Optimization decision-making method and device for edge service cache of power distribution and utilization business
By building a computing business request model and a real-time perception model of computing power resources and load, combined with business prediction information, dynamically adjusting the edge service cache, the problem that the edge service cache of power distribution services in traditional methods cannot adapt to high real-time requirements, and improves the problem that the power distribution business edge service cache is not allowed to adapt to the high real-time requirements. Processing efficiency and resource utilization.
Patent Information
- Application Number
- CN202411293716.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-14
- Publication Date
- 2024-11-22
AI Technical Summary
In traditional distribution network automation systems, the static edge service caching method cannot effectively adapt to the high real-time requirements of distribution services, resulting in high service timeout rates and low node resource utilization rates.
Build a computing service request model and a real-time perception model of computing power resources and computing power load, combine business request prediction information, and dynamically adjust edge service cache to optimize the edge service cache of power distribution services and improve processing efficiency and resource utilization.
By real-time perception of the computing power resources and load information of edge nodes, dynamically adjusting the edge service cache, the node resource utilization rate and business processing efficiency of the distribution network automation system are improved, and the requirements of high real-time are adapted.
Smart Images

Figure CN119011586A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of edge service cache optimization, and in particular relates to an optimization decision method and device for edge service cache of power distribution and utilization business. Background Art
[0002] With the continuous development of the Internet of Things technology and power distribution and utilization technology, the power distribution and utilization Internet of Things that integrates the Internet of Things technology with the power distribution and utilization technology is becoming increasingly popular. The power distribution and utilization Internet of Things takes the Internet of Things technology as the core concept, and applies technologies such as smart sensors, cloud computing, edge computing and 5G in the power distribution and utilization scenario. It can better cope with various new situations such as massive distributed resource access and new forms of electricity and energy consumption. At present, a large number of power system regulation, trading, management and other businesses have been extended to the power distribution and utilization side, which has put forward new requirements for the computing resources of the power distribution and utilization system in terms of total amount, flexibility and accessibility. In the traditional distribution network automation system, the corresponding business requests are only executed in their respective business systems, and there is a lack of attention to service cache. When facing massive businesses, the processing efficiency is low, and business timeouts often occur.
[0003] Faced with this situation, the distribution network automation system based on the cloud-edge system has gradually developed. This cloud-edge system is a cloud-edge collaborative system, which includes multiple edge nodes. By integrating heterogeneously distributed computing resources into a unified resource pool, the distribution network automation system can respond to various types of power distribution and utilization business requests with obvious differences and fluctuations. However, in the distribution network automation system based on the cloud-edge collaborative system, the service cache of the distribution network automation system is usually adjusted based on the static edge service caching method. However, this adjustment method only relies on static computing power equipment parameters and cannot adapt to the high real-time requirements of the distribution network automation business. As a result, the distribution network automation system has a high business timeout rate and low node resource utilization when facing massive businesses. In addition, therefore, there is an urgent need for a new optimization decision method and device for the edge service cache of the power distribution and utilization business to solve the defects of the existing technology. Summary of the invention
[0004] The present invention aims to provide an optimization decision method for the edge service cache of the power distribution and utilization business to solve the above-mentioned technical problems, construct a computing business request model, a real-time perception model of computing power resources and computing power load, and an optimization model for the edge service cache of the power distribution and utilization business, so as to generate an optimization decision plan to optimize the edge service cache of the power distribution and utilization business, thereby improving the processing efficiency of the power distribution and utilization business and the node resource utilization rate of the distribution network automation system.
[0005] In order to solve the above technical problems, an embodiment of the present invention provides an optimization decision method for edge service cache of power distribution and utilization business, including:
[0006] Obtain computing power information and service request information of the current edge node, build computing service request model and real-time perception model of computing power resources and computing power load;
[0007] Obtain historical operation information of the cloud-edge system and obtain business request prediction information of the cloud-edge system;
[0008] According to the computing service request model and the real-time perception model of computing resources and computing load, a power distribution service edge service cache optimization model is constructed in combination with the service request prediction information;
[0009] Solving the power distribution and utilization business edge service cache optimization model to obtain an optimization decision-making solution;
[0010] According to the optimization decision plan, the edge service cache of the power distribution and utilization business is optimized.
[0011] It can be understood that, compared with the prior art, the present invention constructs a computing service request model and a real-time perception model of computing resources and computing load through the computing power information and business request information of the current edge node, and then constructs a power distribution business edge service cache optimization model in combination with the business request prediction information, thereby realizing the optimization of the power distribution business edge service cache. The computing service request model can obtain accurate business request information of the distribution network automation system, and the computing power resource and computing load real-time perception model can obtain the business request capability that each edge node can process, thereby accurately optimizing the power distribution business edge service cache, thereby improving the processing efficiency of the power distribution business; by real-time perception of the computing power resources and computing load information of the edge node, and then combining the business request information, the edge service cache of the edge node can be dynamically adjusted, thereby improving the node resource utilization of the distribution network automation system, thereby ensuring the high processing efficiency of the edge node for business requests; using the historical operation information of the cloud edge system, business request prediction is performed, and data support is provided for service cache optimization, thereby enabling the distribution network automation system to adapt to high real-time requirements and improving the processing efficiency of the distribution network automation system for power distribution business.
[0012] As a preferred solution, the acquisition of computing power information and service request information of the current edge node, and the construction of a computing service request model and a real-time perception model of computing power resources and computing power load specifically include:
[0013] Acquire the service request information of the current edge node, wherein the service request information includes: service request arrival information, service request data volume information, service request transmission information and service request type information;
[0014] Constructing a computing service request model according to the service request arrival information, service request data volume information, service request transmission information, service request type information and computing power information of the current edge node;
[0015] According to the computing power information of the current edge node, a real-time perception model of computing power resources and computing power load is constructed in combination with the computing service request model.
[0016] This preferred solution constructs a computing business request model through the business request information of the edge node, so as to obtain accurate information about the business requests in the distribution network automation system, and obtains the ability of each edge node to process business requests through a real-time perception model of computing resources and computing load, thereby accurately optimizing the edge service cache of the distribution and utilization business, thereby improving the processing efficiency of the distribution and utilization business.
[0017] As a preferred solution, the computing service request model is constructed according to the service request arrival information, service request data volume information, service request transmission information, service request type information and computing power information of the current edge node, specifically including:
[0018] Constructing a service instance function of the edge node according to the computing power resource information and the service request type information of the edge node;
[0019] Obtaining the computational complexity coefficient of the service request, and obtaining the computational delay of the service request by combining the data volume information of the service request and the computing power information of the current edge node;
[0020] Obtaining a communication delay of the service request according to the service request transmission information;
[0021] Obtaining a waiting delay of the service request according to the service request arrival information;
[0022] A service delay function of the edge node is constructed according to the calculation delay, communication delay and waiting delay of the service request.
[0023] A computing service request model is constructed based on the service instance function and the service delay function.
[0024] This preferred solution can accurately describe the processing flow of business requests in the distribution network automation system through the calculation delay, communication delay and waiting delay of the business request; by constructing the service instance function of the edge node through the computing power resource information and business request type information of the edge node, it can accurately describe the processing capability of the business request in the distribution network automation system, thereby obtaining a more accurate calculation business request model.
[0025] As a preferred solution, the computing power information of the current edge node is combined with the computing service request model to build a real-time perception model of computing power resources and computing power load, specifically including:
[0026] The computing power information of the current edge node includes: computing power resource information of the edge node and computing power load status information of the edge node;
[0027] Determine the business request processing capability information of the service instance in the edge node according to the computing power resource information of the edge node and the computing business request model, and then construct a real-time computing power resource representation matrix according to the business request processing capability information;
[0028] Obtaining business processing time information of the service instance in the edge node, and determining the computing load information of the service instance in the edge node in combination with the computing load status information of the edge node, and then constructing a real-time computing load representation matrix according to the computing load information;
[0029] A real-time perception model of computing power resources and computing power load is constructed based on the real-time computing power resource characterization matrix and the real-time computing power load characterization matrix.
