Power load scheduling method and device, electronic equipment and storage medium

Abstract

The invention belongs to the technical field of power dispatching, and discloses a power load dispatching method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining the load information of loads in a to-be-tested region, the load information comprises the number of all loads and the operation conditions of all loads in a preset period, and according to the number and the operation conditions, obtaining the load information of the loads in the to-be-tested region; the method comprises the following steps: dividing a load by using a spectral clustering algorithm to obtain a plurality of load groups, constructing a likelihood function based on a Markov random field based on the operation condition of the plurality of load groups so as to obtain the planned operation condition of the plurality of load groups through Kalman filtering and calculation, and performing operation scheduling on the corresponding load groups according to the planned operation condition; through a likelihood function and Kalman filtering based on a Markov random field, load groups obtained through division based on a spectral clustering algorithm are scheduled, and the scheduling efficiency of a power system is improved.

Description

Technical Field

[0001] The present application relates to the technical field of power dispatching, and in particular to a power load dispatching method, device, electronic device and storage medium. Background Art

[0002] In recent years, the continuous growth of electricity demand and the diversification of power loads have brought new challenges to the stable operation and efficient dispatch of power systems. However, traditional power dispatch methods often rely on empirical judgment or simple models, which are difficult to accurately capture the temporal changes and spatial distribution characteristics of loads, leading to problems such as uneven power distribution and insufficient efficiency.

[0003] Therefore, in order to solve the technical problems that the existing power dispatching methods rely on empirical judgment or simple models and are difficult to accurately capture the temporal changes and spatial distribution characteristics of the load, resulting in uneven power distribution and insufficient power dispatching efficiency, there is an urgent need for a power load dispatching method, device, electronic device and storage medium. Summary of the invention

[0004] The purpose of the present application is to provide a method, device, electronic device and storage medium for dispatching electric loads. By using a likelihood function based on a Markov random field and a Kalman filter, the load groups obtained by dividing based on a spectral clustering algorithm are dispatched for operation, so as to solve the problem that the existing electric power dispatching methods rely on empirical judgment or simple models and are difficult to accurately capture the time series changes and spatial distribution characteristics of the loads, resulting in uneven distribution of electric energy and insufficient power dispatching efficiency. By grouping the loads to achieve optimal dispatching of the power system, the power system can be dispatched from the perspective of time and space, reducing the transmission loss of the power system and improving the dispatching efficiency of the power system.

[0005] In a first aspect, the present application provides a method for dispatching an electric load, comprising: Obtaining load information of loads in the area to be tested; the load information includes the number of all loads and the operating conditions of all loads within a preset period; According to the quantity and the operation status, the load is divided by using a spectral clustering algorithm to obtain a plurality of load groups; Based on the operation conditions of the plurality of load groups, a likelihood function based on a Markov random field is constructed to calculate the planned operation conditions of the plurality of load groups through Kalman filtering; According to the planned operation status, the corresponding load groups are scheduled for operation.

[0006] The electric load dispatching method provided in the present application can realize the optimized dispatching of electric loads. Through the likelihood function based on Markov random field and the application of Kalman filtering, the load groups obtained by the division based on the spectral clustering algorithm are dispatched for operation. This solves the problem that the existing electric power dispatching methods rely on empirical judgment or simple models and are difficult to accurately capture the time series changes and spatial distribution characteristics of the load, resulting in uneven distribution of electric energy and insufficient power dispatching efficiency. By grouping the loads to realize the optimized dispatching of the power system, the power system can be dispatched from the perspective of time and space, reducing the transmission loss of the power system and improving the dispatching efficiency of the power system.

[0007] Optionally, the load is divided according to the quantity and the operation status by using a spectral clustering algorithm to obtain a plurality of load groups, including: constructing an adjacency matrix of the load with respect to the operating condition according to the quantity; Constructing an eigenvector space according to the Laplace matrix corresponding to the adjacency matrix; The load is divided based on the feature vectors in the feature vector space by using a K-means clustering algorithm to obtain a plurality of load groups.

[0008] The electric load scheduling method provided in the present application can realize the optimized scheduling of the electric load. By constructing the eigenvector space based on the Laplace matrix corresponding to the operating conditions, the load is divided into multiple load groups using K-means clustering. The load is scheduled through the load grouping, which is beneficial to improving the scheduling efficiency of the power system.

[0009] Optionally, constructing an adjacency matrix of the load regarding the operating condition according to the quantity includes: constructing an operation matrix of the quantity of the load with respect to the operation condition; Using a difference method, a differential operation matrix corresponding to the operation matrix is ​​established; According to the specific value of each element in the differential operation matrix, an adjacency matrix of the load related to the operation status is constructed.

[0010] Optionally, the load is divided based on the feature vectors in the feature vector space by a K-means clustering algorithm to obtain a plurality of load groups, including: Based on the feature vectors in the feature vector space, the operation status of the load is represented by dimension reduction to obtain load information after dimension reduction; According to the load information after dimensionality reduction, multiple loads are selected from all the loads as multiple initial clustering centers; Based on the initial cluster center, all the loads are grouped by the K-means clustering algorithm to obtain multiple load groups.

