Method for clustering of dynamic zones of a main transformer

Abstract

The application provides a clustering method for dynamic partition of main transformer, comprising: obtaining historical power data of 500kV main transformer of a power grid, and dividing the historical data into a first sample and a second sample; obtaining a distance of historical power change value and a distance of historical power change direction between the main transformers according to the first sample; constructing a first distance function through the distance of historical power change value and the distance of historical power change direction; obtaining a distance threshold of hierarchical clustering according to the first distance function; performing hierarchical clustering on the second sample according to the distance threshold; adjusting the clustering result; and performing partition aggregation on the 500kV main transformer of the power grid according to the adjusted clustering result, and the main transformers in the same partition are considered to supply power to the same 220kV area. The application can utilize real-time historical main transformer power data to perform dynamic clustering analysis, and can dynamically master the 220kV power supply partition mode and change.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric power, in particular to a clustering method for dynamic partition of main transformer. BACKGROUND

[0002] Real-time control of dynamic partition operation of power grid is a major measure to ensure the safe operation and reliable power supply of important large cities, and also helps the future power grid planning to move towards digitalization and intelligentization.

[0003] The existing power grid 500kV transformer capacity expansion planning method is to obtain a 500kV transformer station capacity expansion planning scheme that meets the safety and stability requirements of the power grid according to the boundary conditions of long-term regional load size and distribution prediction, power supply planning, network structure, etc., through power supply balance analysis and related electrical calculation of regional power grid typical operation mode, comprehensive economic and technical comparison, to guide the implementation of new or expanded transformer engineering.

[0004] However, the existing power grid planning method does not consider the dynamic partition change of 500kV main transformer power supply to 220kV area, and cannot timely grasp the power supply complementarity between multiple main transformers in the same partition according to the change trend of 500kV main transformer power supply to 220kV area, resulting in waste of power grid investment planning. SUMMARY

[0005] The purpose of the present application is to provide a clustering method for dynamic partition of main transformer, which can use real-time historical main transformer power data for dynamic clustering analysis, so as to understand the main transformer power supply to 220kV power grid in a shorter time, dynamically grasp the 220kV power supply partition mode and change, and timely adjust the power grid planning project implementation progress according to the change trend, to assist the scientific planning and precise investment of power grid.

[0006] In order to achieve the above purpose, the present application provides a clustering method for dynamic partition of main transformer, comprising:

[0007] Obtain historical power data of 500kV main transformer of power grid, and divide the historical data into first sample and second sample;

[0008] According to the first sample, the distance of historical power change value between 500kV main transformers and the distance of historical power change direction between 500kV main transformers are obtained;

[0009] The distance weighted combination is constructed as a first distance function by the distance of historical power change value and the distance of historical power change direction;

[0010] According to the first distance function, the distance threshold of hierarchical clustering is obtained;

[0011] According to the distance threshold, the second sample is subjected to hierarchical clustering.

[0012] adjusting the clustering result; and

[0013] partitioning and aggregating the 500kV transformers according to the adjusted clustering result, and the 500kV transformers in the same partition are considered to supply power to the same 220kV area.

[0014] Optionally, in the clustering method for dynamic partitioning of the transformers, the historical power data comprises 96 power values of the power grid in one day.

[0015] Optionally, in the clustering method for dynamic partitioning of the transformers, the historical power data is obtained by sampling the historical power curve.

[0016] Optionally, in the clustering method for dynamic partitioning of the transformers, the method for obtaining the distance of the historical power change value between the 500kV transformers and the distance of the historical power change direction between the 500kV transformers according to the first sample comprises:

[0017] performing maximum-minimum normalization on the historical power data of the first sample to obtain the power change value of each 500kV transformer;

[0018] determining the adjacent historical power change direction according to the positive or negative of the difference between the adjacent historical power data;

[0019] obtaining the Euclidean distance of the power change value according to the power change value of each transformer, as the distance of the historical power change value between the 500kV transformers; and

[0020] obtaining the Euclidean distance of the historical power change direction according to the adjacent historical power change direction, as the distance of the historical power change direction between the 500kV transformers.

[0021] Optionally, in the clustering method for dynamic partitioning of the transformers, the method for determining the adjacent historical power change direction according to the positive or negative of the difference between the adjacent historical power data comprises:

[0022] when the difference between the next value and the previous value is positive, the direction from the previous value to the next value is considered to be positive, and the value of the adjacent historical power change direction is assigned as 1;

[0023] when the difference between the next value and the previous value is negative, the direction from the previous value to the next value is considered to be negative, and the value of the adjacent historical power change direction is assigned as -1; and

[0024] when the difference between the next value and the previous value is zero, the value of the adjacent historical power change direction is assigned as 0.