[0030] This preferred solution determines the business request processing capability information and computing load information of the service instance in the edge node, and then constructs a real-time perception model of computing resources and computing load, so that it can perceive the processing capability of the distribution network automation system for business requests in real time, and thus dynamically adjust the edge service cache of the edge node, thereby improving the node resource utilization of the distribution network automation system, thereby ensuring the high processing efficiency of the edge node for business requests.
[0031] As a preferred solution, the computing service request model and the computing resource and computing load real-time perception model are combined with the service request prediction information to build a power distribution service edge service cache optimization model, which specifically includes:
[0032] Obtaining a service instance termination decision vector and a new service instance decision vector of the cloud-edge system;
[0033] According to the service instance termination decision vector and the new service instance decision vector of the cloud edge system, combined with the business request prediction information, the computing business request model and the computing power resource and computing power load real-time perception model, the power distribution business edge service cache optimization objective function is constructed;
[0034] According to the power distribution and utilization business edge service cache optimization objective function, a power distribution and utilization business edge service cache optimization model is constructed in combination with the computing business request model and the computing power resource and computing power load real-time perception model.
[0035] This preferred solution builds an edge service cache optimization model for power distribution and utilization business through service instance termination decision vectors and new service instance decision vectors, combined with business request prediction information, business request calculation model, and computing power resource and computing power load real-time perception model, and provides data support for service cache optimization through business request prediction; optimizes the edge service cache through service instance termination decision vectors and new service instance decision vectors, so that the distribution network automation system can adapt to high real-time requirements and improve the processing efficiency of the distribution network automation system for power distribution and utilization business.
[0036] As a preferred solution, the power distribution business edge service cache optimization objective function is constructed based on the service instance termination decision vector and the new service instance decision vector of the cloud edge system, combined with the business request prediction information, the computing business request model and the computing power resource and computing power load real-time perception model, specifically including:
[0037] The service request prediction information includes: a predicted amount of service requests of edge nodes;
[0038] Obtaining the current number of service requests of the edge node, and determining the predicted value of the computing load of the edge node in combination with the predicted amount of service requests of the edge node and the computing service request model;
[0039] According to the computing load prediction value of the edge node, an imminent computing load representation matrix is constructed in combination with the computing resources and the computing load real-time perception model;
[0040] Determine the computing resource change value of the edge node according to the service instance termination decision vector and the new service instance decision vector of the cloud-edge system in combination with the computing business request model;
[0041] According to the computing power resource change value of the edge node, the imminent computing power resource representation matrix is constructed in combination with the computing power resource and the computing power load real-time perception model;
[0042] According to the imminent computing power load characterization matrix and the imminent computing power resource characterization matrix, combined with a preset similarity characterization algorithm, a power distribution and utilization business edge service cache optimization objective function is constructed.
[0043] This preferred solution predicts the computing load and computing resources of the edge nodes in the distribution network automation system through business request prediction information, and then constructs an optimization objective function for the edge service cache of the power distribution business, so that the objective function fully considers the processing capabilities of the current and future edge nodes, provides an optimization target for dynamically adjusting the edge service cache of the edge nodes, improves the node resource utilization of the distribution network automation system, and thus ensures the high processing efficiency of the edge nodes for business requests.
[0044] As a preferred solution, the power distribution business edge service cache optimization objective function is constructed based on the imminent computing load characterization matrix and the imminent computing resource characterization matrix in combination with a preset similarity characterization algorithm, specifically including:
[0045] Determine a predicted value of the service request processing capacity of the edge node according to the computing power resource change value of the edge node;
[0046] According to the predicted value of the service request processing capability of the edge node, the imminent computing power resource representation matrix is converted to obtain an augmented imminent computing power resource representation matrix;
[0047] Determine a service request computing load prediction value of the edge node according to the computing load prediction value of the edge node;
[0048] According to the service request computing load prediction value of the edge node, the approximate computing load representation matrix is converted to obtain an augmented approximate computing load representation matrix;
[0049] Calculating the matrix similarity of the augmented approximate computing power resource representation matrix and the augmented approximate computing power load representation matrix according to a preset similarity representation algorithm;
[0050] The maximization of the matrix similarity is taken as the optimization goal, and the power distribution business edge service cache optimization objective function is obtained.
[0051] This preferred solution takes into account the spatial shifting effect of computing load unloading of business requests, and then constructs an optimization objective function for the edge service cache of the power distribution and utilization business, so that the objective function fully considers the processing capabilities of current and future edge nodes, provides an optimization target for dynamically adjusting the edge service cache of edge nodes, improves the node resource utilization of the distribution network automation system, and thus ensures the high processing efficiency of edge nodes for business requests.
[0052] As a preferred solution, the power distribution and utilization business edge service cache optimization model specifically includes:
[0053]
[0054]
[0055] Among them, θ is the service instance termination decision vector, is the decision vector for adding a new service instance, E is the number of edge nodes, S is the number of computing services; for It is used to represent the edge nodes E perceived by the cloud nodes in the cloud-edge system. e In the next service cache cycle, the service s i The processing power provided; It is used to represent the edge nodes E perceived by the cloud nodes in the cloud-edge system. e The business category s undertaken by the current service instance i The computing load of C e is the edge node E e The collection of current service instances; e,k is the edge node E e Upload service instance c e,k The computing power resources of the edge node; r0 is the occupied resources of the cloud-edge collaborative service instance in the edge node; To represent the edge node E e New corresponding to s i The number of service instances; For business i The amount of computing resources required; θ e,k A 0-1 variable indicating whether to terminate service instance c e,k , such as deciding to terminate service instance c e,k Then θ e,k The value is 1, otherwise it is 0; R e is the edge node E e The computing resources on board; u(c e,k ) is the service instance c e,k status.
[0056] This preferred solution optimizes the objective function of the edge service cache of the power distribution business, provides an optimization target for dynamically adjusting the edge service cache of the edge node, improves the node resource utilization of the distribution network automation system, and thus ensures the high processing efficiency of the edge node for business requests.
[0057] As a preferred solution, the optimization of the power distribution and utilization business edge service cache according to the optimization decision scheme specifically includes:
[0058] According to the optimization decision method, the decision variable value of the edge node is obtained;
[0059] terminating the service of the corresponding service instance in the edge node according to the decision variable value of the edge node, and releasing computing power resource space;
[0060] A new service instance is generated according to the decision variable value of the edge node, and the released computing resource space is allocated to the new service instance to optimize the edge service cache of the power distribution and utilization business.
[0061] This preferred solution controls the changes of service instances in the edge nodes of the distribution network automation system through decision variable values, can dynamically adjust the edge service cache of the edge nodes, improves the node resource utilization of the distribution network automation system, and thus improves the processing efficiency of the edge nodes for business requests.
[0062] Accordingly, an embodiment of the present invention provides an optimization decision device for power distribution and utilization business edge service cache, characterized in that it includes: a computing business request model and a computing power resource and computing power load real-time perception model construction module, a business request prediction information acquisition module, a power distribution and utilization business edge service cache optimization model construction module, an optimization decision solution acquisition module and a power distribution and utilization business edge service cache optimization module;
[0063] Among them, the computing service request model and computing power resource and computing power load real-time perception model construction module is used to obtain the computing power information and service request information of the current edge node, and construct the computing service request model and the computing power resource and computing power load real-time perception model;
[0064] The business request prediction information acquisition module is used to acquire historical operation information of the cloud-edge system to obtain business request prediction information of the cloud-edge system;
[0065] The power distribution and utilization business edge service cache optimization model construction module is used to construct a power distribution and utilization business edge service cache optimization model based on the computing business request model and the computing power resource and computing power load real-time perception model, combined with the business request prediction information;
[0066] The optimization decision solution acquisition module is used to solve the power distribution and utilization business edge service cache optimization model to obtain an optimization decision solution;
[0067] The power distribution and utilization business edge service cache optimization module is used to optimize the power distribution and utilization business edge service cache according to the optimization decision scheme.
