Power distribution network resource scheduling method and system based on computing power network integration

Abstract

The invention provides a power distribution network resource scheduling method and system based on computing power network integration, and belongs to the technical field of data processing and resource scheduling, and the method comprises the steps: obtaining a plurality of communication signals and a plurality of corresponding service requests, and packaging the communication signals and the service requests into a plurality of standardized data; corresponding response tasks are generated according to the service requests, and task resource requirements of the response tasks are determined; performing resource scheduling on each response task according to each task resource demand and a preset resource scheduling model, and determining resource configuration of each response task; according to each resource configuration, scheduling and allocating a plurality of virtual resources to each response task from a preset resource pool; and according to each response task and a plurality of corresponding virtual resource units, carrying out data processing on each piece of standardized data, so that by implementing the method and the device, the problems of relatively low utilization rate of power distribution network resources and inflexible scheduling in the prior art can be solved.

Description

Technical Field

[0001] This application relates to the fields of data processing and resource scheduling technology, and in particular to a distribution network resource scheduling method and system based on integrated computing power and network. Background Technology

[0002] The large-scale integration of distributed energy resources and diversified loads into new power distribution networks is driving the evolution of power distribution and consumption services towards higher interactivity and stronger real-time performance, posing a fundamental challenge to the traditional architecture that separates communication and computing. In multi-service concurrent scenarios, the communication needs of heterogeneous terminals differ significantly, and existing single communication standards struggle to dynamically adapt to complex network topology changes. Edge computing resources lack flexible scheduling capabilities, making it difficult to guarantee the timeliness and reliability of service responses. Furthermore, the risk of computing resource failure caused by extreme environments further restricts the improvement of power distribution network resilience.

[0003] In current research and application practices, a series of analytical methods and technical frameworks have emerged to address the resource management and allocation issues of converged services in computing networks. One common approach focuses on establishing a correlation model between services and various resources. By collecting and analyzing historical data and operational status, a quantitative evaluation mechanism is formed, generating resource allocation or optimization strategy recommendations accordingly. While this approach has some theoretical value, its analytical process typically emphasizes the assessment of relatively stable resource states, failing to fully consider the complexities arising from the diversity of access methods, the instantaneous fluctuations in task requirements, and the dynamic adjustments to network topology in actual deployment environments. Furthermore, because such methods generally lack effective support for unified pooling and virtualization scheduling of underlying resources, their ability to handle high-concurrency, high-real-time application scenarios is limited, especially in meeting the operational requirements of systems like the new power distribution IoT, which have stringent demands for service determinism and response speed.

[0004] On the other hand, to address the issues of efficient discovery and service-oriented provision of computing resources, the industry has proposed various service-oriented partitioning and access guidance strategies. A representative approach is to abstract network areas with specific service capabilities into logical "service domains" based on the distribution and connectivity of computing nodes within the physical network, and then publish the available computing power attributes of these areas. When a terminal device enters the wireless coverage area of ​​a service domain, it can discover and use the corresponding computing power services based on pre-acquired attribute information. This concept simplifies the user's perception and invocation process of distributed computing power to some extent. However, when facing highly heterogeneous and constantly changing field environments—such as frequently changing link quality, diverse terminal types, and sudden local loads—this type of method is still insufficient in terms of autonomous identification and adaptive selection of access signals. Especially in scenarios requiring intelligent optimization and seamless switching between multiple communication modes (such as 5G, Ethernet, and industrial wireless), simply relying on preset service domain attributes may not fully guarantee the reliability and efficiency of the connection, thus affecting the overall system's agile response and robust operational capabilities. Summary of the Invention

[0005] To address the aforementioned technical problems, this application provides a distribution network resource scheduling method and system based on integrated computing power and network, which can solve the problems of insufficient flexibility in distribution network resource scheduling and low resource utilization in the prior art.

[0006] In a first aspect, embodiments of this application provide a distribution network resource scheduling method based on integrated computing power and network, comprising: Acquire several communication signals and their corresponding service requests; According to the preset protocol format, each of the communication signals is encapsulated to obtain a number of standardized data; Generate corresponding response tasks based on each of the aforementioned business requests, and determine the task resource requirements of each of the aforementioned response tasks; If the current state of the distribution network system meets the preset conditions, then resource scheduling is performed on each of the response tasks according to the resource requirements of each task and the preset resource scheduling model to determine the resource configuration of each response task; otherwise, the resource configuration of each response task is determined according to the resource requirements of each task; wherein, the resource scheduling model is constructed based on the objective function and several constraints. According to the resource configurations, several virtual resources are scheduled and allocated to each response task from a preset resource pool. The resource pool is constructed by converting communication resources, computing resources and storage resources into corresponding virtual resource units. Based on each of the aforementioned response tasks and the corresponding virtual resource units, the standardized data is processed.

[0007] This application proposes a distribution network resource scheduling method based on integrated computing and network capabilities. It pools dispersed communication, computing, and storage resources and dynamically allocates virtual resources through a resource scheduling model, enabling the system to respond quickly to demands in high-concurrency and multi-service scenarios. Compared to traditional static resource allocation models, this embodiment significantly improves resource utilization and the flexibility of distribution network resource scheduling. For example, in the case of large-scale distributed energy access, traditional methods often lead to response delays and resource waste due to resource silos and rigid scheduling mechanisms. This embodiment, however, by monitoring the distribution network status in real time and combining it with an optimization model, can adaptively adjust resource allocation, ensuring low-latency and high-reliability processing of critical services (such as fault detection and load control). This embodiment also encapsulates different types of communication signals, unifying signals from different sources and converting them into a standard format. The encapsulated data can be uniformly understood by the system, ensuring that the entire system can efficiently process signals from different sources, avoiding conflicts and compatibility issues between different protocols, thus facilitating subsequent processing and scheduling. Furthermore, the introduction of virtualization technology reduces hardware dependence, enhances system scalability and fault tolerance, and provides strong technical support for the sustainable development of future smart distribution networks.

[0008] Furthermore, acquiring several communication signals includes: Obtain power line carrier signals via PowerPLC board; Obtain private wireless network signals through the NR-U board; Obtain public network signal through 5G board.

