A power distribution network operation scenario extraction method and device

Abstract

The application provides a power distribution network operation scene extraction method and device, which can be used in photovoltaic power generation or other fields. The method comprises the following steps: taking a predetermined initial cluster number as an iterative cluster number, performing an iterative operation by increasing the iterative cluster number to obtain a clustering result under each iterative cluster number and an iterative loss generated thereby, and stopping the operation until the iterative cluster number reaches a preset maximum cluster number; calculating an initial cluster center, wherein the number of the initial cluster center is equal to the iterative cluster number; calculating the clustering result and the iterative loss generated thereby according to the initial cluster center; calculating a clustering loss under the corresponding iterative cluster number according to the iterative loss generated by the clustering result under each iterative cluster number; and selecting the clustering result corresponding to the smallest clustering loss as a scene extraction result.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of photovoltaic power generation, in particular to a power distribution network operation scene extraction method and device. BACKGROUND

[0002] The penetration rate of distributed photovoltaic in the power distribution network increases year by year. Classifying and extracting typical operation scenes of the power distribution network containing distributed photovoltaic is an important basis for developing operation and planning of the power distribution network containing high penetration rate of distributed photovoltaic. Due to the uncertainty of the distributed photovoltaic power generation, the complexity of extracting the operation scene of the power distribution network containing distributed photovoltaic is increased.

[0003] In the prior art, the typical operation scene of the power distribution network containing distributed photovoltaic is mainly obtained by clustering. Common clustering methods include k-means clustering, fuzzy C-means clustering, hierarchical clustering, etc. However, there are mainly three problems in the prior art when clustering: first, it is difficult to select a suitable initial clustering center for the traditional k-means clustering method; second, when calculating the similarity between the sample point and the clustering center using the Euclidean distance, only the distribution characteristics of the curve are considered, and the overall dynamic characteristics of the curve are not considered; third, the correlation between the distributed photovoltaic power generation and the load consumption power is not considered. SUMMARY

[0004] In view of the problems in the prior art, the embodiments of the present application provide a power distribution network operation scene extraction method and device, which can at least partially solve the problems in the prior art.

[0005] In a first aspect, the present application provides a power distribution network operation scene extraction method, comprising:

[0006] Taking a predetermined initial clustering number as an iterative clustering number, performing an iterative operation by increasing the iterative clustering number to obtain a clustering result under each iterative clustering number and an iterative loss generated thereby, until the iterative clustering number reaches a preset maximum clustering number: calculating an initial clustering center, the number of the initial clustering center being equal to the iterative clustering number; calculating the clustering result and the iterative loss generated thereby according to the initial clustering center;

[0007] Calculating the clustering loss under the corresponding iterative clustering number according to the iterative loss generated by the clustering result under each iterative clustering number;

[0008] Selecting the clustering result corresponding to the smallest clustering loss as the scene extraction result.

[0009] The calculation of the initial clustering center comprises:

[0010] Randomly selecting a point in the pre-established light-load time sequence sample point as a first initial clustering center;

[0011] selecting the rest of the initial clustering centers from the light-load time sequence sample points according to the first initial clustering center.

[0012] wherein the selecting the rest of the initial clustering centers from the light-load time sequence sample points according to the first initial clustering center comprises:

[0013] calculating a comprehensive similarity distance of the non-initial clustering center light-load time sequence sample point to each initial clustering center;

[0014] calculating a probability of the light-load time sequence sample point being selected as a next clustering center according to the comprehensive similarity distance, the probability being proportional to a square of a minimum comprehensive similarity distance in the comprehensive similarity distances of the light-load time sequence sample point to each initial clustering center;

[0015] selecting an initial clustering center from the non-initial clustering center light-load time sequence sample points according to the probability;

[0016] iteratively performing the above operations until the number of selected initial clustering centers equals the iterative clustering number.

[0017] wherein the calculating a comprehensive similarity distance of the non-initial clustering center light-load time sequence sample point to each initial clustering center comprises:

[0018] calculating a load dynamic time warping distance of the light-load time sequence sample point to each initial clustering center according to a load component of the light-load time sequence sample point and a load component of each initial clustering center;

[0019] calculating a photovoltaic dynamic time warping distance of the light-load time sequence sample point to each initial clustering center according to a photovoltaic component of the light-load time sequence sample point and a photovoltaic component of each initial clustering center;

[0020] calculating a light-load correlation distance of the light-load time sequence sample point to each initial clustering center according to a coordinate of the light-load time sequence sample point and a coordinate of each initial clustering center;

[0021] calculating the comprehensive similarity distance according to a preset weight and the load dynamic time warping distance, the photovoltaic dynamic time warping distance and the light-load correlation distance.

[0022] wherein the calculating a clustering result and an iterative loss generated according to the initial clustering center comprises:

[0023] calculating a comprehensive similarity distance of the non-clustering center light-load time sequence sample point to each clustering center;

[0024] select a classification in which a cluster center with a minimum comprehensive similarity distance with the light-load time sequence sample point is located as a classification of the light-load time sequence sample point, to obtain a clustering result of this iteration;

[0025] calculate a coordinate mean of the light-load time sequence sample points in each classification as a coordinate of a new cluster center of the classification;

[0026] calculate an iteration loss according to the coordinates of the light-load time sequence sample points in each classification and the coordinates of the new cluster center;

[0027] perform the above operation iteratively until the iteration loss is less than a preset iteration loss threshold.

[0028] Before performing the iteration operation, the method further includes:

[0029] obtain a photovoltaic measured power and a load measured power, and perform normalization and standardization processing on the photovoltaic measured power and the load measured power;

[0030] obtain coordinates of light-load time sequence sample points according to the processed photovoltaic measured power and load measured power, and each coordinate of the light-load time sequence sample points comprises a photovoltaic component and a load component.

[0031] In a second aspect, the present application provides a power distribution network operation scenario extraction device, comprising:

[0032] an iteration clustering unit, configured to take a predetermined initial cluster number as an iteration cluster number, perform an iteration operation by increasing the iteration cluster number to obtain clustering results under each iteration cluster number and iteration losses generated thereby, until the iteration cluster number reaches a preset maximum cluster number; the iteration clustering unit comprises: an initial cluster center acquisition module, configured to calculate initial cluster centers, the number of the initial cluster centers being equal to the iteration cluster number; and a result loss calculation module, configured to calculate clustering results and iteration losses generated thereby according to the initial cluster centers;

[0033] a clustering loss calculation unit, configured to calculate a clustering loss under a corresponding iteration cluster number according to an iteration loss generated by a clustering result under each iteration cluster number;

[0034] a scenario extraction result acquisition unit, configured to select a clustering result corresponding to a minimum clustering loss as a scenario extraction result.

[0035] In a third aspect, the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the power distribution network operation scenario extraction method of any one of the above-mentioned embodiments when executing the computer program.

[0036] In a fourth aspect, the present application provides a computer readable storage medium storing a computer program, which, when executed by a processor, implements the power distribution network operation scenario extraction method of any of the above embodiments.

[0037] In a fifth aspect, the present application provides a computer program product comprising a computer program, which, when executed by a processor, implements the power distribution network operation scenario extraction method of any of the above embodiments.

[0038] The power distribution network operation scenario extraction method and device provided by the embodiments of the present application can take the pre-determined initial clustering number as the iterative clustering number, perform an iterative operation by increasing the iterative clustering number to obtain the clustering results under each iterative clustering number and the generated iterative loss until the iterative clustering number reaches the pre-set maximum clustering number, calculate the initial clustering center, the number of the initial clustering center being equal to the iterative clustering number, calculate the clustering results and the generated iterative loss according to the initial clustering center, calculate the clustering loss under the corresponding iterative clustering number according to the iterative loss generated by the clustering results under each iterative clustering number, and select the clustering result corresponding to the smallest clustering loss as the scenario extraction result, so that the initial clustering center is automatically calculated and selected, and the overall dynamic characteristics of the curve and the correlation between the photovoltaic power generation power and the load consumption power are fully considered in clustering. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0040] Figure 1 is a flowchart of the power distribution network operation scenario extraction method provided by an embodiment of the present application.

[0041] Figure 2 is a flowchart of calculating the initial clustering center provided by an embodiment of the present application.