[0068] It can be understood that, compared with the prior art, the present device constructs a computing service request model and a real-time perception model of computing resources and computing load through the computing power information and business request information of the current edge node, and then constructs a power distribution business edge service cache optimization model in combination with the business request prediction information, thereby realizing the optimization of the power distribution business edge service cache. The computing service request model can obtain accurate business request information of the distribution network automation system, and the computing power resource and computing load real-time perception model can obtain the business request capability that each edge node can process, thereby accurately optimizing the power distribution business edge service cache, thereby improving the processing efficiency of the power distribution business; by real-time perception of the computing power resources and computing load information of the edge node, combined with the business request information, the edge service cache of the edge node can be dynamically adjusted, thereby improving the node resource utilization of the distribution network automation system, thereby ensuring the high processing efficiency of the edge node for business requests; using the historical operation information of the cloud edge system, business request prediction is performed, providing data support for service cache optimization, thereby enabling the distribution network automation system to adapt to high real-time requirements, thereby improving the processing efficiency of the distribution network automation system for power distribution business. BRIEF DESCRIPTION OF THE DRAWINGS
[0069] Figure 1 : A flowchart of the steps of an optimization decision method for edge service cache of power distribution and utilization business provided by an embodiment of the present invention;
[0070] Figure 2 : A schematic diagram of the structure of an optimization decision device for edge service cache of power distribution and utilization business provided by an embodiment of the present invention;
[0071] Among them, 201: a module for building a computing business request model and a computing resource and computing load real-time perception model; 202: a module for obtaining business request prediction information; 203: a module for building a cache optimization model for an edge service of a power distribution and utilization business; 204: a module for obtaining an optimization decision plan; 205: a module for optimizing the cache of an edge service of a power distribution and utilization business. DETAILED DESCRIPTION
[0072] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0073] The power distribution and utilization Internet of Things takes the Internet of Things technology as its core concept, and applies intelligent sensors, cloud computing, edge computing, 5G and other technologies in the power distribution and utilization scenarios. It can cope with new situations such as the access of massive distributed resources and the continuous emergence of new forms of electricity and energy consumption. It is a feasible evolution direction for new distribution systems. At present, a large number of power system regulation, trading, management and other businesses have been extended to the power distribution and utilization side. Some business links must be processed nearby on the edge side, which puts forward new requirements for the system's computing power resources in terms of total amount, flexibility and accessibility, and gradually makes it a key production factor in the power distribution and utilization Internet of Things. The cloud-edge collaborative system integrates heterogeneous and distributed computing power resources into a unified resource pool, which can dynamically respond to various types of power distribution and utilization business requests with obvious differences and fluctuations, and can give full play to the advantages of business collaboration, data collaboration and resource collaboration between the cloud and the edge, and will become the digital computing power base of the power distribution and utilization Internet of Things.
[0074] The advantage of the cloud-edge computing paradigm is that service instances are deployed closer to data sources and service requests, that is, at the edge, while cloud nodes can be used for monitoring and scheduling, and backup computing power can be provided for the system. Edge service caching is an important method to achieve resource coordination of the power distribution and utilization Internet of Things. It refers to deploying service instances on cloud nodes and edge nodes in the power distribution and utilization Internet of Things to respond to business requests. Static edge service caching solutions cannot adapt to the dynamically changing working environment of the power distribution and utilization Internet of Things, and it is difficult to achieve dynamic adjustment of edge service caches by relying solely on static computing power equipment parameters.
[0075] The traditional distribution network automation system does not adopt the cloud-edge architecture. Each subsystem is independent and vertically isolated, and the demand for the coordination mechanism of computing resources between the cloud and the edge is weak. Therefore, the corresponding business requests are executed in the corresponding business system, and there is no need to pay attention to the service cache. Some solutions attempt to introduce technologies such as containers and microservices into the distribution and power consumption business processing, and try to rely on static parameters to optimize the edge service cache decision, and determine the deployment plan of the service instance in the business design stage, including the number of computing devices, the number of service instances, and the deployment location of the service instance. This method has a long decision cycle and is difficult to follow the trend of computing device status changes. The existing general cloud-edge computing resource coordination methods are mostly targeted at Internet application scenarios, and pay more attention to call loss rate, load balancing rate, etc. On the one hand, they fail to meet the high real-time requirements of some distribution network automation services, and on the other hand, they lack distributed scheduling methods that are compatible with the distribution network structure.
[0076] In summary, the embodiments of the present invention provide an optimization decision method and device for the edge service cache of the power distribution and utilization business. Based on the cloud-edge system, it can effectively support the automation business of the power distribution and utilization Internet of Things system under the background of explosive growth of access objects, increasingly complex operation scenarios, and ever-expanding task scales. It can flatten the spatiotemporal distribution curve of computing power load to cope with the ever-increasing contradiction between computing power supply and demand. It takes into account the real-time status of computing power resource equipment and business sources, can adapt to the differentiated needs of power distribution and utilization businesses, and plays a positive role in improving the high business timeout rate and low cloud edge node resource utilization. The specific content is described in the following embodiments one and two.
[0077] Embodiment 1
[0078] The optimization decision method for edge service cache of power distribution business described in the embodiment of the present invention is implemented based on the power distribution network automation system based on the cloud edge system. The cloud edge system in the power distribution network automation system is described below.
[0079] Specifically, there is a cloud node E0 and a set of edge nodes E={E1,E2,...,E |E|} and computing service set S = {s1,s2,...,s |S|}, each edge node has computing resources, and the edge node E e The computing resources are R e , edge node E e A set of service instances are deployed on the The service instance is used for business requests of the corresponding edge node.
[0080] Please refer to Figure 1 , which is a flowchart of a method for optimizing decision-making of edge service cache for power distribution and utilization business provided by an embodiment of the present invention, including steps S101-S105.
[0081] Step S101: Obtain computing power information and service request information of the current edge node, and build a computing service request model and a real-time perception model of computing power resources and computing power load.
[0082] In this embodiment, the acquisition of computing power information and service request information of the current edge node, and the construction of a computing service request model and a real-time perception model of computing power resources and computing power load specifically include:
[0083] Obtaining service request information of the current edge node, the service request information including: service request arrival information, service request data volume information, service request transmission information and service request type information;
[0084] Constructing a computing service request model according to the service request arrival information, service request data volume information, service request transmission information, service request type information and computing power information of the current edge node;
[0085] According to the computing power information of the current edge node, a real-time perception model of computing power resources and computing power load is constructed in combination with the computing service request model.
[0086] In this embodiment, the computing service request model is constructed according to the service request arrival information, service request data volume information, service request transmission information, service request type information and computing power information of the current edge node, specifically including:
[0087] Constructing a service instance function of the edge node according to the computing power resource information and the service request type information of the edge node;
[0088] Obtaining the computational complexity coefficient of the service request, and obtaining the computational delay of the service request by combining the data volume information of the service request and the computing power information of the current edge node;
[0089] Obtaining a communication delay of the service request according to the service request transmission information;
[0090] Obtaining a waiting delay of the service request according to the service request arrival information;
[0091] Constructing a service delay function of the edge node according to the calculation delay, communication delay and waiting delay of the service request;
[0092] A computing service request model is constructed based on the service instance function and the service delay function.
[0093] In an optional embodiment, the service instance function specifically includes:
[0094] c e,k =.r e,k ,s(c e,k ),u(c e,k ) / (1),
[0095] Among them, c e,k is a service instance; e,k is the computing power resource of the service instance, s(c e,k ) is the executable business type of the service example, u(c e,k ) is the service instance status.
[0096] In an optional embodiment, the service instance function further includes:
[0097] Since the service instances in the cloud-edge system for the same type of business requests all correspond to the same service image, when s(c e,k)=s(c e,x )=s i Sometimes
[0098] Among them, s(c e,k ) is the executable business type of the service example, including: fault location, status monitoring, line loss analysis and EV orderly charging, etc.; u(c e,k ) is the service instance status, including: creating, busy and idle, represented by the values 2, 1 and 0 respectively. The service instance in the creating state cannot provide computing services. The duration of the creating state is a constant d P ;s(c e,x ) is the computing service performed by the service instance; i To calculate the business type; r e,k For service instance c e,k The amount of computing resources, r e,x For service instance c e,x The amount of computing resources; The amount of computing resources required for the service instance; e,x and Two different indexing methods for computing resources of service instances;
[0099] A cloud-edge collaborative service instance needs to be deployed in the edge node, and the resource usage is r0. The cloud-edge collaborative service instance is always in a busy state.