[0011] Optionally, based on the initial cluster center, all the loads are grouped by the K-means clustering algorithm to obtain multiple load groups, including: Step A1, calculating the distances between all the loads except the initial cluster center and the multiple initial cluster centers; Step A2, allocating the other loads to the initial cluster center that is closest to the other loads to form a primary load grouping; Step A3, recalculating the cluster center in the primary load grouping, and redistributing all the other loads outside the cluster center according to the distance between the cluster center and all the other loads outside the cluster center; Step A4, repeating step A3 until the cluster center no longer changes, thereby obtaining multiple load groups.

[0012] Optionally, based on the operation conditions of the plurality of load groups, a likelihood function based on a Markov random field is constructed to calculate the planned operation conditions of the plurality of load groups through Kalman filtering, including: Based on the operation conditions of the plurality of load groups, construct a corresponding Markov random function; Adding a correction term to the Markov random function to construct a likelihood function based on the Markov random field; The likelihood function based on the Markov random field is iterated by Kalman filtering to calculate the planned operation conditions of the multiple load groups.

[0013] The power load scheduling method provided in the present application can realize the optimized scheduling of power loads by adding a correction term to the Markov random function constructed based on the operating conditions to construct a likelihood function based on the Markov random field, and by Kalman filtering, the likelihood function based on the Markov random field is iterated to calculate the planned operating conditions of multiple load groups. Based on the planned operating conditions, the operation of multiple load groups is scheduled, which is beneficial to improving the scheduling efficiency of the power system.

[0014] Optionally, the correction term includes an observation matrix correction term and a noise correction term; adding a correction term to the Markov random function to construct a likelihood function based on a Markov random field includes: Adding the noise correction term to the Markov random function to obtain an optimized Markov random function; According to the optimized Markov random function and the actual operation status of the corresponding load group, combined with the observation matrix correction term, an operation status function is constructed; Based on the noise correction term, the optimized Markov random function, the observation matrix correction term and the operation status function, a likelihood function based on a Markov random field is constructed.

[0015] In a second aspect, the present application provides a power load scheduling device, comprising: An acquisition module is used to acquire load information of loads in the area to be tested; the load information includes the quantity of all loads and the operation status of all loads within a preset period; A partitioning module, configured to partition the load according to the quantity and the operation status by using a spectral clustering algorithm to obtain a plurality of load groups; A calculation module, configured to construct a likelihood function based on a Markov random field based on the operation conditions of the plurality of load groups, so as to calculate the planned operation conditions of the plurality of load groups through Kalman filtering; The scheduling module is used to schedule the operation of the corresponding load groups according to the planned operation status.

[0016] The electric load dispatching device uses Kalman filtering based on the likelihood function of Markov random fields to carry out operation dispatching on the load groups obtained by dividing based on the spectral clustering algorithm, so as to solve the problems that the existing electric power dispatching methods rely on empirical judgment or simple models and are difficult to accurately capture the time series changes and spatial distribution characteristics of the load, resulting in uneven distribution of electric energy and insufficient power dispatching efficiency. By grouping the loads to achieve optimal dispatching of the power system, the power system can be dispatched from the perspective of time and space, reducing the transmission loss of the power system and improving the dispatching efficiency of the power system.

[0017] In a third aspect, the present application provides an electronic device, comprising a processor and a memory, wherein the memory stores a computer program executable by the processor, and when the processor executes the computer program, it runs the steps in the power load scheduling method described above.

[0018] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps in the power load scheduling method described above are executed.

[0019] Beneficial effects: The power load scheduling method, device, electronic device and storage medium provided in the present application, through a likelihood function based on a Markov random field and a Kalman filter, perform operation scheduling on the load groups obtained by dividing based on a spectral clustering algorithm, thereby solving the problem that the existing power scheduling method relies on empirical judgment or simple models and is difficult to accurately capture the timing changes and spatial distribution characteristics of the load, resulting in uneven power distribution and insufficient power scheduling efficiency. By grouping the loads to achieve optimal scheduling of the power system, the power system can be scheduled from the perspective of time and space, reducing the transmission loss of the power system and improving the scheduling efficiency of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A flow chart of a power load scheduling method provided in an embodiment of the present application.

[0021] Figure 2 A schematic diagram of the structure of the power load scheduling device provided in an embodiment of the present application.

[0022] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0023] Explanation of reference numerals: 1. Acquisition module; 2. Division module; 3. Calculation module; 4. Scheduling module; 301. Processor; 302. Memory; 303. Communication bus. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application usually described and shown in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0025] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0026] Please refer to Figure 1 , Figure 1A method for dispatching electric loads in some embodiments of the present application is used to optimize the dispatching of electric loads, comprising the steps of: Step S101, obtaining load information of loads in the area to be tested; the load information includes the number of all loads and the operation status of all loads in a preset period; Step S102, dividing the loads according to the quantity and operation status of (all loads) using a spectral clustering algorithm to obtain a plurality of load groups; Step S103, based on the operation conditions of the multiple load groups, construct a likelihood function based on a Markov random field, so as to calculate the planned operation conditions of the multiple load groups through Kalman filtering; Step S104: according to the planned operation status, the corresponding load groups are scheduled for operation.

