A load access phase optimization method and device based on cluster analysis

Abstract

The application discloses a load access phase optimization method and device based on cluster analysis, and the method comprises the following steps: constructing a cluster analysis matrix according to the monitoring voltage data of different single-phase power consumption load users; performing cluster analysis on the access phase of the single-phase power consumption load users according to the cluster analysis matrix to obtain a single-phase load phase set; based on seasonal characteristics, performing scene cluster analysis on the annual daily load change curve of the users by using a K-means algorithm to obtain a scene load change curve; constructing a multi-objective optimization model based on line loss rate and three-phase unbalanced current; based on an NSGA-II algorithm, solving the access phase of the target single-phase power consumption load user in the multi-objective optimization model according to the single-phase load phase set and the scene load change curve to obtain a target single-phase load access phase scheme. The application can solve the technical problems that the existing wiring mode cannot adapt to the mixed-phase load access situation, and the three-phase unbalanced problem caused by the load access cannot be relieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of load analysis, in particular to a load phase access optimization method and device based on cluster analysis. BACKGROUND

[0002] As an important part of the power system, the distribution network is an important infrastructure in the region and a hub connecting power supply and users. The distribution network can provide high-quality power supply for the economic and social development of the region and the improvement of people's living standards. At the same time, the power quality also restricts the speed of enterprise development and social progress. Therefore, the power supply reliability of the distribution network can not only reflect the satisfaction of the power industry to the power demand of the national economy, but also has become one of the standards for measuring the economic development of a country. With the rapid increase of power demand, the traditional load access wiring mode has been unable to adjust the three-phase imbalance problem caused by the differentiated growth of single-phase load, resulting in poor power quality at the terminal of the distribution network, which cannot meet the user's expected effect, and further affects the line loss rate of the power supply area and the line.

[0003] Low-voltage residential load includes single-phase and three-phase power load. When the load is connected to the power grid, it is limited by monitoring data and cannot accurately measure the three-phase load imbalance. The access of the load may further aggravate the three-phase imbalance of the load, causing serious safety hazards. Moreover, the power supply area is a collection node of load access to the power grid, and is also a window for line loss monitoring and a key indicator for measuring the effectiveness of low-voltage equipment line operation and management of the distribution network. However, in the low-voltage three-phase four-wire system of urban residents and rural power supply systems: because most of the electricity users are single-phase loads or mixed single-phase and three-phase loads, and the load sizes and electricity usage times are different. Therefore, the imbalance current between the three phases in the power grid exists objectively, and this unbalanced power consumption condition is irregular and cannot be predicted in advance. This leads to long-term imbalance of three-phase load in the low-voltage power supply system; at the same time, it affects the line loss rate of the area and threatens the operation reliability and stability of the low-voltage line equipment of the distribution network. SUMMARY

[0004] The present application provides a load phase access optimization method and device based on cluster analysis, which is used to solve the technical problem that the existing wiring mode cannot clearly access the single-phase load, so it cannot adapt to the mixed-phase load access situation, and the three-phase imbalance problem caused by load access cannot be alleviated.

[0005] Therefore, the first aspect of the present application provides a load phase access optimization method based on cluster analysis, comprising:

[0006] calculating a Laplacian matrix according to the monitoring voltage data of different single-phase power load users;

[0007] After singular value decomposition is performed on the Laplacian matrix, three largest eigenvalues in the obtained eigenvectors are selected to construct a cluster analysis matrix;

[0008] According to the cluster analysis matrix, a single-phase load phase set of the single-phase power consumption load user is obtained;

[0009] Based on seasonal characteristics, a K-means algorithm is used to perform scene cluster analysis according to the daily load change curve of the user in a year, and four types of scene load change curves are obtained;

[0010] Based on line loss rate and three-phase unbalanced current, a multi-objective optimization model is constructed, which includes power flow constraint conditions, branch power flow inequality constraint conditions and power grid node voltage constraint conditions;

[0011] Based on an NSGA-II algorithm, the access phase of the target single-phase power consumption load user in the multi-objective optimization model is solved according to the single-phase load phase set and the scene load change curve, and a target single-phase load access phase scheme is obtained.

[0012] Preferably, the calculation of the Laplacian matrix according to the monitoring voltage data of different single-phase power consumption load users further comprises:

[0013] The single-phase power consumption load user is monitored in real time by an electric energy metering device, and monitoring voltage data is obtained.

[0014] Preferably, the calculation of the Laplacian matrix according to the monitoring voltage data of different single-phase power consumption load users comprises:

[0015] A monitoring data matrix is constructed according to the monitoring voltage data of different single-phase power consumption load users;

[0016] After a corresponding adjacency matrix is defined based on the monitoring data matrix, a degree matrix is calculated according to the adjacency matrix;

[0017] The Laplacian matrix is calculated through the degree matrix.

[0018] Preferably, the K-means algorithm is used to perform scene cluster analysis according to the daily load change curve of the user in a year based on seasonal characteristics, and four types of scene load change curves are obtained, which comprises:

[0019] Based on seasonal characteristics, a K-means algorithm is used to configure cluster analysis parameters, which include four cluster types and multiple initial cluster center curves;

[0020] Calculate curve Euclidean distance according to the annual daily load variation curve of the user and the initial clustering center curve, and perform category assignment of the daily load variation curve into 4 clustering categories based on the curve Euclidean distance, to obtain 4 scenario clustering curve sets, each of which includes a plurality of sample curves;

[0021] Calculate the average value of the sample curves in the same time in each of the scenario clustering curve sets to obtain a time average curve, and combine the 4 time average curves to obtain a new clustering center curve;

[0022] Replace the new clustering center curve with the initial clustering center curve, return to the step of calculating the curve Euclidean distance according to the annual daily load variation curve of the user and the initial clustering center curve, until a preset iteration condition is reached, to obtain 4 scenario load variation curves.

