Wind power data clustering method and device, computer device and storage medium

Abstract

The present application relates to the technical field of new energy power generation simulation, and discloses a wind power data clustering method and device, computer equipment and storage medium, the present application introduces the preset period length into the preset particle swarm algorithm to optimize the sample profile coefficient, obtains the improved target sample profile coefficient, so that the improved target profile coefficient can better quantitatively judge the advantages and disadvantages of the classification result. Further, taking the improved target sample profile coefficient as the optimization target, taking the period length and the classification number of the wind power data set to be clustered as the optimization variables, the particle swarm algorithm is used to automatically find the best data classification length and the classification number in the wind power data set to be clustered. Compared with manually specifying the classification length and the classification number, the present application has better classification effect.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of new energy power generation simulation technology, and particularly relates to a wind power data clustering method and device, computer equipment and storage medium. BACKGROUND

[0002] New energy power generation output simulation technology can provide important technical support for new energy planning, operation and power system operation and dispatching.

[0003] The existing new energy power generation output simulation technology mainly focuses on short-term power prediction based on numerical weather prediction, and pursues the accuracy of the prediction. The research on long-term simulation of new energy power generation mainly focuses on wind power data clustering methods. The current mainstream clustering method needs to manually set the cycle length and the number of categories, which may introduce human factors. If the clustering cycle and the number of categories are not suitable, the accuracy of new energy power generation output simulation will be poor, and the use conditions will not be met. SUMMARY

[0004] Therefore, the present application provides a wind power data clustering method, device, computer equipment and storage medium to solve the problem of poor accuracy of new energy power generation output simulation caused by the need for manual setting of cycle length and category number in wind power data clustering.

[0005] In a first aspect, the present application provides a wind power data clustering method for a wind power station, which comprises:

[0006] Obtaining a wind power data set to be clustered and a historical target power generation data set of the wind power station; based on a preset cycle length, processing the wind power data set to be clustered by a preset particle swarm algorithm to obtain a target sample profile coefficient corresponding to the wind power data set to be clustered; based on the preset cycle length, taking the target sample profile coefficient as the optimization target, and using the preset particle swarm algorithm to optimize the wind power data set to be clustered to obtain a target classification number and a target data classification length; based on the target classification number and the target data classification length, using a preset clustering algorithm to cluster the wind power data set to be clustered to obtain a target clustering result.

[0007] The wind power data clustering method provided by this invention incorporates a preset period length into a preset particle swarm optimization algorithm to optimize the initial sample profile coefficients, resulting in improved target sample profile coefficients. These improved target profile coefficients can better quantify the quality of the classification results. Furthermore, using these improved target sample profile coefficients as the optimization objective, and the period length and number of categories in the wind power dataset to be clustered as optimization variables, the particle swarm optimization algorithm automatically finds the optimal data classification length and number of categories in the wind power dataset to be clustered. Compared to manually specifying the classification length and number of categories, this method achieves better classification results.

[0008] In one optional implementation, the wind power dataset to be clustered and the historical target power dataset of the wind farm are obtained, including:

[0009] Obtain the wind power dataset to be clustered and the first historical power generation dataset of the wind farm; preprocess the first historical power generation dataset to obtain the second historical power generation dataset; convert the format of the second historical power generation dataset to generate the historical target power generation dataset.

[0010] This invention improves the accuracy of historical power generation datasets from wind farms by preprocessing and converting them into historical target power generation datasets that meet the required format. Furthermore, it provides data support for subsequent particle swarm optimization algorithms.

[0011] In one optional implementation, based on a preset period length, the target sample profile coefficients corresponding to the wind power dataset to be clustered are obtained through a preset particle swarm optimization algorithm, including:

[0012] Based on the preset particle swarm optimization algorithm and preset Euclidean distance calculation method, the first sample profile coefficient relationship is determined; based on the preset period length and preset Euclidean distance calculation method, the first sample profile coefficient relationship is optimized to obtain the second sample profile coefficient relationship; based on the second sample profile coefficient relationship, the target sample profile coefficients corresponding to the wind power data set to be clustered are determined.

[0013] This invention introduces a preset period length to improve the first contour coefficient relationship, and uses the improved second sample contour coefficient relationship to determine the target sample contour coefficient, so that the target sample contour coefficient can better quantify and judge the quality of the classification result.

[0014] In one optional implementation, the first sample contour coefficient relationship is determined based on a preset particle swarm optimization algorithm and a preset Euclidean distance calculation method, including:

[0015] The wind power dataset to be clustered is processed using a pre-defined particle swarm optimization algorithm to generate a first-class population cluster and a second-class population cluster. The first initial average distance between each wind power data point in the first-class population cluster and other wind power data points in the first-class population cluster is calculated using a pre-defined Euclidean distance calculation method. The second initial average distance between each wind power data point in the first-class population cluster and the second-class population cluster is calculated using the same pre-defined Euclidean distance calculation method. Based on each first initial average distance and each second initial average distance, the first sample profile coefficient relationship is determined.

[0016] This invention utilizes the first and second initial average distances to reflect cohesion and separation, enabling the calculated initial profile coefficients to better quantify the quality of classification results.

[0017] In one optional implementation, a first sample contour coefficient relationship is determined based on each first initial average distance and each second initial average distance, including:

[0018] Based on each first initial average distance, determine the target first initial average distance corresponding to each wind power data; based on each second initial average distance, determine the target second initial average distance corresponding to each wind power data; based on each target first initial average distance and each target second initial average distance, determine the third sample profile coefficient relationship; based on the third sample profile coefficient relationship, determine the first sample profile coefficient relationship.