[0027] The power load dispatching method uses Kalman filtering based on the likelihood function of Markov random fields to perform operation dispatch on the load groups obtained by spectral clustering algorithm, so as to solve the problem that the existing power dispatching method relies on empirical judgment or simple models and is difficult to accurately capture the time series changes and spatial distribution characteristics of the load, resulting in uneven power distribution and insufficient power dispatching efficiency. By grouping the loads to achieve optimal dispatching of the power system, the power system can be dispatched from the perspective of time and space, reducing the transmission loss of the power system and improving the dispatching efficiency of the power system.

[0028] Specifically, in step S101, load information of the loads in the test area is collected at multiple moments (i.e., sampling moments) within a preset period, and the load information includes the quantity of all loads and the operating conditions of all loads within the preset period, wherein the preset period is a time period of a preset length, the operating condition is whether the load is in an operating state or a non-operating state, and the operating condition of the load within the preset period includes the operating condition of the load at multiple moments within the preset period; the preset period can be set according to actual needs.

[0029] For example, the preset cycle includes T moments, and assuming there are n loads, then there are n×T operating conditions of the loads within the T moments.

[0030] Specifically, in step S102, the load is divided according to the quantity and operation status using a spectral clustering algorithm to obtain multiple load groups, including: Based on the quantity, construct an adjacency matrix of the load regarding the operation status; According to the Laplace matrix corresponding to the adjacency matrix, the eigenvector space is constructed; The load is divided based on the feature vectors in the feature vector space by using the K-means clustering algorithm to obtain multiple load groups.

[0031] Specifically, in step S102, according to the quantity, construct an adjacency matrix of the load regarding the operating conditions, including: Construct an operation matrix of the quantity of the load regarding the operating conditions; Adopt the difference method to establish a differential operation matrix corresponding to the operation matrix; According to the specific values of the elements in the differential operation matrix, construct an adjacency matrix of the load regarding the operating conditions.

[0032] In step S102, for the operating conditions at T moments, construct an operation matrix X corresponding to the quantity n of the load (i.e., an n×T matrix). Each element in the operation matrix can be represented by 0 or 1. 0 indicates the non-operating state, and 1 indicates the operating state. Among them, Indicates that the load i (i≤n) is in the operating state at the j-th (j≤T) moment ( is the element in the i-th row and j-th column of X), Indicates that the load i is in the non-operating state at the j-th moment, Indicates the operating conditions of all loads at the j-th moment ( is the vector composed of the elements in the j-th column of X).

[0033] Adopt the difference method to subtract the operating conditions of all loads at any moment from the operating conditions of all loads at the corresponding previous moment to obtain a differential operation matrix. For example, for j < T, subtract the operating conditions of all loads at the (j + 1)-th moment from the operating conditions of all loads at the j-th moment to obtain the differential operation vector at the j-th moment , that is . For j = T, . The differential operation matrix is obtained from the differential operation vectors at T moments.

[0034] Construct an adjacency matrix W of the load regarding the operating conditions. The initial adjacency matrix W is a zero matrix (the size of the adjacency matrix W is also n×T). According to the specific values of the elements in the differential operation matrix, perform iterative updates to determine each element of the adjacency matrix W. When the element in the differential operation matrix of the load i at the j-th moment ( is the element in the i-th row and j-th column of the differential operation matrix), for all loads in the differential operation matrix that satisfy the element ( is the element in the k-th row and j-th column of the differential operation matrix) ( ), let , that is, the value of the element representing the operating condition relationship between the load i and the load k in the adjacency matrix W is used with +1 value to replace ( is the element in the i-th row and k-th column of W), that is, when the elements of the differential operation matrix at the same time are the same, the element values ​​of the corresponding load operation status relationship in the adjacency matrix are the same, and when the elements of the differential operation matrix are all 1, the element values ​​of the corresponding load operation status relationship in the adjacency matrix will be updated to 1.

[0035] Specifically, in step S102, the degree matrix D of the adjacency matrix W is calculated. The degree matrix D is a diagonal matrix whose diagonal elements are ( represents the adjacency matrix of the operating status of load i at time j), and the normalized adjacency matrix is ​​calculated , thus calculating the Laplace matrix , is the unit matrix. The calculation process of the degree matrix is ​​prior art and will not be described in detail here.

[0036] Laplacian Matrix Perform eigenvalue decomposition and use the eigenvectors corresponding to the number of non-zero eigenvalues ​​to construct an eigenvector space, wherein the non-zero eigenvalues ​​can be set according to actual needs.

[0037] Specifically, in step S102, the load is divided based on the feature vectors in the feature vector space by using the K-means clustering algorithm to obtain multiple load groups, including: Based on the eigenvectors in the eigenvector space, the operation status of the load is represented by dimensionality reduction to obtain the load information after dimensionality reduction; According to the load information after dimensionality reduction, multiple loads are selected from all loads as multiple initial clustering centers; Based on the initial cluster centers, all loads are grouped using the K-means clustering algorithm to obtain multiple load groups.

[0038] In step S102, based on the number of non-zero eigenvalues ​​corresponding to the eigenvectors in the eigenvector space, the number of load operation conditions is reduced to a corresponding number through existing dimensionality reduction methods such as principal component analysis (PCA) method or linear discriminant analysis method (LDA) method, thereby obtaining load information after dimensionality reduction. Among them, existing dimensionality reduction methods such as principal component analysis method or linear discriminant analysis method are prior arts and will not be described in detail herein.

