Power load clustering prediction method, prediction system and storage medium

Abstract

The invention discloses a power load clustering prediction method and system and a storage medium, and the method comprises the steps: building a load safety baseline after clustering based on historical power load data, and building a load prediction baseline based on historical power load peak value weighting; and iteratively selecting a clustering center in an area between the load safety baseline and the load prediction baseline to obtain a mean vector, performing cluster division through an Euclidean average distance between each sample point in the area between the load safety baseline and the load prediction baseline and the mean vector, and updating the mean vector according to a cluster division result to obtain the load prediction baseline. And meanwhile, the load prediction base line can be dynamically adjusted, so that an appropriate load prediction base line and a clustering center can be iteratively solved, and power load prediction can be more quickly and accurately carried out in the face of numerous and jumbled historical power load data.

Description

Technical Field

[0001] The present invention belongs to the technical field of power load forecasting and relates to a power load clustering forecasting method, a forecasting system and a storage medium. Background Art

[0002] With the in-depth construction of smart grids, existing technologies can collect and process the huge amount of data in the power system by deploying a large number of intelligent detection devices and systems, and then analyze the data to predict the power load. How to accurately and effectively predict the power load is related to whether the smart grid can operate normally.

[0003] In the prior art, it is usually done by directly monitoring the power load manually and making predictions, or by manually extracting specified load features to perform logistic regression predictions on the power load. Among them, directly monitoring the power load manually and making predictions is time-consuming and labor-intensive and relies on manual experience, and the prediction results are not accurate enough. When faced with huge amounts of power data, it is difficult to fully utilize and mine the power data for logistic regression prediction of power load, and the prediction effect is not good.

[0004] Therefore, in view of the above-mentioned problems existing in the prior art for power load prediction, the present invention discloses a power load clustering prediction method, a prediction system and a storage medium. Summary of the invention

[0005] The purpose of the present invention is to provide a power load clustering prediction method, prediction system and storage medium to solve the problems of low prediction efficiency, low prediction accuracy and unsuitability for complex data sets in the existing manual logistic regression for power load prediction, so as to realize efficient and accurate clustering prediction of power load.

[0006] The present invention is achieved through the following technical solutions:

[0007] A power load clustering prediction method is provided, which obtains historical power load data and clusters them, obtains the initial cluster center of each cluster, establishes a load safety baseline based on the initial cluster center, and calculates the load prediction baseline according to the weighted historical power load peak value; at the same time, randomly selects a new cluster center in the area between the load safety baseline and the load prediction baseline, and calculates the Euclidean average distance between the sample points in the area between the load safety baseline and the load prediction baseline and the new cluster center, repeats the above steps to obtain n cluster centers in the area between the load safety baseline and the load prediction baseline, and calculates a mean vector based on the n cluster centers; calculates the Euclidean average distance between each sample point and the mean vector in the area between the load safety baseline and the load prediction baseline, performs cluster division according to the mean vector closest to the sample point, and updates the mean vector according to the cluster division result until the cluster center converges to obtain the final load prediction baseline.

[0008] In order to better implement the present invention, further, the following steps are specifically included:

[0009] Step 1: cluster the historical power load data by using the K-means clustering method to obtain n initial cluster centers, and establish a load safety baseline based on the n initial cluster centers, so that the load safety baseline can include the n initial cluster centers;

[0010] Step 2: Obtain several historical power load peaks in time period T, calculate the average power load peak value, and establish a load forecast baseline by weighting the average power load peak value;

[0011] Step 3: randomly select a new cluster center in the area between the load safety baseline and the load forecast baseline, and calculate the Euclidean average distance between the sample points in the area between the load safety baseline and the load forecast baseline and the new cluster center, repeat the above steps to obtain n cluster centers in the area between the load safety baseline and the load forecast baseline, and calculate the mean vector based on the n cluster centers;

[0012] Step 4: Calculate the Euclidean average distance between each sample point and the mean vector in the area between the load safety baseline and the load forecast baseline, perform clustering according to the mean vector closest to the sample point, and update the mean vector according to the clustering result;

[0013] Step 5: Determine whether the cluster center converges. If the cluster center converges, stop; if the cluster center diverges, update the load forecast baseline and repeat the above steps 1 to 4 until the cluster center converges.

