Resource scheduling method and system for edge calculation of power distribution network

Abstract

The invention discloses a power distribution network edge calculation-oriented resource scheduling method and system, and the method comprises the steps: obtaining random features corresponding to each power business scene based on the obtained historical wind power data, historical photovoltaic data and historical load data corresponding to each power business scene, obtaining a resource demand corresponding to each power business scene based on the random feature; taking the resource utilization rate as a spring length, determining an elastic coefficient based on a service flexibility value and a decision risk value, and constructing a spring model corresponding to each power service scene; and determining a resource adjustment value corresponding to each power business scene based on the resource demand and the spring model, and performing resource scheduling on each power business scene based on the resource adjustment value. The method can adapt to resource requirements of different power businesses, improves the resource scheduling capability in a complex environment, and provides support for elastic resource management of the power distribution network edge computing nodes.

Description

Technical Field

[0001] The present application relates to the field of resource scheduling, and in particular to a resource scheduling method and system for edge computing of distribution networks. Background Art

[0002] Driven by policy guidance and technological progress, the distribution network is undergoing an unprecedented transformation. Specifically, on the power generation side, distributed power generation of renewable energy such as photovoltaics has achieved explosive growth in the distribution network. This trend has significantly aggravated the uncertainty of current flow and the complexity of dynamic characteristics. On the power consumption side, electric vehicles, as representatives of new loads, have brought a large number of random charging behaviors due to their large-scale access to the power grid. The fluctuation characteristics of these new power sources and loads are superimposed on each other, which makes the actual physical carrier of the task, the edge computing node, have to face a more drastic time variability, volatility and uncertainty environment. Therefore, in this context, how can edge computing nodes explore and implement optimal strategies under limited computing resources and capacity constraints to ensure that various power business needs can be met in a timely and effective manner.

[0003] Existing edge computing nodes mostly use a fixed method to configure power services and resources, which fails to fully adapt to the resource requirements of different power services and makes it difficult to truly reflect the correlation between power task resources and distribution network systems, resulting in insufficient resource scheduling capabilities in complex time-varying environments.

[0004] Application Contents

[0005] The present application provides a resource scheduling method and system for edge computing of distribution networks to adapt to the resource requirements of different power businesses, improve resource scheduling capabilities in complex environments, and provide support for resource elasticity management of edge computing nodes in distribution networks.

[0006] In a first aspect, the present application provides a resource scheduling method for edge computing of a distribution network, comprising:

[0007] Based on the historical wind power data, historical photovoltaic data and historical load data corresponding to each power business scenario, a random feature corresponding to each power business scenario is obtained, and based on the random feature, a resource demand corresponding to each power business scenario is obtained, wherein the random feature includes a wind power prediction error, a photovoltaic prediction error and a load uncertainty;

[0008] Taking the resource utilization rate as the spring length, determining the elasticity coefficient based on the business flexibility value and the decision risk value, and constructing a spring model corresponding to each power business scenario, wherein the business flexibility value is the flexibility value of the edge computing node in resource adjustment, and the decision risk value is the risk value caused by the uncertainty of the power business scenario;

[0009] Based on the resource demand and the spring model, a resource adjustment value corresponding to each power business scenario is determined, and resources are scheduled for each power business scenario based on the resource adjustment value.

[0010] The embodiment of the present application determines the corresponding resource demand based on the random characteristics corresponding to each power business scenario, and can accurately determine the resource demand corresponding to each power business scenario, so as to facilitate the subsequent resource scheduling of each power business scenario based on the resource demand; by taking the resource utilization rate as the spring length, determining the elasticity coefficient based on the business flexibility value and the decision risk value, and constructing a spring model corresponding to each power business scenario, it is possible to construct a spring model that adapts to different resource scheduling requirements under the source and load fluctuation scenario of the distribution network, and provides support for the resource elasticity management of the edge computing nodes of the distribution network; the resource adjustment value corresponding to each power business scenario can be accurately determined through the spring model, and the resource scheduling of each power business scenario can be performed based on the resource adjustment value, which can adapt to the resource requirements of different power businesses and improve the resource scheduling capability in complex environments.

[0011] Furthermore, the resource requirements corresponding to each power business scenario are obtained based on the random features, specifically:

[0012] Constructing a corresponding probability distribution function based on the random feature;

[0013] Obtaining source load data corresponding to the random feature through the probability distribution function;

[0014] By clustering the source-load data, the resource demand corresponding to each power business scenario is obtained.

[0015] In this way, the corresponding probability distribution function is determined by obtaining the random features corresponding to each power business scenario, and then the corresponding resource demand is accurately determined through clustering. The resource demand corresponding to each power business scenario can be accurately determined, which facilitates the subsequent resource scheduling of each power business scenario based on resource demand.

[0016] Furthermore, the formula of the probability distribution function is specifically:

[0017]

[0018]

[0019] In the formula, F(x w ∣x pv ,x r ) is the joint probability distribution function, F(x w ∣x pv ), F(x r ∣x pv ), F(x r∣x pv ) are the distribution functions corresponding to the corresponding random features, is the Copula function corresponding to the corresponding random feature, u1,u2,…,u d is a random number, is the symbol for the numerical partial derivative.