[0025] Optionally, in the clustering method of dynamic partition of main transformer, the method of constructing distance weighted combination as the first distance function by the distance of historical power change value and the distance of historical power change direction comprises:

[0026] dist((s1, d1), (s2, d2)) = w1 x dist1(s1, s2) + w2 x dist2(d1, d2);

[0027] wherein w1>0, w2>0, w1+w2=1, dist((s1, d1), (s2, d2)) is the first distance function, dist1(s1, s2) is the distance of historical power change value, dist2(d1, d2) is the distance of historical power change direction; ω1 is the weight of the distance of historical power change value, ω2 is the weight of the distance of historical power change direction.

[0028] Optionally, in the clustering method of dynamic partition of main transformer, the method of obtaining the distance threshold of hierarchical clustering according to the first distance function comprises:

[0029] forming multiple weights of the distance of historical power change value between (0, 0) and (0, 1) with a step of 0.01, and forming multiple weights of the distance of historical power change direction between (0, 0) and (1, 0) with a step of 0.01;

[0030] arbitrarily combining the weight of the distance of historical power change value and the weight of the distance of historical power change direction under the condition that the sum of the weight of the distance of historical power change value and the weight of the distance of historical power change direction is 1; and

[0031] calculating the first distance according to the combination and the first distance function, and when the value of the first distance is closest to the set value, the first distance at this time is the threshold distance.

[0032] Optionally, in the clustering method of dynamic partition of main transformer, the method of performing hierarchical clustering on the second sample according to the distance threshold comprises:

[0033] obtaining the distance of historical power change value between 500kV main transformers and the distance of historical power change direction between main transformers according to the second sample;

[0034] obtaining the second distance by substituting the distance of historical power change value and the distance of historical power change direction into the first distance function; and

[0035] performing hierarchical clustering on the second sample according to the distance threshold and the second distance.

[0036] Optionally, in the clustering method of dynamic partition of main transformer, the method of hierarchical clustering of the second sample according to the distance threshold and the second distance comprises:

[0037] When the difference between the second distance and the set value is within the distance threshold, the 500kV main transformer corresponding to the historical power data is classified into the same class.

[0038] Optionally, in the clustering method of dynamic partition of main transformer, the method of adjusting the clustering result comprises:

[0039] If a clustering group is composed of at least five 500kV main transformer partitions, the clustering group is split and reduced in dimension.

[0040] In the clustering method of dynamic partition of main transformer provided by the application, real-time historical main transformer power data can be used for dynamic clustering analysis, so that the power supply of the main transformer to the 220kV power grid can be understood in a shorter time, the 220kV power supply partition mode and the change can be dynamically mastered, and the 500kV main transformer dynamic partition aggregation relationship can be mastered, the power supply complementarity between multiple main transformers in the same partition is considered, the power grid planning capacity expansion scale and project implementation schedule are adjusted, the scientific planning and precise investment of the power grid are assisted, and economic management and social benefits are created through innovation. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 is a flowchart of the clustering method of dynamic partition of main transformer according to an embodiment of the application. DETAILED DESCRIPTION

[0042] The specific embodiments of the application will be described in more detail below with reference to the accompanying drawings. The advantages and features of the application will be clearer according to the following description. It should be noted that the drawings are very simplified and use non-precise proportions, and are only used to facilitate and clearly assist the purpose of describing the embodiments of the application.

[0043] In the following, the terms "first", "second", etc. are used to distinguish between similar elements, and do not necessarily describe a specific order or time sequence. It should be understood that these terms as used can be replaced under appropriate circumstances. Similarly, if the method described herein includes a series of steps, and the order of the steps presented herein is not necessarily the only order in which the steps can be performed, and some of the steps described can be omitted and / or some other steps not described herein can be added to the method.

[0044] Please refer to Figure 1 The application provides a clustering method of dynamic partition of main transformer, comprising:

[0045] S11: Obtain historical power data of 500kV grid main transformers, and divide the historical data into a first sample and a second sample;

[0046] S12: Obtain a distance of historical power change values between the 500kV main transformers and a distance of historical power change directions between the 500kV main transformers according to the first sample;

[0047] S13: Construct a distance weighted combination as a first distance function through the distance of historical power change values and the distance of historical power change directions;

[0048] S14: Obtain a distance threshold of hierarchical clustering according to the first distance function;

[0049] S15: Perform hierarchical clustering on the second sample according to the distance threshold;

[0050] S16: Adjust the clustering result; and

[0051] S17: Perform partition aggregation on the 500kV main transformers of the grid according to the adjusted clustering result, and consider that the 500kV main transformers in the same partition supply power to the same 220kV area.