[0023] Preferably, the NSGA-II algorithm is used to solve the access phase of the target single-phase power load user in the multi-objective optimization model according to the single-phase load phase set and the scenario load variation curve, to obtain a target single-phase load access phase scheme, which includes:

[0024] Based on the NSGA-II algorithm, the single-phase load phase set is used as the optimization space, the target single-phase power load user is used as the chromosome, and the initial access phase population of the chromosome is initialized;

[0025] Calculate the node admittance matrix of the distribution network wiring network generated based on the individual of the initial access phase population;

[0026] Based on the optimization space and the scenario load variation curve, the node admittance matrix is used to perform forward power flow calculation on different time sequence scenarios in the target optimization model to obtain a power flow calculation result;

[0027] If the power flow calculation result meets the preset constraint condition, the individual optimization data is calculated according to the preset objective function;

[0028] The individual optimization data is sequentially subjected to cross mutation operation to obtain an updated access phase population, the updated access phase population is used to replace the initial access phase population, and the step of calculating the node admittance matrix of the distribution network wiring network generated based on the individual of the initial access phase population is returned, until a preset genetic generation number is reached, to obtain a target single-phase load access phase scheme.

[0029] The second aspect of the present application provides a load access phase optimization device based on clustering analysis, which includes:

[0030] The matrix calculation module is configured to calculate a Laplacian matrix according to the monitored voltage data of different single-phase power load users.

[0031] a matrix decomposition module, configured to construct a cluster analysis matrix by selecting three largest eigenvalues from eigenvectors obtained after singular value decomposition of the Laplacian matrix;

[0032] a first clustering module, configured to perform cluster analysis on access phases of single-phase power consumption load users according to the cluster analysis matrix, to obtain a single-phase load phase set;

[0033] a second clustering module, configured to perform scene cluster analysis on annual daily load curves of users according to a K-means algorithm based on seasonal characteristics, to obtain four types of scene load curves;

[0034] a model construction module, configured to construct a multi-objective optimization model based on line loss rate and three-phase unbalanced current, the multi-objective optimization model including a power flow constraint condition, a branch power flow inequality constraint condition and a power grid node voltage constraint condition;

[0035] a model solution module, configured to solve an access phase of a target single-phase power consumption load user in the multi-objective optimization model based on an NSGA-II algorithm according to the single-phase load phase set and the scene load curve, to obtain a target single-phase load access phase scheme.

[0036] Preferably, the method further comprises:

[0037] a data acquisition module, configured to monitor single-phase power consumption load users in real time through an electric energy metering device and acquire monitored voltage data.

[0038] Preferably, the matrix calculation module is specifically configured to:

[0039] construct a monitored data matrix according to monitored voltage data of different single-phase power consumption load users;

[0040] define a corresponding adjacency matrix based on the monitored data matrix, and calculate a degree matrix according to the adjacency matrix;

[0041] obtain a Laplacian matrix through the degree matrix.

[0042] Preferably, the second clustering module comprises:

[0043] a parameter configuration sub-module, configured to configure cluster analysis parameters including four cluster categories and multiple initial cluster center curves according to a K-means algorithm based on seasonal characteristics;

[0044] The category assignment submodule is configured to calculate a curve Euclidean distance according to the annual daily load variation curve of a user and the initial clustering center curve, and assign a category to the daily load variation curve based on the curve Euclidean distance, so as to obtain four sets of scenario clustering curves, each of which includes a plurality of sample curves.

[0045] The update calculation submodule is configured to calculate an average value of the sample curves in the same time in each of the scenario clustering curves, so as to obtain a time average curve, and combine the four time average curves to obtain a new clustering center curve.

[0046] The iterative optimization submodule is configured to replace the initial clustering center curve with the new clustering center curve, trigger the category assignment submodule, and obtain four types of scenario load variation curves until a preset iteration condition is reached.

[0047] Preferably, the model solving module comprises:

[0048] The initialization submodule is configured to initialize an initial access phase population of the chromosome based on the NSGA-II algorithm, take the single-phase load phase set as an optimization space, and take the target single-phase power consumption user as the chromosome.

[0049] The admittance calculation submodule is configured to calculate a node admittance matrix of a power distribution network wiring network generated based on the initial access phase population.

[0050] The power flow calculation submodule is configured to perform a forward power flow calculation on different time sequence scenarios in the target optimization model according to the node admittance matrix based on the optimization space and the scenario load variation curve, so as to obtain a power flow calculation result.

[0051] The optimization screening submodule is configured to calculate individual optimization data according to a preset objective function if the power flow calculation result meets a preset constraint condition.

[0052] The update iteration submodule is configured to perform a cross variation operation on the individual optimization data in sequence to obtain an updated access phase population, replace the initial access phase population with the updated access phase population, trigger the admittance calculation submodule, and obtain a target single-phase load access phase scheme until a preset genetic generation number is reached.