[0042] Figure 3 is a flowchart of selecting the remaining initial clustering centers from the remaining sample points according to the first initial clustering center provided by an embodiment of the present application.

[0043] Figure 4 is a flowchart of calculating the comprehensive similarity distance of the non-initial clustering center to each initial clustering center provided by an embodiment of the present application.

[0044] Figure 5is a flowchart of the method for calculating a clustering result and an iteration loss generated according to an initial clustering center provided by an embodiment of the present application.

[0045] Figure 6 is a flowchart of the method for extracting a power distribution network operation scenario provided by an embodiment of the present application.

[0046] Figure 7 is a flowchart of the method for extracting a power distribution network operation scenario provided by an embodiment of the present application.

[0047] Figure 8 is a structural schematic diagram of the device for extracting a power grid operation scenario provided by an embodiment of the present application.

[0048] Figure 9 is a structural schematic diagram of the device for extracting a power grid operation scenario provided by an embodiment of the present application.

[0049] Figure 10 is a structural schematic diagram of the device for extracting a power grid operation scenario provided by an embodiment of the present application.

[0050] Figure 11 is a structural schematic diagram of the device for extracting a power grid operation scenario provided by an embodiment of the present application.

[0051] Figure 12 is a structural schematic diagram of the device for extracting a power grid operation scenario provided by an embodiment of the present application.

[0052] Figure 13 is a structural schematic diagram of the device for extracting a power grid operation scenario provided by an embodiment of the present application.

[0053] Figure 14 is a structural schematic diagram of the device for extracting a power grid operation scenario provided by an embodiment of the present application. DETAILED DESCRIPTION

[0054] To make the objectives, technical solutions and advantages of embodiments of the present application clearer, further detailed descriptions of the embodiments of the present application are given below with reference to the drawings. Here, the illustrative embodiments of the present application and their descriptions are used to explain the present application, but are not used as limitations of the present application. It should be noted that, in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other at will.

[0055] The specific implementation process of the method for extracting a power distribution network operation scenario provided by an embodiment of the present application is described below with the server as an execution subject.

[0056] Figure 1 is a flowchart of the method for extracting a power distribution network operation scenario provided by an embodiment of the present application, as shown in Figure 1 The method for extracting a power distribution network operation scenario provided by an embodiment of the present application includes:

[0057] S101: taking a predetermined initial cluster number as an iterative cluster number;

[0058] Specifically, the initial cluster number can be any positive integer greater than or equal to 2, which is an empirical data and can be determined according to the range of historical classification numbers.

[0059] S102: calculating initial cluster centers, the number of initial cluster centers being equal to the iterative cluster number;

[0060] Specifically, the initial cluster centers are determined one by one until the number of initial cluster centers is equal to the iterative cluster number, wherein each initial cluster center is determined according to the initial cluster centers selected before the initial cluster center is selected, so as to avoid the situation that some initial cluster centers are too concentrated due to completely random selection.

[0061] S103: calculating a clustering result and an iterative loss generated according to the initial cluster centers;

[0062] Specifically, the first clustering result is calculated according to the selected initial cluster centers, and the cluster centers are updated according to each clustering result, and the next clustering result is calculated according to the updated cluster centers. The iterative loss of the current clustering result is calculated each time a clustering result is obtained, until the iterative loss is less than a preset iterative loss threshold. The clustering result corresponding to the iterative loss is taken as the clustering result under the iterative cluster number, wherein the iterative loss threshold is an empirical value.

[0063] S104: increasing the iterative cluster number and determining whether the iterative cluster number reaches a preset maximum cluster number;

[0064] Specifically, the iterative cluster number is increased, and it is determined whether the iterative cluster number reaches a preset maximum cluster number. If the iterative cluster number does not reach the preset maximum cluster number, S102 is entered. If the iterative cluster number reaches the preset maximum cluster number, S105 is entered. The increase of the iterative cluster number each time can be set according to the actual situation. In the case of requiring high accuracy, the iterative cluster number can be increased by one each time, and the clustering under each iterative cluster number is calculated to select the optimal clustering result. In the case of not requiring high accuracy, a larger increase of the iterative cluster number can be selected to select a better result to save the operation time and operation resources. The maximum cluster number is an empirical data and can be determined according to the range of historical classification numbers.

[0065] S105: calculating a clustering loss under the corresponding iterative cluster number according to the iterative loss generated by the clustering result under each iterative cluster number;

[0066] Specifically, the average iterative loss of each light-charge time-series sample point in the clustering results under each iteration number of clustering is taken as the clustering loss under the corresponding number of iterations, which is the average distance from each light-charge time-series sample point to the corresponding cluster center in the clustering results under each iteration number of clustering. This loss can be calculated according to the following formula, where n is the number of sample points without cluster centers, and I SSE Here, f represents the clustering loss for that number of iterations, and f is the iterative loss generated by the clustering results for that number of iterations.

[0067]

[0068] S106: Select the clustering result corresponding to the minimum clustering loss as the scene extraction result.

[0069] Specifically, clustering loss represents the difference between the clustering result and the real situation. The larger the clustering loss, the greater the difference between the clustering result and the real situation. In order to obtain results that are closer to the real situation, the clustering result corresponding to the minimum clustering loss is selected as the scene extraction result. Each cluster center in the clustering result is the classic scene of the corresponding classification.

[0070] The distribution network operation scenario extraction method provided in this embodiment of the invention can calculate the clustering results and the generated iteration loss under each iteration clustering number, calculate the clustering loss under the corresponding iteration clustering number based on the iteration loss under each iteration clustering number, and select the clustering result corresponding to the minimum clustering loss as the scenario extraction result. This realizes the automatic calculation of probability to select a suitable initial clustering center, and solves the problem that traditional clustering algorithms have difficulty in selecting a suitable initial clustering center.

[0071] Figure 2 This is a flowchart of calculating the initial cluster centers provided in an embodiment of the present invention, as follows: Figure 2 As shown, based on the above embodiments, the initial cluster centers are further calculated, including:

[0072] S201: Randomly select a point from the pre-established light-charge time series sample points as the first initial cluster center;

[0073] Specifically, a point is randomly selected from the pre-established light-load time series sample points, and this point serves as the first initial cluster center. The light-load time series sample points include load components and photovoltaic components. The load component of the m-th light-load time series sample point is represented as p. ld,m =[p ld,m,1 ,p ld,m,2 ,…,p ld,m,n ], where p ld,m,i The load component at the m-th photoelectric-load time series sample point is represented by the i-th power value; the photovoltaic component at the m-th photoelectric-load time series sample point is represented by p. pv,m =[ppv,m,1 ,p pv,m,2 ,…,p pv,m,n ], where p pv,m,i This represents the i-th power value of the photovoltaic component at the m-th light-load time series sample point.

[0074] S202: Select the remaining initial cluster centers from the light-load time series sample points based on the first initial cluster center.

[0075] Specifically, the next initial cluster center is selected based on the first initial cluster center, and the next initial cluster center is selected based on the already selected initial cluster centers, until the number of initial cluster centers reaches the number of iterations of this clustering.

[0076] The distribution network operation scenario extraction method provided in this embodiment of the invention can randomly select a point as the first initial cluster center from the pre-established photovoltaic-load time series sample points; and select other initial cluster centers from the photovoltaic-load time series sample points based on the first initial cluster center. This realizes the automatic calculation of probability to select a suitable initial cluster center, and fully considers the overall dynamic characteristics of the curve and the correlation between photovoltaic power generation and load power consumption during clustering.

[0077] Figure 3 This is a flowchart provided by an embodiment of the present invention for selecting remaining initial cluster centers from the remaining sample points based on the first initial cluster center, as shown below. Figure 3 As shown, based on the above embodiments, further, selecting other initial cluster centers from the remaining sample points according to the first initial cluster center includes:

[0078] S301: Calculate the comprehensive similarity distance from the light-charge time series sample points of non-initial cluster centers to each initial cluster center;

[0079] Specifically, the comprehensive similarity distance between each sample point that is not currently selected as an initial cluster center and each currently selected initial cluster center is calculated.

[0080] S302: Calculate the probability that a light-charge time series sample point is selected as the next cluster center based on the comprehensive similarity distance. The probability is proportional to the square of the minimum comprehensive similarity distance among the comprehensive similarity distances from the light-charge time series sample point to each initial cluster center.