[0100] It should be noted that the business request q e,g And arrival time Deadline Calculate the amount of data A e,g When a service request arrives, the edge node can perceive the above information and forward it to the cloud node. e,g ) refers to the business request q e,g The time from the arrival of the edge node to the completion of the service is determined by the waiting delay d W (q e,g ), communication delay d T (q e,g ), calculate the delay d C (q e,g )composition:
[0101] d(q e,g ) = d W (q e,g )+d T (q e,g )+d C (q e,g ) (2),
[0102] Among them, d(qe,g ) is the service delay, d W (q e,g ) is the waiting delay, d T (q e,g ) is the communication delay, d C (q e,g ) is the calculation delay;
[0103] Communication delay refers to the time it takes to offload to the edge node E x Under the condition, at the edge node E e and E x The waiting delay refers to the time consumed by the data transmission process required for business execution from the business request q e,g The time consumed from reaching the edge node to the start of data transmission required by the service; the calculation delay refers to the time consumed in the service execution process, and its starting point is the start execution time of the service. Right now The computing resources of service instances for the same type of business in the cloud-edge system are equal.
[0104] In an optional embodiment, the obtaining of the computational complexity coefficient of the service request and obtaining the computational delay of the service request in combination with the data volume information of the service request and the computing power information of the current edge node specifically includes:
[0105]
[0106] Among them, d C (q e,g ) is the calculation delay; For business request e,g The computational complexity coefficient of the corresponding service indicates the number of instructions required for this type of service to process 1MByte of input data. Q (q e,g )=s Q (q e,h )=s i Time Satisfaction and Q (q e,h ) are business requests q e,g and q e,h The corresponding computing service type, s i To calculate the business type, For business request e,h The computational complexity coefficient of the corresponding service, For i The computational complexity coefficient of the corresponding service; A e,g The amount of data required for the business; For business request e,g The amount of computing resources corresponding to the computing business type.
[0107] In an optional embodiment, obtaining the communication delay of the service request according to the service request transmission information specifically includes:
[0108]
[0109] Among them, d T (q e,g ) is the communication delay, δ max is the maximum delay jitter, which is generally set to 10 milliseconds. e,x is the edge node E e To E x The transmission speed, A e,g is the amount of data required for the business request, sup[d T (q e,g )] is used to indicate business request q e,g The minimum communication delay d T (q e,g ) upper bound.
[0110] In an optional embodiment, obtaining the waiting delay of the service request according to the service request arrival information specifically includes:
[0111] According to business request e,g The time consumed from reaching the edge node to the start of data transmission required by the service is obtained as the service request q e,g The waiting delay.
[0112] In some optional implementations of this embodiment, the computing service request model further includes:
[0113] Using the intrinsic importance η i Characterizing Computing Services i The importance of all corresponding to s i Business request q e,g The inherent importance of i ,Right now:
[0114] When Q (q e,g )=s i When η(q e,g ) = η i (5),
[0115] Among them, q e,g Indicates submission to edge node E e The g-th service request, η(q e,g ) indicates a business request q e,g The inherent importance of Q (q e,g ) indicates a business request qe,g Corresponding computing service type.
[0116] It should be noted that η max The definition of Refers to the maximum value of the corresponding inherent importance among all business requests; when η i =η max When i For important business; when η(q e,g ) = η max That is, η(s Q (q e,g ))=η max When e,g In the embodiment of the present invention, the business importance is divided into five categories: unimportant, less important, generally important, more important and very important, and are assigned values of 1, 3, 5, 7 and 9, respectively. max is 9.
[0117] It should be noted that, the construction of the service delay function of the edge node according to the computing delay, communication delay and waiting delay of the service request refers to combining equations (2), (3) and (4) to form the service delay function of the edge node; the construction of the computing service request model according to the service instance function and the service delay function refers to combining equations (1), (2), (3) and (4) to form a computing service request model, and the computing service request model also includes the content described in equation (5).
[0118] By constructing a computing business request model, the business request information in the distribution network automation system can be accurately described, thereby providing data support for the optimization decision of the edge service cache of the power distribution business, and improving the processing efficiency of the distribution network automation system for business requests.
[0119] In this embodiment, the computing power information of the current edge node is combined with the computing service request model to build a real-time perception model of computing power resources and computing power load, specifically including:
[0120] The computing power information of the current edge node includes: computing power resource information of the edge node and computing power load status information of the edge node;
[0121] Determine the business request processing capability information of the service instance in the edge node according to the computing power resource information of the edge node and the computing business request model, and then construct a real-time computing power resource representation matrix according to the business request processing capability information;
[0122] Obtaining business processing time information of the service instance in the edge node, and determining the computing load information of the service instance in the edge node in combination with the computing load status information of the edge node, and then constructing a real-time computing load representation matrix according to the computing load information;
[0123] A real-time perception model of computing power resources and computing power load is constructed based on the real-time computing power resource characterization matrix and the real-time computing power load characterization matrix.
[0124] In an optional embodiment, the real-time computing resource characterization matrix (RCRM) is used to characterize the processing capability of the cloud-edge system for business requests;
[0125] The real-time computing power resource representation matrix is shown in the following formula (6):
[0126]
[0127] Among them, B e,i Used to indicate the current deployment on the edge node E perceived by the cloud node e The service instance can be used to provide services to the business s in the next service cache cycle. i The processing capacity provided meets the requirements of the following formula (7):
[0128]
[0129] In formula (7), T is the length of the service cache cycle, r e,k For its computing power resources, Used to represent a service instance c e,k Can it be s i The computing resources provided by the business of this type satisfy the following formula (8):
[0130]
[0131] In formula (8), s(c e,k ) is the service instance c e,k The types of business that can be performed.
[0132] In an optional embodiment, the real-time work-load matrix (RWLM) is used to characterize the computing load status of the edge node, that is, the processing progress of the business request being processed by the service instance;
[0133] The real-time computing load characterization matrix is shown in formula (9):
[0134]
[0135] In formula (9), L e,i Used to represent the edge node E perceived by the cloud node e The business category s undertaken by the current service instance i The computing load meets the requirements of the following formula (10):
[0136]
[0137] In formula (10), Used to represent a service instance c e,k Can it be s i Type of business provides computing resources, r e,k For its computing power resources, and Respectively represent service instances c e,k The remaining time required to complete the current business (primary business) and the reserved business (secondary business) is calculated as shown in the following formula (11) and formula (12) respectively:
[0138]
[0139] In formula (11), For service instance c e,k The computational complexity coefficient of For service instance c e,k The amount of data for the business request currently being processed, For service instance c e,k The amount of computing resources, t is the current time, For service instance c e,k The start execution time of the current business request;
[0140] In formula (12), For service instance c e,k The computational complexity coefficient of ; For service instance c e,k The amount of data requested by the business it has reserved, when c e,k If there is no reservation, For service instance c e,k The amount of computing resources.
[0141] It should be noted that the real-time perception model of computing power resources and computing power loads constructed according to the real-time computing power resource characterization matrix and the real-time computing power load characterization matrix refers to the real-time perception model of computing power resources and computing power loads constructed according to formula (6), formula (7), formula (8), formula (9), formula (10), formula (11), formula (12) and their corresponding description contents. The real-time perception model of computing power resources and computing power loads can accurately describe the processing capability of business requests in the distribution network system, thereby improving the accuracy of optimization decisions for the edge service cache of the distribution and utilization business, thereby improving the processing efficiency of business requests.
[0142] Step S102: Acquire historical operation information of the cloud-edge system to obtain business request prediction information of the cloud-edge system.