[0009] This embodiment clarifies the method of acquiring heterogeneous signals through multiple communication boards (such as PowerPLC, NR-U, and 5G boards), solving the problem that a single communication standard in traditional distribution networks cannot adapt to diverse business needs. Power line carrier signals are suitable for long-distance transmission, private wireless network signals ensure the security of local communication, and 5G public network signals provide wide coverage and high-speed support. This multi-source signal acquisition mechanism not only improves the redundancy and reliability of communication links but also enables signals from different sources and of different types to be converted into a unified format. This allows the system to accurately assess the system resources required to process each communication signal, thereby providing reliable reference data for subsequent resource scheduling and improving the accuracy and efficiency of distribution network resource scheduling.

[0010] In one possible implementation, the construction of the resource scheduling model based on the objective function and several constraints includes: The objective function is constructed with the goal of minimizing the communication latency, computational cost, and storage cost of a single response task. Delay constraints are constructed based on the constraint that the sum of communication delay and computation delay is less than or equal to a preset maximum delay tolerance. The computing requirement constraint is constructed with the constraint that the virtual computing power resources of a single response task are greater than or equal to the task resource requirements of a single response task. The resource scheduling model is constructed by combining the objective function, the time delay constraint, and the computational requirement constraint.

[0011] This application provides a method for constructing a resource scheduling model. By building a resource scheduling model that aims to minimize communication latency, computational cost, and storage cost, and introducing latency constraints and computational demand constraints into the model, it achieves refined and scientific resource allocation. Traditional resource scheduling methods often rely on empirical rules or static strategies, making it difficult to cope with dynamically changing service loads and network conditions. The model proposed in this embodiment, however, can dynamically optimize resource allocation while ensuring service quality. For example, in services with high real-time requirements (such as voltage regulation or frequency control), the model can prioritize the allocation of low-latency resources to ensure the timely completion of critical tasks. Simultaneously, the mathematical optimization foundation of the model enables it to remain efficient when handling large-scale problems, further improving the resource utilization rate and flexibility of distribution network resource scheduling.

[0012] In one possible implementation, when a plurality of virtual resources are scheduled and allocated from a preset resource pool to the corresponding response task according to any of the resource configurations, the step of scheduling and allocating the plurality of virtual resources from the preset resource pool to the respective response tasks according to each of the resource configurations includes: Based on the communication latency configuration in the resource configuration, the corresponding virtual pipeline in the resource pool is allocated to the response task; Based on the computing power resource configuration in the resource configuration, the corresponding virtual computing power units in the resource pool are allocated to the response task; Based on the storage resource configuration in the resource configuration, several virtual data volumes corresponding to the resource pool are allocated to the response task.

[0013] This application details the allocation method for virtual resources. By allocating virtual pipelines, virtual computing units, and virtual data volumes to response tasks on demand, it achieves precise and automated resource scheduling. This method avoids the resource mismatch and waste problems common in traditional manual configuration, improving the overall system response speed and resource utilization. For example, in scenarios with sudden load surges, the system can quickly allocate idle computing and storage resources to prevent system overload or crashes. Furthermore, the dynamic allocation of virtualized resources supports elastic scaling, enabling the system to automatically adjust resource scale according to fluctuations in business volume, meeting peak demand while avoiding resource idleness during off-peak periods, thus improving the flexibility of resource scheduling.

[0014] In one possible implementation, the step of converting communication resources, computing resources, and storage resources into corresponding virtual resource units and then constructing the resource pool includes: Based on the total bandwidth in the communication resources, the communication resources are converted into several virtual pipes; Based on the floating-point operation capability and video memory in the computing resources, the computing resources are converted into several virtual computing units; Based on the storage capacity and I / O performance of the storage resources, the storage resources are converted into several virtual data volumes.

[0015] This application provides a method for constructing a resource pool, which unifies communication, computing power, and storage resources into virtual resource units, thus building a highly integrated resource pool. This pooling management approach breaks the limitations of traditional resource silos, enabling different types of resources to be uniformly scheduled and shared, significantly improving the overall utilization efficiency and scheduling flexibility of resources. Simultaneously, the introduction of virtualization technology reduces the system's dependence on hardware, facilitating subsequent upgrades and maintenance.

[0016] Furthermore, the step of converting the communication resources into several virtual pipes based on the total bandwidth of the communication resources includes: For each wireless channel in the communication resources, the allocatable bandwidth of each wireless channel is determined based on the total bandwidth of each wireless channel and the preset modulation efficiency. Each wireless channel is split and converted according to its allocable bandwidth to obtain several wireless virtual channels; For each wired channel in the communication resources, according to a preset aggregation number, the wired channels are aggregated and bound into a wired virtual pipe, and the wired channels are aggregated and converted into a number of wired virtual pipes.

[0017] This application further refines the virtualization conversion process of communication resources, particularly the differentiated processing of wireless and wired channels. By splitting wireless channels into multiple virtual pipes, spectrum resources are fully utilized; while for wired channels, aggregation and binding improve transmission efficiency and reliability. The resource conversion method provided by this application not only optimizes bandwidth utilization but also enhances network flexibility and adaptability. For example, in scenarios with frequent changes in network topology, the system can dynamically adjust the number and configuration of virtual pipes to quickly adapt to new communication needs, improving the flexibility and utilization of subsequent resource scheduling.

[0018] Furthermore, the step of converting the computing power resources into several virtual computing power units based on the floating-point operation capability and video memory in the computing power resources includes: Each NPU and each first GPU in the computing power resources are converted into several first virtual computing units, and each first virtual computing unit is assigned to a real-time computing area, wherein the first GPU is a GPU whose core number meets a preset condition. Each GPU other than the first GPU in the computing resources is converted into several second virtual computing units, and each second virtual computing unit is assigned to a batch computing area; The FPGA logic units in the computing resources are converted into several third virtual computing units, and each of the third virtual computing units is assigned to the emergency computing area.

[0019] This application embodiment achieves classified management and optimized scheduling of computing resources by dividing computing resources into different virtual computing units according to hardware type and allocating them to real-time computing areas, batch computing areas, and emergency computing areas. This partitioned management method ensures that high real-time tasks (such as fault protection) can prioritize the use of high-performance resources, while non-real-time tasks (such as data analysis) can utilize the resources of the batch computing area, thereby achieving reasonable allocation and efficient utilization of resources, improving the utilization efficiency of power distribution network resources and the flexibility of resource scheduling. For example, in the event of a sudden emergency, the emergency computing area can be quickly activated and provide additional computing power support to ensure the stable operation of the system.