[0081] Specifically, since clustering is essentially classifying sample data, one cluster center corresponds to one classification, and the closer two points are, the higher the probability that they are in the same classification, therefore, the sample point farthest from the selected initial cluster center should be selected as the next cluster center as much as possible. Meanwhile, in order to avoid falling into a local optimal solution, instead of directly selecting the point with the maximum comprehensive similarity distance from the selected initial cluster center with the minimum comprehensive similarity distance as the next cluster center, the probability of each sample point being selected is set according to the above comprehensive similarity distance, and the probability is proportional to the square of the minimum comprehensive similarity distance of the comprehensive similarity distances of the light-charge time sequence sample point to each initial cluster center, so that the probability of the sample point with the maximum minimum comprehensive similarity distance being selected as the next initial cluster center is greater.

[0082] S303: selecting one initial cluster center from the light-charge time sequence sample points of the non-initial cluster center according to the probability randomly;

[0083] Specifically, unlike the traditional clustering algorithm that selects the cluster center with equal probability, the probability of each sample point being selected is set according to the calculated probability, and one sample point is randomly selected as the initial cluster center, so that a more suitable cluster center can be selected, and the clustering result greatly different from the real situation caused by the unsuitable selection of the cluster center can be avoided.

[0084] S304: determining whether the number of the selected initial cluster centers is equal to the iterative clustering number.

[0085] Specifically, one initial cluster center is selected each time, and it is determined whether the number of the selected initial cluster centers is equal to the iterative clustering number, if the number of the selected initial cluster centers is not equal to the iterative clustering number, the loop enters S301, and the next initial cluster center is continuously selected until the number of the selected initial cluster centers is equal to the iterative clustering number.

[0086] The power distribution network operation scenario extraction method provided by the embodiment of the application can calculate the comprehensive similarity distances of the light-charge time sequence sample points of the non-initial cluster center to each initial cluster center, calculate the probability of the light-charge time sequence sample point being selected as the next cluster center according to the comprehensive similarity distance, and select one initial cluster center from the light-charge time sequence sample points of the non-initial cluster center according to the probability randomly, so that the probability of selecting a suitable initial cluster center is automatically calculated, and the problem that the traditional clustering algorithm is difficult to select a suitable initial cluster center is solved.

[0087] Figure 4 is a flowchart of calculating the comprehensive similarity distances of the light-charge time sequence sample points of the non-initial cluster center to each initial cluster center provided by the embodiment of the application, as shown in Figure 4Further, based on the above embodiments, the comprehensive similarity distance of the non-initial clustering center's light-load time sequence sample point to each initial clustering center is calculated, including:

[0088] S401: Calculate the load dynamic time warping distance of the light-load time sequence sample point to each initial clustering center according to the load component of the light-load time sequence sample point and the load component of each initial clustering center;

[0089] Specifically, due to the difference of seasons, weather, etc. of the load work, the load component of the light-load time sequence sample point and the load component of the initial clustering center may have different dynamic characteristics. If the Euclidean distance between the above load components is directly calculated, the result may be greater than the actual difference between the two load components. Therefore, the overall dynamic characteristics of the two components should be fully considered, and the load dynamic time warping distance is used to represent the difference between the load component of the light-load time sequence sample point and the load component of the initial clustering center. The load dynamic time warping distance can be calculated by the DTW (Dynamic Time Warping) algorithm, and is expressed as H1(D i ,D c,j ), wherein D i is the i-th sample point, and D c,j is the j-th clustering center.

[0090] For example, the load of the sample point is the load classification detected in summer, and the load is concentrated in the work between six o'clock in the morning and nine o'clock in the evening. The load component of the initial clustering center is the load classification detected in winter, and the load is concentrated in the work between seven o'clock in the morning and ten o'clock in the evening. If the difference between the two load components at the same time is calculated, it will be greater than the actual difference. The overall dynamic characteristics of the two load works should be fully considered, the working time of the two loads is corresponded, and the load dynamic time warping distance between the two load components is calculated.

[0091] S402: Calculate the photovoltaic dynamic time warping distance of the light-load time sequence sample point to each initial clustering center according to the photovoltaic component of the light-load time sequence sample point and the photovoltaic component of each initial clustering center;

[0092] Specifically, due to the difference of seasons, weather, etc., the time when the distributed photovoltaic receives light every day may also be different. Therefore, the photovoltaic dynamic time warping distance is more accurate to represent the difference between the photovoltaic component of the light-load time sequence sample point and the photovoltaic component of the initial clustering center. The photovoltaic dynamic time warping distance can be calculated by the DTW algorithm, and is expressed as H2(D i ,D c,j ), wherein D i is the i-th sample point, and D c,j is the j-th clustering center.

[0093] S403: Calculate the light-load correlation distance between the light-load time sequence sample point and each initial clustering center according to the coordinates of the light-load time sequence sample point and the coordinates of each initial clustering center;

[0094] Specifically, the coordinates of the light-load time sequence sample point include a load component and a photovoltaic component, the coordinates of the initial clustering center also include a load component and a photovoltaic component, and the light-load correlation distance between the light-load time sequence sample point and the initial clustering center can be calculated by the following formula, wherein H3(D i ,D c,j ) is the light-load correlation distance, D i is the i-th sample point, D c,j is the j-th clustering center, and D c,j is the clustering center of the cluster where D i is located:

[0095]

[0096] S404: Calculate the comprehensive similarity distance according to the preset weight and the load dynamic time warping distance, the photovoltaic dynamic time warping distance and the light-load correlation distance.

[0097] Specifically, the comprehensive similarity distance is calculated according to the sum of the product of the load dynamic time warping distance, the photovoltaic dynamic time warping distance and the light-load correlation distance and the corresponding weight, and the weight is an empirical value, which can be expressed by the following formula, wherein H i,j (D i ,D c,j ) is the comprehensive similarity distance, P1 is the weight of the load dynamic time warping distance, P2 is the weight of the photovoltaic dynamic time warping distance, and P3 is the weight of the light-load correlation distance:

[0098] H i,j (D i ,D c,j )=P1H1(D i ,D c,j )+P2H2(D i ,D c,j )+P3H3(D i ,D c,j ) (3)

[0099] The power distribution network operation scenario extraction method provided by the embodiment can calculate the load dynamic time warping distance, the photovoltaic dynamic time warping distance and the light-load correlation distance between the light-load time sequence sample point and the initial clustering center, and calculate the comprehensive similarity distance between the light-load time sequence sample point and the initial clustering center according to each distance and the weight corresponding to the distance, and fully consider the overall dynamic characteristics of the curve and the correlation between the photovoltaic power generation power and the load consumption power when clustering.

[0100] Figure 5 is a flowchart of calculating the clustering result and the iteration loss generated according to the initial clustering center provided by an embodiment of the present application, as shown in the above embodiments, further, the clustering result and the iteration loss generated according to the initial clustering center, comprising: Figure 5

[0101] S501: calculating the comprehensive similarity distance of the non-clustering center light-charge time sequence sample point to each clustering center;

[0102] Specifically, the comprehensive similarity distance of the non-clustering center light-charge time sequence sample point to each clustering center is calculated, and the calculation method of the comprehensive similarity distance is the same as that given in the above embodiment. Figure 4

[0103] S502: selecting the classification of the clustering center where the comprehensive similarity distance of the light-charge time sequence sample point is the smallest as the classification of the light-charge time sequence sample point, and obtaining the clustering result of this iteration;

[0104] Specifically, the smaller the comprehensive similarity distance is, the smaller the difference between the light-charge time sequence sample point and the corresponding clustering center is, and the greater the probability of belonging to the same classification is, therefore, the classification of the clustering center where the comprehensive similarity distance of the light-charge time sequence sample point is the smallest should be selected as the classification of the light-charge time sequence sample point, and the corresponding classification is selected for all non-clustering center light-charge time sequence sample points, and the clustering result of this iteration is obtained.

[0105] S503: calculating the coordinate mean of the light-charge time sequence sample points in each classification as the coordinate of the new clustering center of the classification;

[0106] Specifically, after obtaining the clustering result of this iteration, the current clustering center of each classification may not be the point with the smallest distance sum to all light-charge time sequence sample points in the current classification region, therefore, the coordinate mean of the light-charge time sequence sample points in each classification is calculated as the coordinate of the new clustering center of the classification.