[0020] Furthermore, by clustering the source-load data, the resource requirements corresponding to each power business scenario are obtained, specifically:

[0021] Based on the source-load data, determining the initial resource requirements corresponding to each power business scenario;

[0022] Establishing a similarity matrix based on the similarity between any two of the initial resource requirements, and constructing a corresponding diagonal matrix based on the similarity matrix;

[0023] Based on the diagonal matrix and the similarity matrix, a Laplace matrix is ​​obtained;

[0024] Extracting eigenvalues ​​from the Laplace matrix and forming a eigenmatrix based on eigenvectors corresponding to the eigenvalues;

[0025] By performing k-means clustering on the feature matrix, the resource requirements corresponding to each power business scenario are obtained.

[0026] In this way, by clustering the source-load data, the resource requirements corresponding to each power business scenario can be accurately determined, which facilitates the subsequent resource scheduling of each power business scenario based on resource requirements.

[0027] Furthermore, the calculation formula for obtaining the resource requirements corresponding to each power business scenario is specifically:

[0028] V opt = argmax s(V g ,V m )

[0029]

[0030] Where V opt is the resource requirement, sV g ,V ν / is the similarity between the resource demand subsets, V g 、V m are the resource requirement subsets of the gth and mth partitions, respectively, P(V opt ) is the probability density function, x opt is the resource requirement V opt The corresponding subset, F(x opt ) is xopt The joint probability distribution of is the symbol for the numerical partial derivative.

[0031] Furthermore, the evaluation formula of the business flexibility value is specifically:

[0032] P i =1 / ceil(k v V i +k c C i -0.5);

[0033] Where P i is the service flexibility value; ceil(·) means rounding up; k v , k c Represent the weights of task value and running time respectively, satisfying k v +k c =1; V i is the task value; C i is the time value, where C i =P max ×T i / (T i -t), P max is the maximum value of static flexibility, T i represents the adjusted deadline time of task i, and t is the current time.

[0034] Furthermore, the resource adjustment value corresponding to each power business scenario is determined based on the resource demand and the spring model, specifically:

[0035]

[0036] In the formula, x i (t) is the deformation of the i-th spring at time t, indicating the resource adjustment value; L total Represents the total resource utilization of edge computing nodes; L i (t) is the original length of the i-th spring at time t, representing the resource demand; is the elastic coefficient of the i-th spring at time t, where the resource utilization rate L total The calculation formula is:

[0037]

[0038] Where: R i (t) is the amount of resources allocated to power business scenario i at time t; is the average resource utilization rate; N is the number of power business scenarios; T is the total time period for resource adjustment; R(t) is the resource utilization rate of the edge computing node at time t; Si is the total resource demand of power business i; S i,min (t) and S i,max (t) are the minimum and maximum occupied resources of power business i at time t; w i P is the resource allocation weight for power business scenario i; i is the business flexibility value.

[0039] Furthermore, the spring model includes: an objective function and constraints, and the spring model with optimal resource utilization is established according to the objective function and the constraints, specifically:

[0040] The objective function is expressed as:

[0041]

[0042] Where L(t) is the original length of the composite spring; ΔL(t) is the deformation of the composite spring at time t, which represents the resource adjustment value; is the average spring length, representing resource demand; p is a symbolic variable, when the spring is extended, p = 1, and when it is compressed, p = -1; T is the total time period for the composite spring adjustment; φ(t) is the elastic coefficient of the spring under the node at time t; ΔR(t) is the resource utilization rate of the edge computing node at time t; C is a constant; S min (t) is the minimum resource utilization of the edge computing node at time t; S max (t) is the maximum resource utilization of the edge computing node at time t.

[0043] In a second aspect, the present application provides a resource scheduling system for edge computing of a distribution network, including: a resource scheduling system for edge computing of a distribution network, characterized in that it includes: an acquisition module, a construction module and a scheduling module;

[0044] The acquisition module is used to obtain random features corresponding to each power business scenario based on the acquired historical wind power data, historical photovoltaic data and historical load data corresponding to each power business scenario, and obtain resource requirements corresponding to each power business scenario based on the random features, wherein the random features include wind power prediction error, photovoltaic prediction error and load uncertainty;

[0045] The construction module is used to use the resource utilization rate as the spring length, determine the elasticity coefficient based on the business flexibility value and the decision risk value, and construct a spring model corresponding to each power business scenario, wherein the business flexibility value is the flexibility value of the edge computing node in resource adjustment, and the decision risk value is the risk value caused by the uncertainty of the power business scenario;

[0046] The scheduling module is used to determine the resource adjustment value corresponding to each power business scenario based on the resource demand and the spring model, and perform resource scheduling for each power business scenario based on the resource adjustment value.