[0020] Secondly, embodiments of this application provide a power distribution network resource scheduling system based on integrated computing power and network, including an acquisition module, an encapsulation module, a response module, a resource scheduling module, a resource allocation module, and a data processing module; The acquisition module is used to acquire several communication signals and corresponding service requests; The encapsulation module is used to encapsulate each of the communication signals according to a preset protocol format to obtain several standardized data. The response module is used to generate corresponding response tasks based on each of the business requests, and to determine the task resource requirements of each of the response tasks. The resource scheduling module is used to perform resource scheduling on each response task according to the resource requirements of each task and the preset resource scheduling model if the current state of the distribution network system meets the preset conditions, and to determine the resource configuration of each response task; otherwise, it determines the resource configuration of each response task according to the resource requirements of each task; wherein, the resource scheduling model is constructed based on the objective function and several constraints. The resource allocation module is used to schedule and allocate a number of virtual resources from a preset resource pool to each of the response tasks according to the configuration of each resource. The resource pool is constructed by converting communication resources, computing power resources and storage resources into corresponding virtual resource units. The data processing module is used to process the standardized data according to each response task and the corresponding virtual resource units.

[0021] Furthermore, the process of constructing the resource scheduling model based on the objective function and several constraints includes: The objective function is constructed with the goal of minimizing the communication latency, computational cost, and storage cost of a single response task. Delay constraints are constructed based on the constraint that the sum of communication delay and computation delay is less than or equal to a preset maximum delay tolerance. The computing requirement constraint is constructed with the constraint that the virtual computing power resources of a single response task are greater than or equal to the task resource requirements of a single response task. The resource scheduling model is constructed by combining the objective function, the time delay constraint, and the computational requirement constraint.

[0022] Furthermore, the step of converting communication resources, computing resources, and storage resources into corresponding virtual resource units to construct the resource pool includes: Based on the total bandwidth in the communication resources, the communication resources are converted into several virtual pipes; Based on the floating-point operation capability and video memory in the computing resources, the computing resources are converted into several virtual computing units; Based on the storage capacity and I / O performance of the storage resources, the storage resources are converted into several virtual data volumes. Attached Figure Description

[0023] Figure 1 A flowchart illustrating a power distribution network resource scheduling method based on integrated computing power and network provided in this application embodiment; Figure 2A flowchart illustrating a specific application of a power distribution network resource scheduling method based on integrated computing power and network provided in this application embodiment; Figure 3 A frame-type computing power base architecture for receiving and encapsulating data in a power distribution network resource scheduling method based on integrated computing power and network provided in this application embodiment; Figure 4 This is a schematic diagram illustrating the construction process of a resource pool in a power distribution network resource scheduling method based on integrated computing power and network provided in an embodiment of this application.

[0024] Figure 5 This is a schematic diagram of the structure of a power distribution network resource scheduling system based on the integration of computing power and network, provided for an embodiment of this application. Detailed Implementation

[0025] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0026] It should be noted that the step numbers in this document are only for the convenience of explaining the specific embodiments and are not intended to limit the order in which the steps are performed. In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of that feature.

[0027] Example 1: like Figure 1 As shown, Embodiment 1 provides a distribution network resource scheduling method based on integrated computing power and network, including steps S1-S6: Step S1: Obtain several communication signals and corresponding service requests; Step S2: According to the preset protocol format, each of the communication signals is encapsulated to obtain several standardized data. Step S3: Generate corresponding response tasks based on each of the business requests, and determine the task resource requirements of each of the response tasks; Step S4: If the current state of the distribution network system meets the preset conditions, then resource scheduling is performed on each of the response tasks according to the resource requirements of each task and the preset resource scheduling model to determine the resource configuration of each response task; otherwise, the resource configuration of each response task is determined according to the resource requirements of each task; wherein, the resource scheduling model is constructed based on the objective function and several constraints. Step S5: According to the resource configuration, allocate a number of virtual resources from the preset resource pool to each of the response tasks. The resource pool is constructed by converting communication resources, computing resources and storage resources into corresponding virtual resource units. Step S6: Based on each of the response tasks and the corresponding virtual resource units, perform data processing on each of the standardized data.

[0028] This application proposes a distribution network resource scheduling method based on integrated computing and network capabilities. It pools dispersed communication, computing, and storage resources and dynamically allocates virtual resources through a resource scheduling model, enabling the system to respond quickly to demands in high-concurrency and multi-service scenarios. Compared to traditional static resource allocation models, this embodiment significantly improves resource utilization and the flexibility of distribution network resource scheduling. For example, in the case of large-scale distributed energy access, traditional methods often lead to response delays and resource waste due to resource silos and rigid scheduling mechanisms. This embodiment, however, by monitoring the distribution network status in real time and combining it with an optimization model, can adaptively adjust resource allocation, ensuring low-latency and high-reliability processing of critical services (such as fault detection and load control). This embodiment also encapsulates different types of communication signals, unifying signals from different sources and converting them into a standard format. The encapsulated data can be uniformly understood by the system, ensuring that the entire system can efficiently process signals from different sources, avoiding conflicts and compatibility issues between different protocols, thus facilitating subsequent processing and scheduling. Furthermore, the introduction of virtualization technology reduces hardware dependence, enhances system scalability and fault tolerance, and provides strong technical support for the sustainable development of future smart distribution networks.

[0029] In a preferred embodiment, a specific application flow of the power distribution network resource scheduling method provided in this application is as follows: Figure 2 As shown, the chassis-type computing base connects to multiple signal sources and abstracts and transmits data resources to the resource pool. Then, the resource pool flexibly schedules and efficiently utilizes system resources based on business needs. Next, the software-defined engine makes decisions based on real-time data and task requirements, and dynamically adjusts resources and tasks through control commands. Finally, the scheduling results are transmitted back to the intelligent power distribution network management platform, enabling overall monitoring and management, ensuring the efficient and stable operation of the system, and the results are also fed back to the software-defined engine.

[0030] Furthermore, in step S1, acquiring several communication signals includes: Obtain power line carrier signals via PowerPLC board; Obtain private wireless network signals through the NR-U board; Obtain public network signal through 5G board.

[0031] This embodiment clarifies the method of acquiring heterogeneous signals through multiple communication boards (such as PowerPLC, NR-U, and 5G boards), solving the problem that a single communication standard in traditional distribution networks cannot adapt to diverse business needs. Power line carrier signals are suitable for long-distance transmission, private wireless network signals ensure the security of local communication, and 5G public network signals provide wide coverage and high-speed support. This multi-source signal acquisition mechanism not only improves the redundancy and reliability of communication links but also enables signals from different sources and of different types to be converted into a unified format. This allows the system to accurately assess the system resources required to process each communication signal, thereby providing reliable reference data for subsequent resource scheduling and improving the accuracy and efficiency of distribution network resource scheduling.