[0014] In order to better implement the present invention, further, the step 3 specifically includes:

[0015] Step 3.1, randomly select cluster center K0 in the area between the load safety baseline and the load forecast baseline, calculate the Euclidean average distance between the remaining sample points in the area between the load safety baseline and the load forecast baseline and the cluster center K0; define a distance baseline with cluster center K0 as the center, and select any sample point in the area within the distance baseline range as cluster center K1;

[0016] Step 3.2, calculate the Euclidean average distance between the remaining sample points in the area between the load safety baseline and the load forecast baseline and the cluster center K1, and define the distance baseline with the cluster center K1 as the center;

[0017] Step 3.3, repeat the above steps until the nth cluster center Kn is obtained;

[0018] Step 3.4: Calculate the mean vector based on n cluster centers.

[0019] In order to better implement the present invention, further, the step 4 specifically includes:

[0020] Step 4.1, taking n cluster centers as initial centers, calculating the Euclidean square distance between sample points and each mean vector in the area between the load safety baseline and the load forecast baseline;

[0021] Step 4.2: Cluster the calculated Euclidean square distance for the mean vector closest to the sample point;

[0022] Step 4.3: Update the mean vector according to the result of cluster division.

[0023] In order to better implement the present invention, further, in step 4.3, the formula for updating the mean vector is:

[0024]

[0025] in: represents the updated mean vector; |λ i | represents the sum of the mean vectors of the i-th cluster; x represents the sample points in the i-th cluster; λ i represents the mean vector of the i-th cluster.

[0026] In order to better implement the present invention, further, the step 1 specifically includes:

[0027] Step 1.1, select any cluster center as the center, and establish an initial load safety baseline with an initial radius to include all clusters;

[0028] Step 1.2, obtain the distance between the center and the edge of any cluster, and update the initial radius with the maximum distance;

[0029] Step 1.3, repeat the above steps, continuously iterate and select the center, and obtain the final load safety baseline with the smallest radius and the rule that all clusters are included.

[0030] In order to better implement the present invention, further, the step 2 specifically includes:

[0031] Step 2.1, obtaining m historical power load peaks in any time period, and calculating the power load peak average of the m historical power load peaks;

[0032] Step 2.2, based on the center finally selected in step 1, the load forecast baseline is obtained with the average value of the power load peak as the radius;

[0033] Step 2.3: If the load forecast baseline is within the load safety baseline, reselect m historical power load peaks in another time period and repeat the above steps 2.1-2.2 to update the load forecast baseline until the load safety baseline is within the load forecast baseline.

[0034] A power load clustering prediction system is used to implement a power load clustering prediction method, including a data processing module, a baseline establishment module, a clustering module, and a safety prediction module. The data processing module is used to crawl and store historical power load data, the clustering module is used to extract historical power load data from the data processing module and cluster the historical power load data, the baseline establishment module is used to establish a load safety baseline and a load prediction baseline, and the safety prediction module is used to perform a final power load prediction based on the clustering results, the load safety baseline, and the load prediction baseline.

[0035] In order to better implement the present invention, further, it also includes a data sharing module, which is used to share the data involved in the entire prediction process to power monitoring centers at all levels.

[0036] A storage medium for power load clustering prediction, wherein a computer execution program is stored in the storage medium, and when the computer execution program is executed by a computer, the computer execution program can control a power load clustering prediction system to execute a power load clustering prediction method.