[0107] S504: calculating the iteration loss according to the coordinates of the light-charge time sequence sample points in each classification and the coordinates of the new clustering center;

[0108] Specifically, the iteration loss is calculated according to the coordinates of the light-charge time sequence sample points in each classification and the coordinates of the new clustering center, the comprehensive similarity distance of the light-charge time sequence sample point in each classification to the new clustering center of the corresponding classification is calculated, the sum of the comprehensive similarity distances of all light-charge time sequence sample points in all classifications to the new clustering center of the corresponding classification is taken as the iteration loss of this iteration, which can be represented by the following formula, wherein, f x is the iteration loss of the xth iteration, D j is the jth light-charge time sequence sample point, D​​c,i is the new cluster center of the i-th classification in the x-th iteration.

[0109]

[0110] S505: judging whether the iteration loss is less than a preset iteration loss threshold.

[0111] Specifically, if the iteration loss is less than the preset iteration loss threshold, the operation is completed; if the iteration loss is not less than the preset iteration loss threshold, the cycle enters S501, the clustering result is updated, and the iteration loss is received again until the iteration loss is less than the preset iteration loss threshold.

[0112] The power distribution network operation scenario extraction method provided by the embodiment of the application can calculate the comprehensive similarity distance of the non-cluster center light-load time sequence sample point to each cluster center, classify the non-cluster center light-load time sequence sample point, and update each classification cluster center, so that the overall dynamic characteristics of the curve and the correlation between the photovoltaic power generation power and the load consumption power are fully considered in the clustering.

[0113] Figure 6 is a flowchart of the power distribution network operation scenario extraction method provided by an embodiment of the application, as shown in Figure 6 Based on the above embodiments, further, the power distribution network operation scenario extraction method provided by the application further comprises:

[0114] S601: obtaining photovoltaic measured power and load measured power, and performing normalization and standardization processing on the photovoltaic measured power and the load measured power;

[0115] Specifically, the photovoltaic measured power includes photovoltaic power measured at multiple time periods of each day of the power grid, and the load measured power includes load power measured at multiple time periods of each day of the power grid, the photovoltaic measured power is divided by the annual photovoltaic power peak value to obtain normalized photovoltaic measured power, and the load measured power is divided by the annual load power peak value to obtain normalized load measured power; any standardization method can be selected as needed to standardize the normalized photovoltaic measured power and the normalized load measured power, such as min-mix standardization, z-score standardization, etc.

[0116] S602: obtaining coordinates of light-load time sequence sample points according to the processed photovoltaic measured power and load measured power, each light-load time sequence sample point including a photovoltaic component and a load component.

[0117] Specifically, the light-load time sequence sample point includes a load component and a photovoltaic component, the load component and the photovoltaic component in one light-load time sequence sample point include multiple load measured powers and photovoltaic measured powers of a day respectively, and the load component of the m-th light-load time sequence sample point is represented as pm. ld,m=[p ld,m,1 ,p ld,m,2 ,…,p ld,m,n ], where p ld,m,i p represents the measured power of the i-th load on day m; the photovoltaic component of the m-th photovoltaic-load time series sample point is represented as p. pv,m =[p pv,m,1 ,p pv,m,2 ,…,p pv,m,n ], where p pv,m,i This represents the measured photovoltaic power of the i-th photovoltaic power on the m-th day; the coordinates of the m-th photovoltaic load time series sample point can be represented as D. m =[p pv,m ,p ld,m ],Right now

[0118]

[0119] The distribution network operation scenario extraction method provided in this embodiment of the invention can obtain the measured photovoltaic power and the measured load power, and perform per-unit and standardization processing on the measured photovoltaic power and the measured load power; obtain photovoltaic-load time series sample points based on the processed measured photovoltaic power and measured load power, each photovoltaic-load time series sample point contains photovoltaic component and load component, and fully considers the overall dynamic characteristics of the curve and the correlation between photovoltaic power generation and load power consumption during clustering.

[0120] Figure 7 This is a flowchart of a distribution network operation scenario extraction method provided in an embodiment of the present invention, as follows: Figure 7 As shown, the specific implementation process of the power distribution network operation scenario extraction method provided by the present invention will now be described.

[0121] S701: Obtain the measured photovoltaic power and the measured load power, and perform per-unit and normalization processing on the measured photovoltaic power and the measured load power;

[0122] Specifically, assuming that five measured photovoltaic (PV) powers were obtained on day m, namely 2, 3, 1, 1, and 3, and the annual peak PV power was 5, the measured PV power on that day was normalized to obtain the normalized PV power: 0.4, 0.6, 0.2, 0.2, and 0.6. The normalized PV power was then standardized to obtain the standardized PV power. In this embodiment, min-max standardization was selected: 0.5, 1, 0, 0, and 1.

[0123] S702: The coordinates of the photovoltaic-load time series sample points are obtained based on the processed photovoltaic measured power and load measured power. The coordinates of each photovoltaic-load time series sample point include the photovoltaic component and the load component.

[0124] Specifically, the photovoltaic measured power after the mth day processing is: 0.5, 1, 0, 0, 1, and thus the photovoltaic component of the mth day is [0.5, 1, 0, 0, 1], assuming that the load measured power after the mth day processing is: 0, 0.6, 0.4, 1, 0.4, then the load component of the mth day is [0, 0.6, 0.4, 1, 0.4], and the coordinate of the mth light-load time sequence sample point is:

[0125]

[0126] S703: taking the predetermined initial clustering number as the iterative clustering number;

[0127] Specifically, k c represents the iterative clustering number, and in the embodiment, the initial clustering number is 2, and thus k c = 2.

[0128] S704: randomly selecting a point in the pre-established light-load time sequence sample point as a first initial clustering center;

[0129] Specifically, a point is randomly selected in the pre-established light-load time sequence sample point, and the point is the first initial clustering center, assuming that the coordinate of the point is:

[0130]

[0131] S705: calculating the load dynamic time warping distance between the light-load time sequence sample point and each initial clustering center according to the load component of the light-load time sequence sample point and the load component of each initial clustering center;

[0132] Specifically, the comprehensive similarity distance of the light-load time sequence sample point to each initial clustering center is calculated by the DTW algorithm, and the basic idea of the DTW algorithm is: first, pairing each numerical point in the vector, then calculating the square of the difference between each paired numerical point, summing the square of the difference between all paired points and taking the square root to obtain the dynamic time warping distance, and when pairing, the following needs to be met: a. the first numerical value of the two vectors is paired with each other, and the last numerical value of the two vectors is paired with each other; b. all numerical values must be paired, assuming that the two numerical points currently paired are p i and q i , and p i and q i are not the last numerical points in the vector, then the next pair of numerical points must be p i+1 and q i , p i and q i+1 , and p i+1 and q i+1The two points with the minimum absolute value of the difference. Taking the load component [0, 0.6, 0.4, 1, 0.4] of the mth light-load time sequence sample point and the load component [0, 0.6, 1, 0.4, 0.4] of the first initial clustering center as an example, the load dynamic time warping distance is: (0-0) 2 +(0.6-0.6) 2 +(0.4-0.6) 2 +(1-1) 2 +(0.4-0.4) 2 +(0.4-0.4) 2 =0.04. H1(D i ,D c,j ) represents the light-load time sequence sample point and the distance center load dynamic time warping distance, wherein D i is the ith sample point, and D c,j is the jth clustering center, then H1(D m ,D c,1 )=0.04.

[0133] S706: Calculate the light-load time sequence sample point and each initial clustering center photovoltaic dynamic time warping distance according to the photovoltaic component of the light-load time sequence sample point and the photovoltaic component of each initial clustering center;

[0134] Specifically, also taking the load component [0.5, 1, 0, 0, 1] of the mth light-load time sequence sample point and the load component [0, 0.7, 1, 0.3, 0] of the first initial clustering center as an example, the photovoltaic dynamic time warping distance is: (0.5-0) 2 +(0.5-0.7) 2 +(1-1) 2 +(0-0.3) 2 +(0-0) 2 +(1-0) 2 =1.78. H2(D i ,D c,j ) represents the light-load time sequence sample point and the distance center photovoltaic dynamic time warping distance, wherein D i is the ith sample point, and D c,j is the jth clustering center, then H2(D m ,D c,1 )=1.78.

