Load prediction method, apparatus and device for power distribution area, and medium

Abstract

The invention discloses a load prediction method, device and equipment for a power distribution area, and a medium. The method comprises the following steps: obtaining target load data before a day to be predicted; performing per-unit processing on the target load data to obtain target per-unit load data; constructing a per-unit load similar day matrix based on the target per-unit load data; clustering the target per-unit load data based on the per-unit load similar day matrix to obtain a load curve clustering result; and predicting load data of the day to be tested based on a load curve clustering result. Visibly, through per-unit processing, similarity quantification, clustering analysis and prediction of the target load data, the accuracy of load prediction of the day to be predicted can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution network, and in particular to a load prediction method, device, equipment and medium for a power distribution area. Background Art

[0002] As a power supply unit, the operation status of the distribution transformer area directly affects the power supply quality. Heavy overload operation of equipment is one of the main causes of power outages, so the load forecasting of the area is an important part of the distribution network operation and maintenance.

[0003] However, with the large-scale access of distributed photovoltaic, wind power and other power sources on the load side, the characteristics of the load side have changed significantly. The active characteristics of the load side are obvious, the uncertainty has increased, and the target load data contains a large amount of redundant information after integration, which leads to overfitting in the training process of the existing load forecasting model, reducing the accuracy of the prediction results. Summary of the invention

[0004] The present application proposes a method, device, equipment and medium for load prediction in a distribution substation area to improve the accuracy of load prediction in a distribution substation area.

[0005] In a first aspect, a load forecasting method for a distribution station area is provided, comprising:

[0006] Obtain the target load data before the test day;

[0007] Performing per-unit normalization processing on the target load data to obtain target per-unit load data;

[0008] Constructing a per-unit load similarity day matrix based on the target per-unit load data;

[0009] Clustering the target per-unit load data based on the per-unit load similarity day matrix to obtain a load curve clustering result;

[0010] The load data of the day to be tested is predicted based on the load curve clustering result.

[0011] In a second aspect, a load prediction device for a distribution station area is provided, comprising:

[0012] An acquisition module is used to acquire target load data before the day to be tested;

[0013] A processing module, used for performing per-unit processing on the target load data to obtain target per-unit load data;

[0014] A construction module, used for constructing a per-unit load similarity day matrix based on the target per-unit load data;

[0015] A clustering module, used for clustering the target per-unit load data based on the per-unit load similarity day matrix to obtain a load curve clustering result;

[0016] A prediction module is used to predict the load data of the test day based on the load curve clustering result.

[0017] Optionally, in some embodiments of the present application, the acquisition module includes:

[0018] A first acquisition submodule is used to acquire the historical load data before the test day;

[0019] A first determination submodule, configured to determine a confidence interval based on the historical load data, wherein the confidence interval is used to identify abnormal data points on the load curve of each day in the historical load data;

[0020] The cleaning submodule is used to clean the historical load data based on the confidence interval to obtain the target load data before the test day.

[0021] Optionally, in some embodiments of the present application, the first determining submodule includes:

[0022] An estimation unit, used for estimating the conditional expectation of the historical load data based on a kernel regression function;

[0023] A calculation unit, used for calculating the data standard deviation of the historical load data;

[0024] The first determining unit is used to determine the confidence interval based on the conditional expectation and the data standard deviation.

[0025] Optionally, in some embodiments of the present application, the building blocks include:

[0026] An extraction submodule, used for extracting the load data of the day before the test day from the target per unit load data;

[0027] A second acquisition submodule is used to acquire load data of the same period as the load data of the day before the day to be measured from the target per unit load data to obtain multiple sequence curves;

[0028] A calculation submodule, used for calculating the correlation coefficient between the sequence curves;

[0029] A construction submodule is used to construct the per-unit load similarity day matrix based on each of the correlation coefficients.

[0030] Optionally, in some embodiments of the present application, the clustering module includes:

[0031] An initialization submodule, used to initialize the target per unit load data to obtain an attraction degree matrix and an attribution degree matrix;

[0032] The updating submodule is used to alternately update the attraction degree matrix and the attribution degree matrix until the attraction degree and the attribution degree meet preset conditions, thereby obtaining the load curve clustering result.

[0033] Optionally, in some embodiments of the present application, the prediction module includes:

[0034] A second determination submodule is used to determine the load curve to be predicted from the load curve clustering result;

[0035] The prediction submodule is used to predict the load data of the day to be measured based on the load curve to be predicted.

[0036] Optionally, in some embodiments of the present application, the prediction submodule includes:

[0037] A prediction unit, configured to predict the initial load data of the day to be measured based on the load curve to be predicted;

[0038] A second determining unit is used to determine a minimum value and a maximum value of the daily load data of the day before the day to be measured in the target load data;

[0039] The denormalization unit is used to denormalize the initial load data based on the minimum value of the daily load data and the maximum value of the daily load data to obtain the load data of the day to be measured.