[0047] The embodiment of the present application determines the corresponding resource demand based on the random characteristics corresponding to each power business scenario, and can accurately determine the resource demand corresponding to each power business scenario, so as to facilitate the subsequent resource scheduling of each power business scenario based on the resource demand; by taking the resource utilization rate as the spring length, determining the elasticity coefficient based on the business flexibility value and the decision risk value, and constructing a spring model corresponding to each power business scenario, it is possible to construct a spring model that adapts to different resource scheduling requirements under the source and load fluctuation scenario of the distribution network, and provides support for the resource elasticity management of the edge computing nodes of the distribution network; the resource adjustment value corresponding to each power business scenario can be accurately determined through the spring model, and the resource scheduling of each power business scenario can be performed based on the resource adjustment value, which can adapt to the resource requirements of different power businesses and improve the resource scheduling capability in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a flow chart of an embodiment of a resource scheduling method for edge computing of a distribution network provided by the present application;

[0049] Figure 2 This is a schematic diagram of the adjustment area of ​​edge computing resources occupied by the power business scenario provided in this application;

[0050] Figure 3 It is a flow chart of an embodiment of a resource scheduling system for distribution network edge computing provided by the present application. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0052] It should be understood that the step numbers used in this article are only for the convenience of description and are not intended to limit the order in which the steps are executed.

[0053] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an" and "the" are intended to include plural forms unless the context clearly indicates otherwise.

[0054] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0055] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0056] The distribution network is undergoing an unprecedented transformation. Specifically, on the power generation side, the surge in renewable energy such as photovoltaics has exacerbated the uncertainty of current flow and the complexity of dynamic characteristics. On the power consumption side, the large-scale access of electric vehicles has brought a large number of random charging behaviors, so edge computing nodes face more intense challenges of time, fluctuations and uncertainty. In this context, edge computing nodes need to explore the optimal strategy under limited resources, and meeting the needs of power business has become a technical difficulty. However, existing technologies use a fixed method to configure power business and resources, which cannot adapt to power business needs and have insufficient resource scheduling capabilities.

[0057] Next, the nouns involved in this application are analyzed:

[0058] Clustering algorithms are core technologies in data mining and pattern recognition. They group objects in a data set so that objects in the same group have a high degree of similarity, while objects in different groups have a low degree of similarity. These algorithms can be widely used in image processing, market analysis, bioinformatics and other fields. Through clustering, people can understand the structure and characteristics of data, discover patterns and laws hidden in the data, and thus conduct more effective data analysis and decision-making. Common clustering algorithms include K-means clustering, hierarchical clustering and DBSCAN.

[0059] Based on this, the embodiments of the present application provide a resource scheduling method and system for distribution network edge computing, which can adapt to the resource requirements of different power services, improve the resource scheduling capabilities in complex environments, and provide support for resource elasticity management of distribution network edge computing nodes.

[0060] An embodiment of the present application provides a resource scheduling method for edge computing of a distribution network, which is specifically illustrated by the following embodiments. First, the resource scheduling method for edge computing of a distribution network in an embodiment of the present application is described.

[0061] The resource scheduling method for edge computing of distribution network provided in the embodiment of the present application relates to the field of resource scheduling. The resource scheduling method for edge computing of distribution network provided in the embodiment of the present application can be applied to the terminal, can also be applied to the server side, and can also be software running in the terminal or the server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can realize the application of a component selection recommendation method, etc., but is not limited to the above forms.

[0062] The present application can also be used in numerous general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0063] Embodiment 1

[0064] Please refer to Figure 1 , Figure 1 It is a flow chart of an embodiment of a resource scheduling method for edge computing of a distribution network provided by the present application, including steps S101 to S103;

[0065] Step S101: obtaining random features corresponding to each power business scenario based on the acquired historical wind speed data, historical photovoltaic data, and historical load data corresponding to each power business scenario, and obtaining resource requirements corresponding to each power business scenario based on the random features, wherein the random features include wind power prediction error, photovoltaic prediction error, and load uncertainty;

[0066] It is understandable that the historical wind power data, historical photovoltaic data and historical load data corresponding to each power business scenario can be obtained through existing technologies. Specifically, wind speed statistics can be obtained by, but not limited to, accessing energy meteorological data platforms, government research institutions, wind power companies and industry associations; photovoltaic power generation data can be obtained by, but not limited to, accessing energy meteorological data platforms, government websites, public data platforms, photovoltaic companies and research institutions; and historical power load data can be collected through, but not limited to, load-side management platforms. After obtaining the historical wind speed data, historical photovoltaic data and historical load data corresponding to each power business scenario, by analyzing the historical wind speed data, historical photovoltaic data and historical load data, the random characteristics of the wind power prediction error, photovoltaic prediction error and load uncertainty in the collection time period can be determined. It should be noted that the means of data analysis are not the focus of this application, so it will not be repeated here.