[0143] In this embodiment, the acquiring of historical operation information of the cloud-edge system to obtain the service request prediction information of the cloud-edge system specifically includes:
[0144] Obtain historical operation information of the cloud-edge system, where the historical operation information includes: the number of historical business requests;
[0145] According to the number of historical business requests, the probability distribution of the number of business requests arriving is obtained, and the corresponding expectation is obtained as the predicted value;
[0146] The predicted value is used as the business request prediction information of the cloud-edge system.
[0147] It should be noted that there are currently many existing methods and technologies for obtaining prediction information through historical operation information. For example, the business request prediction information in the embodiment of the present invention can also be achieved through artificial intelligence methods, so the process will not be described in detail here.
[0148] Step S103: constructing a power distribution business edge service cache optimization model based on the computing business request model and the computing power resource and computing power load real-time perception model in combination with the business request prediction information.
[0149] In this embodiment, the computing service request model and the computing resource and computing load real-time perception model are combined with the service request prediction information to build a power distribution service edge service cache optimization model, which specifically includes:
[0150] Obtaining a service instance termination decision vector and a new service instance decision vector of the cloud-edge system;
[0151] According to the service instance termination decision vector and the new service instance decision vector of the cloud edge system, combined with the business request prediction information, the computing business request model and the computing power resource and computing power load real-time perception model, the power distribution business edge service cache optimization objective function is constructed;
[0152] According to the power distribution and utilization business edge service cache optimization objective function, a power distribution and utilization business edge service cache optimization model is constructed in combination with the computing business request model and the computing power resource and computing power load real-time perception model.
[0153] In an optional embodiment, the decision variables in the edge service cache of the power distribution business include two types. The first type is to decide whether to terminate the deployed service instance, and the second type is to decide how many new service instances of a certain type are added to the edge node. In this embodiment of the present invention, the service instance termination decision is represented by a vector θ (the vector θ is the service instance termination decision vector), and the vector Represents the decision to add a new service instance (vector That is, the decision vector for adding a new service instance).
[0154] Specifically, the vector θ and the vector As shown in the following formula (13) and formula (14):
[0155]
[0156] In formula (13), θ e,k A 0-1 variable indicating whether to terminate service instance c e,k , such as deciding to terminate service instance c e,k Then θ e,k The value is 1, otherwise it is 0; C e is the set of service instances, and E is the set of edge nodes;
[0157] In formula (14), is an integer variable, representing the edge node E e New corresponding to s i The number of service instances (the number depends on the optimization result), which is an integer greater than 0, E is the set of edge nodes, and S is the set of computing services.
[0158] In this embodiment, according to the service instance termination decision vector and the new service instance decision vector of the cloud edge system, combined with the business request prediction information, the computing business request model and the computing power resource and computing power load real-time perception model, the power distribution business edge service cache optimization objective function is constructed, which specifically includes:
[0159] The service request prediction information includes: a predicted amount of service requests of edge nodes;
[0160] Obtaining the current number of service requests of the edge node, and determining the predicted value of the computing load of the edge node in combination with the predicted amount of service requests of the edge node and the computing service request model;
[0161] According to the computing load prediction value of the edge node, an imminent computing load representation matrix is constructed in combination with the computing resources and the computing load real-time perception model;
[0162] Determine the computing resource change value of the edge node according to the service instance termination decision vector and the new service instance decision vector of the cloud-edge system in combination with the computing business request model;
[0163] According to the computing power resource change value of the edge node, the imminent computing power resource representation matrix is constructed in combination with the computing power resource and the computing power load real-time perception model;
[0164] According to the imminent computing power load characterization matrix and the imminent computing power resource characterization matrix, combined with a preset similarity characterization algorithm, a power distribution and utilization business edge service cache optimization objective function is constructed.
[0165] It should be noted that the upcoming work-load matrix (UWLM) is used to describe the number of business requests arriving in the next service cache cycle, and then estimate the additional computing load in the next service cache cycle; the upcoming computility resource matrix (UCRM) is used to describe the time d required to deploy a new service instance. P The corresponding loss of effective computing resources.
[0166] In an optional embodiment, the computing load prediction value of the edge node is combined with the computing resources and the computing load real-time perception model to construct an approximate computing load representation matrix, which specifically includes:
[0167] The predicted computing load value of the edge node is added to the real-time computing load representation matrix to obtain the approximate computing load representation matrix:
[0168]
[0169] In formula (15), UWLM is the imminent computing load characterization matrix, RWLM is the real-time computing load characterization matrix, Characterize the edge node E predicted by the cloud node e In the next service cache cycle, the corresponding service s i The computing load, Satisfies the following formula (16):
[0170]
[0171] In formula (16), |C e | is the service instance set C e The number of service instances in ne,i To come to the edge node E e Business i The number of For business i The computational complexity coefficient of For business i The average amount of data that needs to be processed.
[0172] In an optional embodiment, the constructing of an imminent computing resource representation matrix based on the computing resource change value of the edge node and the computing resource and computing load real-time perception model specifically includes:
[0173] The computing power resource change value of the edge node is added to the real-time computing power resource representation matrix to obtain the approximate computing power resource representation matrix:
[0174]
[0175] In formula (17), UCRM is the representation matrix of the approximate computing power resources, RCRM is the representation matrix of the real-time computing power resources, Characterizes that due to the adjustment of service instances, at the edge node E e The corresponding business s i Changes in computing resources, Satisfies the following formula (18):
[0176]
[0177] In formula (18), T is the length of the service cache period, d P It takes time to deploy new service instances. Used to represent the edge node E e Newly added corresponding to business s i The number of service instances; For business i The amount of computing resources required, |C e | is the service instance set C e The number of service instances in Used to describe the service instance c e,k Can it be s i Provide computing resources for business requests of this type; e,k A 0-1 variable indicating whether to terminate service instance c e,k , such as deciding to terminate service instance c e,k Then θ e,k The value is 1, otherwise it is 0; r e,k For service instance c e,k The amount of computing resources.
[0178] In this embodiment, the power distribution business edge service cache optimization objective function is constructed based on the imminent computing load characterization matrix and the imminent computing resource characterization matrix in combination with a preset similarity characterization algorithm, specifically including:
[0179] Determine a predicted value of the service request processing capacity of the edge node according to the computing power resource change value of the edge node;
[0180] According to the predicted value of the service request processing capability of the edge node, the imminent computing power resource representation matrix is converted to obtain an augmented imminent computing power resource representation matrix;
[0181] Determine a service request computing load prediction value of the edge node according to the computing load prediction value of the edge node;
[0182] According to the service request computing load prediction value of the edge node, the approximate computing load representation matrix is converted to obtain an augmented approximate computing load representation matrix;
[0183] Calculating the matrix similarity of the augmented approximate computing power resource representation matrix and the augmented approximate computing power load representation matrix according to a preset similarity representation algorithm;
[0184] The maximization of the matrix similarity is taken as the optimization goal, and the power distribution business edge service cache optimization objective function is obtained.
[0185] It should be noted that the augmented upcoming computility resource matrix (AUCRM) and the augmented upcoming work-load matrix (AUWLM) are used to characterize the overall computing resources and computing load of the system respectively.