[0039] The load is regarded as a point, and multiple loads are randomly selected from the load information after dimensionality reduction as multiple initial clustering centers. Other loads are assigned to the initial clustering center with the smallest distance from it, forming a load grouping with the initial clustering center as the main center (that is, the clustering center of the load grouping is the initial clustering center). After each grouping is completed, the clustering center of each load grouping must be redivided (the division method is to select the load closest to the grouping center in the load grouping as the new clustering center), and based on the divided clustering center, the other loads of each load grouping are redistributed (the distribution method is to distribute other loads to the clustering center with the smallest distance from it), until all loads in the redistributed load grouping are no longer redistributed, forming multiple load groups.

[0040] Specifically, in step S102, based on the initial cluster center, all loads are grouped by using the K-means clustering algorithm to obtain multiple load groups, including: Step A1, calculating the distances between all loads except the initial cluster center and multiple initial cluster centers; Step A2, allocating other loads to the nearest initial cluster center to form a primary load grouping; Step A3, recalculating the cluster center in the primary load grouping, and redistributing all other loads outside the cluster center according to the distance between the cluster center and all other loads outside the cluster center; Step A4, repeating step A3 until the cluster center no longer changes, and obtaining multiple load groups.

[0041] In step S102, the distance between other loads and the initial cluster center in the load information after dimensionality reduction is calculated, and grouping is performed with the initial cluster center as the central node. The initial cluster center with the smallest distance to other loads is extracted from all initial cluster centers, and other loads are allocated to the group where the initial cluster center with the smallest distance to it is located. After each allocation is completed, it is necessary to recalculate the load closest to the center point as the cluster center of the load group according to the position of each load in each load group, and recalculate the distance between other loads except the cluster center and multiple cluster centers, and allocate other loads to the group where the cluster center with the smallest distance to it is located. Repeat the steps of calculating the cluster center and allocating the load until the cluster center no longer changes, or the load of each load group no longer changes, and the load grouping finally calculated is determined as multiple load groups.

[0042] Specifically, in step S103, based on the operation conditions of multiple load groups, a likelihood function based on a Markov random field is constructed to calculate the planned operation conditions of multiple load groups through Kalman filtering, including: Based on the operation status of multiple load groups, a corresponding Markov random function is constructed; Adding a correction term to the Markov random function to construct a likelihood function based on the Markov random field; Through Kalman filtering, the likelihood function based on Markov random field is iterated to calculate the planned operation status of multiple load groups.

[0043] In step S103, the operation of all loads in the timing is regarded as a Markov chain, and the above normalized adjacency matrix is ​​used as the state transfer matrix to obtain the corresponding Markov random function, which is specifically: ; in, It is the operation matrix at time T+j (time T is the current time).

[0044] Through the Markov random function, from the operation matrix at time T (current time) Initially, through iteration, the operation status of the load at times T+1, T+2, ... to T+j can be calculated.

[0045] Specifically, the correction term includes a measurement matrix correction term and a noise correction term; in step S103, a correction term is added to the Markov random function to construct a likelihood function based on the Markov random field, including: Add a noise correction term to the Markov random function to obtain an optimized Markov random function; According to the optimized Markov random function and the actual operation status of the corresponding load grouping, combined with the observation matrix correction term, the operation status function is constructed; Based on the noise correction term, the optimized Markov random function, the observation matrix correction term and the operation status function, a likelihood function based on the Markov random field is constructed.

[0046] In step S103, the iteration of the operation status cannot ignore the influence of noise, and the zero mean noise in the iteration process is considered. , add the noise correction term to the Markov random function to obtain the optimized Markov random function. The optimized Markov random function is specifically: ; in, for The noise correction term at time t (i.e., zero-mean noise).

[0047] By using the observation matrix correction term, the relationship expression between the optimized Markov random function and the actual operation status of the corresponding load group is constructed to obtain the operation status function. The specific operation status function is: ; in, For observable operating conditions (operating conditions Including observable operation and unobservable operating conditions); is the measurement matrix correction term, , is the observation matrix, is the observation matrix The degree matrix of is a diagonal matrix whose diagonal elements are , is the observation matrix The element corresponding to load i at time j. The observation matrix can be set according to actual needs.

[0048] For the observation matrix , whose diagonal elements When , it means that load i is the anchor point. When , it means that load i is not an anchor point. When load k and load i are assigned to the same load group, let the observation matrix Elements in Among them, the anchor point is the load closest to the cluster center in each load group, as well as the load with the smallest horizontal coordinate, the load with the largest horizontal coordinate, the load with the smallest vertical coordinate, and the load with the largest vertical coordinate among all loads. That is, there are always K+4 anchor points, and K is the number of load groups.

[0049] In some embodiments, after multiple load groups are obtained, the load closest to the cluster center can be selected in each load group as the anchor point, and the load with the smallest horizontal coordinate, the load with the largest horizontal coordinate, the load with the smallest vertical coordinate, and the load with the largest vertical coordinate can be selected from all loads as anchor points. After determining the anchor point, algorithms such as the XGBoost algorithm can be used to predict the spatial positions of loads outside the anchor point and the cluster center.