[0067] After obtaining the random features corresponding to each power business scenario, it is necessary to obtain the resource requirements corresponding to each power business scenario based on the random features. Specifically, first, the corresponding probability distribution function is constructed based on the random features, that is, considering the impact of the fluctuation of demand-side load and new energy on the uncertainty of power business, and the scenario is constructed through the probability density functions of wind power prediction error, photovoltaic prediction error and load uncertainty, and the correlation modeling of high-dimensional data is performed. Finally, the joint probability distribution function of source-load scenarios considering the correlation of multi-dimensional variables is obtained, wherein the probability density functions are all functions that obey their own normal distributions, and the relevant formula is:

[0068]

[0069] In the formula, F(x w ) is the probability density function of wind power prediction error, x w is the wind power prediction error, μ w and σ w are the mean and standard deviation of wind power forecast error, F(x pv ) is the probability density function of photovoltaic prediction error, x pv is the photovoltaic prediction error, μ pv and σ pv are the mean and standard deviation of photovoltaic prediction errors, respectively, and F(x r ) is the probability density function of the load uncertainty, x r is the photovoltaic prediction error, μ r and σ r are the mean and standard deviation of the uncertainty in demand side load, respectively.

[0070] It should be noted that the probability distribution function used in this application can be but is not limited to the vine structure Copula function, wherein the specific construction method of the vine structure Copula function is: (1) using the Kendall rank correlation coefficient to calculate the correlation coefficient between the three random features of all nodes in the first layer (i.e., wind power prediction error, photovoltaic prediction error and load uncertainty), and determine the selection of the node with the largest average correlation coefficient The node of is taken as the root node, and the tree structure is of D-vine structure type, that is, starting from the edge with the largest correlation coefficient, connecting sequentially to establish a sequential tree structure. The calculation formula of the correlation coefficient γ is: In the formula, γ is the correlation coefficient between two random features, N c is the number of identical pairs between two random features, N d is the number of different order pairs between two random features, n is the total number of random features, γ ij is feature x i and x j The Kendall rank correlation coefficient between |γ| is calculated. The closer the value of |γ| is to 1, the stronger the correlation between random features is. The closer the value of |γ| is to 0, the weaker the correlation between random features is. (2) For each edge of the first-layer tree, the parameter value of the binary Copula function is estimated by maximizing the log-likelihood function, and the Akaike information criteria (AIC) is used to assist in determining the optimal binary Copula function. The calculation formula of the maximum likelihood estimation method is: N AIC =-2ln(L)+2k; where N AIC is the maximum likelihood estimate, L is the maximum likelihood estimate function value of the Copula function, k is the number of parameters of the model, where N AIC The smaller the value, the better the model fit. AIC The smallest Copula function is taken as the optimal Copula function. (3) Use the Copula function obtained in step (2) to sample and obtain the corresponding node value u of the second layer tree j , repeat steps 2 and 3 to determine the second-level tree structure and Copula function, where the sampling formula is specifically: In the formula, u j is the corresponding node value, is the conditional inverse function of the Copula function, v i is a random number drawn from the unit interval (0,1), u iis the root node of the previous layer. (4) Repeat steps (1), (2) and (3) until all tree structures and Copula functions under the vine structure are obtained, and the vine Copula function model is obtained. (5) Calculate the CvM) distance of each Copula function to evaluate the fitting effect of each model and determine the optimal vine structure Copula model. The relevant formula for calculating the CVM distance is: In the formula, C n (u) is the constructed Copula function, C e (u) is the empirical Copula function, and the CvM distance is the distance between the target Copula function and the empirical Copula function. The smaller the distance is, the higher the fitting accuracy of the model is.

[0071] It should be noted that the vine structure Copula function is a high-dimensional probability distribution modeling method that can well describe the correlation between variables.

[0072] It should be noted that CVM refers to the Cramér-Von Mises statistic, which is a statistical method used to test whether the data obeys a specific distribution (such as normal distribution) or to evaluate the goodness of model fit.

[0073] Secondly, the source load data corresponding to the random feature is obtained through the probability distribution function; specifically, the random numbers u1, u2, ..., u1 are independently and uniformly distributed in the range of [0, 1] based on the Monte Carlo method. d , a vector composed of random numbers u=[u1,u2,…,u d ], and then based on The source-charge data considering the correlation can be obtained. Repeat M times to obtain the source-charge data x corresponding to the M groups of random features. 1,y ,x 2,y ,…,x d,y ,y=1,2,…,M, where the formula of the probability distribution function is specifically:

[0074]

[0075] In the formula, F(x w ∣x pv ,x r ) is the joint probability distribution function, F(x w ∣x pv ), F(x r ∣x pv ), F(x r ∣x pv ) are the distribution functions corresponding to the corresponding random features, is the Copula function corresponding to the corresponding random feature, u1,u2,…,ud is a random number, is the symbol for the numerical partial derivative.