[0135] S707: Calculate the light-load time sequence sample point and each initial clustering center light-load correlation distance according to the coordinates of the light-load time sequence sample point and the coordinates of each initial clustering center;

[0136] Specifically, taking the coordinates of the mth light-charge time series sample point and the coordinates of the load component of the first initial clustering center as an example, the light-charge correlation distance is: [(0.5-0)-(0.5-0)] 2 +[(1-0.6)-(0.7-0.6)] 2 +[(0-0.4)-(1-1)] 2 +[(0-1)-(0.3-0.4)] 2 +[(1-0.4)-(0-0.4)] 2 =2.06. H3(D i ,D c,j ) represents the light-charge correlation distance between the light-charge time series sample point and the distance center, where D i is the ith sample point, D c,j is the jth clustering center, then H3(D m ,D c,1 )=2.06.

[0137] S708: Calculate the comprehensive similarity distance according to the preset weight and the load dynamic time warping distance, the photovoltaic dynamic time warping distance and the light-charge correlation distance;

[0138] Specifically, H i,c (D i ,D c,j ) represents the comprehensive similarity distance, where P1 is the weight of the load dynamic time warping distance, P2 is the weight of the photovoltaic dynamic time warping distance, and P3 is the weight of the light-charge correlation distance. It is assumed that P1=0.4, P2=0.4, and P3=0.2, then H m,1 (D m ,D c,1 )=P1H1(D m ,D c,1 )+P2H2(D m ,D c,1 )+P3H3(D m ,D c,1 )=1.14.

[0139] S709: Calculate the probability that the light-charge time series sample point is selected as the next initial clustering center according to the comprehensive similarity distance, which is proportional to the square of the minimum comprehensive similarity distance of the comprehensive similarity distance of the light-charge time series sample point to each initial clustering center;

[0140] Specifically, taking the square sum of the comprehensive similarity distance of all non-cluster center light-charge time sequence sample points to the first initial cluster center as 10 as an example, since there is only one cluster center, the first cluster center, the minimum comprehensive similarity distance of the light-charge time sequence sample points to each initial cluster center is the comprehensive similarity distance to the first cluster center. Since the comprehensive similarity distance of the mth light-charge time sequence sample point to the first cluster center is 1.14, the probability of the mth light-charge time sequence sample point being selected as the next cluster center is

[0141] S710: randomly selecting an initial cluster center from the non-initial cluster center light-charge time sequence sample points according to the probability;

[0142] Specifically, an initial cluster center is selected from the non-initial cluster center light-charge time sequence sample points according to the probability calculated in the previous step.

[0143] S711: determining whether the number of selected initial cluster centers is equal to the number of iterative clusters.

[0144] Specifically, it is determined whether the number of selected initial cluster centers is equal to the number of iterative clusters. If the number of selected initial cluster centers is not equal to the number of iterative clusters, the next initial cluster center is selected in S705, and the process is repeated until the number of selected initial cluster centers is equal to the number of iterative clusters, and S712 is entered.

[0145] S712: calculating the comprehensive similarity distance of the non-cluster center light-charge time sequence sample points to each cluster center;

[0146] Specifically, the comprehensive similarity distance of the non-cluster center light-charge time sequence sample points to each cluster center can be calculated by the DTW algorithm.

[0147] S713: selecting the classification of the cluster center with the minimum comprehensive similarity distance to the light-charge time sequence sample point as the classification of the light-charge time sequence sample point, and obtaining the clustering result of this iteration;

[0148] Specifically, assuming that there are three cluster centers, H m,1 (D m ,D c,1 )=1.44, H m,2 (D m ,D c,2 )=0.78, H m,3 (D m ,D c,3) = 0.32, the third cluster center is selected as the classification of the mth light-charged time sequence sample point, and the classification of all light-charged time sequence sample points is selected, and the clustering result of this iteration is obtained.

[0149] S714: The coordinate mean of the light-charged time sequence sample points in each classification is calculated as the coordinate of the new cluster center of the classification.

[0150] Specifically, assuming that there are two light-charged time sequence sample points in a classification, the coordinates of the two points are and The new cluster center coordinates of the classification are

[0151] S715: The iteration loss is calculated according to the coordinates of the light-charged time sequence sample points in each classification and the coordinates of the new cluster center.

[0152] Specifically, the comprehensive similarity distance between the light-charged time sequence sample points in each classification and the new cluster center of the corresponding classification is calculated according to the coordinates of the light-charged time sequence sample points in each classification and the coordinates of the new cluster center, and the sum of the comprehensive similarity distances between the light-charged time sequence sample points in all classifications and the new cluster center of the corresponding classification is taken as the iteration loss of this iteration.

[0153] S716: Determine whether the iteration loss is less than the preset iteration loss threshold.

[0154] Specifically, if the iteration loss is less than the preset iteration loss threshold, the operation is completed; if the iteration loss is not less than the preset iteration loss threshold, the loop enters S712, the clustering result is updated, and the iteration loss is received again until the iteration loss is less than the preset iteration loss threshold, and then S717 is entered. The iteration loss threshold is an empirical data.

[0155] S717: Increase the iteration clustering number and determine whether the iteration clustering number reaches the preset maximum clustering number.

[0156] Specifically, the iteration clustering number is increased, and it is determined whether the iteration clustering number reaches the preset maximum clustering number. If the iteration clustering number does not reach the preset maximum clustering number, S704 is entered; if the iteration clustering number reaches the preset maximum clustering number, S718 is entered.

[0157] S718: The clustering loss corresponding to the iteration clustering number is calculated according to the iteration loss generated by the clustering result under each iteration clustering number.

[0158] Specifically, assuming that the initial cluster number in the embodiment is 2, the maximum cluster number is 6, the iteration loss when the final cluster center is updated in each iteration cluster number is 9, 6, 6, 4, 5 respectively, and there are 100 light-charge time sequence sample points, the cluster loss in each iteration cluster number can be calculated as 0.092, 0.062, 0.063, 0.042, 0.053 respectively.

[0159] S719: Select the clustering result corresponding to the smallest cluster loss as the scene extraction result.

[0160] Specifically, the clustering result corresponding to the smallest cluster loss is selected as the scene extraction result, when the iteration cluster number is 5, the cluster loss is the smallest, which is 0.042, and therefore the clustering result when the iteration cluster number is 5 is selected as the scene extraction result, each cluster center is a classical scene, and the scene extraction result has five classifications.

[0161] The power distribution network operation scene extraction method provided by the embodiment of the application can take the initially determined initial cluster number as the iteration cluster number, perform iteration operation by increasing the iteration cluster number to obtain the clustering result in each iteration cluster number and the iteration loss generated, until the iteration cluster number reaches the preset maximum cluster number, calculate the initial cluster center, the number of the initial cluster center is equal to the iteration cluster number, calculate the clustering result and the iteration loss generated according to the initial cluster center, calculate the cluster loss in the corresponding iteration cluster number according to the iteration loss generated by the clustering result in each iteration cluster number, and select the clustering result corresponding to the smallest cluster loss as the scene extraction result, so that the initial cluster center is automatically calculated and selected, and the overall dynamic characteristics of the curve and the correlation between the photovoltaic power generation power and the load consumption power are fully considered in clustering.

[0162] Based on the same inventive concept, the embodiment of the application also provides a power grid operation scene extraction device, which can be used to implement the method described in the above embodiments, as described in the following embodiments. Since the principle of solving the problem of the power grid operation scene extraction device is similar to that of the power grid operation scene extraction method, the implementation of the power grid operation scene extraction device can be referred to the implementation of the software performance benchmark determination method, and the repeated parts will not be described here. The term "unit" or "module" used below can be a combination of software and / or hardware that implements a predetermined function. Although the system described in the following embodiments is preferably implemented in software, hardware or a combination of software and hardware is also possible and is conceived.