[0186] In an optional embodiment, the converting of the imminent computing power resource representation matrix according to the predicted value of the service request processing capability of the edge node to obtain an augmented imminent computing power resource representation matrix specifically includes:
[0187] According to the spatial translation effect characteristics of the computing load of the service request unloading and the predicted value of the service request processing capacity of the edge node, the imminent computing resource representation matrix is transformed to obtain an augmented imminent computing resource representation matrix:
[0188]
[0189] In formula (19), AUCRM is the augmented approximate computing resource representation matrix, Represents the edge node E perceived by the cloud node eIn the next service cache cycle, the service s i Provides processing power to meet Represents the cloud node's perception of the entire system that can be provided to the business s in the next service cache cycle i Provides processing power to meet
[0190] In an optional embodiment, the converting of the imminent computing load characterization matrix according to the service request computing load prediction value of the edge node to obtain an augmented imminent computing load characterization matrix specifically includes:
[0191] According to the spatial translation effect characteristics of the computing load of the service request unloading and the predicted value of the computing load of the service request of the edge node, the imminent computing load characterization matrix is transformed to obtain an augmented imminent computing load characterization matrix:
[0192]
[0193] In formula (20), Characterize the edge node E predicted by the cloud node e Ignore task offloading and assume the corresponding service s in the next service cache cycle. i The computing load meets Characterize the cloud node prediction to obtain the corresponding business s borne by the whole system in the next service cache cycle i The computing load meets
[0194] In an optional embodiment, the matrix similarity of the augmented approximate computing power resource representation matrix and the augmented approximate computing power load representation matrix is calculated according to a preset similarity representation algorithm, specifically including:
[0195] Based on the Euclidean similarity, the matrix similarity of the augmented approximate computing power resource representation matrix and the augmented approximate computing power load representation matrix is calculated:
[0196]
[0197] In formula (21), is the matrix similarity, Represents the edge node E perceived by the cloud node e In the next service cache cycle, the service s i Provides processing power to meet Characterize the edge node E predicted by the cloud node e Ignore task offloading and assume the corresponding service s in the next service cache cycle. iThe computing load meets |E| is the number of elements in the edge node set, and |S| is the number of elements in the computing service set.
[0198] In some optional implementations of this embodiment, the power distribution and utilization business edge service cache optimization model specifically includes:
[0199] Construct constraints for optimizing the cache of edge services for power distribution and utilization business, including: edge node computing resource constraints, running service instance status constraints, and new instance constraints.
[0200] The edge node computing resource constraint means that the sum of computing resources occupied by all service instances and schedulers cannot be higher than the total computing resources of the edge node:
[0201]
[0202] In formula (22), the first term on the left side of the inequality sign is the adjustment of the front edge node E e The computing resource usage of the edge cache solution is the second item, which is the computing resource adjustment of the current edge cache solution. e | is the service instance set C e The number of service instances in r e,k For service instance c e,k The computing power resources of r0 are as follows: r1 is the computing power resources required to deploy cloud-edge collaborative service instances in edge nodes. Used to represent the edge node E e Newly added corresponding to business s i The number of service instances; For business i The amount of computing resources required, θ e,k A 0-1 variable indicating whether to terminate service instance c e,k , such as deciding to terminate service instance c e,k Then θ e,k The value is 1, otherwise it is 0; r e,k For service instance c e,k The amount of computing resources, R e is the edge node E e The computing resources carried on it.
[0203] The running service instance state constraint means that the service instance that is executing the computing service cannot be terminated:
[0204]
[0205] In formula (23), θ e,k A 0-1 variable indicating whether to terminate service instance c e,k , such as deciding to terminate service instance c e,k Then θe,k The value is 1, otherwise it is 0, u(c e,k ) is the service instance status.
[0206] New instance constraints are It cannot be a negative number and must be less than the multiple of edge node resources to service instance resources:
[0207]
[0208] In formula (24), Used to represent the edge node E e Newly added corresponding to business s i The number of service instances; R e is the edge node E e r0 is the computing power resources installed on the edge node, r1 is the amount of resources required to deploy the cloud-edge collaborative service instance in the edge node, For business i The amount of computing resources required.
[0209] In this embodiment, the power distribution and utilization business edge service cache optimization model specifically includes:
[0210]
[0211] Among them, θ is the service instance termination decision vector, is the decision vector for adding a new service instance, E is the number of edge nodes, S is the number of computing services; for It is used to represent the edge nodes E perceived by the cloud nodes in the cloud-edge system. e In the next service cache cycle, the service s i The processing power provided; It is used to represent the edge nodes E perceived by the cloud nodes in the cloud-edge system. e The business category s undertaken by the current service instance i The computing load of C e is the edge node E e The collection of current service instances; e,k is the edge node E e Upload service instance c e,k The computing power resources of the edge node; r0 is the occupied resources of the cloud-edge collaborative service instance in the edge node; To represent the edge node E e New corresponding to s i The number of service instances; For business i The amount of computing resources required; θ e,k A 0-1 variable indicating whether to terminate service instance c e,k, such as deciding to terminate service instance c e,k Then θ e,k The value is 1, otherwise it is 0; R e is the edge node E e The computing resources on board; u(c e,k ) is the service instance c e,k status.
[0212] Step S104: solving the power distribution and utilization business edge service cache optimization model to obtain an optimization decision solution.
[0213] In this embodiment, the power distribution and utilization business edge service cache optimization model belongs to quadratic integer programming, which has the characteristics of fast solution speed and mature solution method. It can be solved by solvers such as Gurobi to obtain an optimized decision plan. Among them, affected by the power distribution and utilization business edge service cache optimization objective function, there may be computing power resources on each edge node where service instances are not deployed, but these resources will be utilized in the business request offloading link.
[0214] Step S105: Optimizing the power distribution and utilization business edge service cache according to the optimization decision plan.
[0215] In this embodiment, the optimization of the power distribution and utilization business edge service cache according to the optimization decision scheme specifically includes:
[0216] According to the optimization decision method, the decision variable value of the edge node is obtained;
[0217] terminating the service of the corresponding service instance in the edge node according to the decision variable value of the edge node, and releasing computing power resource space;
[0218] A new service instance is generated according to the decision variable value of the edge node, and the released computing resource space is allocated to the new service instance to optimize the edge service cache of the power distribution and utilization business.
[0219] This embodiment constructs a computing service request model and a real-time perception model of computing resources and computing load through the computing power information and business request information of the current edge node, and then constructs a power distribution business edge service cache optimization model in combination with the business request prediction information, thereby optimizing the power distribution business edge service cache. The computing service request model can obtain accurate business request information of the distribution network automation system, and the computing power resource and computing load real-time perception model obtains the business request capability that each edge node can process, thereby accurately optimizing the power distribution business edge service cache, thereby improving the processing efficiency of the power distribution business; by real-time perception of the computing power resources and computing load information of the edge node, combined with the business request information, the edge service cache of the edge node can be dynamically adjusted, thereby improving the node resource utilization rate of the distribution network automation system, thereby ensuring the high processing efficiency of the edge node for business requests; using the historical operation information of the cloud edge system, business request prediction is performed to provide data support for service cache optimization, thereby enabling the distribution network automation system to adapt to high real-time requirements and improving the processing efficiency of the distribution network automation system for power distribution business.
[0220] Embodiment 2
[0221] Please refer to Figure 2 , which is a schematic diagram of a structure of an optimization decision device for an edge service cache of a power distribution and utilization business provided by an embodiment of the present invention, comprising: a computing service request model and a computing resource and computing load real-time perception model construction module 201, a service request prediction information acquisition module 202, a power distribution and utilization business edge service cache optimization model construction module 203, an optimization decision solution acquisition module 204 and a power distribution and utilization business edge service cache optimization module 205;
[0222] Among them, the computing service request model and computing power resource and computing power load real-time perception model construction module 201 is used to obtain the computing power information and service request information of the current edge node, and construct the computing service request model and the computing power resource and computing power load real-time perception model.
[0223] In this embodiment, the computing service request model and computing power resources and computing power load real-time perception model construction module 201 includes: computing service request model and computing power resources and computing power load real-time perception model construction unit;
[0224] The computing service request model and computing power resource and computing power load real-time perception model construction unit is used to obtain the service request information of the current edge node, and the service request information includes: service request arrival information, service request data volume information, service request transmission information and service request type information;
[0225] Constructing a computing service request model according to the service request arrival information, service request data volume information, service request transmission information, service request type information and computing power information of the current edge node;
[0226] According to the computing power information of the current edge node, a real-time perception model of computing power resources and computing power load is constructed in combination with the computing service request model.
[0227] In this embodiment, the computing service request model and computing power resource and computing power load real-time perception model construction unit includes: a computing service request model construction subunit;
[0228] The computing service request model building subunit is used to build a service instance function of the edge node according to the computing power resource information and service request type information of the edge node;
[0229] Obtaining the computational complexity coefficient of the service request, and obtaining the computational delay of the service request by combining the data volume information of the service request and the computing power information of the current edge node;
[0230] Obtaining a communication delay of the service request according to the service request transmission information;
[0231] Obtaining a waiting delay of the service request according to the service request arrival information;
[0232] Constructing a service delay function of the edge node according to the calculation delay, communication delay and waiting delay of the service request;
[0233] A computing service request model is constructed based on the service instance function and the service delay function.