[0032] In a preferred embodiment, such as Figure 3 As shown, this embodiment proposes an innovative multimodal access chassis-based computing power platform architecture, which aims to improve the flexibility of data access and the efficiency of computing power processing in power distribution networks. The architecture consists of three main functional modules: physical layer signal reception, data access module, and high-speed backplane interconnection, which together constitute the basic hardware platform of the system, providing support for subsequent resource scheduling and intelligent decision-making. This chassis-based computing power platform architecture includes the following three parts.

[0033] (1) Physical layer signal reception The physical layer is the first step in the architecture, responsible for receiving signals from different communication networks and transmitting them to the upper-layer processing modules. Each communication path's access method is implemented through dedicated boards. At the physical layer, the hardware base receives and transmits data from different types of signal sources through various access boards. The physical layer signal reception design ensures that the system can support multiple different communication protocols and provides stable input for subsequent data processing. The specific design is as follows: A. Power Line Carrier Signal Access: The power line carrier signal is accessed through a PowerPLC board. The board uses a modem chip, supports the CENELEC-A band, and is suitable for signal transmission in power systems, effectively resisting noise interference in power line communication. After demodulation, the power line carrier signal is converted into a standard digital signal for use by subsequent data processing modules.

[0034] B. Private Wireless Network Signal Access: The private wireless network signal is accessed through the NR-U board, which supports dual RF front-ends of 1.4GHz / 1.8GHz and has a receiving sensitivity of ≤-110dBm, adapting to signal interference in the wireless environment and ensuring stable access even in complex environments. The received wireless signal is modulated, demodulated, and decoded into a data format conforming to the system protocol, ready to enter the protocol adaptation engine for further processing.

[0035] C. Public network signal access is achieved through a 5G board. The 5G board has a built-in pool of 32 SIM cards, supporting NSA / SA dual-mode. It automatically identifies and selects the appropriate network access method through the SIM card, ensuring a continuous and stable public network connection. After accessing the public network, the data is converted into an adapted data stream through the communication interface and enters the next processing stage of the system.

[0036] (2) Data access module The data access module encapsulates various types of received data according to a unified protocol format and transmits it through an interface. This module supports multiple standard communication protocols and provides efficient data processing capabilities.

[0037] The data access module uses a multi-protocol adaptation engine to convert and adapt data from different signal sources. Its core function is: • RS485 / Modbus interface: Supports connection to traditional industrial equipment, especially suitable for common monitoring devices, sensors, transformers, etc. in power systems. RS485 is a commonly used industrial standard that can stably transmit data over long distances and in environments with high interference. The Modbus protocol is also widely used in industrial automation and is suitable for data communication between different types of devices.

[0038] • Zigbee / LoRa interface: Supports data access for wireless sensor networks. LoRa is particularly suitable for long-distance, low-power sensor data transmission, while Zigbee is suitable for self-organizing networks within a smaller range and is widely used in IoT devices and wireless sensor networks.

[0039] By supporting these standard communication protocols, the data access module can ensure broad compatibility with a variety of devices and sensors, and can receive and process signals from different terminals in real time.

[0040] The adaptation engine encapsulates different data into the IEEE 1888 format, which includes source address, destination address, data type, timestamp, and payload data, ensuring standardization and compatibility of data transmission.

[0041] The data access module is designed with system scalability in mind, allowing for the addition, removal, or replacement of modules to meet different communication needs. Its support for RS485 / Modbus and Zigbee / LoRa protocols enables the system to adapt to various industrial equipment and IoT terminals, and the addition of new devices in the future will not affect the operation of the existing system.

[0042] (3) High-speed interconnection of backplane The high-speed interconnect module on the backplane ensures high-speed data transmission between the various hardware modules in the system, providing the necessary bandwidth to support efficient data flow. Each hardware module is connected via a PCIe 4.0 x16 bus, providing a high-speed bandwidth of ≥64Gbps to support parallel transmission of large-scale data. This high-speed connection ensures high-speed interconnection and data transmission between modules, providing a stable and reliable communication channel for the system and meeting the needs of real-time computing and big data processing. This bus can adapt to the high bandwidth requirements of computing tasks, especially in multi-task processing and large-scale parallel computing applications, significantly improving the system's response speed and computing efficiency.

[0043] To further reduce data transmission latency, Cut-Through switching technology is employed to optimize the data transmission process and reduce the waiting time during packet forwarding. In traditional switching technologies, switching devices only begin forwarding after receiving the entire data packet, which introduces significant latency. Conversely, Cut-Through switching technology begins forwarding immediately upon the arrival of the first few bytes of the data packet, without waiting for the entire packet to be received. When data enters the switch or switching module via the high-speed bus, the device does not wait for the complete data packet to be received but begins forwarding when the first few bytes arrive, significantly reducing waiting time.

[0044] Building upon the high-speed interconnect on the backplane, the computing boards in the hardware layer work closely with the NPU unit and GPU cluster via a high-speed bus. This collaborative work at the hardware layer ensures parallel execution and real-time response of computing tasks through efficient data transmission and processing. In this embodiment, the roles of the NPU unit and GPU cluster are as follows: • NPU Unit: The NPU unit is primarily responsible for edge inference tasks, processing data from various sensors in real time. It executes inference tasks through hardware acceleration, providing real-time analysis results with low latency and high efficiency. After receiving data, the NPU unit quickly performs inference calculations and transmits the results back to the system for immediate decision-making, such as in intelligent device monitoring and fault detection.

[0045] • GPU Clusters: GPU clusters are responsible for training deep learning models. They utilize multiple GPUs to process large-scale datasets in parallel, optimize model parameters, and provide more accurate inference models for the NPU units. Through parallel computing, GPU clusters can accelerate the training process for large datasets and rapidly iterate deep learning models.

[0046] The module's high-speed interconnect and low-latency design ensure smooth data exchange between the NPU and GPU, avoids bottlenecks in data transmission, maximizes computing power and real-time performance, and provides hardware support for subsequent resource scheduling.

[0047] In a preferred embodiment, in step S3, the generation of the response task is primarily accomplished by the business-aware module. The business-aware module is responsible for receiving requests from the business layer and identifying their resource requirements. Input business requests are request data that the system can receive and process. By parsing the business requests, the system derives the requirements for resources such as communication, computing power, and storage. Business requests include: • Input service requests: The system receives different types of service requests. For example, V2G (Vehicle to Grid) task requests may require high-priority computing resources and low-latency communication capabilities.