[0163] Figure 8 is a structural schematic diagram of the power grid operation scene extraction device provided by an embodiment of the application, as Figure 8 shown, the power grid operation scene extraction device provided by the embodiment of the application comprises:

[0164] The iterative clustering unit 8100 is configured to take a predetermined initial clustering number as an iterative clustering number, perform an iterative operation by increasing the iterative clustering number to obtain a clustering result and an iterative loss generated under each iterative clustering number, until the iterative clustering number reaches a preset maximum clustering number; and includes: an initial clustering center acquisition module 8110 configured to calculate initial clustering centers, the number of the initial clustering centers being equal to the iterative clustering number; and a result loss calculation module 8120 configured to calculate a clustering result and an iterative loss generated according to the initial clustering centers.

[0165] Specifically, the initial clustering number can be any positive integer greater than or equal to 2, which is an empirical data and can be determined according to the range of historical classification numbers. The initial clustering center acquisition module 8110 determines the initial clustering centers one by one until the number of the initial clustering centers is equal to the iterative clustering number, wherein each initial clustering center is determined according to the initial clustering centers selected before the initial clustering center is selected, so as to avoid the situation that some initial clustering centers are too concentrated caused by completely random selection; and the result loss calculation module 8120 calculates a clustering result according to the initial clustering centers and calculates an iterative loss generated according to the clustering result.

[0166] The clustering loss calculation unit 8200 is configured to calculate a clustering loss under a corresponding iterative clustering number according to an iterative loss generated under a clustering result under each iterative clustering number.

[0167] Specifically, the clustering loss calculation unit 8200 takes the average iterative loss of each light-charge time sequence sample point in the clustering result under each iterative clustering number, i.e., the average distance from each light-charge time sequence sample point in the clustering result under each iterative clustering number to the clustering center of the corresponding classification, as the clustering loss under the corresponding iterative clustering number.

[0168] The scene extraction result acquisition unit 8300 is configured to select a clustering result corresponding to the smallest clustering loss as a scene extraction result.

[0169] Specifically, the clustering loss represents the difference between the clustering result and the true situation, the larger the clustering loss, the greater the difference between the clustering result and the true situation, in order to obtain a result closer to the true situation, the scene extraction result acquisition unit 8300 selects a clustering result corresponding to the smallest clustering loss as a scene extraction result, and each clustering center in the clustering result is a classic scene corresponding to the classification.

[0170] The power distribution network operation scene extraction device provided by the embodiment of the present application comprises: an iterative clustering unit 8100, a clustering loss calculation unit 8200 and a scene extraction result acquisition unit 8300, wherein the iterative clustering unit 8100 comprises: an initial clustering center acquisition module 8110 and a result loss calculation module 8120, and the initial clustering center acquisition module 8110 can automatically calculate a probability to select a suitable initial clustering center, thereby solving the problem that a suitable initial clustering center is difficult to be selected by a traditional clustering algorithm.

[0171] Figure 9 FIG. 1 is a structural schematic diagram of the power grid operation scene extraction device provided by an embodiment of the present application, as shown in the figure, the initial clustering center acquisition module 8110 comprises: Figure 9

[0172] A first initial clustering center acquisition module 8111 is configured to randomly select a point in the pre-established light-load time sequence sample point as a first initial clustering center.

[0173] Specifically, the first initial clustering center acquisition module 8111 randomly selects a point in the pre-established light-load time sequence sample point, and the point is the first initial clustering center. The light-load time sequence sample point comprises a load component and a photovoltaic component.

[0174] An initial clustering center calculation module 8112 is configured to select the remaining initial clustering centers from the light-load time sequence sample point according to the first initial clustering center.

[0175] Specifically, the initial clustering center calculation module 8112 selects the next initial clustering center according to the first initial clustering center, and selects the next initial clustering center according to the selected initial clustering center, until the number of initial clustering centers reaches the iterative clustering number of this clustering.

[0176] The power distribution network operation scene extraction device provided by the embodiment of the present application, the initial clustering center acquisition module 8110 comprises: the first initial clustering center acquisition module 8111 and the initial clustering center calculation module 8112, which realizes automatic calculation of a probability to select a suitable initial clustering center, and fully considers the overall dynamic characteristics of the curve and the correlation between the photovoltaic power generation power and the load consumption power when clustering.

[0177] Figure 10 FIG. 2 is a structural schematic diagram of the power grid operation scene extraction device provided by an embodiment of the present application, as shown in the figure, the initial clustering center calculation module 8112 comprises: Figure 10

[0178] A first comprehensive similarity distance calculation module 8112.1 is configured to calculate the comprehensive similarity distance from the light-load time sequence sample point of a non-initial clustering center to each initial clustering center.

[0179] ​​Specifically, the first comprehensive similarity distance calculation module 8112.1 calculates the comprehensive similarity distance of each sample point that is not currently selected as an initial clustering center to each currently selected initial clustering center.

[0180] The probability calculation module 8112.2 is configured to calculate the probability of the light-load time sequence sample point being selected as the next clustering center according to the comprehensive similarity distance, and the probability is proportional to the square of the minimum comprehensive similarity distance of the comprehensive similarity distance of the light-load time sequence sample point to each initial clustering center.

[0181] Specifically, since clustering is essentially classification of sample data, one clustering center corresponds to one classification, and the closer two points are, the higher the probability that they belong to the same classification is. Therefore, the sample point farthest from the selected initial clustering center should be selected as the next clustering center as much as possible. Meanwhile, in order to avoid falling into a local optimal solution, the probability calculation module 8112.2 does not directly select the point with the maximum comprehensive similarity distance to the initial clustering center with the minimum comprehensive similarity distance as the next clustering center, but sets the probability of each sample point being selected according to the comprehensive similarity distance, and the probability is proportional to the square of the minimum comprehensive similarity distance of the comprehensive similarity distance of the light-load time sequence sample point to each initial clustering center, so that the sample point with the larger minimum comprehensive similarity distance has a higher probability of being selected as the next initial clustering center.

[0182] The selection module 8112.3 is configured to select an initial clustering center from the light-load time sequence sample points that are not initial clustering centers according to the probability.

[0183] Specifically, unlike the traditional clustering algorithm that selects clustering centers with equal probability, the selection module 8112.3 sets the probability of each sample point being selected according to the calculated probability, and randomly selects a sample point as an initial clustering center, which can select a more suitable clustering center and avoid obtaining a clustering result that greatly deviates from the true situation due to the selection of an unsuitable clustering center.

[0184] The power grid operation scenario extraction device provided by the embodiment of the present application comprises an initial clustering center calculation module 8112, which comprises a first comprehensive similarity distance calculation module 8112.1, a probability calculation module 8112.2, and a selection module 8112.3, and can automatically calculate the probability and select a suitable initial clustering center, and fully consider the overall dynamic characteristics of the curve and the correlation between the photovoltaic power generation power and the load consumption power during clustering.

[0185] Figure 11 is a structural schematic diagram of the power grid operation scenario extraction device provided by an embodiment of the present application, as shown in Figure 11 The first comprehensive similarity distance calculation module 8112.1 comprises:

[0186] The load dynamic time warping distance calculation module 8112.1a calculates the load dynamic time warping distance between the light-load time sequence sample point and each initial clustering center according to the load component of the light-load time sequence sample point and the load component of each initial clustering center.

[0187] Specifically, the load dynamic time warping distance calculation module 8112.1a calculates the load dynamic time warping distance between the light-load time sequence sample point and each initial clustering center according to the DTW algorithm.

[0188] The photovoltaic dynamic time warping distance calculation module 8112.1b calculates the photovoltaic dynamic time warping distance between the light-load time sequence sample point and each initial clustering center according to the photovoltaic component of the light-load time sequence sample point and the photovoltaic component of each initial clustering center.

[0189] Specifically, the photovoltaic dynamic time warping distance calculation module 8112.1b calculates the photovoltaic dynamic time warping distance between the light-load time sequence sample point and each initial clustering center according to the DTW algorithm.

[0190] The light-load correlation distance calculation module 8112.1c calculates the light-load correlation distance between the light-load time sequence sample point and each initial clustering center according to the coordinates of the light-load time sequence sample point and the coordinates of each initial clustering center.

[0191] Specifically, the coordinates of the light-load time sequence sample point include the load component and the photovoltaic component, and the coordinates of the initial clustering center also include the load component and the photovoltaic component. The light-load correlation distance calculation module 8112.1c calculates the light-load correlation distance according to the above formula (2).

