopts the following technical solution.

[0018] The first aspect of this invention provides an adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional service tags, comprising the following steps: The power grid secondary operation and maintenance master station system is containerized, and microservice decoupling and container modeling are performed on the power grid secondary operation and maintenance master station system. Based on the containerization partitioning results of the power grid secondary operation and maintenance master station system, power grid business is containerized and a multi-dimensional business tag system is defined, and standardized business tags are defined for each service. The system senses the real-time status of the power grid and inputs it into the intelligent dispatcher. The intelligent dispatcher calculates the dynamic weight of the service tag according to the weight mapping rules, calculates the score of each node, generates the optimal node, and dispatches the corresponding deployable unit to the optimal node.

[0019] Preferably, the containerized power grid secondary operation and maintenance master station system includes: The power grid secondary operation and maintenance master station system is divided into: data acquisition and monitoring module, power grid analysis and control module, historical data and archiving module, graphics and model management module, integrated application module and platform support service module, and the specific services in each module are defined at the container level.

[0020] Preferably, the multi-dimensional business tagging system includes: tag categories, tag keys, optional values, and scheduling impact; The tag categories include: business criticality, real-time requirements, computing characteristics, high availability mode, data affinity, and security partitioning.

[0021] Preferably, the scheduling process of the intelligent scheduler includes: Users submit deployable unit definitions with business tags; The business-aware intelligent scheduler listens for deployable unit creation events; Entering the pre-selection stage includes: forcibly filtering cluster nodes based on the business tags of deployable units to obtain nodes that pass the pre-selection; Entering the optimization stage includes: obtaining the real-time status of the power grid through real-time status perception, the intelligent dispatcher calculating dynamic weights according to the weight mapping rules, and then scoring the pre-selected nodes to generate the optimal node; Perform the binding and deployment of the unit to the optimal node.

[0022] Preferably, the mandatory filtering in the pre-selection phase includes: Security partition mandatory isolation: The security partition label of the deployable unit must exactly match this label of the node; otherwise, it will be excluded. Hardware capability matching: Filter by computing characteristic tags and retain nodes with corresponding tags; Real-time performance guarantee: Filter the tags that meet real-time requirements and retain nodes that meet network performance requirements.

[0023] Preferably, the intelligent scheduler calculates the dynamic weights according to the weight mapping rules, including: The intelligent scheduler maintains a weight mapping rule base, which defines a nonlinear relationship between the power grid state and the scheduling strategy weights. The intelligent scheduler takes the real-time power grid state as input and changes or maintains the weights of service tags based on the nonlinear relationship.

[0024] Preferably, in the fault-priority weight mapping rule: when a power grid fault is detected, the weight of the business criticality label is increased; and the weight of the calculation characteristic label is decreased accordingly.

[0025] Preferably, the calculation of the score for each node includes: Calculate the business criticality score, real-time performance and data affinity score, and high availability and balance score, and sum them according to dynamic weights to obtain the score for each node; The business criticality score, real-time performance and data affinity score, and high availability and balance score are obtained through corresponding tags.

[0026] Preferably, core nodes are identified through node tags or real-time monitoring data for business criticality scoring; Real-time performance and data affinity scores are obtained by calculating the network latency between the node and the data source specified by the data affinity label. By using the high availability mode label, for services in primary-standby mode, strict anti-affinity ensures that the primary and standby replicas are distributed across different physical nodes or availability zones; for services in active-active mode, soft anti-affinity is used to distribute instances to achieve load balancing and fault isolation.

[0027] The second aspect of the present invention provides an adaptive intelligent scheduling system based on real-time power grid status and multi-dimensional business tags, which executes an adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional business tags according to the first aspect, characterized in that it includes: a bottom infrastructure layer, a containerized secondary system platform layer, and a core business application layer. The underlying infrastructure layer is used to provide a unified containerized runtime environment, computing and storage resources for the execution of the adaptive intelligent scheduling method; The containerized secondary system platform layer is used to carry and implement the core logic of the adaptive intelligent scheduling method, including the dynamic weight adaptive scheduler, container network and service governance components required to implement the adaptive intelligent scheduling method. The core business application layer is used to provide specific scheduling objects and decision-making basis for the adaptive intelligent scheduling method. It contains various power grid secondary microservices that are strictly deployed according to security partitions and carry multi-dimensional business tags. These microservices are the targets and input sources for the adaptive intelligent scheduling method to perform business perception and dynamic response. Compared with the prior art, the beneficial effects of the present invention include at least the following: 1. Dynamic Business-Driven Scheduling: For the first time, the semantics of power grid business are systematically introduced into container scheduling decisions, and the real-time operating status of the power grid is incorporated as a core decision factor into container scheduling, achieving a leap from "static resource scheduling" to "dynamic business assurance." The scheduling strategy can adaptively adjust with changes in the power grid situation, significantly improving the level of intelligence.