[0040] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the load prediction method for the distribution station area when executing the computer program.

[0041] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the load prediction method for the distribution station area are implemented.

[0042] The present application provides a load prediction method, device, equipment and medium for a distribution station area, which obtains target load data before a day to be measured; performs per-unit processing on the target load data to obtain target per-unit load data; constructs a per-unit load similar day matrix based on the target per-unit load data; clusters the target per-unit load data based on the per-unit load similar day matrix to obtain a load curve clustering result; and predicts the load data of the day to be measured based on the load curve clustering result. In the load forecasting scheme for the distribution area provided in the present application, the target load data is normalized and converted into a dimensionless relative value, which is conducive to eliminating the difference in load scale between different dates, making the target load data easier to compare and analyze; then, a per-unit load similar day matrix is ​​constructed based on the target per-unit load data, and the similarity between the load curves of different dates is quantified, which provides a basis for subsequent cluster analysis; then, through cluster analysis, similar load curves are classified into one category, and dates with similar load curves are identified, thereby improving the accuracy of the prediction; finally, based on the load curve clustering results, the load data of the test day is predicted, which can make full use of the load curve clustering results and improve the reliability and accuracy of the prediction. It can be seen that through the normalization, similarity quantification, cluster analysis and prediction of the target load data, the accuracy of the load prediction on the test day can be effectively improved, providing strong data support for the dispatching and planning of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 An application environment diagram of the load prediction method for a distribution station area provided in an embodiment of the present application;

[0045] Figure 2 A flow chart of a method for load prediction of a distribution station area provided in an embodiment of the present application;

[0046] Figure 3 A schematic diagram of a curve of target load data provided in an embodiment of the present application;

[0047] Figure 4 A structural block diagram of a load prediction device for a distribution station area provided in an embodiment of the present application;

[0048] Figure 5 A structural block diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] In addition, described feature, structure or characteristic can be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present application. However, those skilled in the art will appreciate that the technical scheme of the present application can be put into practice without one or more of the specific details, or other methods, components, devices, steps, etc. can be adopted. In other cases, known methods, devices, realizations or operations are not shown or described in detail to avoid blurring the various aspects of the application.

[0051] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0052] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0053] The load prediction method for a distribution station area provided by the embodiment of the present invention can be applied in the following aspects: Figure 1In the application environment. The computer device 110 communicates with the server 120 through the network 130. The computer device 110 can obtain the target load data before the test day; perform per-unit processing on the target load data to obtain the target per-unit load data; construct a per-unit load similar day matrix based on the target per-unit load data; cluster the target per-unit load data based on the per-unit load similar day matrix to obtain a load curve clustering result; predict the load data of the test day based on the load curve clustering result, and display it through the computer device 110. In the present invention, by normalizing the target load data and converting it into a dimensionless relative value, it is beneficial to eliminate the difference in load scale between different dates, making the target load data easier to compare and analyze; then, based on the target normalized load data, a normalized load similar day matrix is ​​constructed to quantify the similarity between load curves of different dates, providing a basis for subsequent cluster analysis; then, through cluster analysis, similar load curves are classified into one category, and dates with similar load curves are identified, thereby improving the accuracy of the prediction; finally, based on the load curve clustering results, the load data of the test day is predicted, which can make full use of the load curve clustering results and improve the reliability and accuracy of the prediction. It can be seen that through normalization, similarity quantification, cluster analysis and prediction of the target load data, the accuracy of the load prediction of the test day can be effectively improved, providing strong data support for the dispatching and planning of the power system. Among them, the computer device 110 can be, but not limited to, various smart phones 110-1, tablet computers 110-2 and laptop computers 110-3. The computer device 110 can also be an intelligent distribution transformer terminal. The present invention is described in detail below through specific embodiments.

[0054] See also Figure 2 As shown, Figure 2 A flow chart of a method for load prediction of a distribution area provided in an embodiment of the present invention is provided. The method can be applied to both a terminal and a server. This embodiment is illustrated by applying it to a server. The method for load prediction of a distribution area includes the following steps:

[0055] 101: Obtain the target load data before the test day.

[0056] The target load data may be a load data set used to predict the load data of the day to be tested. The load data set includes a load curve for each day within a preset time period, and each load curve may include a preset number of sampling data points.

[0057] The smart energy meter continuously collects and uploads the measured power consumption data (5min / 15min / 1h) to the smart distribution transformer terminal for aggregation within one day, i.e. 24h. For example, Figure 3As shown in the figure, we can obtain the load data set of the smart electric energy meter in a certain distribution area in Shandong Province 30 days before the test date. Figure 3 It can be seen that the load data set includes the load curve of each day in 30 days, and each load curve includes 96 data points.