[0050] In summary, based on the noise correction term, the optimized Markov random function, the observation matrix correction term and the operation status function, the likelihood function based on the Markov random field is constructed. The likelihood function based on the Markov random field is specifically: ; Where p is a given condition ( ) appears under The probability of S is zero mean noise. The covariance matrix of , where the covariance matrix S is equivalent to the operating matrix at the current moment Variance ,Right now , the superscript T is the transposition symbol.

[0051] In step S103, the likelihood function based on the Markov random field is iterated by Kalman filtering to maximize the likelihood function of the Markov random field (i.e., calculate the maximum value of p), thereby determining the value corresponding to the maximum value. , and obtain the planned operation status of multiple load groups. Among them, Kalman filtering is an existing technology and will not be described in detail here.

[0052] Specifically, in step S104, operation scheduling is performed on the corresponding load groups according to the planned operation status of each load group.

[0053] As can be seen from the above, the power load scheduling method obtains the load information of the loads in the test area, and the load information includes the number of all loads and the operating conditions of all loads within a preset period. According to the number and the operating conditions, the load is divided by using the spectral clustering algorithm to obtain multiple load groups. Based on the operating conditions of the multiple load groups, a likelihood function based on a Markov random field is constructed to calculate the planned operating conditions of the multiple load groups through Kalman filtering, and the corresponding load groups are scheduled for operation according to the planned operating conditions; thus, through the likelihood function based on the Markov random field and the Kalman filter, the load groups obtained by the spectral clustering algorithm are scheduled for operation, so as to solve the problem that the existing power scheduling method relies on empirical judgment or simple models and is difficult to accurately capture the time series changes and spatial distribution characteristics of the load, resulting in uneven power distribution and insufficient power scheduling efficiency. By grouping the loads to achieve optimal scheduling of the power system, the power system can be scheduled from the perspective of time and space, reducing the transmission loss of the power system and improving the scheduling efficiency of the power system.

[0054] refer to Figure 2 The present application provides a power load scheduling device for optimizing the scheduling of power loads, including: Acquisition module 1, used to acquire load information of loads in the area to be tested; the load information includes the number of all loads and the operation status of all loads within a preset period; A partitioning module 2 is used to partition the loads according to the quantity and operation status of (all loads) by using a spectral clustering algorithm to obtain multiple load groups; The calculation module 3 is used to construct a likelihood function based on the Markov random field based on the operation conditions of the multiple load groups, so as to calculate the planned operation conditions of the multiple load groups through Kalman filtering; The scheduling module 4 is used to schedule the operation of the corresponding load groups according to the planned operation status.

[0055] The electric load dispatching device uses Kalman filtering based on the likelihood function of Markov random fields to carry out operation dispatching on the load groups obtained by dividing based on the spectral clustering algorithm, so as to solve the problems that the existing electric power dispatching methods rely on empirical judgment or simple models and are difficult to accurately capture the time series changes and spatial distribution characteristics of the load, resulting in uneven distribution of electric energy and insufficient power dispatching efficiency. By grouping the loads to achieve optimal dispatching of the power system, the power system can be dispatched from the perspective of time and space, reducing the transmission loss of the power system and improving the dispatching efficiency of the power system.

[0056] Specifically, when the acquisition module 1 is executed, the load information of the loads in the test area is collected at multiple moments (i.e., sampling moments) within a preset period, and the load information includes the number of all loads and the operating conditions of all loads within the preset period, wherein the preset period is a time period of a preset length, the operating condition is whether the load is in an operating state or a non-operating state, and the operating condition of the load within the preset period includes the operating condition of the load at multiple moments within the preset period; the preset period can be set according to actual needs.

[0057] For example, the preset cycle includes T moments, and assuming there are n loads, then there are n×T operating conditions of the loads within the T moments.

[0058] Specifically, when the division module 2 divides the load according to the quantity and operation status and obtains multiple load groups by using the spectral clustering algorithm, it executes: Based on the quantity, construct an adjacency matrix of the load regarding the operation status; According to the Laplace matrix corresponding to the adjacency matrix, the eigenvector space is constructed; The load is divided based on the feature vectors in the feature vector space by using the K-means clustering algorithm to obtain multiple load groups.

[0059] Specifically, when the partitioning module 2 constructs the adjacency matrix of the load regarding the operation status according to the quantity, it executes: Construct an operational matrix of the number of loads with respect to operational situations; Using the difference method, a differential operation matrix corresponding to the operation matrix is ​​established; According to the specific value of each element in the differential operation matrix, an adjacency matrix of the load regarding the operation status is constructed.

[0060] When the partitioning module 2 is executing, for the operating conditions at T moments, a running matrix X corresponding to the number n of loads is constructed (i.e., an n×T matrix). Each element in the running matrix can be represented by 0 or 1, where 0 indicates a non-running state and 1 indicates a running state. Among them, indicates that load i (i≤n) is in a running state at the j-th (j≤T) moment ( is the element in the i-th row and j-th column of X), indicates that load i is in a non-running state at the j-th moment, indicates the operating conditions of all loads at the j-th moment ( is the vector composed of the elements in the j-th column of X).

[0061] Using the difference method, the operating conditions of all loads at any moment are subtracted from the operating conditions of all loads at the corresponding previous moment to obtain a differential running matrix. For example, for j<T, the operating conditions of all loads at the (j + 1)-th moment are subtracted from the operating conditions of all loads at the j-th moment to obtain the differential running vector at the j-th moment , that is, , for j = T, . The differential running matrix is obtained from the differential running vectors at T moments.