[0028] 2. Precise protection in fault scenarios: When a power grid fault occurs, the system can automatically increase the scheduling priority of core control services to ensure that they can quickly obtain the necessary resources, thus winning valuable time for rapid fault handling and greatly enhancing the power grid's emergency response capabilities.

[0029] 3. Situational awareness optimization of resource efficiency: Under different power grid operating conditions (such as peak and normal), the resource allocation strategy is dynamically adjusted to ensure that infrastructure resources always serve the most urgent business needs, thereby achieving global optimization of resource utilization efficiency.

[0030] 4. Enhanced intrinsic safety: By implementing a mandatory pre-selection strategy for security partition labels, strict logical isolation is achieved at the scheduling level, meeting the regulatory requirements for security protection of power monitoring systems and reducing security risks.

[0031] 5. Enhanced System Elasticity: By combining high-availability business models with anti-affinity strategies, redundant deployment of core business processes is automatically achieved. When a node fails, the scheduler can intelligently and quickly reschedule the business to a compliant node based on business tags, achieving self-healing.

[0032] 6. Platform Universality and Flexibility: The tag system and method proposed in this invention are universal, applicable to Kubernetes and adaptable to other container platforms such as Rancher. The tag system is easily extensible, allowing for the addition of new dimensions as business grows. Attached Figure Description

[0033] Figure 1 This is a flowchart of the dynamic weighted adaptive scheduler based on the real-time state of the power grid, provided in accordance with an embodiment of the present invention. Figure 2 This is a schematic diagram of the containerized architecture and service partitioning of the power grid secondary system provided in accordance with an embodiment of the present invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0035] Embodiment 1 of the present invention provides an adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional service tags, comprising the following steps: Step 1: Containerize the power grid secondary operation and maintenance master station system. In order to realize intelligent scheduling based on business awareness, the traditional power grid secondary operation and maintenance master station system is decoupled and containerized through microservices.

[0036] Preferably, but not limitingly, the present invention follows the principles of "high cohesion, low coupling" and "single responsibility" and divides the power grid secondary operation and maintenance master station system into six core functional modules, including: data acquisition and monitoring module, power grid analysis and control module, historical data and archiving module, graphics and model management module, integrated application module and platform support service module, and defines the specific services in each module at the container level.

[0037] For example, but not restrictively, the containerized names, functions, and key characteristics of each service are shown in the table below:

[0038] Step 2: Based on the containerized partitioning results of the power grid secondary operation and maintenance master station system in Step 1, containerized modeling of power grid services and definition of a multi-dimensional service tag system are performed. Specifically, the power grid secondary operation and maintenance master station system is decomposed into microservices, and standardized service tags are defined for each service. It is worth noting that this is the data foundation for achieving intelligent scheduling.

[0039] Preferably, but not limitingly, the multi-dimensional business tagging system includes: tag categories, tag keys, optional values, and scheduling impact; more preferably, but not limitingly, the tag categories include: business criticality, real-time requirements, computing characteristics, high availability mode, data affinity, and security partitioning.

[0040] For example, but not restrictively, the core tag system is shown in the table below:

[0041] As a further example, but not a limitation, the relay protection fault recording analysis service based on the core tagging system is defined as: yaml labels: grid.business / criticality: mission-critical grid.business / latency: <50ms grid.business / compute-type: cpu-intensive grid.ha / mode: active-standby grid.data / locality: real-time-db grid.security / zone: production-control It is worth noting that, as one of the outstanding substantive features of this invention, this invention proposes a set of "semantic tagging system" specifically for describing power grid containerization services. It is not just a simple technical tag, but a set of "dictionaries" or "protocols" that transform the abstract requirements of power grid services (such as "mission accomplished" and "<10ms latency") into instructions that can be understood and executed by the container platform.