[0058] In actual application scenarios, the measured historical load data is inevitably missing and distorted, so the historical load data needs to be preprocessed first. The impact on the similarity of loads when individual historical load data is missing is small, so it is possible to choose not to process the missing data. Distorted data may be caused by temporary inrush interference to the measuring device (i.e., smart point energy meter) due to unexpected power outages, protection device action, etc., resulting in normal data outliers that cannot reflect the actual electricity consumption behavior of users in the substation area, so the distorted data needs to be removed. That is, in one embodiment, the acquisition of the target load data before the day to be measured includes:

[0059] Obtaining historical load data before the day to be tested;

[0060] A confidence interval is determined based on the historical load data.

[0061] The confidence interval is used to identify abnormal data points on the daily load curve in the historical load data.

[0062] The historical load data is cleaned based on the confidence interval to obtain the target load data before the test day.

[0063] The historical load data may be load data obtained from a smart energy meter within a preset time period (e.g., 30 days) that may contain abnormal data points. The load data may include load curves for each day, and each load curve includes a preset number of data points (e.g., 96). For example, Figure 3 As shown in the figure, we can obtain the load data set of the smart electric energy meter in a certain distribution area in Shandong Province 30 days before the test date. Figure 3 It can be seen that the load data set includes the load curve of each day in 30 days, and each load curve includes 96 data points.

[0064] The average value (mean) and standard deviation of all data points of all load curves in the historical load data can be calculated; then, the confidence interval is calculated based on the average value and standard deviation; finally, if there are data points in the historical load data that fall outside the confidence interval, they are regarded as abnormal data points and are eliminated.

[0065] In one embodiment, the confidence interval can be expressed according to the following formula:

[0066]

[0067] in, represents the average value (mean), Z represents the critical value of the confidence level, n represents the number of data points included in the historical load data, and σ represents the standard deviation. The confidence level can be predetermined, for example, when the confidence level is 90%, the critical value can be 1.645; when the confidence level is 95%, the critical value can be 1.96; when the confidence level is 99%, the critical value can be 2.576.

[0068] In one embodiment, determining the confidence interval based on the historical load data includes:

[0069] estimating the conditional expectation of the historical load data based on a kernel regression function;

[0070] Calculating the data standard deviation of the historical load data;

[0071] The confidence interval is determined based on the conditional expectation and the standard deviation of the data.

[0072] The kernel regression function may be Nadaraya-Watson kernel regression (Nadaraya-Watson), which is a nonparametric method for estimating the conditional expectation of a random variable. It can be expressed according to the following formula:

[0073]

[0074] Where m(x) represents the predicted value at data point x, that is, the conditional expectation estimate of the response variable, n represents the number of data points in the historical load data, and W i (x) represents the weight of the i-th data point at x, y i Represents the response variable value of the i-th data point x, that is, the observed value of the predicted variable.

[0075] Definition: m(x) = S·y i is the linear smoothing operator of the regression function m(x), then the linear operator S is expressed as:

[0076]

[0077] The linear operator is used to calculate the value of the data point x i The distance between the target point x (the data point selected to be estimated) is used to weight y i (i.e., the value of the response variable), the linear operator S is used to transform y i The values ​​are weighted according to their proximity to the target point x to calculate the predicted value at the target point x. During the smoothing process, each y i The values ​​are assigned a certain weight according to their distance from the target point x, and the weight is determined by the data point x. iThe distance from the target point x is determined.

[0078] The weight of the i-th data point at x can be calculated according to the following formula:

[0079]

[0080] Among them, K h is a kernel function with bandwidth h, x i represents the explanatory variable value of the ith data point x, that is, the observed value of the input variable (i.e., the ith data point x) used to predict the response variable. For any kernel function K h , which can ensure that m(x) has stable consistency. The kernel function is a non-negative function symmetric about zero, which is used to calculate x and x i The similarity or distance between them. The kernel function can be Gaussian kernel, Epanechnikov kernel, uniform kernel, etc. It should be noted that

[0081] Assume that the relationship between historical load data and characteristic curve m(x) is:

[0082] y i =m(x i )+ε i

[0083] Among them, ε i is the error term of historical load data, and approaches zero, so we can get the confidence interval of m(x):

[0084]

[0085] Among them, σ is the standard deviation of the data, z α / 2 The critical value represents the confidence level, and n represents the number of data points in the historical load data. The confidence level can be predetermined or calculated based on the standard deviation of the data. For example, when the confidence level is 90%, the critical value can be 1.645; when the confidence level is 95%, the critical value can be 1.96; when the confidence level is 99%, the critical value can be 2.576.

[0086] In one embodiment, the data standard deviation can be expressed as:

[0087]

[0088] For example, when σ is 0.1, the confidence level is 1-σ=0.9, and z can be found from the preset normal distribution table. α / 2 is 1.645.