[0037] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0038] The present invention establishes a load safety baseline after clustering historical power load data, establishes a load prediction baseline based on weighted historical power load peak values, and iteratively selects cluster centers in the area between the load safety baseline and the load prediction baseline to obtain a mean vector. Clustering is performed by using the Euclidean average distance between each sample point and the mean vector in the area between the load safety baseline and the load prediction baseline, and the mean vector is updated according to the clustering result. At the same time, the load prediction baseline can be dynamically adjusted, and then a suitable load prediction baseline and cluster centers are iteratively calculated, so as to ensure that power load prediction can be performed more quickly and accurately when faced with complex historical power load data. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a flowchart of the power load clustering prediction method;

[0040] Figure 2 This is the architecture diagram of the power load clustering prediction system;

[0041] Figure 3 Schematic diagram for establishing load safety baseline. DETAILED DESCRIPTION

[0042] The following detailed descriptions are all illustrative and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0043] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the present invention explicitly states otherwise, singular forms are also intended to include plural forms. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0044] For the convenience of description, if the words "up", "down", "left" and "right" appear in the present invention, they only indicate that they are consistent with the up, down, left and right directions of the drawings themselves, and do not limit the structure. They are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0045] Terminology explanation section: The terms "install", "connect", "connect", "fixed" and the like in the present invention should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integrated connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be an internal connection between two elements, or an interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances.

[0046] Embodiment 1:

[0047] A power load clustering prediction method in this embodiment obtains historical power load data and clusters it, and obtains the initial cluster center of each cluster, establishes a load safety baseline based on the initial cluster center, and calculates the load prediction baseline based on the weighted historical power load peak; at the same time, randomly selects a new cluster center in the area between the load safety baseline and the load prediction baseline, and calculates the Euclidean average distance between the sample points in the area between the load safety baseline and the load prediction baseline and the new cluster center, repeats the above steps to obtain n cluster centers in the area between the load safety baseline and the load prediction baseline, and calculates the mean vector based on the n cluster centers; calculates the Euclidean average distance between each sample point and the mean vector in the area between the load safety baseline and the load prediction baseline, performs cluster division according to the mean vector closest to the sample point, and updates the mean vector according to the cluster division result until the cluster centers converge to obtain the final load prediction baseline.

[0048] like Figure 1 As shown, the specific steps include:

[0049] Step 1: cluster the historical power load data by using the K-means clustering method to obtain n initial cluster centers, and establish a load safety baseline based on the n initial cluster centers, so that the load safety baseline can include the n initial cluster centers;

[0050] Step 2: Obtain several historical power load peaks in time period T, calculate the average power load peak value, and establish a load forecast baseline by weighting the average power load peak value;

[0051] Step 3: randomly select a new cluster center in the area between the load safety baseline and the load forecast baseline, and calculate the Euclidean average distance between the sample points in the area between the load safety baseline and the load forecast baseline and the new cluster center, repeat the above steps to obtain n cluster centers in the area between the load safety baseline and the load forecast baseline, and calculate the mean vector based on the n cluster centers;

[0052] Step 4: Calculate the Euclidean average distance between each sample point and the mean vector in the area between the load safety baseline and the load forecast baseline, perform clustering according to the mean vector closest to the sample point, and update the mean vector according to the clustering result;

[0053] Step 5: Determine whether the cluster center converges. If the cluster center converges, stop; if the cluster center diverges, update the load forecast baseline and repeat the above steps 1 to 4 until the cluster center converges.

[0054] like Figure 3 As shown, the step 1 specifically includes:

[0055] Step 1.1, select any cluster center as the center, and establish an initial load safety baseline with an initial radius to include all clusters;

[0056] Step 1.2, obtain the distance between the center and the edge of any cluster, and update the initial radius with the maximum distance;

[0057] Step 1.3, repeat the above steps, continuously iterate and select the center, and obtain the final load safety baseline with the smallest radius and the rule that all clusters are included.

[0058] The step 2 specifically includes:

[0059] Step 2.1, obtaining m historical power load peaks in any time period, and calculating the power load peak average of the m historical power load peaks;

[0060] Step 2.2, based on the center finally selected in step 1, the load forecast baseline is obtained with the average value of the power load peak as the radius;

[0061] Step 2.3: If the load forecast baseline is within the load safety baseline, reselect m historical power load peaks in another time period and repeat the above steps 2.1-2.2 to update the load forecast baseline until the load safety baseline is within the load forecast baseline.