[0048] • Output resource requirements: The system automatically calculates and outputs the corresponding resource requirements based on business requests, such as latency and computing power requirements. For example, a high-priority V2G task may require computing power greater than 5 TFLOPS and latency less than 30ms.

[0049] Through the business demand awareness module, the system can dynamically adjust resource allocation according to business requests to ensure that real-time and computing requirements are met.

[0050] In one possible implementation, step S4, which involves constructing the resource scheduling model based on the objective function and several constraints, includes: The objective function is constructed with the goal of minimizing the communication latency, computational cost, and storage cost of a single response task. Delay constraints are constructed based on the constraint that the sum of communication delay and computation delay is less than or equal to a preset maximum delay tolerance. The computing requirement constraint is constructed with the constraint that the virtual computing power resources of a single response task are greater than or equal to the task resource requirements of a single response task. The resource scheduling model is constructed by combining the objective function, the time delay constraint, and the computational requirement constraint.

[0051] This application provides a method for constructing a resource scheduling model. By building a resource scheduling model that aims to minimize communication latency, computational cost, and storage cost, and introducing latency constraints and computational demand constraints into the model, it achieves refined and scientific resource allocation. Traditional resource scheduling methods often rely on empirical rules or static strategies, making it difficult to cope with dynamically changing service loads and network conditions. The model proposed in this embodiment, however, can dynamically optimize resource allocation while ensuring service quality. For example, in services with high real-time requirements (such as voltage regulation or frequency control), the model can prioritize the allocation of low-latency resources to ensure the timely completion of critical tasks. Simultaneously, the mathematical optimization foundation of the model enables it to remain efficient when handling large-scale problems, further improving the resource utilization rate and flexibility of distribution network resource scheduling.

[0052] In a preferred embodiment, the current state of the distribution network system can be monitored in real time. When the distribution network is in a peak load period or the resource requirements of multiple response tasks change, a resource scheduling model needs to be introduced to perform resource scheduling for each response task, determining or updating the resource configuration of each response task. The resource scheduling model schedules resources within the system by optimizing an objective function. The core objective of the scheduling is to minimize the total cost of communication, computation, and storage. The specific objective function is as follows: in, For communication delay, For calculating consumption (e.g., TFLOPS). Storage consumption (such as IOPS and storage capacity) is considered. The weights of various resources are adjusted using coefficients α, β, and γ to adapt to different types of business needs and priorities. By minimizing this objective function, the system can effectively balance resource usage and improve overall resource efficiency.

[0053] Scheduling also needs to follow certain constraints to ensure that the required computing power and storage capacity are met. The constraints are as follows: Latency Constraint: The sum of total communication latency and computation latency must meet the Maximum Service Level Agreement (SLA) requirements. in, Communication latency refers to the time required for a task to complete the communication process, which usually involves the data transmission process. For computational latency, it refers to the computation time required for a task to be processed, which usually involves operations such as data processing and computational inference. The Maximum Service Level Protocol (MSLP) defines the maximum latency tolerance that the system can provide. In other words, the total latency of a task (communication latency + computation latency) must be below this threshold to ensure that the task meets real-time requirements. This constraint ensures that tasks will not fail due to excessive latency, thus guaranteeing timely task response.

[0054] Computational requirements constraint: The system must ensure that it meets the computing resources required by the business. For each virtual computing unit (vPU), the system will allocate sufficient computing power according to task requirements. in, This refers to a Virtual Processing Unit (vPU), where each vPU corresponds to a certain computing power (e.g., TFLOPS). The system allocates multiple vPUs to meet the computing requirements of a task. The computing power required for a task is typically expressed in TFLOPS (trillion floating-point operations per second). The system ensures that the computing needs of all tasks are met by scheduling resources. This constraint ensures sufficient computing resources and prevents tasks from being delayed or failing due to insufficient computing power.

[0055] In one possible implementation, in step S5, when a plurality of virtual resources are scheduled and allocated from a preset resource pool to the corresponding response task according to any of the resource configurations, the scheduling and allocation of the plurality of virtual resources from the preset resource pool to the respective response tasks according to each of the resource configurations includes: Based on the communication latency configuration in the resource configuration, the corresponding virtual pipeline in the resource pool is allocated to the response task; Based on the computing power resource configuration in the resource configuration, the corresponding virtual computing power units in the resource pool are allocated to the response task; Based on the storage resource configuration in the resource configuration, several virtual data volumes corresponding to the resource pool are allocated to the response task.

[0056] This application details the allocation method for virtual resources. By allocating virtual pipelines, virtual computing units, and virtual data volumes to response tasks on demand, it achieves precise and automated resource scheduling. This method avoids the resource mismatch and waste problems common in traditional manual configuration, improving the overall system response speed and resource utilization. For example, in scenarios with sudden load surges, the system can quickly allocate idle computing and storage resources to prevent system overload or crashes. Furthermore, the dynamic allocation of virtualized resources supports elastic scaling, enabling the system to automatically adjust resource scale according to fluctuations in business volume, meeting peak demand while avoiding resource idleness during off-peak periods, thus improving the flexibility of resource scheduling.

[0057] In a preferred embodiment, transmission latency is primarily reduced through bandwidth allocation optimization, and real-time performance is optimized by increasing bandwidth or employing low-latency paths based on communication latency. Different computing resources (such as GPUs and NPUs) are allocated to corresponding response tasks based on computational consumption, where real-time tasks may be allocated high-performance computing resources, and batch tasks may be allocated lower-performance resources. Storage capacity and IOPS are adjusted through virtual data volumes (vVolumes). Storage-intensive tasks are allocated higher IOPS and storage capacity, while lightweight tasks are allocated lower resources. Storage capacity and IOPS are adjusted through virtual data volumes (vVolumes) based on storage consumption. Storage-intensive tasks are allocated higher IOPS and storage capacity, while lightweight tasks are allocated lower resources.

[0058] In one possible implementation, step S5, which involves converting communication resources, computing resources, and storage resources into corresponding virtual resource units and then constructing the resource pool, includes: Based on the total bandwidth in the communication resources, the communication resources are converted into several virtual pipes; Based on the floating-point operation capability and video memory in the computing resources, the computing resources are converted into several virtual computing units; Based on the storage capacity and I / O performance of the storage resources, the storage resources are converted into several virtual data volumes.