[0089] 102: Perform per-unit normalization processing on the target load data to obtain target per-unit load data.

[0090] Among them, the per-unit processing can be to convert the target load data into a dimensionless relative value form, which can eliminate the dimensional influence of the target load data, make the load curves of each day more intuitive and fair, and is conducive to analyzing the load curves at different times or conditions. The target per-unit load data can be the load data obtained after the target load data is per-unit processed.

[0091] Determine the benchmark value, which can be the average daily load (the average value of all data points within 24 hours a day), the daily maximum load, the minimum load or other representative values ​​calculated based on the target load data. Based on the benchmark value, the target load data is normalized to obtain the target normalized load data.

[0092] In one embodiment, the load curve of each day in the target load data is normalized according to the following formula:

[0093]

[0094] Where x′ is the normalized load curve, x is the data point of the load curve in the target load data, and x max is the maximum load in the load curve, x min is the minimum load in the load curve.

[0095] In one embodiment, the load curve of each day in the target load data is normalized according to the following formula:

[0096]

[0097] Among them, x' is the normalized load curve, x is each data point of the load curve in the target load data, and a is the reference value.

[0098] 103: Construct a per-unit load similarity day matrix based on the target per-unit load data.

[0099] The per-unit load similarity day matrix can be used to quantify the similarity between the per-unit load curves of each day in the target per-unit load data. Each element in the per-unit load similarity day matrix represents the similarity between the per-unit load curves of two different dates.

[0100] A similarity measurement method (such as Pearson correlation coefficient, cosine similarity, Euclidean distance, etc.) can be used to calculate the similarity between the per-unit load curves of each pair of days; thereby generating a per-unit load similarity day matrix based on each similarity. The rows and columns of the per-unit load similarity day matrix can represent dates, and each element represents the similarity between the corresponding row and column dates.

[0101] In one embodiment, constructing a per-unit load similarity day matrix based on the target per-unit load data includes:

[0102] Extracting the load data of the day before the day to be measured from the target per unit load data;

[0103] Obtaining load data of the same period as the load data of the day before the day to be measured from the target per unit load data, and obtaining a plurality of sequence curves;

[0104] Calculating the correlation coefficient between each of the sequence curves;

[0105] The per-unit load similarity day matrix is ​​constructed based on the correlation coefficients.

[0106] The load data of the day before the day to be measured represents a load curve in the target per unit load data corresponding to the predicted period of the day to be measured in the per unit load curve of the day before the day to be measured. The predicted period can be 24 hours or a specific period of a day, which is not specifically limited here.

[0107] After extracting the load data of the day before the day to be tested, the load data of the same period can be obtained based on the load data of the previous day, thereby obtaining multiple sequence curves, calculating the correlation coefficients between the sequence curves, that is, the similarity distances between the sequence curves, and constructing a per-unit load similarity matrix based on the correlation coefficients.

[0108] The correlation coefficient R between the sequence curves can be calculated according to the following formula xy :

[0109]

[0110] Among them, x i represents the data point (i.e. load value) of the i-th point in the X-th sequence curve, y i represents the data point (i.e., load value) of the i-th point in the Y-th sequence curve, n represents the number of data points in each sequence curve, represents the mean of all data points in the Xth sequence curve, represents the mean of all data points in the Yth sequence curve, R xy is the correlation coefficient between the xth day (i.e. the Xth sequence curve) and the yth day (i.e. the Yth sequence curve).

[0111] 104: Clustering the target per-unit load data based on the per-unit load similarity day matrix to obtain a load curve clustering result.

[0112] The load curve clustering result may be a clustering result indicating that per-unit load curves having similar load characteristics are grouped together.

[0113] A plurality of clusters can be obtained from the per-unit load similarity matrix through a predetermined clustering algorithm, and the corresponding target per-unit load curve in each cluster is classified into one category to obtain a load curve clustering result. The predetermined clustering algorithm can be K-means clustering, hierarchical clustering, or density-based clustering algorithm (DBSCAN).

[0114] In one embodiment, clustering the target per-unit load data based on the per-unit load similarity day matrix to obtain a load curve clustering result includes:

[0115] Initializing the target per unit load data to obtain an attraction degree matrix and an attribution degree matrix;

[0116] The attraction degree matrix and the attribution degree matrix are updated alternately until the attraction degree and the attribution degree meet preset conditions, thereby obtaining the load curve clustering result.

[0117] Among them, the attraction matrix is ​​used to measure the potential of each per-unit load curve to become a cluster center, indicating the attraction of the per-unit load curve as a per-unit load curve as a cluster center (i.e., the correlation coefficient between the two per-unit load curves). Therefore, the attraction matrix can be obtained through the per-unit load similarity day matrix. The diagonal elements of the attraction matrix are usually set to 0, because the per-unit load curve as a cluster center cannot become its own cluster center. The attribution matrix is ​​used to determine which cluster center each per-unit load curve is most likely to belong to, indicating the degree of attribution of the per-unit load curve to a certain class of per-unit load curves and the class of per-unit load curves as cluster centers.