[0076] Finally, by clustering the source-load data, the resource requirements corresponding to each power business scenario are obtained; specifically, based on the source-load data, the initial resource requirements corresponding to each power business scenario are determined; based on the similarity between any two of the initial resource requirements, a similarity matrix S is established. D , and construct the corresponding diagonal matrix S based on the similarity matrix node Specifically, for the M partition subsets V1, V2, ..., V M , define the corresponding resource demand membership matrices as H1, H2, …, H M , the element H of the mth matrix μ,ν,m Used to describe whether the resource demand μ in each partition belongs to partition U ν (ν=1,2,…,I): In order to obtain the maximum similarity of subset partitioning, we need to select a method to partition two nodes into the same subset. The ratio of two nodes being partitioned into the same subset in all subset partitioning is defined as the node similarity matrix S of the node pair. node , where the similarity matrix can be obtained by Solve, where H = [H1, H2, …, H M ], which is the cascade matrix of all node membership matrices in the target area; based on the diagonal matrix S node and the similarity matrix S D , that is, L = S D -S node , obtain the Laplace matrix L; extract the eigenvalues ​​in the Laplace matrix L, and form a feature matrix A based on the eigenvectors corresponding to the eigenvalues; perform k-means clustering on the feature matrix A to divide each resource demand into its own partition, and then obtain the resource demand V corresponding to each power business scenario opt , wherein the resource demand V corresponding to each power business scenario is obtained opt The calculation formula is as follows:

[0077] V opt = argmax s(V g ,V m )

[0078]

[0079] Where V opt is the resource requirement, sV g ,V ν / is the similarity between the resource demand subsets, Vg 、V m are the resource demand subsets of the gth and mth partitions, respectively, P(V opt ) is the probability density function, x opt is the resource requirement V opt The corresponding subset, F(x opt ) is x opt The joint probability distribution of is the symbol for the numerical partial derivative.

[0080] In this way, the corresponding probability distribution function is determined by obtaining the random features corresponding to each power business scenario, and then the corresponding resource demand is accurately determined through clustering. The resource demand corresponding to each power business scenario can be accurately determined, which facilitates the subsequent resource scheduling of each power business scenario based on resource demand.

[0081] Step S102: Taking the resource utilization rate as the spring length, determining the elasticity coefficient based on the business flexibility value and the decision risk value, and constructing a spring model corresponding to each power business scenario, wherein the business flexibility value is the flexibility value of the edge computing node in resource adjustment, and the decision risk value is the risk value caused by the uncertainty of the power business scenario;

[0082] It is understandable that after obtaining the resource demand, it is necessary to first calculate the resource utilization rate of the power business scenario, and use the resource utilization rate of the power business scenario as the spring length, and together with the subsequent elastic coefficient, construct the spring model corresponding to each power business scenario to schedule resources. Specifically, after the power business enters the edge computing node processing queue, it can use the controller to adjust its occupied resources, and participate in the overall resource utilization optimization under the premise of ensuring the safe operation of the power business. Among them, the power business scenario occupies the edge computing resource adjustment area, such as Figure 2 As shown in the figure: t i,a S is the starting time when the power service i enters the edge computing node processing queue; i is the total resource demand of power business i, t i,b is the fastest completion time of power business i; t i,d is the latest start time for power business i to ensure the completion of demand; t i,c is the cut-off time of power business i, the blue area abcda is the maximum edge computing resource adjustment area for power business i, where the polyline abc is the upper boundary of the maximum adjustment area, indicating the fastest adjustment process of edge computing resources; the polyline adc is the lower boundary of the maximum adjustment area, indicating the slowest adjustment process of edge computing resources. The corresponding mathematical expression is: Where P i,c Adjust the response rate of edge computing resources for power business i; Ei is the edge computing resource capacity for power business i. Through the above method, the resource adjustment area of ​​all power businesses can be obtained, and the dynamic resource adjustment of all individuals in the edge computing node can be realized, which is physically similar to an equivalent machine. In order to make the resource utilization rate of each time period as even as possible, the objective function of resource adjustment is defined as minimizing the variance of resource utilization rate of each time period, and then determining the resource utilization rate, where the calculation formula of resource utilization rate R(t) is:

[0083]

[0084] In the formula, R i (t) is the amount of resources allocated to power business scenario i at time t, satisfying the resource adjustment area constraint; is the average resource utilization rate; N is the number of power business scenarios; T is the total time period for resource adjustment; R(t) is the resource utilization rate of the edge computing node at time t; S i is the total resource demand of power business i; S i,min (t) and S i,max (t) are the minimum and maximum occupied resources of power business i at time t; w i Assign weights to resources in power business scenario i. The business with lower flexibility is more likely to be allocated resources. i is the business flexibility value.

[0085] It is understandable that the flexibility value of the edge computing node in resource adjustment needs to be evaluated first to determine the elasticity coefficient of the spring model. The evaluation formula of the business flexibility value is specifically:

[0086] P i =1 / ceil(k v V i +k c C i -0.5);

[0087] Where P i is the service flexibility value; ceil(·) means rounding up; k v , k c Represent the weights of task value and running time respectively, satisfying k v +k c =1; V i is the task value; C i is the time value, indicating the distance to the deadline, where C i =P max ×T i / (T i -t), P max is the maximum value of static flexibility, Ti represents the adjusted deadline time of task i, and t is the current time.