[0053] As can be seen from the above technical solutions, the embodiments of the present application have the following advantages:

[0054] In the application, a load access phase optimization method based on cluster analysis is provided, comprising: calculating a Laplace matrix according to the monitored voltage data of different single-phase power consumption load users; constructing a cluster analysis matrix by selecting three largest eigenvalues from the obtained eigenvectors after singular value decomposition of the Laplace matrix; performing cluster analysis on the access phase of the single-phase power consumption load users according to the cluster analysis matrix to obtain a single-phase load phase set; based on seasonal characteristics, performing scene cluster analysis according to the daily load change curve of the users in a year by using a K-means algorithm to obtain four types of scene load change curves; constructing a multi-objective optimization model based on line loss rate and three-phase unbalanced current, the multi-objective optimization model comprising a power flow constraint condition, a branch power flow inequality constraint condition and a power grid node voltage constraint condition; based on an NSGA-II algorithm, solving the target single-phase power consumption load user's access phase in the multi-objective optimization model according to the single-phase load phase set and the scene load change curve to obtain a target single-phase load access phase scheme.

[0055] The load access phase optimization method based on cluster analysis provided by the application first performs targeted analysis and calculation on the access phase of the single-phase power consumption load users according to the monitored power consumption data, and determines the access phase of the single-phase power consumption load users; then, in order to reduce the calculation amount in the line loss analysis process, scene cluster analysis is performed on the daily load change curve in a year to obtain several types of scene load change curves based on seasonal changes, which greatly reduces the calculation amount of the line loss analysis; and the multi-objective optimization model constructed based on the line loss rate and the three-phase unbalanced current can make the final target single-phase load access phase scheme meet the requirements of power grid operation, and the scheme is more practical and targeted. The scheme of the embodiment can not only determine the specific single-phase access phase, but also provide an optimized access phase scheme, which can better meet the actual mixed-phase load access demand. Therefore, the application can solve the technical problems that the existing wiring mode cannot determine the single-phase load access phase, so it cannot adapt to the mixed-phase load access situation, and the three-phase unbalance problem caused by load access cannot be alleviated. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 A flowchart of a load access phase optimization method based on cluster analysis provided by the embodiment of the application is shown in the figure;

[0057] Figure 2 Another flowchart of a load access phase optimization method based on cluster analysis provided by the embodiment of the application is shown in the figure;

[0058] Figure 3 A structural diagram of a load access phase optimization device based on cluster analysis provided by the embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0059] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0060] For the convenience of understanding, please refer to Figure 1 The present application provides an embodiment of a load access phase optimization method based on cluster analysis, which comprises:

[0061] Because electrical equipment has inherent properties of three-phase and single-phase, the influence of single-phase electrical equipment and three-phase electrical equipment on the power grid after access is different. The access of three-phase electrical equipment is already fixed, determined by the geographical properties of the user, while the load access of single-phase electrical equipment will affect the power flow of the power grid and cause the three-phase imbalance of the power grid, leading to the lack of rationality of the operation of the power system.

[0062] For three-phase unbalanced current, the power department has almost no effective solution except for reasonably distributing the load. The unbalanced current in the power grid will increase the copper loss of the line and the transformer, also increase the iron loss of the transformer, reduce the output of the transformer and even affect the safe operation of the transformer, cause the imbalance of three-phase voltage, and further affect the power supply quality of the customer. Therefore, the present application analyzes the load access point and access phase and proposes a more targeted optimization wiring scheme.

[0063] Step 101, calculate the Laplace matrix according to the monitored voltage data of different single-phase electrical load users.

[0064] The monitored voltage data can be obtained through the smart electric energy meter. Each user can obtain multiple data at different times, and based on multiple users, the monitored voltage data matrix can be obtained, based on which the corresponding Laplace matrix can be calculated.

[0065] Because of the lack of access phase data of the single-phase electrical equipment of the user, and the time-consuming and laborious on-site verification by artificial, it is necessary to accurately identify the access phase of the existing user through theoretical analysis. The smart meter installed on the user side has the functions of electric energy metering and voltage monitoring, and the spectral clustering algorithm is established on the basis of the spectral graph theory in graph theory, and its essence is to transform the clustering problem into the optimal partition problem of the graph.

[0066] Step 102, after singular value decomposition of the Laplace matrix, three largest eigenvalues are selected from the obtained eigenvectors to construct a cluster analysis matrix.

[0067] Each user can select three largest eigenvalues from the eigenvector, and then all users n participating in the analysis can form a 3xn clustering analysis matrix. The clustering analysis matrix is mainly used for subsequent clustering analysis.

[0068] Step 103, clustering analysis is performed on the access phase of the single-phase power load user according to the clustering analysis matrix, and a single-phase load phase set is obtained.

[0069] In actual power grid operation, the phase of the single-phase load access phase cannot be directly obtained, resulting in great uncertainty in the load access of the power grid. Therefore, the single-phase load phase of the single-phase power load user is obtained by calculation and analysis in this embodiment. The clustering analysis method can be selected according to actual needs, for example, K-means algorithm; the specific is not limited.

[0070] Step 104, based on the seasonal characteristics, K-means algorithm is used to perform scene clustering analysis according to the daily load change curve of the user in a year, and four types of scene load change curves are obtained.