[0042] A significant difference from existing technologies is that existing technologies (such as Kubernetes) use generic tags without business semantics (e.g., app: v1), primarily for grouping and selection. In contrast, the tags in this invention (e.g., grid.business / criticality: mission-critical) inherently contain business logic and scheduling strategies. The significant advancements compared to existing technologies include at least the following: in subsequent steps, when the scheduler sees this tag, it directly understands a series of implicit requirements such as "highest priority scheduling required, preemption allowed, and deployment on the most reliable node."

[0043] It is particularly worth emphasizing that dividing the power grid secondary operation and maintenance master station system into six core functional modules in step 1, and defining six tag categories and tag keys, optional values, and scheduling effects in step 2 based on step 1, are all preferred but non-limiting implementation methods. Those skilled in the art will understand that adopting the core concept of this invention, dividing it into more or fewer modules, more or fewer tag categories, or using tags not exemplified in this invention, all fall under the category of utilizing the tag system architecture and definition method of this invention, are all used to establish their mapping relationship with power grid business attributes, and all fall within the scope of this invention.

[0044] Step 3: Sensing the real-time status of the power grid, the intelligent dispatcher calculates the dynamic weight of the service tag according to the weight mapping rules, generates the optimal node, and schedules the corresponding deployable unit to the optimal node.

[0045] Specifically, as one of the core concepts of this invention, it employs a dynamic weighted adaptive scheduler. Its innovation lies in introducing "real-time grid status" as the third-dimensional dynamic input for scheduling decisions, and achieving online adaptive adjustment of the scheduling strategy through "weight mapping rules." Its workflow is as follows: Figure 1 As shown, the workflow of the dynamic weighted adaptive scheduler based on the real-time state of the power grid is illustrated.

[0046] Preferably, but not restrictively, the core scheduling process in step 3 specifically includes: Step 3.1: The user submits a Pod definition with business tags.

[0047] Step 3.2: The business-aware intelligent scheduler listens for Pod creation events.

[0048] Step 3.3, proceed to the pre-selection stage, which includes: performing mandatory filtering on cluster nodes based on the business tags of the Pods to obtain the nodes that pass the pre-selection.

[0049] Further preferably, but not limitingly, the mandatory filtering includes: Security partition mandatory isolation: The security partition label key of a Pod, i.e., grid.security / zone, must exactly match this label on the node; otherwise, it will be excluded. This is the highest priority hard constraint.

[0050] Hardware capability matching: For Pods whose compute feature tag key value is GPU required, i.e., grid.business / compute-type: gpu-required, only nodes with the tag gpu: "true" will be retained.

[0051] Real-time performance guarantee: For Pods with a real-time requirement tag key value of less than 10ms, i.e., grid.business / latency:<10ms, only nodes equipped with high-performance network devices are retained, such as, but not limited to, SR-IOV (Single Root I / O Virtualization).

[0052] Step 3.4, proceeding to the optimization stage, includes: obtaining the real-time status of the power grid through real-time status perception, the intelligent dispatcher calculating dynamic weights according to the weight mapping rules, and then scoring the pre-selected nodes to generate the optimal node.

[0053] Further preferred, but not limiting, the real-time status awareness includes: the scheduler acquiring dynamic status indicators in real time from the power grid EMS (Energy Management System), SCADA (Supervisory Control and Data Acquisition) system, etc., through standard API interfaces, such as, but not limited to, RESTful APIs, including but not limited to: grid_fault_alarm: Grid fault alarm signal (Boolean value); load_rate: Power flow load rate (percentage) of critical transmission sections; total_load: Total system load (megawatts).

[0054] Further preferred, but not limiting, the calculation of dynamic weights includes: maintaining a weight mapping rule base within the scheduler, which defines the nonlinear relationship between the power grid state and the scheduling strategy weights.

[0055] For example, but not limited to, in the example of fault-priority rules: when grid_fault_alarm == true, significantly increase the weight W_criticality of the grid.business / criticality tag, for example, but not limited to, from 1.0 to 3.0; and correspondingly decrease the weight of non-core tags such as grid.business / compute-type.

[0056] Further preferred, but not limiting, the scoring of the pre-selected nodes includes: scoring the pre-selected nodes using dynamically adjusted weights.

[0057] Furthermore, the core algorithm for scoring the pre-selected nodes includes: Pods with a mission-critical score (S_criticality) tend to be scheduled to "core nodes" with stronger resource guarantees and a more stable historical operation. This can be identified through the node label `node-type: core` or real-time monitoring data.