[0059] This application provides a method for constructing a resource pool, which unifies communication, computing power, and storage resources into virtual resource units, thus building a highly integrated resource pool. This pooling management approach breaks the limitations of traditional resource silos, enabling different types of resources to be uniformly scheduled and shared, significantly improving the overall utilization efficiency and scheduling flexibility of resources. Simultaneously, the introduction of virtualization technology reduces the system's dependence on hardware, facilitating subsequent upgrades and maintenance.

[0060] Furthermore, the step of converting the communication resources into several virtual pipes based on the total bandwidth of the communication resources includes: For each wireless channel in the communication resources, the allocatable bandwidth of each wireless channel is determined based on the total bandwidth of each wireless channel and the preset modulation efficiency. Each wireless channel is split and converted according to its allocable bandwidth to obtain several wireless virtual channels; For each wired channel in the communication resources, according to a preset aggregation number, the wired channels are aggregated and bound into a wired virtual pipe, and the wired channels are aggregated and converted into a number of wired virtual pipes.

[0061] This application further refines the virtualization conversion process of communication resources, particularly the differentiated processing of wireless and wired channels. By splitting wireless channels into multiple virtual pipes, spectrum resources are fully utilized; while for wired channels, aggregation and binding improve transmission efficiency and reliability. The resource conversion method provided by this application not only optimizes bandwidth utilization but also enhances network flexibility and adaptability. For example, in scenarios with frequent changes in network topology, the system can dynamically adjust the number and configuration of virtual pipes to quickly adapt to new communication needs, improving the flexibility and utilization of subsequent resource scheduling.

[0062] Furthermore, the step of converting the computing power resources into several virtual computing power units based on the floating-point operation capability and video memory in the computing power resources includes: Each NPU and each first GPU in the computing power resources are converted into several first virtual computing units, and each first virtual computing unit is assigned to a real-time computing area, wherein the first GPU is a GPU whose core number meets a preset condition. Each GPU other than the first GPU in the computing resources is converted into several second virtual computing units, and each second virtual computing unit is assigned to a batch computing area; The FPGA logic units in the computing resources are converted into several third virtual computing units, and each of the third virtual computing units is assigned to the emergency computing area.

[0063] This application embodiment achieves classified management and optimized scheduling of computing resources by dividing computing resources into different virtual computing units according to hardware type and allocating them to real-time computing areas, batch computing areas, and emergency computing areas. This partitioned management method ensures that high real-time tasks (such as fault protection) can prioritize the use of high-performance resources, while non-real-time tasks (such as data analysis) can utilize the resources of the batch computing area, thereby achieving reasonable allocation and efficient utilization of resources, improving the utilization efficiency of power distribution network resources and the flexibility of resource scheduling. For example, in the event of a sudden emergency, the emergency computing area can be quickly activated and provide additional computing power support to ensure the stable operation of the system.

[0064] In a preferred embodiment, such as Figure 4 As shown, by virtualizing and pooling communication, computing, and storage resources, flexible scheduling and efficient utilization of system resources are achieved. Specifically, the abstraction, partitioning, and allocation of resources are implemented through the following steps and technologies: (1) Resource abstraction The core objective of resource abstraction is to transform communication resources, computing resources, and storage resources into independent virtual resource units, facilitating unified management, scheduling, and dynamic allocation. This virtualization is achieved through virtual pipes (vPipe), virtual computing units (vPU), and virtual data volumes (vVolume). Communication virtualization uses vPipe to virtualize bandwidth, latency, and QoS, dynamically allocating wireless spectrum or wired channel resources. Computing virtualization uses vPU to dynamically allocate computing power, including floating-point operations per second (TFLOPS) and video memory (Mem). Storage virtualization manages storage resources through vVolume, providing virtualization support based on IOPS and storage capacity.

[0065] (2) Communication virtualization: virtual pipe vPipe Communication resource virtualization uses virtual pipes (vPipe) to abstract and allocate communication resources, ensuring that the system can dynamically adjust bandwidth, latency, and QoS according to demand. The specific steps are as follows: • Wireless spectrum allocation: Calculate the allocatable bandwidth: Allocable bandwidth = Total bandwidth × (1 - Guard interval) × Modulation efficiency For example, the 20MHz NR-U band is divided into 4×5MHz vPipes, and each virtual pipe (vPipe) is allocated according to different bandwidth, latency and QoS requirements.

[0066] • Wired channel bonding: Aggregates multiple PLC channels into a single high-reliability virtual pipeline (vPipe). For example, bonding 4 PLC channels into a virtual pipeline can control the packet loss rate to **<10⁻6 **This ensures high reliability of data transmission.

[0067] (3) Computing power virtualization: Virtual computing power unit (vPU) Computing virtualization uses virtual computing units (vPUs) to allocate and isolate computing resources. The computing resource partitioning is as follows: • Real-time zone: Allocates NPU and 2-core GPU, and uses NVIDIA MIG technology to achieve resource isolation, suitable for tasks with low latency and high real-time requirements.

[0068] • Batch Zone: Allocates remaining GPU cores and uses Kubernetes quotas for resource management, suitable for large-scale data processing tasks.

[0069] Emergency Zone: Reserves FPGA logic units to support the computing needs of special tasks through dynamic reprogramming technology.

[0070] (4) Storage virtualization: Virtual data volume vVolume Storage virtualization abstracts and manages storage resources through virtual data volumes (vVolumes). vVolumes provide virtual storage units that support virtualization of storage capacity and I / O performance. These virtual storage units dynamically adjust their capacity and I / O performance based on task requirements to meet the needs of different storage tasks. For example, a vVolume might be configured with IOPS = 10K and storage capacity = 500GB.

[0071] Furthermore, after completing step S5, resource isolation technology and priority management mechanisms can be used to allocate more resources to high-priority tasks and ensure that different tasks can be executed without interfering with each other. For example, high-priority tasks (such as emergency fault handling) will receive more resources and can be scheduled first when computing resources are scarce, while low-priority tasks (such as batch data analysis) can be allocated according to the remaining resources. During execution, if new business requirements arise or system resources fluctuate (for example, a communication link becomes bottlenecked, or computing resources are overloaded), the system will perform real-time resource scheduling and adjustment, dynamically adjusting resource allocation to prioritize the execution of real-time tasks.