[0192] The distance integration module 8112.1d calculates the comprehensive similarity distance according to the preset weight and the load dynamic time warping distance, the photovoltaic dynamic time warping distance, and the light-load correlation distance.

[0193] Specifically, the distance integration module 8112.1d calculates the comprehensive similarity distance according to the sum of the product of the load dynamic time warping distance, the photovoltaic dynamic time warping distance, and the light-load correlation distance and the corresponding weight according to formula (3). The weight is an empirical value.

[0194] The power distribution network operation scenario extraction device provided by the embodiment of the application includes a first comprehensive similarity distance calculation module 8112.1, which includes a load dynamic time warping distance calculation module 8112.1a, a photovoltaic dynamic time warping distance calculation module 8112.1b, a light-load correlation distance calculation module 8112.1c, and a distance integration module 8112.1d. The overall dynamic characteristics of the curve and the correlation between the photovoltaic power and the load consumption power are fully considered during clustering.

[0195] Figure 12is a structural schematic diagram of the power grid operation scenario extraction device provided by an embodiment of the present application, as shown in Figure 12 The result loss calculation module 8120 includes:

[0196] The second comprehensive similarity distance calculation module 8121 is configured to calculate the comprehensive similarity distance of the non-cluster center light-charge time sequence sample point to each cluster center.

[0197] Specifically, the second comprehensive similarity distance calculation module 8121 calculates the comprehensive similarity distance of the non-cluster center light-charge time sequence sample point to each cluster center, and the calculation method of the comprehensive similarity distance is the same as that given in the first embodiment. Figure 4

[0198] The cluster result calculation module 8122 is configured to select the classification of the cluster center with the minimum comprehensive similarity distance of the light-charge time sequence sample point as the classification of the light-charge time sequence sample point, to obtain the clustering result of this iteration.

[0199] Specifically, the smaller the comprehensive similarity distance, the smaller the difference between the light-charge time sequence sample point and the corresponding cluster center, and the greater the probability of belonging to the same classification. Therefore, the cluster result calculation module 8122 selects the classification of the cluster center with the minimum comprehensive similarity distance of the light-charge time sequence sample point as the classification of the light-charge time sequence sample point, and all non-cluster center light-charge time sequence sample points select the corresponding classification, to obtain the clustering result of this iteration.

[0200] The cluster center updating module 8123 is configured to calculate the coordinate mean of the light-charge time sequence sample points in each classification as the coordinate of the new cluster center of the classification.

[0201] Specifically, after obtaining the clustering result of this iteration, the current cluster center of each classification may not be the point with the minimum distance sum to all light-charge time sequence sample points in the current classification region. Therefore, the cluster center updating module 8123 calculates the coordinate mean of the light-charge time sequence sample points in each classification as the coordinate of the new cluster center of the classification.

[0202] The iteration loss calculation module 8124 is configured to calculate the iteration loss according to the coordinates of the light-charge time sequence sample points in each classification and the coordinates of the new cluster center.

[0203] Specifically, the iteration loss calculation module 8124 calculates the comprehensive similarity distance of the light-charge time sequence sample points in each classification to the new cluster center of the corresponding classification according to the coordinates of the light-charge time sequence sample points in each classification and the coordinates of the new cluster center, and takes the sum of the comprehensive similarity distances of the light-charge time sequence sample points in all classifications to the new cluster center of the corresponding classification as the iteration loss of this iteration, which can be calculated by formula (4). ​

[0204] The power distribution network operation scenario extraction device provided by the embodiment of the present application includes a second comprehensive similarity distance calculation module 8121, a clustering result calculation module 8122, a clustering center updating module 8123 and an iteration loss calculation module 8124, and the overall dynamic characteristics of the curve and the correlation between the photovoltaic power generation power and the load consumption power are fully considered during clustering.

[0205] Figure 13 FIG. 1 is a structural schematic diagram of the power grid operation scenario extraction device provided by an embodiment of the present application, as shown in the figure, Figure 13 The power grid operation scenario extraction device provided by the embodiment of the present application further includes:

[0206] A preprocessing unit 1310 is configured to acquire photovoltaic measured power and load measured power, and perform normalization and standardization processing on the photovoltaic measured power and the load measured power.

[0207] Specifically, the photovoltaic measured power includes photovoltaic power measured at multiple time periods of each day of the power grid, and the load measured power includes load power measured at multiple time periods of each day of the power grid. The preprocessing unit 1310 divides the photovoltaic measured power by the annual photovoltaic power peak value to obtain normalized photovoltaic measured power, and divides the load measured power by the annual load power peak value to obtain normalized load measured power. Any standardization method can be selected as needed to standardize the normalized photovoltaic measured power and the normalized load measured power, such as min-mix standardization, z-score standardization, etc.

[0208] A sample point generation unit 1320 is configured to obtain coordinates of photovoltaic-load time sequence sample points according to the processed photovoltaic measured power and the processed load measured power. Each photovoltaic-load time sequence sample point includes a photovoltaic component and a load component.

[0209] Specifically, the photovoltaic-load time sequence sample point includes a load component and a photovoltaic component. The load component in one photovoltaic-load time sequence sample point includes multiple load measured powers of one day, and the photovoltaic component includes multiple photovoltaic measured powers of one day.

[0210] The power distribution network operation scenario extraction device provided by the embodiment of the present application further includes the preprocessing unit 1310 and the sample point generation unit 1320, which realizes automatic calculation of probability, selection of a suitable initial clustering center, and full consideration of the overall dynamic characteristics of the curve and the correlation between the photovoltaic power generation power and the load consumption power during clustering.

[0211] Figure 14 FIG. 2 is a structural schematic diagram of an electronic device provided by an embodiment of the present application, as shown in the figure, Figure 14As shown, the electronic device can include a processor 1401, a communications interface 1402, a memory 1403, and a communications bus 1404, wherein the processor 1401, the communications interface 1402, and the memory 1403 complete mutual communication through the communications bus 1404. The processor 1401 can invoke a logical instruction in the memory 1403 to execute the following method: taking a predetermined initial cluster number as an iteration cluster number, performing an iteration operation by increasing the iteration cluster number to obtain a clustering result under each iteration cluster number and an iteration loss generated, until the iteration cluster number reaches a preset maximum cluster number: calculating an initial cluster center, the number of the initial cluster center being equal to the iteration cluster number; calculating the clustering result and the iteration loss generated according to the initial cluster center; calculating a clustering loss under the corresponding iteration cluster number according to the iteration loss generated by the clustering result under each iteration cluster number; and selecting a clustering result corresponding to the smallest clustering loss as a scene extraction result.

[0212] In addition, the logical instruction in the memory 1403 described above can be implemented in the form of a software functional unit and sold or used as an independent product, and can be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0213] The embodiment of the present application discloses a computer program product, the computer program product comprises a computer program stored on a computer readable storage medium, the computer program comprises program instructions, when the program instructions are executed by a computer, the computer can execute the method provided by each method embodiment described above, for example, comprising: taking a predetermined initial cluster number as an iterative cluster number, performing an iterative operation by increasing the iterative cluster number to obtain a clustering result under each iterative cluster number and an iterative loss generated, until the iterative cluster number reaches a preset maximum cluster number; calculating an initial cluster center, the number of the initial cluster center is equal to the iterative cluster number; calculating the clustering result and the iterative loss generated according to the initial cluster center; calculating a clustering loss under the corresponding iterative cluster number according to the iterative loss generated by the clustering result under each iterative cluster number; selecting the clustering result corresponding to the smallest clustering loss as a scene extraction result.

[0214] The embodiment of the present application provides a computer readable storage medium, the computer readable storage medium stores a computer program, the computer program enables the computer to execute the method provided by each method embodiment described above, for example, comprising: taking a predetermined initial cluster number as an iterative cluster number, performing an iterative operation by increasing the iterative cluster number to obtain a clustering result under each iterative cluster number and an iterative loss generated, until the iterative cluster number reaches a preset maximum cluster number; calculating an initial cluster center, the number of the initial cluster center is equal to the iterative cluster number; calculating the clustering result and the iterative loss generated according to the initial cluster center; calculating a clustering loss under the corresponding iterative cluster number according to the iterative loss generated by the clustering result under each iterative cluster number; selecting the clustering result corresponding to the smallest clustering loss as a scene extraction result.