[0071] In order to calculate the line loss of the whole year of the transformer area, the power curve of the user needs to be calculated, and then the loss power value at all times in a year is obtained. In actual engineering application, it is extremely difficult to obtain the power curve of all users. The metering equipment on the user side does not have the function of continuous monitoring, and usually monitors the instantaneous power value every hour. The load of the transformer area is taken as a year, and the time interval of the data sample is set to 1 hour. Since the change of the load has continuity, the hourly load curve can be used as the continuous load power curve, so there are 8760 scene power value data samples of all users at that time in a year, that is, 8760 scenes. In the process of model solving, it is necessary to calculate for each scene, which will result in a huge amount of calculation and high planning complexity. Therefore, the scene reduction method is used to reduce the number of uncertain scenes, that is, the clustering analysis of the daily load change curve; finally, four clusters of seasons are obtained, each including multiple scene load change curves and a center curve.

[0072] Step 105, a multi-objective optimization model is constructed based on the line loss rate and the three-phase unbalanced current, and the multi-objective optimization model includes a power flow constraint condition, a branch power flow inequality constraint condition and a power grid node voltage constraint condition.

[0073] Because the power users in the transformer area are mostly single-phase loads or mixed single-phase and three-phase loads, and the loads are different in size and the power is different in time, the unbalanced current among the three phases in the power grid exists objectively, and this power unbalance condition is irregular and cannot be predicted in advance, which leads to long-term imbalance of three-phase loads in the low-voltage power supply system, and further affects the line active loss and voltage drop. Therefore, a multi-objective optimization model is established with the line loss and the three-phase unbalanced current as the optimization objectives. The model needs to configure necessary constraints according to the operation needs of the power grid, so that the model solution conforms to the operation characteristics of the actual power grid.

[0074] Step 106, based on the NSGA-II algorithm, the access phase of the target single-phase load user in the multi-objective optimization model is solved according to the single-phase load phase set and the scene load change curve, and the target single-phase load access phase scheme is obtained.

[0075] The NSGA-II algorithm can select the optimal scheme through iterative optimization. In this embodiment, the NSGA-II algorithm is used to solve the multi-objective optimization model, and the model solution is the optimized single-phase load access phase scheme. The single-phase phase set is the selection space of the optimization process, and the scene change curve is the change of the load under multiple different time sequences, so that the target single-phase load access phase scheme not only conforms to the actual power grid operation characteristics, but also reduces the line loss and alleviates the three-phase imbalance.

[0076] The load access phase optimization method based on clustering analysis provided by the embodiment of the application first analyzes and calculates the access phase of the single-phase load user according to the monitored power consumption data, and determines the access phase of the single-phase load user; then, in order to reduce the calculation amount in the line loss analysis process, the annual daily load change curve is subjected to scene clustering analysis, and several types of scene load change curves based on seasonal changes are obtained, which greatly reduces the calculation amount of the line loss analysis; and the multi-objective optimization model constructed based on the line loss rate and the three-phase unbalanced current can make the final target single-phase load access phase scheme conform to the operation requirements of the power grid, and the scheme is more practical and targeted. The scheme of the embodiment can not only determine the specific single-phase access phase, but also provide an optimized access phase scheme, which can better meet the actual mixed-phase load access demand. Therefore, the embodiment of the application can solve the technical problems that the existing wiring mode cannot determine the single-phase load access phase, so it cannot adapt to the mixed-phase load access situation, and the three-phase imbalance problem caused by the load access cannot be alleviated.

[0077] For ease of understanding, please refer to Figure 2 The embodiment two of the load access phase optimization method based on clustering analysis provided by the application comprises:

[0078] Step 201, real-time monitoring of single-phase power load users by electric energy metering device, and acquiring monitoring voltage data.

[0079] Step 202, constructing a monitoring data matrix according to the monitoring voltage data of different single-phase power load users.

[0080] Step 203, after defining the corresponding adjacency matrix based on the monitoring data matrix, calculating the degree matrix according to the adjacency matrix.

[0081] Step 204, calculating the Laplacian matrix through the degree matrix.

[0082] The monitoring voltage data can be a data sequence in hours as a unit, which may vary due to different sampling frequencies, but generally the monitoring voltage data should be uniform, and the monitoring voltage data is obtained by the electric meter, and generally the voltage amplitude is extracted. Each power load user can collect multiple monitoring voltage data, and the data integration can construct the monitoring data matrix U:

[0083]

[0084] Wherein, U MN is the Nth monitoring voltage data of the Mth power load user.

[0085] The adjacency matrix of the monitoring data matrix U is defined as:

[0086]

[0087] Wherein,

[0088]

[0089] σ is the Gaussian function parameter, which is set to 0.1 by default.

[0090] The degree matrix is calculated according to the adjacency matrix:

[0091]

[0092] Wherein, D MM = 0, if M≠N.

[0093] The Laplacian matrix is calculated through the degree matrix:

[0094]

[0095] Wherein, D is the degree matrix, and S is the adjacency matrix.

[0096] Step 205, after singular value decomposition of the Laplacian matrix, three largest eigenvalues are selected from the obtained eigenvectors to construct a clustering analysis matrix.

[0097] The process of singular value decomposition of Laplacian matrix is as follows:

[0098] P = QΛQ -1

[0099] wherein the diagonal elements of matrix A are in descending order of listed eigenvalues, Q is an eigenvalue matrix composed of eigenvalues corresponding to eigenvalues, and it is assumed that the eigenvalues corresponding to the three largest eigenvalues of the decomposed Laplacian matrix are q1, q2 and q3; then a clustering analysis matrix can be constructed as: Q' M×3 = [q1, q2, q3], wherein M is the number of users.