[0072] In summary, existing distribution network systems suffer from poor flexibility, insufficient compatibility, and weak real-time response capabilities in resource scheduling and management. Traditional systems struggle to handle the access of multiple signal sources and dynamic resource allocation, often leading to resource waste or shortages, impacting the stability and efficiency of the distribution network. This embodiment improves the system's compatibility with different communication signal sources through a multimodal access architecture, ensuring stable signal access. Resource pooling virtualization technology abstracts communication, computing power, and storage resources into virtual units, dynamically allocating them on demand to ensure efficient resource utilization and avoid waste. The software-defined engine improves task processing efficiency and response speed through intelligent scheduling and real-time feedback. Overall, this embodiment significantly improves the resource utilization, real-time response capability, and system stability of the distribution network, supports more complex multi-task processing and dynamic adjustments, and promotes the intelligent development of the distribution network.

[0073] Example 2: like Figure 5 As shown, Embodiment 2 provides a power distribution network resource scheduling system based on integrated computing power and network, including an acquisition module 10, an encapsulation module 20, a response module 30, a resource scheduling module 40, a resource allocation module 50, and a data processing module 60; The acquisition module 10 is used to acquire several communication signals and corresponding service requests; The encapsulation module 20 is used to encapsulate each of the communication signals according to a preset protocol format to obtain several standardized data. The response module 30 is used to generate corresponding response tasks according to each of the service requests, and to determine the task resource requirements of each of the response tasks. The resource scheduling module 40 is used to perform resource scheduling on each response task according to the resource requirements of each task and the preset resource scheduling model if the current state of the distribution network system meets the preset conditions, and to determine the resource configuration of each response task; otherwise, it determines the resource configuration of each response task according to the resource requirements of each task; wherein, the resource scheduling model is constructed based on the objective function and several constraints. The resource allocation module 50 is used to schedule and allocate a number of virtual resources from a preset resource pool to each of the response tasks according to the configuration of each resource. The resource pool is constructed by converting communication resources, computing power resources and storage resources into corresponding virtual resource units. The data processing module 60 is used to process the standardized data according to each of the response tasks and the corresponding virtual resource units.

[0074] Furthermore, the acquisition module 10 acquires several communication signals, including: Obtain power line carrier signals via PowerPLC board; Obtain private wireless network signals through the NR-U board; Obtain public network signal through 5G board.

[0075] In one possible implementation, the construction of the resource scheduling model based on the objective function and several constraints includes: The objective function is constructed with the goal of minimizing the communication latency, computational cost, and storage cost of a single response task. Delay constraints are constructed based on the constraint that the sum of communication delay and computation delay is less than or equal to a preset maximum delay tolerance. The computing requirement constraint is constructed with the constraint that the virtual computing power resources of a single response task are greater than or equal to the task resource requirements of a single response task. The resource scheduling model is constructed by combining the objective function, the time delay constraint, and the computational requirement constraint.

[0076] In one possible implementation, when a plurality of virtual resources are scheduled and allocated from a preset resource pool to the corresponding response task according to any of the resource configurations, the resource allocation module 50, according to each of the resource configurations, schedules and allocates the plurality of virtual resources from the preset resource pool to each of the response tasks, including: Based on the communication latency configuration in the resource configuration, the corresponding virtual pipeline in the resource pool is allocated to the response task; Based on the computing power resource configuration in the resource configuration, the corresponding virtual computing power units in the resource pool are allocated to the response task; Based on the storage resource configuration in the resource configuration, several virtual data volumes corresponding to the resource pool are allocated to the response task.

[0077] In one possible implementation, the step of converting communication resources, computing resources, and storage resources into corresponding virtual resource units and then constructing the resource pool includes: Based on the total bandwidth in the communication resources, the communication resources are converted into several virtual pipes; Based on the floating-point operation capability and video memory in the computing resources, the computing resources are converted into several virtual computing units; Based on the storage capacity and I / O performance of the storage resources, the storage resources are converted into several virtual data volumes.

[0078] Furthermore, the step of converting the communication resources into several virtual pipes based on the total bandwidth of the communication resources includes: For each wireless channel in the communication resources, the allocatable bandwidth of each wireless channel is determined based on the total bandwidth of each wireless channel and the preset modulation efficiency. Each wireless channel is split and converted according to its allocable bandwidth to obtain several wireless virtual channels; For each wired channel in the communication resources, according to a preset aggregation number, the wired channels are aggregated and bound into a wired virtual pipe, and the wired channels are aggregated and converted into a number of wired virtual pipes.

[0079] Furthermore, the step of converting the computing power resources into several virtual computing power units based on the floating-point operation capability and video memory in the computing power resources includes: Each NPU and each first GPU in the computing power resources are converted into several first virtual computing units, and each first virtual computing unit is assigned to a real-time computing area, wherein the first GPU is a GPU whose core number meets a preset condition. Each GPU other than the first GPU in the computing resources is converted into several second virtual computing units, and each second virtual computing unit is assigned to a batch computing area; The FPGA logic units in the computing resources are converted into several third virtual computing units, and each of the third virtual computing units is assigned to the emergency computing area.

[0080] This application proposes a distribution network resource scheduling system based on integrated computing and network capabilities. It pools dispersed communication, computing, and storage resources and dynamically allocates virtual resources through a resource scheduling model, enabling the system to respond quickly to demands in high-concurrency and multi-service scenarios. Compared to traditional static resource allocation models, this embodiment significantly improves resource utilization and the flexibility of distribution network resource scheduling. For example, in the case of large-scale distributed energy integration, traditional methods often lead to response delays and resource waste due to resource silos and rigid scheduling mechanisms. This embodiment, however, by monitoring the distribution network status in real time and combining it with an optimization model, can adaptively adjust resource allocation, ensuring low-latency and high-reliability processing of critical services (such as fault detection and load control). This embodiment also encapsulates different types of communication signals, unifying signals from different sources and converting them into a standard format. The encapsulated data can be uniformly understood by the system, ensuring that the entire system can efficiently process signals from different sources, avoiding conflicts and compatibility issues between different protocols, thus facilitating subsequent processing and scheduling. Furthermore, the introduction of virtualization technology reduces hardware dependence, enhances the system's scalability and fault tolerance, and provides strong technical support for the sustainable development of future smart distribution networks.

[0081] For a more detailed explanation of the working principle and procedures of this embodiment, please refer to the relevant description in Embodiment 1.

[0082] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application for those skilled in the art.