[0058] Real-time performance and data affinity score S_locality: Network latency between the compute node and the data source specified by grid.data / locality (such as the node where the real-time database service resides) (obtained from the monitoring system). The lower the latency, the higher the score. Application Pod affinity, such as "Fault Analysis Service" being preferentially deployed in the same topology domain as "Real-time Database Service".

[0059] High availability and load balancing score S_ha: For services with grid.ha / mode: active-standby, strict Pod anti-affinity ensures that the primary and standby replicas are distributed across different physical nodes or availability zones. For active-active services, soft anti-affinity is used to distribute instances as widely as possible to achieve load balancing and fault isolation.

[0060] For example, but not restrictively, the final score for the pre-selected nodes is calculated using the following formula: Score = (W_criticality * S_criticality) + (W_locality * S_locality) +(W_ha * S_ha) + ... in: S_xx are the raw scores based on labels, namely S_criticality is the raw score for business criticality, S_locality is the raw score for real-time performance and data affinity, and S_ha is the raw score for high availability and balance. W_xx are dynamically adjusted weight coefficients, W_criticality is the dynamically adjusted weight coefficient for business criticality score, W_locality is the dynamically adjusted weight coefficient for real-time performance and data affinity score, and W_ha is the dynamically adjusted weight coefficient for high availability and balance score.

[0061] This formula shows that the real-time status of the power grid directly affects the final ranking of nodes, thus producing the optimal node.

[0062] Step 3.5: Perform the binding Pod scheduling to the optimal node.

[0063] It is worth noting that, as one of the prominent substantive features of this invention, it provides a dynamic weighted adaptive scheduling mechanism based on "real-time power grid status awareness," and on this basis, a custom scheduler that can "sense the pulse of the power grid" and "intelligently respond" is provided. Its core innovation lies in the secondary upgrade of the decision-making logic: from "resource surplus scheduling" to "business demand scheduling," and then to "situation-aware dynamic business scheduling."

[0064] A significant difference from existing technologies lies in the fact that existing schedulers make decisions based on: Node.Remaining CPU > Pod.Requested CPU (static resource view). In contrast, the scheduler of this invention makes decisions based on: [Node.Security Partition == Pod.Security Partition] AND [(W_criticality * S_criticality) + (W_latency * S_latency) + ... ], where W_xx is a function that dynamically changes with the real-time state of the power grid (dynamic service view). Specifically, the significant advancements brought by this invention compared to existing technologies include at least the introduction of the core concept of "weight mapping rules," which defines non-linear mapping relationships such as "increasing the weight of critical services when a power grid fault alarm is triggered."

[0065] It is particularly worth emphasizing that implementing the pre-selection phase with mandatory isolation of security partitions, hardware capability matching, and real-time basic guarantees, and using dynamic weight adjustment calculation of node scores based on business criticality scores, real-time and data affinity scores, and high availability and balance scores are all preferred but non-limiting implementation methods. Those skilled in the art will understand that, using the core concept of this invention, any dynamically changing weight coefficient obtained by a function that dynamically changes with the real-time state of the power grid (i.e., the dynamic workflow of the scheduler), and weighting the static business scores to achieve dynamic adaptation of the final decision (i.e., adaptive algorithm) all fall within the scope of this invention.

[0066] like Figure 2 As shown, Embodiment 2 of the present invention provides an adaptive intelligent scheduling system based on real-time power grid status and multi-dimensional business tags, which runs the adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional business tags as described in Embodiment 1, including: a bottom infrastructure layer, a containerized secondary system platform layer, and a core business application layer.

[0067] The underlying infrastructure layer is used to provide a unified containerized operating environment, computing and storage resources for the execution of the adaptive intelligent scheduling method, and is the physical entity that carries the scheduling decision in the method; The containerized secondary system platform layer is used to carry and implement the core logic of the adaptive intelligent scheduling method, specifically including core components such as the dynamic weight adaptive scheduler, container network and service governance required to implement the method. The core business application layer is used to provide specific scheduling objects and decision-making basis for the adaptive intelligent scheduling method. Its internal network secondary microservices, which are strictly deployed according to security partitions and carry multi-dimensional business tags, are the target and input source for the method to perform "business perception" and "dynamic response".

[0068] Embodiment 3 of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded onto the processor, it implements an adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional business tags as described in Embodiment 1.

[0069] Embodiment 4 of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements an adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional business tags as described in Embodiment 1.