[0062] The step 3 specifically includes:

[0063] Step 3.1, randomly select cluster center K0 in the area between the load safety baseline and the load forecast baseline, calculate the Euclidean average distance between the remaining sample points in the area between the load safety baseline and the load forecast baseline and the cluster center K0; define a distance baseline with cluster center K0 as the center, and select any sample point in the area within the distance baseline range as cluster center K1;

[0064] Step 3.2, calculate the Euclidean average distance between the remaining sample points in the area between the load safety baseline and the load forecast baseline and the cluster center K1, and define the distance baseline with the cluster center K1 as the center;

[0065] Step 3.3, repeat the above steps until the nth cluster center Kn is obtained;

[0066] Step 3.4: Calculate the mean vector based on n cluster centers.

[0067] The step 4 specifically includes:

[0068] Step 4.1, taking n cluster centers as initial centers, calculating the Euclidean square distance between sample points and each mean vector in the area between the load safety baseline and the load forecast baseline;

[0069] Step 4.2: Cluster the calculated Euclidean square distance for the mean vector closest to the sample point;

[0070] Step 4.3: Update the mean vector according to the result of cluster division.

[0071] Furthermore, in step 4.3, the formula for updating the mean vector is:

[0072]

[0073] in: represents the updated mean vector; |λ i | represents the sum of the mean vectors of the i-th cluster; x represents the sample points in the i-th cluster; λ i represents the mean vector of the i-th cluster.

[0074] Embodiment 2:

[0075] A power load clustering forecasting system, such as Figure 2 As shown, a method for implementing power load clustering prediction includes a data processing module, a baseline establishment module, a clustering module, and a safety prediction module. The data processing module is used to crawl and store historical power load data, the clustering module is used to extract historical power load data from the data processing module and cluster the historical power load data, the baseline establishment module is used to establish a load safety baseline and a load prediction baseline, and the safety prediction module is used to perform a final power load prediction based on the clustering results, the load safety baseline, and the load prediction baseline.

[0076] Furthermore, it also includes a data sharing module, which is used to share the data involved in the entire prediction process with power monitoring centers at all levels.

[0077] The other parts of this embodiment are the same as those of Embodiment 1, and thus will not be described in detail.

[0078] Embodiment 3:

[0079] A storage medium for power load clustering prediction, wherein a computer execution program is stored in the storage medium, and when the computer execution program is executed by a computer, the computer execution program can control a power load clustering prediction system to execute a power load clustering prediction method.

[0080] The other parts of this embodiment are the same as those of Embodiment 1 or 2, and thus will not be described in detail.

[0081] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in any form. Any simple modifications or equivalent changes made to the above embodiments based on the technical essence of the present invention fall within the protection scope of the present invention.

Claims

1. A power load clustering prediction method, characterized in that: Obtain historical power load data and cluster them, and obtain the initial cluster center of each cluster, establish a load safety baseline based on the initial cluster center, and calculate the load forecast baseline based on the weighted historical power load peak; at the same time, randomly select a new cluster center in the area between the load safety baseline and the load forecast baseline, and calculate the Euclidean average distance between the sample points in the area between the load safety baseline and the load forecast baseline and the new cluster center, repeat the above steps to obtain n cluster centers in the area between the load safety baseline and the load forecast baseline, and calculate the mean vector based on the n cluster centers; calculate the Euclidean average distance between each sample point and the mean vector in the area between the load safety baseline and the load forecast baseline, perform cluster division according to the mean vector closest to the sample point, and update the mean vector according to the cluster division result until the cluster center converges to obtain the final load forecast baseline.