[0100] Step 206, clustering analysis is performed on the access phase of the single-phase power load user according to the clustering analysis matrix, and a single-phase load phase set is obtained.

[0101] Clustering analysis is a method of adaptively dividing some unlabeled data into several categories, which belongs to an unsupervised learning method. The same data after division has similar characteristics, and the classification result is more and more accurate after continuous optimization. The clustering method selected in this embodiment is the K-means clustering algorithm. The specific process is as follows: first, determine a k value, that is, we hope that the data set will be clustered into k sets. In this embodiment, in order to realize three-phase recognition, three different phase types are set; then, randomly select k data points from the data set as clustering centers; then, for each data point in the matrix, calculate the distance between it and each clustering center, and divide it into the phase type to which the clustering center belongs if it is close to the clustering center; then, after all data are divided into good sets, there are k sets in total. Then, the centroid of each set is recalculated. If the distance between the newly calculated centroid and the original centroid is less than a certain threshold value (indicating that the position of the newly calculated centroid changes little and tends to be stable, or in other words, converges), we can consider that the clustering has reached the expected result, and the algorithm terminates. If the new centroid and the original centroid change greatly, the distance and the subsequent steps need to be iteratively calculated. The target load user can be attributed to a specific phase type according to the distance calculation method, that is, the target access phase. In addition, one of the following distances can be selected for clustering analysis, that is, Euclidean distance, Manhattan distance, Chebyshev distance, cosine distance, etc. The essence of distance is to describe the similarity between two data, so some useful correlation coefficients can also be selected according to actual needs, which will not be described here.

[0102] Step 207, based on the seasonal characteristics, the K-means algorithm is used to configure the clustering analysis parameters, and the clustering analysis parameters include four clustering types and multiple initial clustering center curves.

[0103] Step 208, calculate the curve Euclidean distance according to the annual daily load variation curve of the user and the initial clustering center curve, and perform category assignment of the daily load variation curve to 4 clustering categories based on the curve Euclidean distance, to obtain 4 scene clustering curve sets, each of which includes multiple sample curves.

[0104] Step 209, calculate the average value of the sample curves in the same time in each scene clustering curve set, to obtain an average value curve, and combine the 4 average value curves to obtain a new clustering center curve.

[0105] Step 210, replace the initial clustering center curve with the new clustering center curve, return to step 208 until the preset iteration condition is reached, to obtain 4 scene load variation curves.

[0106] The above has been explained, if the line loss power consumption is calculated for each scene, there are 8760 data samples of power values at the same time of all users in a year, that is, 8760 scenes, and in the model solving process, calculation needs to be performed for each scene, which will cause a very large amount of calculation. Therefore, the K-means algorithm is adopted to reduce the scenes of the daily load variation curve, to simplify the calculation process and reduce the calculation amount.

[0107] The seasonal characteristics mainly manifest in the load variation in 4 different seasons, so the clustering category is defined as 4, and the initial clustering center can be defined as Calculate the Euclidean distance between each daily load variation curve and the four initial clustering center curves, and take the nearest center curve as the corresponding category, and classify the daily load variation curve into the corresponding category. The Euclidean distance calculation process is as follows:

[0108]

[0109] Among them, is the Euclidean distance, is the i-th k-dimensional daily load variation curve, is the initial clustering center curve.

[0110] Based on the curve Euclidean distance, the daily load variation curve is classified into 4 clustering categories, each clustering center represents a category, and each category includes multiple sample curves, the average value of all sample curves in each scene at the same time is calculated, that is, the average value curve of each cluster is calculated, which is taken as the updated new clustering center curve, and 4 new clustering center curves are obtained, that is, The average value curve calculation process is as follows:

[0111]

[0112] In the formula, δ jFor the jth scene, s is the scene category.

[0113] The new cluster center curve replaces the initial cluster center curve to continue the Euclidean distance calculation and category configuration analysis until the iteration result is stable, i.e., the iteration cluster result fluctuates within a preset range without a large mutation, and it is considered that the cluster analysis is completed, and the classification result is obtained, i.e., four categories of scene load change curves.

[0114] In step 211, a multi-objective optimization model is constructed based on the line loss rate and the three-phase unbalanced current, and the multi-objective optimization model includes a power flow constraint condition, a branch power flow inequality constraint condition and a power grid node voltage constraint condition.

[0115] The multi-objective optimization model is defined as:

[0116] G=P LOSS +ΔI

[0117] Wherein, P LOSS is the line loss rate, and ΔI is the three-phase unbalanced current.

[0118] Specifically, the line loss rate P LOSS can be calculated by the following formula:

[0119]

[0120] Wherein, L is the total number of branches of the distribution network, s and l are respectively the scene category and the branch, is the power loss of the lth branch under the s th scene.

[0121] And the three-phase balanced current can be calculated by the following formula:

[0122]

[0123] Wherein, I si is the corresponding load phase current of the ith node under the s th scene, is the three-phase average current of the corresponding node, n and i are respectively the total number of nodes n of the network and the ith node.

[0124] The power flow constraint condition is:

[0125]

[0126] Wherein, P si (t) and Q si (t) are respectively the active power and the reactive power injected by the ith node at the tth moment under the s th scene, U si (t) and U sj (t) are respectively the voltage amplitudes of the ith node and the jth node at the tth moment under the s th scene, G ij , BGijand B ij respectively are conductance and susceptance of branch ij of active distribution network; δ ij (t) is the phase angle difference of voltage between node i and node j at the tth moment.