[0070] To more clearly illustrate the outstanding substantive features of this invention and the significant advancements it brings to the prior art, an application example of implementing this invention is described below. Specifically, taking the deployment of a "relay protection fault recording and analysis service" in a Kubernetes cluster as an example, an adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional service tags is introduced, including the parts on defining service workloads and scheduling decision-making processes.

[0071] For example, but not limitingly, in the business workload definition section, the YAML file is configured as follows: apiVersion: apps / v1 kind: Deployment metadata: name: relay-protection-analyzer spec: replicas: 1 template: metadata: labels: app: relay-protection-analyzer grid.business / criticality: "high" grid.business / latency: "<5s" grid.business / compute-type: "cpu-intensive" grid.ha / mode: "active-standby" grid.data / locality: "real-time-db" grid.security / zone: "production-control" grid.dynamic / scale-on-fault: "true" spec: schedulerName: grid-dynamic-scheduler containers: - name: fault-analyzer image: grid / relay-protection-analyzer:v2.1 resources: requests: CPU: "2000m" memory: "4Gi" For example, but not limitingly, the dispatch decision-making process, taking a transient short-circuit fault on a 500kV line of a power grid as an example, includes the following steps: Step 1: Event triggering, including: the SCADA system detects a fault and broadcasts to the cluster: grid_fault_alarm= true.

[0072] Step 2: Dynamic weight adjustment, including: the scheduler matches the fault-emergency-rule and immediately adjusts the weights: criticality weight: 1.0 → 3.0, real-time weight: 1.0 → 1.5, calculation type weight: 1.0 → 0.7.

[0073] Step 3: Workload scaling, including: HPA automatically expands the number of replicas from 1 to 2 based on the fault event label.

[0074] Step 4: Pre-selection phase, including: security partition verification: excluding nodes in non-production control areas; resource requirement filtering: excluding nodes with insufficient CPU; pre-selection results: multiple nodes such as node-core-01 enter the optimization phase.

[0075] Step 5: In the optimization stage, dynamic weight scoring is performed. Taking node-core-01 as an example, the same applies to the other nodes.

[0076] Business criticality score: 90 × 3.0 = 270 points Real-time performance score: 95 × 1.5 = 142.5 points Data affinity score: 90 × 1.0 = 90 points Computational characteristic score: 85 × 0.7 = 59.5 points Final score: 562 points After comparison, node-core-01 scored the highest with 562, so node-core-01 was ultimately selected. It's understandable that using node-core-01 as an example and the final node is merely illustrative; in practical engineering, after the same dynamic scoring, the highest score would be selected as the final node.

[0077] Step Six: Decision Making and Scheduling The scheduler schedules the fault analysis Pod to node-core-01.

[0078] Furthermore, the comparison of scheduling effects between the traditional scheduler and the dynamic scheduler of this invention is shown in the table below:

[0079] This embodiment fully demonstrates the technical advantages of the present invention: through a dynamic weighting mechanism, a "green channel" is provided for core business at critical moments, reducing fault analysis time from minutes to seconds, and directly improving the efficiency and reliability of power grid fault handling.

[0080] In summary, after a detailed description of the specific embodiments and application examples of this invention, those skilled in the art can clearly understand that, as one of the prominent substantive features of this invention, the "design of the tag system" and the "intelligent scheduling method based on the tag system" complement each other and are inseparable in constituting the core concept of this invention. Specifically, this invention constructs a novel secondary system architecture that, in a containerized environment, can both inherit the traditional core requirements of the power grid (security, reliability, and real-time) and enjoy the elasticity and automation benefits of cloud-native technologies by creating a tag system that maps the characteristics of power grid services and designing an intelligent scheduler that can understand this system.

[0081] It is worth noting that in the embodiments of the present invention, "steps + numbers" is only an expression for clearly describing the specific implementation of the adaptive intelligent scheduling method based on the real-time status of the power grid and multi-dimensional business tags, and is not an absolute restriction on the order of the steps. Under the guidance of the core concept of the present invention, changing the order of these steps to achieve the same or similar technical effects all fall within the scope of the present invention.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. An adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional business tags, characterized in that, Includes the following steps: The power grid secondary operation and maintenance master station system is containerized, and microservice decoupling and container modeling are performed on the power grid secondary operation and maintenance master station system. Based on the containerization partitioning results of the power grid secondary operation and maintenance master station system, power grid business is containerized and a multi-dimensional business tag system is defined, and standardized business tags are defined for each service. The system senses the real-time status of the power grid and inputs it into the intelligent dispatcher. The intelligent dispatcher calculates the dynamic weight of the service tag according to the weight mapping rules, calculates the score of each node, generates the optimal node, and dispatches the corresponding deployable unit to the optimal node.