[0088] It should be noted that the task value V i Set to the static flexibility set when initializing the task, which means that the greater the task value, the lower the flexibility, and the time value C i It is related to the current running time t and the task deadline adjustment time. Similarly, the shorter the deadline, the lower the task flexibility. When t is close to T i When C i The larger the value, the P calculated in the formula i The smaller it is, the lower the flexibility; when t exceeds T i , the task does not participate in the resource adjustment response, ensuring that the current task can be executed and completed within the delay; further, k v , k c These two static parameters determine the flexibility of the task; i The static flexibility value is adopted. Therefore, the change of flexibility value in the formula is mainly determined by C i Determine, and the coefficient k c Then the flexibility adjustment range will be controlled. Obviously, when k c = 0, the flexibility obtained at this time is the static flexibility set by the user; and when k c As it gradually increases, the task flexibility is gradually determined by the earliest critical moment of operation. Therefore, k v , k c The two weight parameters need to be set according to the actual application.

[0089] After evaluating the business flexibility value of the edge computing node, it is also necessary to evaluate the benefit risk value caused by the uncertainty of the power business scenario through CvaR as a basis for resource adjustment strategy. Assume that ω(x, y) represents the function related to the decision vector x, y represents the random characteristics of the uncertainty factor of the power business resource demand and its probability density function is p(y) = P(V opt ), then the risk B with confidence level α CVaR It is expressed as: Among them, B VaR is the VaR value with the corresponding confidence level α; E{·} represents the expected value. To facilitate direct solution, the present invention uses the function F α (x,ε) for B CVaR To express: In the formula, ε is the calculation B CVaR The auxiliary variable y is the set of random features y; there is an integral operation in the formula, and the integral term can be discretized to obtain the expected value, that is, In the formula, p n(y) is the probability of the nth random feature y appearing; N y is the total number of samples of the discretized random feature y.

[0090] After the business flexibility value and decision risk value are determined, the elasticity coefficient can be comprehensively determined based on the business flexibility value and the decision risk value, and the resource utilization rate is used as the spring length to construct a single-time point spring model corresponding to each power business scenario; the resource utilization curve R(t) under the normal operation of a power business is known. For a single time node, the original length of the spring is the resource utilization rate of the business at the current moment, and the processing time of the entire business can be regarded as a combination of multiple different springs; Where L(t) is the original length of the spring at the node at time t, φ(t) is the elastic coefficient of the spring at the node at time t, β is the decision maker's preference for risk, the larger the β, the higher the decision maker's risk aversion, β = 0 means risk neutrality, P i is the business flexibility value, B VaR is the VaR value with the corresponding confidence level of α, i.e., the risk value. Since there is an error between the actual scenario probability distribution and the initial scenario probability obtained from historical data, this application also needs to construct a probability scenario fuzzy set Ω of power business resource demand based on the 1-norm and ∞-norm n To constrain the error range of the spring length parameter probability L(t) in each typical scenario, In the formula, Ω n is the probability scene fuzzy set, p s is the probability of a typical scenario s; is the initial probability of scene s; and Respectively represent p s for The positive offset state and negative offset state of and Respectively represent p s for The positive and negative offsets are all positive values; θ1 and θ are the maximum deviation values ​​of the typical scene probability under the 1-norm and ∞-norm restrictions, respectively.

[0091] It is understandable that there will be resource competition or cooperation between power business scenarios. Therefore, the single-time point spring model cannot obtain accurate resource scheduling values. Therefore, the composite spring model can be used to accurately simulate the resource scheduling process of the actual power business scenario. Different spring elastic coefficients can express the intensity of competition or cooperation. The composite spring can generate any required nonlinear elastic deformation characteristics, thereby expressing the dynamic adjustment of the power business to its own resource acquisition strategy at different time points. The composite spring is composed of multiple springs, in which each spring itself is horizontally connected without considering the weight, one end is connected to the wall, and the other end moves as far as possible along the direction of the spring to the resource boundary under the action of external force, which means that each power business scenario itself tries to obtain the maximum resources. Each power business scenario can decide whether to compete for resources with other power business scenarios based on its own flexibility. Under the action of the combined force, each spring reaches a state of equilibrium under the current edge computing node resource allocation. Then, each spring further adjusts their resource allocation according to the completion of the business scenario until the power business scenario is completed.

[0092] Step S103: determining a resource adjustment value corresponding to each power business scenario based on the resource demand and the spring model, and performing resource scheduling for each power business scenario based on the resource adjustment value.

[0093] It is understandable that the resource adjustment value corresponding to each power business scenario can be determined according to the resource demand and the mathematical model of the spring model, wherein the mathematical model of the spring model is shown in the following formula:

[0094]

[0095] In the formula, x i (t) is the deformation of the i-th spring at time t, indicating the resource adjustment value; L total Represents the total resource utilization of edge computing nodes; L i (t) is the original length of the i-th spring at time t, representing resource utilization; is the elastic constant of the i-th spring at time t.