[0062] Construct an adjacency matrix W of the loads regarding the operating conditions. The initial adjacency matrix W is a zero matrix (the size of the adjacency matrix W is also n×T). According to the specific values of the elements in the differential running matrix, iterative updates are performed to determine each element of the adjacency matrix W. When the element in the differential running matrix of load i at the j-th moment ( is the element in the i-th row and j-th column of the differential running matrix), for all loads in the differential running matrix that satisfy the element ( is the element in the k-th row and j-th column of the differential running matrix), the load ( ), let , that is, replace the value of the element representing the operating condition relationship between load i and load k in the adjacency matrix W with the value of + 1 ( is the element in the i-th row and k-th column of W), that is, when the elements of the differential running matrix at the same moment are the same, the values of the elements representing the operating condition relationships of the corresponding loads in the adjacency matrix are the same, and when all the elements of the differential running matrix are 1, the values of the elements representing the operating condition relationships of the corresponding loads in the adjacency matrix will be updated to 1.

[0063] Specifically, when the partitioning module 2 is executed, it calculates the degree matrix D of the adjacency matrix W. The degree matrix D is a diagonal matrix whose diagonal elements are ( represents the adjacency matrix of the operating status of load i at time j), and the normalized adjacency matrix is ​​calculated , thus calculating the Laplace matrix , is the unit matrix. The calculation process of the degree matrix is ​​prior art and will not be described in detail here.

[0064] Laplacian Matrix Perform eigenvalue decomposition and use the eigenvectors corresponding to the number of non-zero eigenvalues ​​to construct an eigenvector space, wherein the non-zero eigenvalues ​​can be set according to actual needs.

[0065] Specifically, when the partitioning module 2 partitions the load based on the feature vectors in the feature vector space by using the K-means clustering algorithm to obtain multiple load groups, the partitioning module 2 executes: Based on the eigenvectors in the eigenvector space, the operation status of the load is represented by dimensionality reduction to obtain the load information after dimensionality reduction; According to the load information after dimensionality reduction, multiple loads are selected from all loads as multiple initial clustering centers; Based on the initial cluster centers, all loads are grouped using the K-means clustering algorithm to obtain multiple load groups.

[0066] When the partitioning module 2 is executed, based on the number of non-zero eigenvalues ​​corresponding to the eigenvectors in the eigenvector space, the number of load operation conditions is reduced to a corresponding number through existing dimensionality reduction methods such as the principal component analysis (PCA) method or the linear discriminant analysis method (LDA) method, thereby obtaining load information after dimensionality reduction. Among them, existing dimensionality reduction methods such as the principal component analysis method or the linear discriminant analysis method are prior arts and are not described in detail herein.

[0067] The load is regarded as a point, and multiple loads are randomly selected from the load information after dimensionality reduction as multiple initial clustering centers. Other loads are assigned to the initial clustering center with the smallest distance from it, forming a load grouping with the initial clustering center as the main center (that is, the clustering center of the load grouping is the initial clustering center). After each grouping is completed, the clustering center of each load grouping must be redivided (the division method is to select the load closest to the grouping center in the load grouping as the new clustering center), and based on the divided clustering center, the other loads of each load grouping are redistributed (the distribution method is to distribute other loads to the clustering center with the smallest distance from it), until all loads in the redistributed load grouping are no longer redistributed, forming multiple load groups.

[0068] Specifically, when the division module 2 groups all loads based on the initial cluster center and obtains multiple load groups through the K-means clustering algorithm, it executes: Step A1, calculating the distances between all loads except the initial cluster center and the multiple initial cluster centers; Step A2, allocating other loads to the nearest initial cluster center to form a primary load grouping; Step A3, recalculating the cluster center in the primary load grouping, and redistributing all other loads outside the cluster center according to the distance between the cluster center and all other loads outside the cluster center; Step A4, repeating step A3 until the cluster center no longer changes, and obtaining multiple load groups.

[0069] When the partitioning module 2 is executed, it calculates the distance between other loads and the initial cluster center in the load information after dimensionality reduction, groups them with the initial cluster center as the central node, extracts the initial cluster center with the smallest distance from other loads from all initial cluster centers, and distributes other loads to the group where the initial cluster center with the smallest distance is located. After each distribution is completed, it is necessary to recalculate the load closest to the center point as the cluster center of the load group according to the position of each load in each load group, recalculate the distance between other loads except the cluster center and multiple cluster centers, distribute other loads to the group where the cluster center with the smallest distance is located, repeat the steps of calculating the cluster center and distributing the load until the cluster center no longer changes, or the load of each load group no longer changes, and the load grouping finally calculated is determined as multiple load groups.

[0070] Specifically, when the calculation module 3 constructs a likelihood function based on a Markov random field based on the operation conditions of multiple load groups to calculate the planned operation conditions of multiple load groups through Kalman filtering, it executes: Based on the operation conditions of multiple load groups, a corresponding Markov random function is constructed; Adding a correction term to the Markov random function to construct a likelihood function based on the Markov random field; Through Kalman filtering, the likelihood function based on Markov random field is iterated to calculate the planned operation status of multiple load groups.