[0118] Affinity Propagation (AP) clustering algorithm is an unsupervised clustering algorithm. The Affinity Propagation (AP) clustering algorithm can be used to cluster the target per-unit load data based on the per-unit load similarity day matrix to obtain the load curve clustering result. During the iteration of the AP algorithm, the attraction matrix will be updated according to the update of the attribution matrix, and the update of the attribution matrix will guide the per-unit load curve to select the most appropriate clustering center. The AP algorithm iteratively updates the attribution matrix and the attraction matrix until convergence or reaching the preset number of iterations, and determines the clustering center of each per-unit load curve according to the attribution matrix, thereby obtaining the load curve clustering result.

[0119] For example, assuming that the target per unit load data x for the 30 days before the test day is 24×30, where x(i,j) represents the jth data point on the i-th day, the element R in the per-unit load similarity day matrix, r(x,y) represents the correlation coefficient between the x-th day and the y-th day, and r(x,y)=r(y,x). The AP algorithm transmits two message matrices, the attraction matrix (R) and the attribution matrix (A), alternately. The attraction matrix R is composed of the attraction γ(i,k) between each per-unit load curve. The representative degree of the cluster center (i.e., the attribution update formula) can be expressed as:

[0120]

[0121] Among them, γ t+1 (i, k) represents the degree of belonging of per-unit load curve i to a certain per-unit load curve k as the cluster center in the t+1th iteration, r(i, k) is the element in the attraction matrix, which represents the attraction of per-unit load curve i to a certain per-unit load curve k, α t (i, k′) represents the degree of belonging of per-unit load curve i to a per-unit load curve k′ as the cluster center in the tth iteration, and k′ is the index of any per-unit load curve other than k, which is used to find the maximum degree of belonging in the degree update formula, that is, to find α t The largest result (i, k') + r (i, k') is taken, and r (i, k') is subtracted from it to determine the new attraction information of i relative to k. When the iteration starts, the attribution matrix A in the message matrix is ​​initially a zero matrix, that is,

[0122]

[0123] It represents the difference between the correlation coefficient between i and k and the correlation coefficient between i and other data k'. Therefore, the AP algorithm initially regards all per-unit load curves as cluster centers. In the iterative process after initialization, the negative value in the attribution matrix A will correct the correlation coefficient between the per-unit load curve i and other per-unit load curves k', thereby generating the merging and transfer of cluster centers.

[0124] 105: Predicting the load data of the test day based on the load curve clustering result.

[0125] After obtaining the load curve clustering results, it is necessary to calculate the per-unit load curves of the same class. The target class in the corresponding load curve clustering results (i.e., the class of the per-unit load curve on the same day as the day to be tested) can be determined according to the day to be tested, so as to use the preset prediction model (such as a machine learning model) to predict all per-unit load curves of the target class and obtain the load data of the day to be tested.

[0126] In one embodiment, the predicting the load data of the test day based on the load curve clustering result includes:

[0127] Determining a load curve to be predicted from the load curve clustering results;

[0128] The load data of the day to be measured is predicted based on the load curve to be predicted.

[0129] The target class in the corresponding load curve clustering result (i.e., the class of the per-unit load curve on the same day as the day to be measured) can be determined according to the day to be measured, and each per-unit curve in the target class is used as the load curve to be predicted; then, each load curve to be predicted is denormalized to obtain the decoupled load curve to be predicted; finally, a preset prediction model (such as a machine learning model) is used to predict each decoupled load curve to be predicted to obtain the load data of the day to be measured.

[0130] If the load reference value of the day to be measured is the same as that of the previous day of the same type, the load curve to be predicted can be restored using the reference value of the per unit load curve of the same type of day. For example, the load curves to be predicted can be denormalized to be expressed as:

[0131]

[0132] Where x′ represents the load value of the data point of the load curve to be predicted after denormalization, x represents the load per unit value of the data point of the load curve to be predicted, and x min is the minimum value of the load curve to be predicted, x max is the maximum value of the load curve to be predicted.

[0133] In one embodiment, predicting the load data of the day to be measured based on the load curve to be predicted includes:

[0134] Predicting the initial load data of the day to be tested based on the load curve to be predicted;

[0135] Determine the minimum value and the maximum value of the daily load data of the day before the day to be measured in the target load data;

[0136] The initial load data is denormalized based on the minimum value of the daily load data and the maximum value of the daily load data to obtain the load data of the day to be measured.

[0137] The initial load data represents the predicted per unit load curve of the test day.