[0234] In this embodiment, the computing service request model and computing power resource and computing power load real-time perception model construction unit includes: a computing power resource and computing power load real-time perception model construction subunit;
[0235] The computing power resource and computing power load real-time perception model construction subunit is used for the computing power information of the current edge node, including: computing power resource information of the edge node and computing power load status information of the edge node;
[0236] Determine the business request processing capability information of the service instance in the edge node according to the computing power resource information of the edge node and the computing business request model, and then construct a real-time computing power resource representation matrix according to the business request processing capability information;
[0237] Obtaining business processing time information of the service instance in the edge node, and determining the computing load information of the service instance in the edge node in combination with the computing load status information of the edge node, and then constructing a real-time computing load representation matrix according to the computing load information;
[0238] A real-time perception model of computing power resources and computing power load is constructed based on the real-time computing power resource characterization matrix and the real-time computing power load characterization matrix.
[0239] The business request prediction information acquisition module 202 is used to acquire historical operation information of the cloud-edge system to obtain business request prediction information of the cloud-edge system.
[0240] The power distribution and utilization business edge service cache optimization model construction module 203 is used to construct a power distribution and utilization business edge service cache optimization model based on the computing business request model and the real-time perception model of computing resources and computing load, combined with the business request prediction information.
[0241] In this embodiment, the power distribution and utilization business edge service cache optimization model construction module 203 includes: a power distribution and utilization business edge service cache optimization model construction unit;
[0242] The power distribution and utilization business edge service cache optimization model construction unit is used to obtain the service instance termination decision vector and the new service instance decision vector of the cloud edge system;
[0243] According to the service instance termination decision vector and the new service instance decision vector of the cloud edge system, combined with the business request prediction information, the computing business request model and the computing power resource and computing power load real-time perception model, the power distribution business edge service cache optimization objective function is constructed;
[0244] According to the power distribution and utilization business edge service cache optimization objective function, a power distribution and utilization business edge service cache optimization model is constructed in combination with the computing business request model and the computing power resource and computing power load real-time perception model.
[0245] In this embodiment, the power distribution and utilization business edge service cache optimization model construction unit includes: a power distribution and utilization business edge service cache optimization objective function construction subunit;
[0246] The power distribution and utilization business edge service cache optimization objective function construction subunit is used to predict the business request information according to the business request prediction information including: the predicted amount of business requests of edge nodes;
[0247] Obtaining the current number of service requests of the edge node, and determining the predicted value of the computing load of the edge node in combination with the predicted amount of service requests of the edge node and the computing service request model;
[0248] According to the computing load prediction value of the edge node, an imminent computing load representation matrix is constructed in combination with the computing resources and the computing load real-time perception model;
[0249] Determine the computing resource change value of the edge node according to the service instance termination decision vector and the new service instance decision vector of the cloud-edge system in combination with the computing business request model;
[0250] According to the computing power resource change value of the edge node, the imminent computing power resource representation matrix is constructed in combination with the computing power resource and the computing power load real-time perception model;
[0251] According to the imminent computing power load characterization matrix and the imminent computing power resource characterization matrix, combined with a preset similarity characterization algorithm, a power distribution and utilization business edge service cache optimization objective function is constructed.
[0252] In this embodiment, the power distribution and utilization business edge service cache optimization objective function construction subunit includes: a power distribution and utilization business edge service cache optimization objective function construction component;
[0253] The power distribution and utilization business edge service cache optimization objective function construction component is used to determine the predicted value of the business request processing capacity of the edge node according to the computing power resource change value of the edge node;
[0254] According to the predicted value of the service request processing capability of the edge node, the imminent computing power resource representation matrix is converted to obtain an augmented imminent computing power resource representation matrix;
[0255] Determine a service request computing load prediction value of the edge node according to the computing load prediction value of the edge node;
[0256] According to the service request computing load prediction value of the edge node, the approximate computing load representation matrix is converted to obtain an augmented approximate computing load representation matrix;
[0257] Calculating the matrix similarity of the augmented approximate computing power resource representation matrix and the augmented approximate computing power load representation matrix according to a preset similarity representation algorithm;
[0258] The maximization of the matrix similarity is taken as the optimization goal, and the power distribution business edge service cache optimization objective function is obtained.
[0259] In this embodiment, the power distribution and utilization business edge service cache optimization model construction module 203 includes: a power distribution and utilization business edge service cache optimization model unit;
[0260] The power distribution and utilization business edge service cache optimization model unit includes:
[0261]
[0262] Among them, θ is the service instance termination decision vector, is the decision vector for adding a new service instance, E is the number of edge nodes, S is the number of computing services; for It is used to represent the edge nodes E perceived by the cloud nodes in the cloud-edge system. e In the next service cache cycle, the service s i The processing power provided; It is used to represent the edge nodes E perceived by the cloud nodes in the cloud-edge system. e The business category s undertaken by the current service instance i The computing load of C e is the edge node E e The collection of current service instances; e,k is the edge node E e Upload service instance c e,k The computing power resources of the edge node; r0 is the occupied resources of the cloud-edge collaborative service instance in the edge node; To represent the edge node E e New corresponding to s i The number of service instances; For business i The amount of computing resources required; θ e,k A 0-1 variable indicating whether to terminate service instance c e,k , such as deciding to terminate service instance c e,k Then θ e,k The value is 1, otherwise it is 0; R e is the edge node E e The computing resources on board; u(c e,k ) is the service instance c e,k status.
[0263] The optimization decision solution acquisition module 204 is used to solve the power distribution and utilization business edge service cache optimization model to obtain an optimization decision solution.
[0264] The power distribution and utilization business edge service cache optimization module 205 is used to optimize the power distribution and utilization business edge service cache according to the optimization decision scheme.
[0265] In this embodiment, the power distribution and utilization business edge service cache optimization module 205 includes: a power distribution and utilization business edge service cache optimization unit;
[0266] The power distribution and utilization business edge service cache optimization unit is used to obtain the decision variable value of the edge node according to the optimization decision method;
[0267] terminating the service of the corresponding service instance in the edge node according to the decision variable value of the edge node, and releasing computing power resource space;
[0268] A new service instance is generated according to the decision variable value of the edge node, and the released computing resource space is allocated to the new service instance to optimize the edge service cache of the power distribution and utilization business.
[0269] In summary, the embodiment of the present invention constructs a computing service request model and a real-time perception model of computing resources and computing load through the computing power information and business request information of the current edge node, and then constructs a power distribution business edge service cache optimization model in combination with the business request prediction information, thereby realizing the optimization of the power distribution business edge service cache. The computing service request model can obtain accurate business request information of the distribution network automation system, and the computing power resource and computing load real-time perception model obtains the business request capability that each edge node can process, thereby accurately optimizing the power distribution business edge service cache, thereby improving the processing efficiency of the power distribution business; by real-time perception of the computing power resources and computing load information of the edge node, combined with the business request information, the edge service cache of the edge node can be dynamically adjusted, thereby improving the node resource utilization of the distribution network automation system, thereby ensuring the high processing efficiency of the edge node for business requests; using the historical operation information of the cloud edge system, business request prediction is performed to provide data support for service cache optimization, thereby enabling the distribution network automation system to adapt to high real-time requirements and improving the processing efficiency of the distribution network automation system for power distribution business.
[0270] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.
Claims
1. A method for optimizing decision-making of edge service cache for power distribution and utilization business, characterized in that: include: Obtain computing power information and service request information of the current edge node, build computing service request model and real-time perception model of computing power resources and computing power load; Obtain historical operation information of the cloud-edge system and obtain business request prediction information of the cloud-edge system; According to the computing service request model and the real-time perception model of computing resources and computing load, a power distribution service edge service cache optimization model is constructed in combination with the service request prediction information; Solving the power distribution and utilization business edge service cache optimization model to obtain an optimization decision-making solution; According to the optimization decision plan, the edge service cache of the power distribution and utilization business is optimized.