[0096] It is understandable that during scheduling, it is also necessary to control the time of resource scheduling in the power business scenario. The elastic potential energy can be used to characterize the delay constraint of the power business scenario. Since a power business scenario needs to obtain certain resources within the delay to ensure completion, the corresponding spring model can represent that the elastic potential energy of the composite spring composed of multiple different springs is certain. Set the initial value to 0. For example, the elastic potential energy of the spring at a certain time node increases, and the spring at subsequent time nodes needs to be compressed to reduce the corresponding elastic potential energy. The elastic potential energy of the composite spring of the present application can be a negative value. When the spring at a certain time node is stretched, it means that the resource utilization rate after adjustment is improved, and the elastic potential energy is positive. When the spring at a certain time node is compressed, it means that the resource utilization rate after adjustment is reduced, and the elastic potential energy is negative. When the accumulated elastic potential energy reaches a certain negative value, it means that it no longer participates in resource adjustment, and the spring performs potential energy compensation to ensure business completion. In order to make the resource utilization rate of each period as even as possible, the objective function of the composite spring deformation is defined as minimizing the variance of the total length of the spring in each period, wherein the spring model includes: an objective function and constraints, and the spring model with the optimal resource utilization rate is established according to the objective function and the constraints. The calculation process of the objective function is:

[0097]

[0098] Where L(t) is the original length of the composite spring; ΔL(t) is the deformation of the composite spring at time t, which represents the resource adjustment value; is the average spring length; p is a symbolic variable, when the spring is extended, p = 1, and when it is compressed, p = -1; T is the total time period for the composite spring adjustment; φ(t) is the elastic coefficient of the spring under the node at time t; ΔR(t) is the resource utilization rate of the edge computing node at time t; C is a constant; S min (t) is the minimum resource utilization of the edge computing node at time t; S max (t) is the maximum resource utilization of the edge computing node at time t.

[0099] The embodiment of the present application determines the corresponding resource demand based on the random characteristics corresponding to each power business scenario, and can accurately determine the resource demand corresponding to each power business scenario, so as to facilitate the subsequent resource scheduling of each power business scenario based on the resource demand; by taking the resource utilization rate as the spring length, determining the elasticity coefficient based on the business flexibility value and the decision risk value, and constructing a spring model corresponding to each power business scenario, it is possible to construct a spring model that adapts to different resource scheduling requirements under the source and load fluctuation scenario of the distribution network, and provides support for the resource elasticity management of the edge computing nodes of the distribution network; the resource adjustment value corresponding to each power business scenario can be accurately determined through the spring model, and the resource scheduling of each power business scenario can be performed based on the resource adjustment value, which can adapt to the resource requirements of different power businesses and improve the resource scheduling capability in complex environments.

[0100] Embodiment 2

[0101] Please refer to Figure 3 , Figure 3 It is a structural diagram of an embodiment of a resource scheduling system for edge computing of a distribution network provided by the present application, including: an acquisition module 100, a construction module 200 and a scheduling module 300;

[0102] The acquisition module 100 is used to obtain random features corresponding to each power business scenario based on the historical wind power data, historical photovoltaic data and historical load data corresponding to each power business scenario, and obtain resource requirements corresponding to each power business scenario based on the random features, wherein the random features include wind power prediction error, photovoltaic prediction error and load uncertainty;

[0103] The construction module 200 is used to use the resource utilization rate as the spring length, determine the elasticity coefficient based on the business flexibility value and the decision risk value, and construct a spring model corresponding to each power business scenario, wherein the business flexibility value is the flexibility value of the edge computing node in resource adjustment, and the decision risk value is the risk value caused by the uncertainty of the power business scenario;

[0104] The scheduling module 300 is used to determine the resource adjustment value corresponding to each power business scenario based on the resource demand and the spring model, and perform resource scheduling for each power business scenario based on the resource adjustment value.

[0105] The information interaction, execution process, etc. between the modules in the above-mentioned resource scheduling system for edge computing of distribution network are based on the same concept as the embodiment of the resource scheduling method for edge computing of distribution network in the first aspect of the present invention, and the technical effects achieved are basically the same. For specific contents, please refer to the description in the first embodiment of the method of the present invention, and will not be repeated here.

[0106] The device embodiments described above are merely illustrative, wherein the modules described as separate components may or may not be physically separated, i.e., may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the method of this embodiment.

[0107] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the scope of protection of the present application.

[0108] It is particularly pointed out that for those skilled in the art, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A resource scheduling method for edge computing of distribution network, characterized in that: include: Based on the historical wind power data, historical photovoltaic data and historical load data corresponding to each power business scenario, a random feature corresponding to each power business scenario is obtained, and based on the random feature, a resource demand corresponding to each power business scenario is obtained, wherein the random feature includes a wind power prediction error, a photovoltaic prediction error and a load uncertainty; Taking the resource utilization rate as the spring length, determining the elasticity coefficient based on the business flexibility value and the decision risk value, and constructing a spring model corresponding to each power business scenario, wherein the business flexibility value is the flexibility value of the edge computing node in resource adjustment, and the decision risk value is the risk value caused by the uncertainty of the power business scenario; Based on the resource demand and the spring model, a resource adjustment value corresponding to each power business scenario is determined, and resources are scheduled for each power business scenario based on the resource adjustment value.

2. The resource scheduling method for edge computing of distribution network according to claim 1, characterized in that: The resource requirements corresponding to each power business scenario are obtained based on the random features, specifically: Constructing a corresponding probability distribution function based on the random feature; Obtaining source load data corresponding to the random feature through the probability distribution function, wherein the probability distribution function is a Copula function; By clustering the source-load data, the resource demand corresponding to each power business scenario is obtained.