[0138] Assuming that the load pattern within 24 hours will have a necessary connection with the load pattern of the next day, the load data of the day to be tested can be predicted by determining the peak value of the load curve of the day before the day to be tested (that is, the maximum and minimum values ​​of the daily load data).

[0139] For example, taking the load data of the day to be tested for 24 hours as an example, take the load data of the day to be tested Y 24×m The matrix Z composed of the load curves to be predicted for 24 hours before m-1 data (m-1)×24 , where m represents the number of days before the test day, the rows of the matrix Z represent the data points of each hour of the load curve to be predicted for each day, and the columns represent the data points of the same period from 1 to m-1 on the day before the test day. The matrix Z can be expressed as:

[0140]

[0141] Vector V is the per-unit load curve for 24 hours on the day before the test date and matrix Y 24×m , the correlation coefficient between the other per-unit load curves, vector V can be expressed as:

[0142] V=[r1,r2,…r m-1 ]

[0143] In the formula, taking r1 as an example, r1 represents the correlation coefficient between the 24-hour per unit load curve of the day before the test day and the 24-hour per unit load curve of the two days before the test day, and so on. Based on this, the matrix Z can be weighted based on the size of the correlation coefficient, and the initial load data within 24 hours of the test day can be predicted according to the following formula:

[0144] F=V×Z / sum(V)

[0145] Where F represents the initial load data, sum(V) represents the sum of all elements in vector V [r1, r2, …r m-1 ] is used to normalize the weights to ensure the rationality of the prediction.

[0146] Denormalizing the initial load data based on the minimum value of the daily load data and the maximum value of the daily load data can be expressed as:

[0147]

[0148] Where x′ represents the load value of the data point of the initial load data after denormalization, x represents the load value of the data point of the initial load data, and x min is the minimum value of the daily load data of the day before the test day, x max It is the maximum value of the daily load data of the day before the day to be tested.

[0149] In this application, the historical load data stored (such as 30 days) in the intelligent distribution transformer terminal of the distribution substation is collected as sample data (i.e., target load data), and the sample data is cleaned by the kernel smoothing method. Because the sample data is small, it is considered that after the sample is normalized, it is split into a per-unit curve and a base value to calculate and improve the accuracy of the prediction result. Then, the correlation coefficient matrix of the historical load data is constructed by the correlation coefficient method, and the per-unit curve of similar days is obtained by clustering after replacing the affine propagation similarity matrix with the correlation coefficient matrix, and then the base value of the day to be measured is obtained by weighted summation, and finally the per-unit curve of the day to be measured is obtained after the per-unit curve and the base value are denormalized. The load of the distribution substation area can be quickly predicted to avoid the occurrence of power outages, and solve the problem that load overload often occurs in the peak tourist season in some areas, resulting in frequent power outages.

[0150] The above is the load forecasting process for the distribution station area of ​​this application.

[0151] As described above, the present application provides a load prediction method, device, equipment and storage medium for a distribution station area, which obtains the target load data before the day to be measured; normalizes the target load data to obtain the target per-unit load data; constructs a per-unit load similar day matrix based on the target per-unit load data; clusters the target per-unit load data based on the per-unit load similar day matrix to obtain a load curve clustering result; and predicts the load data of the day to be measured based on the load curve clustering result. In the load forecasting scheme for the distribution area provided in the present application, the target load data is normalized and converted into a dimensionless relative value, which is conducive to eliminating the difference in load scale between different dates, making the target load data easier to compare and analyze; then, a per-unit load similar day matrix is ​​constructed based on the target per-unit load data, and the similarity between the load curves of different dates is quantified, which provides a basis for subsequent cluster analysis; then, through cluster analysis, similar load curves are classified into one category, and dates with similar load curves are identified, thereby improving the accuracy of the prediction; finally, based on the load curve clustering results, the load data of the test day is predicted, which can make full use of the load curve clustering results and improve the reliability and accuracy of the prediction. It can be seen that through the normalization, similarity quantification, cluster analysis and prediction of the target load data, the accuracy of the load prediction on the test day can be effectively improved, providing strong data support for the dispatching and planning of the power system.

[0152] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0153] In one embodiment, a load prediction device for a power distribution area is provided, and the load prediction device for the power distribution area corresponds to the load prediction method for the power distribution area in the above embodiment. Figure 4 As shown, the load prediction device of the distribution substation area includes:

[0154] An acquisition module 201 is used to acquire target load data before the test day;

[0155] The processing module 202 is used to perform per-unit processing on the target load data to obtain target per-unit load data;

[0156] A construction module 203 is used to construct a per-unit load similarity day matrix based on the target per-unit load data;

[0157] A clustering module 204 is used to cluster the target per-unit load data based on the per-unit load similarity day matrix to obtain a load curve clustering result;

[0158] The prediction module 205 is used to predict the load data of the test day based on the load curve clustering result.