2. The optimization decision method for edge service cache of power distribution and utilization business according to claim 1, characterized in that: The acquisition of computing power information and service request information of the current edge node, and the construction of a computing service request model and a real-time perception model of computing power resources and computing power load specifically include: Acquire the service request information of the current edge node, wherein the service request information includes: service request arrival information, service request data volume information, service request transmission information and service request type information; Constructing a computing service request model according to the service request arrival information, service request data volume information, service request transmission information, service request type information and computing power information of the current edge node; According to the computing power information of the current edge node, a real-time perception model of computing power resources and computing power load is constructed in combination with the computing service request model.
3. The optimization decision method for edge service cache of power distribution and utilization business according to claim 2, characterized in that: The step of constructing a computing service request model according to the service request arrival information, service request data volume information, service request transmission information, service request type information, and computing power information of the current edge node specifically includes: Constructing a service instance function of the edge node according to the computing power resource information and the service request type information of the edge node; Obtaining the computational complexity coefficient of the service request, and obtaining the computational delay of the service request by combining the data volume information of the service request and the computing power information of the current edge node; Obtaining a communication delay of the service request according to the service request transmission information; Obtaining a waiting delay of the service request according to the service request arrival information; Constructing a service delay function of the edge node according to the calculation delay, communication delay and waiting delay of the service request; A computing service request model is constructed based on the service instance function and the service delay function.
4. The optimization decision method for edge service cache of power distribution and utilization business according to claim 1, 2 or 3, characterized in that: The step of building a real-time perception model of computing resources and computing load based on the computing power information of the current edge node and in combination with the computing service request model specifically includes: The computing power information of the current edge node includes: computing power resource information of the edge node and computing power load status information of the edge node; Determine the business request processing capability information of the service instance in the edge node according to the computing power resource information of the edge node and the computing business request model, and then construct a real-time computing power resource representation matrix according to the business request processing capability information; Obtaining business processing time information of the service instance in the edge node, and determining the computing load information of the service instance in the edge node in combination with the computing load status information of the edge node, and then constructing a real-time computing load representation matrix according to the computing load information; A real-time perception model of computing power resources and computing power load is constructed based on the real-time computing power resource characterization matrix and the real-time computing power load characterization matrix.
5. The optimization decision method for edge service cache of power distribution and utilization business according to claim 1, characterized in that: The method of constructing a power distribution and utilization business edge service cache optimization model based on the computing business request model and the computing power resource and computing power load real-time perception model in combination with the business request prediction information specifically includes: Obtaining a service instance termination decision vector and a new service instance decision vector of the cloud-edge system; According to the service instance termination decision vector and the new service instance decision vector of the cloud edge system, combined with the business request prediction information, the computing business request model and the computing power resource and computing power load real-time perception model, the power distribution business edge service cache optimization objective function is constructed; According to the power distribution and utilization business edge service cache optimization objective function, a power distribution and utilization business edge service cache optimization model is constructed in combination with the computing business request model and the computing power resource and computing power load real-time perception model.
6. The optimization decision method for edge service cache of power distribution and utilization business according to claim 5, characterized in that: The method of constructing a power distribution and utilization business edge service cache optimization objective function based on the service instance termination decision vector and the new service instance decision vector of the cloud edge system, combined with the business request prediction information, the computing business request model, and the computing power resource and computing power load real-time perception model, specifically includes: The service request prediction information includes: a predicted amount of service requests of edge nodes; Obtaining the current number of service requests of the edge node, and determining the predicted value of the computing load of the edge node in combination with the predicted amount of service requests of the edge node and the computing service request model; According to the computing load prediction value of the edge node, an imminent computing load representation matrix is constructed in combination with the computing resources and the computing load real-time perception model; Determine the computing resource change value of the edge node according to the service instance termination decision vector and the new service instance decision vector of the cloud-edge system in combination with the computing business request model; According to the computing power resource change value of the edge node, the imminent computing power resource representation matrix is constructed in combination with the computing power resource and the computing power load real-time perception model; According to the imminent computing power load characterization matrix and the imminent computing power resource characterization matrix, combined with a preset similarity characterization algorithm, a power distribution and utilization business edge service cache optimization objective function is constructed.
7. The optimization decision method for edge service cache of power distribution and utilization business according to claim 6, characterized in that: The constructing of the power distribution and utilization business edge service cache optimization objective function according to the imminent computing power load characterization matrix and the imminent computing power resource characterization matrix in combination with a preset similarity characterization algorithm specifically includes: Determine a predicted value of the service request processing capacity of the edge node according to the computing power resource change value of the edge node; According to the predicted value of the service request processing capability of the edge node, the imminent computing power resource representation matrix is converted to obtain an augmented imminent computing power resource representation matrix; Determine a service request computing load prediction value of the edge node according to the computing load prediction value of the edge node; According to the service request computing load prediction value of the edge node, the approximate computing load representation matrix is converted to obtain an augmented approximate computing load representation matrix; Calculating the matrix similarity of the augmented approximate computing power resource representation matrix and the augmented approximate computing power load representation matrix according to a preset similarity representation algorithm; The maximization of the matrix similarity is taken as the optimization goal, and the power distribution business edge service cache optimization objective function is obtained.
8. The method for optimizing decision-making of edge service cache for power distribution and utilization business according to claim 5, characterized in that: The power distribution and utilization business edge service cache optimization model specifically includes: Among them, θ is the service instance termination decision vector, is the decision vector for adding a new service instance, E is the number of edge nodes, S is the number of computing services; for It is used to represent the edge nodes E perceived by the cloud nodes in the cloud-edge system. e In the next service cache cycle, the service s i The processing power provided; It is used to represent the edge nodes E perceived by the cloud nodes in the cloud-edge system. e The business category s undertaken by the current service instance i The computing load of C e is the edge node E e The collection of current service instances; e,k is the edge node E e Upload service instance c e,k The computing power resources of the edge node; r0 is the occupied resources of the cloud-edge collaborative service instance in the edge node; To represent the edge node E e New corresponding to s i The number of service instances; r si For business i The amount of computing resources required; θ e,k A 0-1 variable indicating whether to terminate service instance c e,k , such as deciding to terminate service instance c e,k Then θ e,k The value is 1, otherwise it is 0; R e is the edge node E e The computing resources on board; u(c e,k ) is the service instance c e,k status.
9. The method for optimizing decision-making of edge service cache for power distribution and utilization business according to claim 1, characterized in that: The optimizing the edge service cache of the power distribution and utilization business according to the optimization decision scheme specifically includes: According to the optimization decision method, the decision variable value of the edge node is obtained; terminating the service of the corresponding service instance in the edge node according to the decision variable value of the edge node, and releasing computing power resource space; A new service instance is generated according to the decision variable value of the edge node, and the released computing resource space is allocated to the new service instance to optimize the edge service cache of the power distribution and utilization business.
10. An optimization decision device for edge service cache of power distribution and utilization business, characterized in that: include: Computing business request model and computing power resource and computing power load real-time perception model construction module, business request prediction information acquisition module, power distribution business edge service cache optimization model construction module, optimization decision plan acquisition module and power distribution business edge service cache optimization module; Among them, the computing service request model and computing power resource and computing power load real-time perception model construction module is used to obtain the computing power information and service request information of the current edge node, and construct the computing service request model and the computing power resource and computing power load real-time perception model; The business request prediction information acquisition module is used to acquire historical operation information of the cloud-edge system to obtain business request prediction information of the cloud-edge system; The power distribution and utilization business edge service cache optimization model construction module is used to construct a power distribution and utilization business edge service cache optimization model based on the computing business request model and the computing power resource and computing power load real-time perception model, combined with the business request prediction information; The optimization decision solution acquisition module is used to solve the power distribution and utilization business edge service cache optimization model to obtain an optimization decision solution; The power distribution and utilization business edge service cache optimization module is used to optimize the power distribution and utilization business edge service cache according to the optimization decision scheme.