[0215] Those skilled in the art should understand that embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0216] The present application is described with reference to flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce a machine that implements the flowcharts and / or block diagrams. Figure 1one or more processes and / or blocks Figure 1 an apparatus for performing the functions specified in the flowchart or multiple flows and / or blocks.

[0217] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flowchart or multiple flows and / or blocks. Figure 1 one or more processes and / or blocks Figure 1 an apparatus for performing the functions specified in the flowchart or multiple flows and / or blocks.

[0218] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flowchart or multiple flows and / or blocks. Figure 1 one or more processes and / or blocks Figure 1 an apparatus for performing the functions specified in the flowchart or multiple flows and / or blocks.

[0219] In the description of the specification, the description of the terms "one embodiment", "one specific embodiment", "some embodiments", "for example", "exemplary", "specific exemplary", or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in one or more embodiments or examples.

[0220] The above-described specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above-described specific embodiments are only specific embodiments of the present application and are not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for extracting power distribution network operation scenarios, characterized in that, include: A predetermined initial cluster number is used as the iterative cluster number. The iterative operation is performed by increasing the iterative cluster number to obtain the clustering results and iterative losses under each iterative cluster number, until the iterative cluster number reaches the preset maximum cluster number. The initial cluster centers are calculated, and the number of initial cluster centers is equal to the iterative cluster number. The clustering results and iterative losses are calculated based on the initial cluster centers. Calculate the clustering loss for the corresponding number of iterations based on the iteration loss generated by the clustering results at each iteration number; The clustering result corresponding to the minimum clustering loss is selected as the scene extraction result, and each cluster center in the clustering result is the classic scene of the corresponding classification. The calculation of the initial cluster centers includes: A point is randomly selected from the pre-established light-charge time series sample points as the first initial cluster center; Based on the first initial cluster center, other initial cluster centers are selected from the light-load time series sample points; The step of selecting other initial cluster centers from the light-load time-series sample points based on the first initial cluster center includes: Calculate the comprehensive similarity distance from the light-charge time-series sample points that are not in the initial cluster centers to each initial cluster center; The probability of the light-charge time series sample point being selected as the next initial cluster center is calculated based on the comprehensive similarity distance. The probability is proportional to the square of the minimum comprehensive similarity distance among the comprehensive similarity distances from the light-charge time series sample point to each initial cluster center. An initial cluster center is randomly selected from the light-charge time-series sample points that are not the initial cluster center according to the aforementioned probability; The above operation is performed iteratively until the number of initial cluster centers selected equals the number of iterative clusters. The calculation of the comprehensive similarity distance from the light-charge time-series sample points of non-initial cluster centers to each initial cluster center includes: The dynamic time curvature distance of the load between the light-load time series sample points and each initial cluster center is calculated based on the load components of the light-load time series sample points and the load components of each initial cluster center. The photovoltaic dynamic time curvature distance between the light-load time series sample points and each initial cluster center is calculated based on the photovoltaic components of the light-load time series sample points and the photovoltaic components of each initial cluster center. The light-load correlation distance between the light-load time series sample points and each initial cluster center is calculated based on the coordinates of the light-load time series sample points and the coordinates of each initial cluster center. The comprehensive similarity distance is calculated based on preset weights and the load dynamic time bending distance, the photovoltaic dynamic time bending distance, and the light-load correlation distance.

2. The method for extracting power distribution network operation scenarios according to claim 1, characterized in that, The calculation of clustering results and the resulting iterative loss based on the initial cluster centers includes: Calculate the comprehensive similarity distance from the light-charge time-series sample points of non-cluster centers to each cluster center; The clustering center with the smallest comprehensive similarity distance to the light-load time series sample points is selected as the classification of the light-load time series sample points, and the clustering result of this iteration is obtained; Calculate the mean coordinates of the light-load time series sample points in each category, and use them as the coordinates of the new cluster centers for the category; The iterative loss is calculated based on the coordinates of the light-load time-series sample points in each category and the coordinates of the new cluster centers. The above operation is performed iteratively until the iteration loss is less than the preset iteration loss threshold.

3. The method for extracting power distribution network operation scenarios according to claim 1, characterized in that, Before performing the iteration operation, the following also applies: Obtain the measured photovoltaic power and the measured load power, and then perform per-unit and normalization processing on the measured photovoltaic power and the measured load power; The coordinates of the photovoltaic-load time series sample points are obtained based on the processed photovoltaic measured power and the load measured power. The coordinates of each photovoltaic-load time series sample point include photovoltaic components and load components.

4. A distribution network operation scenario extraction device, characterized in that, include: An iterative clustering unit is used to take a predetermined initial cluster number as the iterative cluster number, and perform iterative operations by increasing the iterative cluster number to obtain the clustering results and iterative losses at each iterative cluster number, until the iterative cluster number reaches a preset maximum cluster number; it includes: an initial cluster center acquisition module, used to calculate the initial cluster centers, the number of which is equal to the iterative cluster number; and a result loss calculation module, used to calculate the clustering results and iterative losses based on the initial cluster centers; The clustering loss calculation unit is used to calculate the clustering loss for the corresponding number of iterations based on the iteration loss generated by the clustering results at each iteration number. The scene extraction result acquisition unit is used to select the clustering result corresponding to the minimum clustering loss as the scene extraction result, and each cluster center in the clustering result is the classic scene of the corresponding classification. The initial cluster center acquisition module includes: The first initial cluster center acquisition module is used to randomly select a point from the pre-established light-charge time series sample points as the first initial cluster center; An initial cluster center calculation module is used to select other initial cluster centers from the light-load time-series sample points based on the first initial cluster center; The initial cluster center calculation module includes: The first comprehensive similarity distance calculation module is used to calculate the comprehensive similarity distance from the light-charge time series sample points that are not in the initial cluster centers to each initial cluster center; The probability calculation module is used to calculate the probability that the light-charge time series sample point is selected as the next cluster center based on the comprehensive similarity distance. The probability is proportional to the square of the minimum comprehensive similarity distance among the comprehensive similarity distances from the light-charge time series sample point to each initial cluster center. The selection module is used to randomly select an initial cluster center from the light-charge time-series sample points that are not the initial cluster centers according to the probability. The first comprehensive similarity distance calculation module includes: The load dynamic time bending distance calculation module calculates the load dynamic time bending distance between the light-load time series sample points and each initial cluster center based on the load components of the light-load time series sample points and the load components of each initial cluster center. The photovoltaic dynamic time bending distance calculation module calculates the photovoltaic dynamic time bending distance between the light-load time series sample points and each initial cluster center based on the photovoltaic components of the light-load time series sample points and the photovoltaic components of each initial cluster center. The light-load correlation distance calculation module calculates the light-load correlation distance between the light-load time series sample points and each initial cluster center based on the coordinates of the light-load time series sample points and the coordinates of each initial cluster center. The distance integration module calculates the comprehensive similarity distance based on preset weights and the load dynamic time bending distance, the photovoltaic dynamic time bending distance, and the light-load correlation distance.

5. The power distribution network operation scenario extraction device according to claim 4, characterized in that, The result loss calculation module includes: The second comprehensive similarity distance calculation module is used to calculate the comprehensive similarity distance from the light-charge time series sample points of non-cluster centers to each cluster center; The clustering result calculation module is used to select the cluster center with the smallest comprehensive similarity distance to the light-load time series sample points as the classification of the light-load time series sample points, and obtain the clustering result of this iteration; The cluster center update module is used to calculate the mean coordinates of the light-load time series sample points in each category, and use them as the coordinates of the new cluster centers for the category. The iterative loss calculation module is used to calculate the iterative loss based on the coordinates of the light-load time-series sample points in each classification and the coordinates of the new cluster center.

6. The power distribution network operation scenario extraction device according to claim 4, characterized in that, Also includes: The preprocessing unit is used to obtain the measured photovoltaic power and the measured load power, and to perform per-unit and normalization processing on the measured photovoltaic power and the measured load power. The sample point generation unit is used to obtain the coordinates of the photovoltaic-load time series sample points based on the processed photovoltaic measured power and the load measured power. The coordinates of each photovoltaic-load time series sample point include photovoltaic components and load components.

7. 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, it implements the method according to any one of claims 1 to 3.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 3.

9. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method described in any one of claims 1 to 3.

Images