[0127] The branch power flow inequality constraint condition is:

[0128]

[0129] wherein, is the active power flow of the branch connected between node i and node j at the tth moment under the s th typical day scenario, P ij , respectively are the lower and upper limits of the active power flow allowed by the branch.

[0130] The grid node voltage constraint condition is:

[0131]

[0132] wherein, U sti is the voltage value of node i at the tth moment under the s th scenario, U and respectively are the lower and upper limits of the voltage allowed by the node.

[0133] Step 212, based on the NSGA-II algorithm, taking the single-phase load phase set as the optimization space, taking the target single-phase power load user as the chromosome, and initializing the initial access phase population of the chromosome;

[0134] Step 213, calculating the node admittance matrix of the distribution network wiring network generated based on the initial access phase population;

[0135] Step 214, based on the optimization space and the scenario load variation curve, performing forward power flow calculation on different time sequence scenarios in the target optimization model according to the node admittance matrix, to obtain the power flow calculation result;

[0136] Step 215, if the power flow calculation result meets the preset constraint condition, then calculating the individual optimization data according to the preset objective function;

[0137] Step 216, performing cross mutation operation on the individual optimization data in turn to obtain the updated access phase population, replacing the initial access phase population with the updated access phase population, and returning to step 213 until the preset genetic generation number is reached, to obtain the target single-phase load access phase scheme.

[0138] The NSGA-II algorithm adopts a fast non-dominated sorting algorithm, has low calculation complexity, adopts a crowding degree and a crowding degree comparison operator to replace a shared radius that needs to be specified, and uses the crowding degree as a winning standard in the same level comparison after fast sorting, so that individuals in a quasi-Pareto domain can be extended to the entire Pareto domain and uniformly distributed, and the diversity of the population is maintained; the elite strategy is introduced, the sampling space is expanded, the loss of the best individual is prevented, and the operation speed and robustness of the algorithm are improved.

[0139] The preset constraint condition is a power flow constraint condition, a branch power flow inequality constraint condition, and a power grid node voltage constraint condition in the target optimization model. The preset genetic algebra can be set according to actual conditions or experience, and is not specifically limited. Since the daily load change curve in the whole year is reduced to four typical scenes, the annual load data needs to be converted according to the weight in the power flow calculation process.

[0140] In addition, if the power flow calculation result does not satisfy the preset constraint condition, the result of the preset target function is assigned to infinity. All individuals are quickly non-dominated sorted based on the individual optimization data, the excellent individuals are reserved by the elite strategy, and the excellent individuals of the population are selected by the tournament method for crossover and mutation, and finally, the excellent individuals and the new population are combined to form a new population, that is, the access phase population is updated. The optimization population is iteratively optimized in combination with the optimization space of the single-phase access phase set and the power grid constraint condition, until the optimal access phase scheme is obtained.

[0141] The load access phase optimization method based on clustering analysis provided in the embodiments of the application first analyzes and calculates the access phase of the single-phase power load user according to the monitored power consumption data, and determines the access phase of the single-phase power load user; then, in order to reduce the calculation amount in the line loss analysis process, the daily load change curve in the whole year is subjected to scene clustering analysis, several types of scene load change curves based on seasonal changes are obtained, and the calculation amount of the line loss analysis is greatly reduced; and the multi-objective optimization model constructed based on the line loss rate and the three-phase unbalanced current can make the final target single-phase load access phase scheme meet the operation requirements of the power grid, and the scheme is more practical and targeted. The scheme of the embodiments of the application can not only determine the specific single-phase access phase, but also provide an optimized access phase scheme, and can better meet the actual mixed-phase load access demand. Therefore, the embodiments of the application can solve the technical problems that the existing wiring mode cannot determine the single-phase load access phase, and therefore cannot adapt to the mixed-phase load access situation, and the three-phase unbalance problem caused by the load access cannot be alleviated.

[0142] For ease of understanding, please refer to Figure 3 The application provides an embodiment of a load access phase optimization device based on clustering analysis, which comprises:

[0143] The matrix calculation module 301 is configured to calculate a Laplacian matrix according to monitored voltage data of different single-phase power load users.

[0144] The matrix decomposition module 302 is configured to select three largest eigenvalues from the obtained eigenvectors after singular value decomposition of the Laplacian matrix to construct a cluster analysis matrix.

[0145] The first clustering module 303 is configured to perform cluster analysis on the access phase of the single-phase power load user according to the cluster analysis matrix to obtain a single-phase load phase set.

[0146] The second clustering module 304 is configured to perform scene cluster analysis on the annual daily load curve of the user according to the seasonal characteristics and the K-means algorithm to obtain four types of scene load curves.

[0147] The model construction module 305 is configured to construct a multi-objective optimization model based on line loss rate and three-phase unbalanced current, and the multi-objective optimization model includes a power flow constraint condition, a branch power flow inequality constraint condition and a power grid node voltage constraint condition.

[0148] The model solving module 306 is configured to solve the access phase of the target single-phase power load user in the multi-objective optimization model according to the single-phase load phase set and the scene load curve based on the NSGA-II algorithm to obtain a target single-phase load access phase scheme.

[0149] Further, the method further comprises the following steps.

[0150] The data acquisition module 307 is configured to monitor the single-phase power load user in real time through the electric energy metering device and acquire monitored voltage data.