2. A power load clustering prediction method according to claim 1, characterized in that: The specific steps include: Step 1: cluster the historical power load data by using the K-means clustering method to obtain n initial cluster centers, and establish a load safety baseline based on the n initial cluster centers, so that the load safety baseline can include the n initial cluster centers; Step 2: Obtain several historical power load peaks in time period T, calculate the average power load peak value, and establish a load forecast baseline by weighting the average power load peak value; Step 3: randomly select a new cluster center in the area between the load safety baseline and the load forecast baseline, and calculate the Euclidean average distance between the sample points in the area between the load safety baseline and the load forecast baseline and the new cluster center, repeat the above steps to obtain n cluster centers in the area between the load safety baseline and the load forecast baseline, and calculate the mean vector based on the n cluster centers; Step 4: Calculate the Euclidean average distance between each sample point and the mean vector in the area between the load safety baseline and the load forecast baseline, perform clustering according to the mean vector closest to the sample point, and update the mean vector according to the clustering result; Step 5: Determine whether the cluster center converges. If the cluster center converges, stop; if the cluster center diverges, update the load forecast baseline and repeat the above steps 1 to 4 until the cluster center converges.

3. A power load clustering prediction method according to claim 2, characterized in that: The step 3 specifically includes: Step 3.1, randomly select cluster center K0 in the area between the load safety baseline and the load forecast baseline, calculate the Euclidean average distance between the remaining sample points in the area between the load safety baseline and the load forecast baseline and the cluster center K0; define a distance baseline with cluster center K0 as the center, and select any sample point in the area within the distance baseline range as cluster center K1; Step 3.2, calculate the Euclidean average distance between the remaining sample points in the area between the load safety baseline and the load forecast baseline and the cluster center K1, and define the distance baseline with the cluster center K1 as the center; Step 3.3, repeat the above steps until the nth cluster center Kn is obtained; Step 3.4: Calculate the mean vector based on n cluster centers.

4. A power load clustering prediction method according to claim 3, characterized in that: The step 4 specifically includes: Step 4.1, taking n cluster centers as initial centers, calculating the Euclidean square distance between sample points and each mean vector in the area between the load safety baseline and the load forecast baseline; Step 4.2: Cluster the calculated Euclidean square distance for the mean vector closest to the sample point; Step 4.3: Update the mean vector according to the result of cluster division.

5. A power load clustering prediction method according to claim 4, characterized in that: In step 4.3, the formula for updating the mean vector is: in: represents the updated mean vector; |λ i | represents the sum of the mean vectors of the i-th cluster; x represents the sample points in the i-th cluster; λ i represents the mean vector of the i-th cluster.

6. A power load clustering prediction method according to claim 5, characterized in that: The step 1 specifically includes: Step 1.1, select any cluster center as the center, and establish an initial load safety baseline with an initial radius to include all clusters; Step 1.2, obtain the distance between the center and the edge of any cluster, and update the initial radius with the maximum distance; Step 1.3, repeat the above steps, continuously iterate and select the center, and obtain the final load safety baseline with the smallest radius and the rule that all clusters are included.

7. A power load clustering prediction method according to claim 6, characterized in that: The step 2 specifically includes: Step 2.1, obtaining m historical power load peaks in any time period, and calculating the power load peak average of the m historical power load peaks; Step 2.2, based on the center finally selected in step 1, the load forecast baseline is obtained with the average value of the power load peak as the radius; Step 2.3: If the load forecast baseline is within the load safety baseline, reselect m historical power load peaks in another time period and repeat the above steps 2.1-2.2 to update the load forecast baseline until the load safety baseline is within the load forecast baseline.

8. A power load clustering prediction system, used to implement the power load clustering prediction method according to any one of claims 1 to 7, characterized in that: It includes a data processing module, a baseline establishment module, a clustering module, and a safety prediction module. The data processing module is used to crawl and store historical power load data. The clustering module is used to extract historical power load data from the data processing module and cluster the historical power load data. The baseline establishment module is used to establish a load safety baseline and a load prediction baseline. The safety prediction module is used to make a final power load prediction based on the clustering results, the load safety baseline, and the load prediction baseline.

9. The power load clustering prediction system according to claim 8, characterized in that: It also includes a data sharing module, which is used to share the data involved in the entire prediction process with power monitoring centers at all levels.

10. A storage medium for clustered prediction of power loads, characterized in that: The storage medium stores a computer execution program, and when the computer execution program is executed by a computer, the computer execution program can control the power load clustering prediction system to execute the power load clustering prediction method.

Images