[0071] When the calculation module 3 is executed, the operation of all loads in the timing is regarded as a Markov chain, and the above normalized adjacency matrix is ​​used as the state transfer matrix to obtain the corresponding Markov random function. The Markov random function is specifically: ; in, It is the operation matrix at time T+j (time T is the current time).

[0072] Through the Markov random function, from the operation matrix at time T (current time) Initially, through iteration, the operation status of the load at times T+1, T+2, ... to T+j can be calculated.

[0073] Specifically, the correction term includes an observation matrix correction term and a noise correction term; when the calculation module 3 adds the correction term to the Markov random function to construct a likelihood function based on the Markov random field, it executes: Add a noise correction term to the Markov random function to obtain an optimized Markov random function; According to the optimized Markov random function and the actual operation status of the corresponding load grouping, combined with the observation matrix correction term, the operation status function is constructed; Based on the noise correction term, the optimized Markov random function, the observation matrix correction term and the operation status function, a likelihood function based on the Markov random field is constructed.

[0074] When the calculation module 3 is executed, the iteration of the operation status cannot ignore the influence of noise, and the zero mean noise in the iteration process is considered. , add the noise correction term to the Markov random function to obtain the optimized Markov random function. The optimized Markov random function is specifically: ; in, for The noise correction term at time t (i.e., zero-mean noise).

[0075] By using the observation matrix correction term, the relationship expression between the optimized Markov random function and the actual operation status of the corresponding load group is constructed to obtain the operation status function. The specific operation status function is: ; in, For observable operating conditions (operating conditions Including observable operation and unobservable operating conditions); is the observation matrix correction term, , is the observation matrix, is the observation matrix The degree matrix of is a diagonal matrix whose diagonal elements are , is the observation matrix The element corresponding to load i at time j. The observation matrix can be set according to actual needs.

[0076] For the observation matrix , whose diagonal elements When , it means that load i is the anchor point. When , it means that load i is not an anchor point. When load k and load i are assigned to the same load group, let the observation matrix Elements in Among them, the anchor point is the load closest to the cluster center in each load group, as well as the load with the smallest horizontal coordinate, the load with the largest horizontal coordinate, the load with the smallest vertical coordinate, and the load with the largest vertical coordinate among all loads. That is, there are always K+4 anchor points, and K is the number of load groups.

[0077] In some embodiments, after multiple load groups are obtained, the load closest to the cluster center can be selected in each load group as the anchor point, and the load with the smallest horizontal coordinate, the load with the largest horizontal coordinate, the load with the smallest vertical coordinate, and the load with the largest vertical coordinate can be selected from all loads as anchor points. After determining the anchor point, algorithms such as the XGBoost algorithm can be used to predict the spatial positions of loads outside the anchor point and the cluster center.

[0078] In summary, based on the noise correction term, the optimized Markov random function, the observation matrix correction term and the operation status function, the likelihood function based on the Markov random field is constructed. The likelihood function based on the Markov random field is specifically: ; Where p is a given condition ( ) appears under The probability of S is zero mean noise. The covariance matrix of , where the covariance matrix S is equivalent to the operating matrix at the current moment Variance ,Right now , the superscript T is the transposition symbol.

[0079] When the calculation module 3 is executed, the likelihood function based on the Markov random field is iterated through Kalman filtering to maximize the likelihood function of the Markov random field (i.e., calculate the maximum value of p), thereby determining the value corresponding to the maximum value. , and obtain the planned operation status of multiple load groups. Among them, Kalman filtering is an existing technology and will not be described in detail here.

[0080] Specifically, when the scheduling module 4 is executed, it schedules the operation of the corresponding load group according to the planned operation status of each load group.

[0081] As can be seen from the above, the electric load dispatching device obtains the load information of the loads in the test area, the load information includes the number of all loads and the operating conditions of all loads within a preset period, and divides the loads according to the number and the operating conditions by using the spectral clustering algorithm to obtain multiple load groups. Based on the operating conditions of the multiple load groups, a likelihood function based on a Markov random field is constructed to calculate the planned operating conditions of the multiple load groups through Kalman filtering, and the corresponding load groups are dispatched according to the planned operating conditions; thus, through the likelihood function based on the Markov random field and the use of Kalman filtering, the load groups obtained by the division based on the spectral clustering algorithm are dispatched, so as to solve the problem that the existing power dispatching method relies on empirical judgment or simple models and is difficult to accurately capture the time series changes and spatial distribution characteristics of the loads, resulting in uneven power distribution and insufficient power dispatching efficiency. By grouping the loads to achieve optimal dispatching of the power system, the power system can be dispatched from the perspective of time and space, reducing the transmission loss of the power system and improving the dispatching efficiency of the power system.

[0082] Please refer to Figure 3 , Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application, the present application provides an electronic device, including: a processor 301 and a memory 302, the processor 301 and the memory 302 are interconnected and communicate with each other through a communication bus 303 and / or other forms of connection mechanisms (not shown), the memory 302 stores a computer program executable by the processor 301, when the electronic device is running, the processor 301 executes the computer program to execute the power load scheduling method in any optional implementation of the above embodiment, to achieve the following functions: obtain load information of loads in a test area, the load information includes the number of all loads and the operating conditions of all loads within a preset period, divide the loads according to the number and the operating conditions, use a spectral clustering algorithm to obtain multiple load groups, construct a likelihood function based on a Markov random field based on the operating conditions of the multiple load groups, calculate the planned operating conditions of the multiple load groups through Kalman filtering, and schedule the operation of the corresponding load groups according to the planned operating conditions.