[0151] Further, the matrix calculation module 301 is specifically configured to:

[0152] construct a monitored data matrix according to the monitored voltage data of different single-phase power load users;

[0153] After defining the corresponding adjacency matrix based on the monitored data matrix, a degree matrix is calculated according to the adjacency matrix;

[0154] The Laplacian matrix is calculated through the degree matrix.

[0155] Further, the second clustering module 304 comprises the following.

[0156] The parameter configuration sub-module 3041 is configured to configure cluster analysis parameters including four cluster categories and a plurality of initial cluster center curves based on the seasonal characteristics and the K-means algorithm.

[0157] The category assignment submodule 3042 is configured to calculate a curve Euclidean distance according to the annual daily load variation curve of the user and the initial clustering center curve, and perform category assignment of the daily load variation curve in four clustering categories based on the curve Euclidean distance, to obtain four sets of scenario clustering curves, each set of scenario clustering curves including a plurality of sample curves.

[0158] The update calculation submodule 3043 is configured to calculate the average value of the sample curves in the same time in each set of scenario clustering curves to obtain a time average curve, and combine the four time average curves to obtain a new clustering center curve.

[0159] The iterative optimization submodule 3044 is configured to replace the initial clustering center curve with the new clustering center curve to trigger the category assignment submodule 3042 until a preset iteration condition is reached, to obtain four types of scenario load variation curves.

[0160] Further, the model solving module 306 includes:

[0161] The initialization submodule 3061 is configured to take the single-phase load phase set as the optimization space, take the target single-phase power consumption user as the chromosome, and initialize the initial access phase population of the chromosome based on the NSGA-II algorithm.

[0162] The admittance calculation submodule 3062 is configured to calculate the node admittance matrix of the distribution network wiring network generated based on the initial access phase population.

[0163] The power flow calculation submodule 3063 is configured to perform forward power flow calculation on different time scenarios in the target optimization model according to the node admittance matrix based on the optimization space and the scenario load variation curve, to obtain a power flow calculation result.

[0164] The optimization screening submodule 3064 is configured to calculate individual optimization data according to a preset objective function if the power flow calculation result meets a preset constraint condition.

[0165] The update iteration submodule 3065 is configured to perform cross-over mutation operation on the individual optimization data in sequence to obtain an updated access phase population, replace the initial access phase population with the updated access phase population, and trigger the admittance calculation submodule 3062 until a preset genetic generation number is reached, to obtain a target single-phase load access phase scheme.

[0166] In several embodiments provided in the present application, it should be understood that the disclosed apparatus and method can be implemented by other manners. For example, the apparatus embodiments described above are merely illustrative, for example, the division of the units is merely a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units or components shown or discussed can be indirect coupling or communication connection through some interfaces, apparatuses or units, and can be electrical, mechanical or other forms.

[0167] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0168] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0169] When the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for executing all or part of the steps of the method described in each embodiment of the present application by a computer device (which can be a personal computer, a server, or a network device, etc.). The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (English full name: Read-Only Memory, English abbreviation: ROM), a random access memory (English full name: Random Access Memory, English abbreviation: RAM), a magnetic disk or an optical disk, and various program code storage media.

[0170] The above-described embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A load connection optimization method based on cluster analysis, characterized in that, include: Calculate the Laplace matrix based on the monitoring voltage data of different single-phase electrical load users; After performing singular value decomposition on the Laplacian matrix, three largest eigenvalues ​​are selected from the obtained eigenvectors to construct a clustering analysis matrix; Based on the clustering analysis matrix, clustering analysis is performed on the access phases of single-phase power load users to obtain the single-phase load phase set; Based on seasonal characteristics, the K-means algorithm was used to perform scene clustering analysis based on the daily load change curves of users throughout the year, resulting in four types of scene load change curves. A multi-objective optimization model is constructed based on line loss rate and three-phase unbalanced current. The multi-objective optimization model includes power flow constraints, branch power flow inequality constraints, and grid node voltage constraints. Based on the NSGA-II algorithm, the access phase of the target single-phase load user in the multi-objective optimization model is solved according to the single-phase load phase set and the scenario load change curve, so as to obtain the target single-phase load access phase scheme.

2. The load access optimization method based on cluster analysis according to claim 1, characterized in that, The calculation of the Laplace matrix based on the monitoring voltage data of different single-phase electrical load users also includes, prior to: The system monitors single-phase power load users in real time using electricity metering devices and obtains monitoring voltage data.

3. The load access optimization method based on cluster analysis according to claim 1, characterized in that, The calculation of the Laplace matrix based on the monitoring voltage data of different single-phase power load users includes: A monitoring data matrix is ​​constructed based on the monitoring voltage data of different single-phase electricity load users; After defining the corresponding adjacency matrix based on the monitoring data matrix, the degree matrix is ​​calculated based on the adjacency matrix. The Laplace matrix is ​​obtained by calculating the degree matrix.