[0052] In the embodiment of the application, the historical power data includes 96 power values of the grid in one day. The historical power data can be obtained by sampling the historical power curve. For example, taking daily data of a day as an example, first, the power curve of the 500kV main transformer of the studied regional grid is sampled, the sampling frequency can be 15 minutes, and the power change of adjacent time points is obtained by subtracting 96 data of 24 hours in a day, s represents the difference value of the power values of adjacent time points, d represents the change direction of the power curve of adjacent time points, 1 represents increase, and -1 represents decrease.

[0053] Preferably, the method for obtaining the distance of historical power change values between the 500kV main transformers and the distance of historical power change directions between the 500kV main transformers according to the first sample comprises: performing maximum-minimum normalization processing on the historical power data of the first sample to obtain the power change value of each main transformer; determining the adjacent historical power change direction according to the positive and negative of the difference value of the adjacent historical power data; obtaining the Euclidean distance of the power change value as the distance of historical power change values between the 500kV main transformers according to the power change value of each 500kV main transformer; and obtaining the Euclidean distance of the historical power change direction as the distance of historical power change directions between the 500kV main transformers according to the adjacent historical power change direction.

[0054] Preferably, the method for determining the direction of the adjacent historical power change according to the positive or negative of the difference between the adjacent historical power data comprises: when the difference between the next value and the previous value is positive, it is considered that the direction from the previous value to the next value is positive, and at this time the value of the direction of the adjacent historical power change is assigned as 1; when the difference between the next value and the previous value is negative, it is considered that the direction from the previous value to the next value is negative, and at this time the value of the direction of the adjacent historical power change is assigned as -1; when the difference between the next value and the previous value is zero, at this time the value of the direction of the adjacent historical power change is assigned as 0.

[0055] Preferably, the method for constructing the norm weighted combination of the distance of the historical power change value and the distance of the direction of the historical power change as the first distance function comprises:

[0056] dist((s1, d1), (s2, d2)) = w1 x dist1(s1, s2) + w2 x dist2(d1, d2);

[0057] wherein w1>0, w2>0, w1+w2=1, dist((s1, d1), (s2, d2)) is the first distance function, dist1(s1, s2) is the distance of the historical power change value, dist2(d1, d2) is the distance of the direction of the historical power change; ω1 is the weight of the distance of the historical power change value, and ω2 is the weight of the distance of the direction of the historical power change.

[0058] Preferably, the method for obtaining the distance threshold of hierarchical clustering according to the first distance function comprises: forming a plurality of weights of the distance of the historical power change value between (0, 0) and (0, 1) with a step of 0.01, and simultaneously forming a plurality of weights of the distance of the direction of the historical power change between (0, 0) and (1, 0) with a step of 0.01; under the condition that the sum of the weight of the distance of the historical power change value and the weight of the distance of the direction of the historical power change is 1, the weight of the distance of the historical power change value and the weight of the distance of the direction of the historical power change are combined arbitrarily, that is, the following weight combinations (w1=0, w2=1), (w1=0.01, w2=0.99), …, (w1=1, w2=0) are formed; and, the first distance is calculated according to the combination and the first distance function, and when the value of the first distance is closest to the set value, the first distance at this time is the threshold distance.

[0059] Preferably, the method for performing hierarchical clustering on the second sample according to the distance threshold comprises: obtaining the distance of the historical power change value between the main variables and the distance of the historical power change direction between the 500kV main variables according to the second sample; obtaining the second distance by substituting the distance of the historical power change value and the distance of the direction of the historical power change of the second sample into the first distance function; and performing hierarchical clustering on the second sample according to the distance threshold and the second distance.

[0060] The method of hierarchical clustering of the second samples according to the distance threshold and the second distance comprises: when the difference between the second distance and the set value is within the distance threshold, the 500kV main transformer corresponding to the historical power data is classified into the same class.

[0061] Preferably, the method of adjusting the clustering result comprises: if a clustering group (500kV main transformer) is composed of at least five 500kV main transformers, the components of the clustering are split and reduced in dimension. If a clustering result is composed of N i ≥5 elements, the components of the clustering are split and reduced in dimension. For example, for a class composed of five elements, assuming that the class is composed of two small classes, one small class is composed of three elements, and the other small class is composed of two elements, the structure of the dimension reduction adjustment is to divide the class into two classes, one class has four elements, and the other class has three elements.

[0062] In summary, in the clustering method of the main transformer dynamic partition provided in the embodiment of the present application, real-time historical main transformer power data can be used for dynamic clustering analysis, so that the power supply of the 500kV main transformer to the 220kV power grid can be understood in a shorter time, the 220kV power supply partition mode and the change can be dynamically mastered, and through mastering the dynamic partition aggregation relationship of the 500kV main transformer, the power supply complementarity between multiple main transformers in the same partition can be considered, the power grid planning capacity expansion scale and the project implementation progress can be adjusted, and the scientific planning and accurate investment of the power grid can be assisted.