[0083] An embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the power load scheduling method in any optional implementation of the above embodiment is executed to achieve the following functions: load information of loads in a test area is obtained, the load information includes the number of all loads and the operating conditions of all loads within a preset period, and the loads are divided according to the number and the operating conditions by using a spectral clustering algorithm to obtain a plurality of load groups. Based on the operating conditions of the plurality of load groups, a likelihood function based on a Markov random field is constructed to calculate the planned operating conditions of the plurality of load groups through Kalman filtering, and the corresponding load groups are scheduled for operation according to the planned operating conditions. Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable red-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0084] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0085] In addition, the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0086] Furthermore, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0087] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0088] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for dispatching electric loads, for optimizing dispatching electric loads, characterized in that: Includes steps: Obtaining load information of loads in the area to be tested; the load information includes the number of all loads and the operating conditions of all loads within a preset period; According to the quantity and the operation status, the load is divided by using a spectral clustering algorithm to obtain a plurality of load groups; Based on the operation conditions of the plurality of load groups, a likelihood function based on a Markov random field is constructed to calculate the planned operation conditions of the plurality of load groups through Kalman filtering; According to the planned operation status, the corresponding load groups are scheduled for operation.

2. The power load dispatching method according to claim 1, characterized in that: According to the quantity and the operation status, the load is divided by using a spectral clustering algorithm to obtain a plurality of load groups, including: constructing an adjacency matrix of the load with respect to the operating condition according to the quantity; Constructing an eigenvector space according to the Laplace matrix corresponding to the adjacency matrix; The load is divided based on the feature vectors in the feature vector space by using a K-means clustering algorithm to obtain a plurality of load groups.

3. The power load dispatching method according to claim 2, characterized in that: Constructing an adjacency matrix of the load with respect to the operating condition according to the quantity, including: constructing an operation matrix of the quantity of the load with respect to the operation condition; Using a difference method, a differential operation matrix corresponding to the operation matrix is ​​established; According to the specific value of each element in the differential operation matrix, an adjacency matrix of the load related to the operation status is constructed.

4. The power load dispatching method according to claim 2, characterized in that: The load is divided based on the feature vectors in the feature vector space by using a K-means clustering algorithm to obtain a plurality of load groups, including: Based on the feature vectors in the feature vector space, the operation status of the load is represented by dimension reduction to obtain load information after dimension reduction; According to the load information after dimensionality reduction, multiple loads are selected from all the loads as multiple initial clustering centers; Based on the initial cluster center, all the loads are grouped by the K-means clustering algorithm to obtain multiple load groups.

5. The power load dispatching method according to claim 4, characterized in that: Based on the initial cluster center, all the loads are grouped by the K-means clustering algorithm to obtain multiple load groups, including: Step A1, calculating the distances between all the loads except the initial cluster center and the multiple initial cluster centers; Step A2, allocating the other loads to the initial cluster center that is closest to the other loads to form a primary load grouping; Step A3, recalculating the cluster center in the primary load grouping, and redistributing all the other loads outside the cluster center according to the distance between the cluster center and all the other loads outside the cluster center; Step A4, repeating step A3 until the cluster center no longer changes, thereby obtaining multiple load groups.

6. The power load dispatching method according to claim 1, characterized in that: Based on the operation conditions of the plurality of load groups, a likelihood function based on a Markov random field is constructed to calculate the planned operation conditions of the plurality of load groups through Kalman filtering, including: Based on the operation conditions of the plurality of load groups, construct a corresponding Markov random function; Adding a correction term to the Markov random function to construct a likelihood function based on the Markov random field; The likelihood function based on the Markov random field is iterated by Kalman filtering to calculate the planned operation conditions of the multiple load groups.

7. The power load dispatching method according to claim 6, characterized in that: The correction term includes an observation matrix correction term and a noise correction term; A correction term is added to the Markov random function to construct a likelihood function based on the Markov random field, including: Adding the noise correction term to the Markov random function to obtain an optimized Markov random function; According to the optimized Markov random function and the actual operation status of the corresponding load group, combined with the observation matrix correction term, an operation status function is constructed; Based on the noise correction term, the optimized Markov random function, the observation matrix correction term and the operation status function, a likelihood function based on a Markov random field is constructed.

8. An electric load dispatching device for optimizing the dispatching of electric loads, characterized in that: include: An acquisition module is used to acquire load information of loads in the area to be tested; the load information includes the quantity of all loads and the operation status of all loads within a preset period; A partitioning module, configured to partition the load according to the quantity and the operation status by using a spectral clustering algorithm to obtain a plurality of load groups; A calculation module, used for constructing a likelihood function based on a Markov random field based on the operation conditions of the plurality of load groups, so as to calculate the planned operation conditions of the plurality of load groups through Kalman filtering; The scheduling module is used to schedule the operation of the corresponding load groups according to the planned operation status.

9. An electronic device, characterized in that: It comprises a processor and a memory, wherein the memory stores a computer program executable by the processor, and when the processor executes the computer program, it runs the steps in the power load scheduling method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the power load scheduling method according to any one of claims 1 to 7 are executed.

Images