3. The resource scheduling method for edge computing of distribution network according to claim 2, characterized in that: The formula of the probability distribution function is specifically: In the formula, F(x w ∣x pv ,x r ) is the joint probability distribution function, F(x w ∣x pv ), F(x r ∣x pv ), F(x r ∣x pv ) are the distribution functions corresponding to the corresponding random features, is the Copula function corresponding to the corresponding random feature, u1,u2,…,u d is a random number, is the symbol for the numerical partial derivative.

4. The resource scheduling method for edge computing of distribution network according to claim 2, characterized in that: By clustering the source-load data, the resource requirements corresponding to each power business scenario are obtained, specifically: Based on the source-load data, determining the initial resource requirements corresponding to each power business scenario; Establishing a similarity matrix based on the similarity between any two of the initial resource requirements, and constructing a corresponding diagonal matrix based on the similarity matrix; Based on the diagonal matrix and the similarity matrix, a Laplace matrix is ​​obtained; Extracting eigenvalues ​​from the Laplace matrix and forming a eigenmatrix based on eigenvectors corresponding to the eigenvalues; By performing k-means clustering on the feature matrix, the resource requirements corresponding to each power business scenario are obtained.

5. The resource scheduling method for edge computing of distribution network according to claim 4, characterized in that: The calculation formula for obtaining the resource requirements corresponding to each power business scenario is specifically: V opt =argmax s(V g ,V m ) Where V opt is the resource requirement, s(V g ,V ν ) is the similarity between the resource demand subsets, V g 、V m are the resource requirement subsets of the gth and mth partitions, respectively, P(V opt ) is the probability density function, x opt is the resource requirement V opt The corresponding subset, F(x opt ) is x opt The joint probability distribution of is the symbol for the numerical partial derivative.

6. The resource scheduling method for edge computing of distribution network according to claim 1, characterized in that: The evaluation formula of the business flexibility value is specifically: P i =1 / ceil(k v V i +k c C i -0.5); Where P i is the service flexibility value; ceil(·) means rounding up; k v , k c Represent the weights of task value and running time respectively, satisfying k v +k c =1; V i is the task value; C i is the time value, where C i =P max ×T i / (T i -t), P max is the maximum value of static flexibility, T i represents the adjusted deadline time of task i, and t is the current time.

7. The resource scheduling method for edge computing of distribution network according to claim 1, characterized in that: The resource adjustment value corresponding to each power business scenario is determined based on the resource demand and the spring model, specifically: In the formula, x i (t) is the deformation of the i-th spring at time t, indicating the resource adjustment value; L total Represents the total resource utilization of edge computing nodes; L i (t) is the original length of the i-th spring at time t, representing resource utilization; is the elastic coefficient of the i-th spring at time t, where the resource utilization rate L total The calculation formula is: Where: R i (t) is the amount of resources allocated to power business scenario i at time t; is the average resource utilization rate; N is the number of power business scenarios; T is the total time period for resource adjustment; R(t) is the resource utilization rate of the edge computing node at time t; S i is the total resource demand of power business i; S i,min (t) and S i,max (t) are the minimum and maximum occupied resources of power business i at time t; w i P is the resource allocation weight for power business scenario i; i is the business flexibility value.

8. The resource scheduling method for edge computing of distribution network according to claim 1, characterized in that: The spring model includes: an objective function and constraints. The spring model with optimal resource utilization is established according to the objective function and the constraints, specifically: The objective function is expressed as: Where L(t) is the original length of the composite spring; ΔL(t) is the deformation of the composite spring at time t, which represents the resource adjustment value; is the average spring length, representing resource demand; p is a symbolic variable, when the spring is extended, p = 1, and when it is compressed, p = -1; T is the total time period for the composite spring adjustment; φ(t) is the elastic coefficient of the spring under the node at time t; ΔR(t) is the resource utilization rate of the edge computing node at time t; C is a constant; S min (t) is the minimum resource utilization of the edge computing node at time t; S max (t) is the maximum resource utilization of the edge computing node at time t.

9. A resource scheduling system for edge computing of distribution network, characterized in that: include: Get modules, build modules, and schedule modules; The acquisition module is used to obtain random features corresponding to each power business scenario based on the acquired historical wind power data, historical photovoltaic data and historical load data corresponding to each power business scenario, and obtain resource requirements corresponding to each power business scenario based on the random features, wherein the random features include wind power prediction error, photovoltaic prediction error and load uncertainty; The construction module is used to use the resource utilization rate as the spring length, determine the elasticity coefficient based on the business flexibility value and the decision risk value, and construct a spring model corresponding to each power business scenario, wherein the business flexibility value is the flexibility value of the edge computing node in resource adjustment, and the decision risk value is the risk value caused by the uncertainty of the power business scenario; The scheduling module is used to determine the resource adjustment value corresponding to each power business scenario based on the resource demand and the spring model, and perform resource scheduling for each power business scenario based on the resource adjustment value.

Images