2. The adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional service tags according to claim 1, characterized in that: The containerized power grid secondary operation and maintenance master station system includes: The power grid secondary operation and maintenance master station system is divided into: data acquisition and monitoring module, power grid analysis and control module, historical data and archiving module, graphics and model management module, integrated application module and platform support service module, and the specific services in each module are defined at the container level.

3. The adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional service tags according to claim 2, characterized in that: The multi-dimensional business tagging system includes: tag categories, tag keys, optional values, and scheduling impact; The tag categories include: business criticality, real-time requirements, computing characteristics, high availability mode, data affinity, and security partitioning.

4. The adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional service tags according to claim 3, characterized in that: The scheduling process of the intelligent scheduler includes: Users submit deployable unit definitions with business tags; The business-aware intelligent scheduler listens for deployable unit creation events; Entering the pre-selection stage includes: forcibly filtering cluster nodes based on the business tags of deployable units to obtain nodes that pass the pre-selection; Entering the optimization stage includes: obtaining the real-time status of the power grid through real-time status perception, the intelligent dispatcher calculating dynamic weights according to the weight mapping rules, and then scoring the pre-selected nodes to generate the optimal node; Perform the binding and deployment of the unit to the optimal node.

5. The adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional service tags according to claim 4, characterized in that: The mandatory filtering in the pre-selection phase includes: Security partition mandatory isolation: The security partition label of the deployable unit must exactly match this label of the node; otherwise, it will be excluded. Hardware capability matching: Filter by computing characteristic tags and retain nodes with corresponding tags; Real-time performance guarantee: Filter the tags that meet real-time requirements and retain nodes that meet network performance requirements.

6. An adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional service tags according to claim 4 or 5, characterized in that: The intelligent scheduler calculates dynamic weights according to weight mapping rules, including: The intelligent scheduler maintains a weight mapping rule base, which defines the nonlinear relationship between the power grid state and the scheduling strategy weights. The intelligent dispatcher takes the real-time state of the power grid as input and changes or maintains the weight of the service tags based on the nonlinear relationship.

7. The adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional service tags according to claim 6, characterized in that: In the fault-priority weighting mapping rule: when a power grid fault is detected, the weight of the business criticality label is increased; and the weight of the calculation characteristic label is decreased accordingly.

8. An adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional service tags according to claim 4 or 5, characterized in that: The calculation of the score for each node includes: Calculate the business criticality score, real-time performance and data affinity score, and high availability and balance score, and sum them according to dynamic weights to obtain the score for each node; The business criticality score, real-time performance and data affinity score, and high availability and balance score are obtained through corresponding tags.

9. The adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional service tags according to claim 8, characterized in that: Core nodes are identified through node labels or real-time monitoring data for business criticality scoring. Real-time performance and data affinity scores are obtained by calculating the network latency between the node and the data source specified by the data affinity label. By using the high availability mode label, for services in primary-standby mode, strict anti-affinity ensures that the primary and standby replicas are distributed across different physical nodes or availability zones; for services in active-active mode, soft anti-affinity is used to distribute instances to achieve load balancing and fault isolation.

10. An adaptive intelligent scheduling system based on real-time power grid status and multi-dimensional service tags, executing the adaptive intelligent scheduling method based on real-time power grid status and multi-dimensional service tags according to any one of claims 1 to 9, characterized in that, include: The underlying infrastructure layer, the containerized secondary system platform layer, and the core business application layer; The underlying infrastructure layer is used to provide a unified containerized runtime environment, computing and storage resources for the execution of the adaptive intelligent scheduling method; The containerized secondary system platform layer is used to carry and implement the core logic of the adaptive intelligent scheduling method, including the dynamic weight adaptive scheduler, container network and service governance components required to implement the adaptive intelligent scheduling method. The core business application layer is used to provide specific scheduling objects and decision-making basis for the adaptive intelligent scheduling method. It contains various power grid secondary microservices that are strictly deployed according to security partitions and carry multi-dimensional business tags. These microservices are the targets and input sources for the adaptive intelligent scheduling method to perform business perception and dynamic response.

Images