4. The load access optimization method based on cluster analysis according to claim 1, characterized in that, Based on seasonal characteristics, the K-means algorithm is used to perform scene clustering analysis based on the daily load change curves of users throughout the year, resulting in four types of scene load change curves, including: Based on seasonal characteristics, the K-means algorithm is used to configure clustering analysis parameters, which include four clustering categories and multiple initial cluster center curves. The Euclidean distance between the daily load change curve of the user throughout the year and the initial cluster center curve is calculated, and the daily load change curve is classified into four cluster categories based on the Euclidean distance, resulting in four scene cluster curve sets. Each scene cluster curve set includes multiple sample curves. Calculate the average value of the sample curves in each scene cluster curve set at the same time to obtain a time mean curve, and merge the four time mean curves to obtain a new cluster center curve; Replace the initial cluster center curve with the new cluster center curve, and return to the step of calculating the Euclidean distance between the user's daily load change curve throughout the year and the initial cluster center curve, until the preset iteration condition is met, and obtain the load change curves of the four scenarios.

5. The load access optimization method based on cluster analysis according to claim 1, characterized in that, The method based on the NSGA-II algorithm, using the single-phase load phase set and the scenario load change curve, solves for the access phase of the target single-phase load user in the multi-objective optimization model, obtaining the target single-phase load access phase scheme, including: Based on the NSGA-II algorithm, the set of single-phase load phases is used as the search space, the target single-phase power load users are used as chromosomes, and the initial access phase population of the chromosomes is initialized. Calculate the node admittance matrix of the distribution network connection network generated based on the individuals of the initial access phase population; Based on the optimization space and the scenario load change curve, forward power flow calculation is performed on different time-series scenarios in the target optimization model according to the node admittance matrix to obtain the power flow calculation results. If the power flow calculation results satisfy the preset constraints, then individual optimization data are calculated according to the preset objective function; The individual optimization data are sequentially subjected to crossover and mutation operations to obtain an updated access phase population. The updated access phase population replaces the initial access phase population, and the step of calculating the node admittance matrix of the distribution network generated by the individuals based on the initial access phase population is returned until a preset number of generations is reached to obtain the target single-phase load access phase scheme.

6. A load access phase optimization device based on cluster analysis, characterized in that, include: The matrix calculation module is used to calculate the Laplace matrix based on the monitoring voltage data of different single-phase electrical load users; The matrix decomposition module is used to select three largest eigenvalues ​​from the obtained eigenvectors after performing singular value decomposition on the Laplacian matrix to construct a clustering analysis matrix. The first clustering module is used to perform clustering analysis on the access phases of single-phase power load users according to the clustering analysis matrix to obtain a set of single-phase load phases; The second clustering module is used to perform scene clustering analysis based on seasonal characteristics and the K-means algorithm according to the daily load change curve of users throughout the year, resulting in four types of scene load change curves. The model building module is used to build a multi-objective optimization model based on line loss rate and three-phase unbalanced current. The multi-objective optimization model includes power flow constraints, branch power flow inequality constraints, and grid node voltage constraints. The model solving module is used to solve the access phase of the target single-phase load user in the multi-objective optimization model based on the NSGA-II algorithm, according to the single-phase load phase set and the scenario load change curve, so as to obtain the target single-phase load access phase scheme.

7. The load access phase optimization device based on cluster analysis according to claim 6, characterized in that, Also includes: The data acquisition module is used to monitor single-phase power load users in real time through the power metering device and acquire the monitoring voltage data.

8. The load access optimization device based on cluster analysis according to claim 6, characterized in that, The matrix calculation module is specifically used for: A monitoring data matrix is ​​constructed based on the monitoring voltage data of different single-phase electricity load users; After defining the corresponding adjacency matrix based on the monitoring data matrix, the degree matrix is ​​calculated based on the adjacency matrix. The Laplace matrix is ​​obtained by calculating the degree matrix.

9. The load access phase optimization device based on cluster analysis according to claim 6, characterized in that, The second clustering module includes: The parameter configuration submodule is used to configure clustering analysis parameters based on seasonal characteristics using the K-means algorithm. The clustering analysis parameters include four clustering categories and multiple initial cluster center curves. The category assignment submodule is used to calculate the Euclidean distance between the daily load change curve of the user throughout the year and the initial cluster center curve, and to assign the daily load change curve to four cluster categories based on the Euclidean distance, thereby obtaining four scene cluster curve sets, each of which includes multiple sample curves. The update calculation submodule is used to calculate the average value of the sample curves in each scene cluster curve set at the same time to obtain a time mean curve, and merge the four time mean curves to obtain a new cluster center curve; The iterative optimization submodule is used to replace the initial cluster center curve with the new cluster center curve, trigger the category allocation submodule, until the preset iteration conditions are met, and obtain the load change curves of the four scenarios.

10. The load access phase optimization device based on cluster analysis according to claim 6, characterized in that, The model solving module includes: The initialization submodule is used to initialize the initial access phase population of the chromosome based on the NSGA-II algorithm, using the set of single-phase load phases as the search space, the target single-phase power load users as chromosomes, and the initial access phase population of the chromosomes. The admittance calculation submodule is used to calculate the node admittance matrix of the distribution network connection network generated based on the individuals of the initial access phase population; The power flow calculation submodule is used to perform forward power flow calculations on different time-series scenarios in the target optimization model based on the optimization space and the scenario load change curve, according to the node admittance matrix, to obtain the power flow calculation results. The optimization filtering submodule is used to calculate individual optimization data according to a preset objective function if the power flow calculation results meet the preset constraints. The update iteration submodule is used to perform crossover and mutation operations on the individual optimization data in sequence to obtain an updated access phase population. The updated access phase population replaces the initial access phase population, and the admittance calculation submodule is triggered until a preset number of generations is reached to obtain the target single-phase load access phase scheme.

Images