[0063] The above is only the preferred embodiment of the present application, and does not limit the present application in any way. Any person skilled in the art can make any form of equivalent replacement, modification or change to the technical solutions and technical contents disclosed in the present application without departing from the scope of the technical solutions of the present application, which still belongs to the protection scope of the present application.

Claims

1. A clustering method for dynamic partitioning of primary variables, characterized in that, include: Historical power data of the 500kV main transformer of the power grid is obtained, and the historical power data is divided into a first sample and a second sample. The distance between the historical power change values ​​and the distance between the historical power change directions of the 500kV main transformers are obtained based on the first sample. A distance-weighted combination is constructed using the distance between the historical power change values ​​and the distance between the historical power change directions as a first distance function; The distance threshold for hierarchical clustering is obtained based on the first distance function; The second sample is hierarchically clustered based on the distance threshold; Adjust the clustering results; as well as Based on the adjusted clustering results, the 500kV main transformers of the power grid are clustered by region, and the 500kV main transformers in the same region are considered to supply power to the same 220kV area.

2. The clustering method for dynamic partitioning of the main transformer as described in claim 1, characterized in that, The historical power data includes 96 points of grid power values ​​within a single day.

3. The clustering method for dynamic partitioning of the main transformer as described in claim 2, characterized in that, Historical power data is obtained by sampling historical power curves.

4. The clustering method for dynamic partitioning of the main transformer as described in claim 2, characterized in that, The method for determining the distance between the historical power change values ​​and the distance in the direction of the historical power change between the 500kV main transformers based on the first sample includes: The historical power data of the first sample is normalized by maximum and minimum values ​​to obtain the power change value of each 500kV main transformer. The direction of change in adjacent historical power data is determined by the sign of the difference between them. The Euclidean distance of the power change values ​​is obtained based on the power change values ​​of each of the 500kV main transformers, and is used as the distance between the historical power change values ​​of the 500kV main transformers; and The Euclidean distance of the historical power change direction is obtained based on the adjacent historical power change direction, and is used as the distance between the historical power change directions of the 500kV main transformer.

5. The clustering method for dynamic partitioning of the main transformer as described in claim 4, characterized in that, Methods for determining the direction of adjacent historical power changes based on the sign of the difference between adjacent historical power data include: When the difference between the next value and the previous value is positive, the direction from the previous value to the next value is considered positive, and the value of the adjacent historical power change direction is assigned as 1. When the difference between the next value and the previous value is negative, the direction from the previous value to the next value is considered negative, and the value of the adjacent historical power change direction is assigned as -1; and When the difference between the next value and the previous value is zero, the value of the adjacent historical power change direction is assigned to 0.

6. The clustering method for dynamic partitioning of the main transformer as described in claim 5, characterized in that, The method for constructing a distance-weighted combination as a first distance function using the distance between the historical power change values ​​and the distance between the historical power change directions includes: ； in, , Let be the first distance function. The distance to the historical power change value. The distance in the direction of historical power change; The weighting of the distance to historical power change values. The weight of the distance in the direction of historical power change.

7. The clustering method for dynamic partitioning of the main transformer as described in claim 6, characterized in that, The method for obtaining the distance threshold for hierarchical clustering based on the first distance function includes: Using a step size of 0.01 and The weights are calculated based on the distances between multiple historical power change values, and the values ​​are adjusted in steps of 0.

01. and The weights of the distances between the directions of multiple historical power changes; Under the condition that the sum of the weights of the distance to the historical power change value and the distance to the historical power change direction is 1, any combination of the weights of the distance to the historical power change value and the distance to the historical power change direction can be made; and The first distance is calculated based on the combination and the first distance function. When the value of the first distance is closest to the set value, the first distance is the threshold distance.

8. The clustering method for dynamic partitioning of the main transformer as described in claim 1, characterized in that, The method for hierarchical clustering of the second sample based on the distance threshold includes: The distance between the historical power change values ​​and the distance between the historical power change directions of the 500kV main transformers are obtained based on the second sample. The second distance is obtained by substituting the distance of the historical power change value and the distance of the historical power change direction into the first distance function; and The second sample is hierarchically clustered based on the distance threshold and the second distance.

9. The clustering method for dynamic partitioning of the main transformer as described in claim 8, characterized in that, The method for hierarchical clustering of the second sample based on the distance threshold and the second distance includes: When the difference between the second distance and the set value is within the distance threshold, the 500kV main transformer corresponding to the historical power data is classified into the same category.

10. The clustering method for dynamic partitioning of the main transformer as described in claim 1, characterized in that, Methods for adjusting clustering results include: If a cluster component consists of at least five 500kV main transformers, then the components of that cluster are split and dimensionality reduced.

Images