[0159] In this embodiment, the target load data before the day to be measured is obtained; the target load data is normalized to obtain the target normalized load data; a normalized load similar day matrix is ​​constructed based on the target normalized load data; the target normalized load data is clustered based on the normalized load similar day matrix to obtain the load curve clustering result; and the load data of the day to be measured is predicted based on the load curve clustering result. In the load prediction scheme for the distribution station area provided in the present application, the target load data is normalized to obtain a dimensionless relative value, which is conducive to eliminating the difference in load scale between different dates and making the target load data easier to compare and analyze; then, a normalized load similar day matrix is ​​constructed based on the target normalized load data, which quantifies the similarity between load curves of different dates and provides a basis for subsequent clustering analysis; then, through clustering analysis, similar load curves are classified into one category, and dates with similar load curves are identified, thereby improving the accuracy of the prediction; finally, the load data of the day to be measured is predicted based on the load curve clustering result, which can make full use of the load curve clustering result and improve the reliability and accuracy of the prediction. It can be seen that through the standardization, similarity quantification, cluster analysis and prediction of target load data, the accuracy of load forecasting on the test day can be effectively improved, providing strong data support for the dispatching and planning of the power system.

[0160] Optionally, in some embodiments of the present application, the acquisition module includes:

[0161] A first acquisition submodule is used to acquire the historical load data before the test day;

[0162] A first determination submodule, configured to determine a confidence interval based on the historical load data, wherein the confidence interval is used to identify abnormal data points on the load curve of each day in the historical load data;

[0163] The cleaning submodule is used to clean the historical load data based on the confidence interval to obtain the target load data before the test day.

[0164] Optionally, in some embodiments of the present application, the first determining submodule includes:

[0165] An estimation unit, used for estimating the conditional expectation of the historical load data based on a kernel regression function;

[0166] A calculation unit, used for calculating the data standard deviation of the historical load data;

[0167] The first determining unit is used to determine the confidence interval based on the conditional expectation and the data standard deviation.

[0168] Optionally, in some embodiments of the present application, the building blocks include:

[0169] An extraction submodule, used for extracting the load data of the day before the test day from the target per unit load data;

[0170] A second acquisition submodule is used to acquire load data of the same period as the load data of the day before the day to be measured from the target per unit load data to obtain multiple sequence curves;

[0171] A calculation submodule, used for calculating the correlation coefficient between the sequence curves;

[0172] A construction submodule is used to construct the per-unit load similarity day matrix based on each of the correlation coefficients.

[0173] Optionally, in some embodiments of the present application, the clustering module includes:

[0174] An initialization submodule, used to initialize the target per unit load data to obtain an attraction degree matrix and an attribution degree matrix;

[0175] The updating submodule is used to alternately update the attraction degree matrix and the attribution degree matrix until the attraction degree and the attribution degree meet preset conditions, thereby obtaining the load curve clustering result.

[0176] Optionally, in some embodiments of the present application, the prediction module includes:

[0177] A second determination submodule is used to determine the load curve to be predicted from the load curve clustering result;

[0178] The prediction submodule is used to predict the load data of the day to be measured based on the load curve to be predicted.

[0179] Optionally, in some embodiments of the present application, the prediction submodule includes:

[0180] A prediction unit, configured to predict the initial load data of the day to be measured based on the load curve to be predicted;

[0181] A second determining unit is used to determine a minimum value and a maximum value of the daily load data of the day before the day to be measured in the target load data;

[0182] The denormalization unit is used to denormalize the initial load data based on the minimum value of the daily load data and the maximum value of the daily load data to obtain the load data of the day to be measured.

[0183] In one embodiment, a computer device is provided, the internal structure diagram of which can be as follows: Figure 5 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server through a network connection. When the computer program is executed by the processor, the functions or steps of a load prediction method for a distribution station area are implemented.

[0184] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following steps are implemented:

[0185] Obtain the target load data before the day to be tested; normalize the target load data to obtain the target per-unit load data; construct a per-unit load similar day matrix based on the target per-unit load data; cluster the target per-unit load data based on the per-unit load similar day matrix to obtain the load curve clustering result; predict the load data of the day to be tested based on the load curve clustering result.

[0186] In this embodiment, by normalizing the target load data and converting it into a dimensionless relative value, it is helpful to eliminate the difference in load scale between different dates, making the target load data easier to compare and analyze; then, based on the target normalized load data, a normalized load similar day matrix is ​​constructed to quantify the similarity between load curves of different dates, providing a basis for subsequent cluster analysis; then, through cluster analysis, similar load curves are classified into one category, and dates with similar load curves are identified, thereby improving the accuracy of the prediction; finally, based on the load curve clustering results, the load data of the test day is predicted, which can make full use of the load curve clustering results and improve the reliability and accuracy of the prediction. It can be seen that through normalization, similarity quantification, cluster analysis and prediction of the target load data, the accuracy of the load prediction of the test day can be effectively improved, providing strong data support for the dispatching and planning of the power system.