Claims

1. A distribution network resource scheduling method based on integrated computing power and network, characterized in that, include: Acquire several communication signals and their corresponding service requests; According to a preset protocol format, each of the communication signals is encapsulated to obtain several standardized data. Generate corresponding response tasks based on each of the aforementioned business requests, and determine the task resource requirements of each of the aforementioned response tasks; If the current state of the distribution network system meets the preset conditions, then resource scheduling is performed on each of the response tasks according to the resource requirements of each task and the preset resource scheduling model to determine the resource configuration of each response task; otherwise, the resource configuration of each response task is determined according to the resource requirements of each task; wherein, the resource scheduling model is constructed based on the objective function and several constraints. According to the resource configurations, several virtual resources are scheduled and allocated to each response task from a preset resource pool. The resource pool is constructed by converting communication resources, computing resources and storage resources into corresponding virtual resource units. Based on each of the aforementioned response tasks and the corresponding virtual resource units, the standardized data is processed.

2. The distribution network resource scheduling method based on integrated computing power and network as described in claim 1, characterized in that, The acquisition of several communication signals includes: Obtain power line carrier signals via PowerPLC board; Obtain private wireless network signals through the NR-U board; Obtain public network signal through 5G board.

3. The distribution network resource scheduling method based on integrated computing power and network as described in claim 1, characterized in that, The process of constructing the resource scheduling model based on the objective function and several constraints includes: The objective function is constructed with the goal of minimizing the communication latency, computational cost, and storage cost of a single response task. Delay constraints are constructed based on the constraint that the sum of communication delay and computation delay is less than or equal to a preset maximum delay tolerance. The computing requirement constraint is constructed with the constraint that the virtual computing power resources of a single response task are greater than or equal to the task resource requirements of a single response task. The resource scheduling model is constructed by combining the objective function, the time delay constraint, and the computational requirement constraint.

4. The distribution network resource scheduling method based on integrated computing power and network as described in claim 1, characterized in that, When a plurality of virtual resources are scheduled and allocated from a preset resource pool to the corresponding response task according to any of the resource configurations, the step of scheduling and allocating the plurality of virtual resources from the preset resource pool to the respective response tasks according to each of the resource configurations includes: Based on the communication latency configuration in the resource configuration, the corresponding virtual pipeline in the resource pool is allocated to the response task; Based on the computing power resource configuration in the resource configuration, the corresponding virtual computing power units in the resource pool are allocated to the response task; Based on the storage resource configuration in the resource configuration, several virtual data volumes corresponding to the resource pool are allocated to the response task.

5. The distribution network resource scheduling method based on integrated computing power and network as described in claim 1, characterized in that, The process of converting communication resources, computing resources, and storage resources into corresponding virtual resource units to construct the resource pool includes: Based on the total bandwidth in the communication resources, the communication resources are converted into several virtual pipes; Based on the floating-point operation capability and video memory in the computing resources, the computing resources are converted into several virtual computing units; Based on the storage capacity and I / O performance of the storage resources, the storage resources are converted into several virtual data volumes.

6. The distribution network resource scheduling method based on integrated computing power and network as described in claim 5, characterized in that, The step of converting the communication resources into several virtual pipes based on the total bandwidth of the communication resources includes: For each wireless channel in the communication resources, the allocatable bandwidth of each wireless channel is determined based on the total bandwidth of each wireless channel and the preset modulation efficiency. Each wireless channel is split and converted according to its allocable bandwidth to obtain several wireless virtual channels; For each wired channel in the communication resources, according to a preset aggregation number, several wired channels are aggregated and bound into a single wired virtual pipe, and each wired channel is converted into several wired virtual pipes.

7. The distribution network resource scheduling method based on integrated computing power and network as described in claim 5, characterized in that, The step of converting the computing power resources into several virtual computing power units based on the floating-point operation capability and video memory in the computing power resources includes: Each NPU and each first GPU in the computing power resources are converted into several first virtual computing units, and each first virtual computing unit is assigned to a real-time computing area, wherein the first GPU is a GPU whose core number meets a preset condition. Each GPU other than the first GPU in the computing resources is converted into several second virtual computing units, and each second virtual computing unit is assigned to a batch computing area; The FPGA logic units in the computing resources are converted into several third virtual computing units, and each of the third virtual computing units is assigned to the emergency computing area.

8. A power distribution network resource scheduling system based on integrated computing power and network, characterized in that, It includes an acquisition module, an encapsulation module, a response module, a resource scheduling module, a resource allocation module, and a data processing module; The acquisition module is used to acquire several communication signals and corresponding service requests; The encapsulation module is used to encapsulate each of the communication signals according to a preset protocol format to obtain several standardized data. The response module is used to generate corresponding response tasks based on each of the business requests, and to determine the task resource requirements of each of the response tasks. The resource scheduling module is used to perform resource scheduling on each response task according to the resource requirements of each task and the preset resource scheduling model if the current state of the distribution network system meets the preset conditions, and to determine the resource configuration of each response task; otherwise, it determines the resource configuration of each response task according to the resource requirements of each task; wherein, the resource scheduling model is constructed based on the objective function and several constraints. The resource allocation module is used to schedule and allocate a number of virtual resources from a preset resource pool to each of the response tasks according to the configuration of each resource. The resource pool is constructed by converting communication resources, computing power resources and storage resources into corresponding virtual resource units. The data processing module is used to process the standardized data according to each response task and the corresponding virtual resource units.

9. A power distribution network resource scheduling system based on integrated computing power and network as described in claim 8, characterized in that, The process of constructing the resource scheduling model based on the objective function and several constraints includes: The objective function is constructed with the goal of minimizing the communication latency, computational cost, and storage cost of a single response task. Delay constraints are constructed based on the constraint that the sum of communication delay and computation delay is less than or equal to a preset maximum delay tolerance. The computing requirement constraint is constructed with the constraint that the virtual computing power resources of a single response task are greater than or equal to the task resource requirements of a single response task. The resource scheduling model is constructed by combining the objective function, the time delay constraint, and the computational requirement constraint.

10. A power distribution network resource scheduling system based on integrated computing power and network as described in claim 8, characterized in that, The process of converting communication resources, computing resources, and storage resources into corresponding virtual resource units to construct the resource pool includes: Based on the total bandwidth in the communication resources, the communication resources are converted into several virtual pipes; Based on the floating-point operation capability and video memory in the computing resources, the computing resources are converted into several virtual computing units; Based on the storage capacity and I / O performance of the storage resources, the storage resources are converted into several virtual data volumes.

Images