[0187] In one embodiment, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0188] Obtain the target load data before the day to be tested; normalize the target load data to obtain the target per-unit load data; construct a per-unit load similar day matrix based on the target per-unit load data; cluster the target per-unit load data based on the per-unit load similar day matrix to obtain the load curve clustering result; predict the load data of the day to be tested based on the load curve clustering result.

[0189] In this embodiment, by normalizing the target load data and converting it into a dimensionless relative value, it is beneficial to eliminate the difference in load scale between different dates, making the target load data easier to compare and analyze; then, based on the target normalized load data, a normalized load similar day matrix is ​​constructed to quantify the similarity between load curves of different dates, providing a basis for subsequent cluster analysis; then, through cluster analysis, similar load curves are classified into one category, and dates with similar load curves are identified, thereby improving the accuracy of the prediction; finally, based on the load curve clustering results, the load data of the test day is predicted, which can make full use of the load curve clustering results and improve the reliability and accuracy of the prediction. It can be seen that through the normalization of the target load data, similarity quantification, cluster analysis and prediction, the accuracy of the load prediction of the test day can be effectively improved, providing strong data support for the dispatching and planning of the power system. It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device can be referred to the relevant descriptions of the server side and the client side in the aforementioned method embodiment. To avoid repetition, they are not described one by one here.

[0190] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0191] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0192] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A method for load prediction in a distribution station area, characterized in that: include: Obtain target load data before the day to be tested; Performing per-unit normalization processing on the target load data to obtain target per-unit load data; Constructing a per-unit load similarity day matrix based on the target per-unit load data; Clustering the target per-unit load data based on the per-unit load similarity day matrix to obtain a load curve clustering result; The load data of the day to be tested is predicted based on the load curve clustering result.

2. The load prediction method for a distribution station area according to claim 1, characterized in that: The obtaining of target load data before the test day includes: Obtaining historical load data before the day to be tested; Determining a confidence interval based on the historical load data, wherein the confidence interval is used to identify abnormal data points on the load curve of each day in the historical load data; The historical load data is cleaned based on the confidence interval to obtain the target load data before the test day.

3. The load prediction method for a distribution station area according to claim 2, characterized in that: The determining of the confidence interval based on the historical load data comprises: estimating the conditional expectation of the historical load data based on a kernel regression function; Calculating the data standard deviation of the historical load data; The confidence interval is determined based on the conditional expectation and the standard deviation of the data.

4. The load prediction method for a distribution station area according to claim 1, characterized in that: The step of constructing a per-unit load similarity day matrix based on the target per-unit load data includes: Extracting the load data of the day before the day to be measured from the target per unit load data; Obtaining load data of the same period as the load data of the day before the day to be measured from the target per unit load data, and obtaining a plurality of sequence curves; Calculating the correlation coefficient between each of the sequence curves; The per-unit load similarity day matrix is ​​constructed based on the correlation coefficients.

5. The load prediction method for a distribution station area according to claim 1, characterized in that: The clustering of the target per-unit load data based on the per-unit load similarity day matrix to obtain a load curve clustering result includes: Initializing the target per unit load data to obtain an attraction degree matrix and an attribution degree matrix; The attraction degree matrix and the attribution degree matrix are updated alternately until the attraction degree and the attribution degree meet preset conditions, thereby obtaining the load curve clustering result.

6. The load prediction method for a distribution station area according to claim 1, characterized in that: The predicting the load data of the test day based on the load curve clustering result includes: Determining a load curve to be predicted from the load curve clustering results; The load data of the day to be measured is predicted based on the load curve to be predicted.

7. The load prediction method for a distribution station area according to claim 6, characterized in that: The step of predicting the load data of the day to be measured based on the load curve to be predicted includes: Predicting the initial load data of the day to be tested based on the load curve to be predicted; Determine the minimum value and the maximum value of the daily load data of the day before the day to be measured in the target load data; The initial load data is denormalized based on the minimum value of the daily load data and the maximum value of the daily load data to obtain the load data of the day to be measured.

8. A load prediction device for a distribution station area, characterized in that: include: An acquisition module is used to acquire target load data before the day to be tested; A processing module, used for performing per-unit processing on the target load data to obtain target per-unit load data; A construction module, used for constructing a per-unit load similarity day matrix based on the target per-unit load data; A clustering module, used for clustering the target per-unit load data based on the per-unit load similarity day matrix to obtain a load curve clustering result; A prediction module is used to predict the load data of the test day based on the load curve clustering result.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the load prediction method for the distribution station area according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the load prediction method for a distribution station area according to any one of claims 1 to 7 are implemented.

Images