[0019] This invention utilizes the first initial average distance and the second initial average distance to reflect cohesion and separation, so that the sample profile coefficients calculated according to the first sample profile coefficient relationship can better quantify the quality of the classification results.

[0020] In one optional implementation, based on a preset period length and a preset Euclidean distance calculation method, the first sample contour coefficient relationship is optimized to obtain a second sample contour coefficient relationship, including:

[0021] A preset period length is introduced into the preset Euclidean distance calculation method to obtain the target Euclidean distance calculation method; using the target Euclidean distance calculation method, the contour coefficient relationship of the first sample is optimized to obtain the contour coefficient relationship of the second sample.

[0022] This invention introduces a preset period length to improve the contour coefficient, so that the improved target contour coefficient can better quantify and judge the quality of the classification results.

[0023] In one optional implementation, based on a preset period length and using the target sample profile coefficient as the optimization objective, a preset particle swarm optimization algorithm is used to optimize the wind power dataset to be clustered, resulting in the target number of classifications and the target data classification length, including:

[0024] Obtain a preset range of the number of categories and a range of the period length based on a preset period length; using the preset range of the number of categories and the range of the period length as constraints, and the target sample profile coefficient as the optimization target, use a preset particle swarm optimization algorithm to optimize the wind power dataset to be clustered, and obtain the target number of categories and the target data category length.

[0025] This invention incorporates a preset period length into a preset particle swarm optimization (PSO) algorithm to optimize the sample profile coefficient, resulting in an improved target sample profile coefficient. This improved target profile coefficient can better quantify the quality of classification results. Furthermore, using this improved target sample profile coefficient as the optimization objective, and the period length and number of categories in the wind power dataset to be clustered as optimization variables, the PSO algorithm automatically finds the optimal data classification length and number of categories in the wind power dataset to be clustered. Compared to manually specifying the classification length and number of categories, this method achieves better classification results. Simultaneously, the preset range of the number of categories and the range of the period length serve as constraints on the optimization process, further improving the classification performance.

[0026] In one optional implementation, the first historical power generation dataset is preprocessed to obtain a second historical power generation dataset, including:

[0027] Null and outlier data are identified in the first historical power generation dataset; the null and outlier data are then interpolated using a linear interpolation method to obtain the second historical power generation dataset.

[0028] This invention improves the accuracy of the second historical power generation dataset by interpolating the first historical power generation dataset.

[0029] In one alternative implementation, the method further includes:

[0030] Based on the target clustering results, the first sample silhouette coefficient is obtained by calculating the second sample silhouette coefficient relationship. The target clustering results are evaluated using the first sample silhouette coefficient, and the target clustering results are adjusted according to the evaluation results until the target clustering results that meet the conditions are obtained.

[0031] This invention improves the clustering effect of the target clustering results by repeatedly calculating the sample silhouette coefficient.

[0032] In one alternative implementation, the method further includes:

[0033] Obtain the preset number of categories and preset data length; based on the preset number of categories and preset data length, use the preset clustering algorithm to perform clustering processing on the wind power dataset to be clustered, and obtain the preset clustering results; compare the preset clustering results with the target clustering results that meet the conditions, and determine the clustering effect of the target clustering results that meet the conditions based on the comparison results.

[0034] This invention uses a preset number of categories and a preset data length to determine the clustering effect of the target clustering result that meets the conditions. This can further verify the clustering effect of the wind power data clustering method provided by this invention.

[0035] In a second aspect, the present invention provides a wind power data clustering device for wind farms; the wind power data clustering device includes:

[0036] The system comprises three modules: an acquisition module for acquiring the wind power dataset to be clustered and the historical target power dataset of wind farms; a first processing module for processing the wind power dataset to be clustered using a preset particle swarm optimization algorithm based on a preset period length to obtain the target sample profile coefficients; a second processing module for optimizing the wind power dataset to be clustered using a preset particle swarm optimization algorithm based on a preset period length and the target sample profile coefficients to obtain the target number of categories and the target data category length; and a third processing module for clustering the wind power dataset to be clustered using a preset clustering algorithm based on the target number of categories and the target data category length to obtain the target clustering results.

[0037] In one alternative implementation, the acquisition module includes:

[0038] The first acquisition submodule is used to acquire the wind power dataset to be clustered and the first historical power generation dataset of the wind farm; the preprocessing submodule is used to preprocess the first historical power generation dataset to obtain the second historical power generation dataset; the format conversion submodule is used to convert the format of the second historical power generation dataset to generate the historical target power generation dataset.

[0039] In one optional implementation, the first processing module includes:

[0040] The first determining submodule is used to determine the relationship of the first sample contour coefficients based on a preset particle swarm algorithm and a preset Euclidean distance calculation method.

[0041] The optimization submodule is used to optimize the contour coefficient relationship of the first sample based on the preset period length and the preset Euclidean distance calculation method to obtain the contour coefficient relationship of the second sample.

[0042] The second determination submodule is used to determine the target sample profile coefficients corresponding to the wind power power dataset to be clustered based on the second sample profile coefficient relationship.

[0043] Thirdly, the present invention provides a computer device, including: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the wind power data clustering method of the first aspect or any corresponding embodiment described above.

[0044] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the wind power data clustering method of the first aspect or any corresponding embodiment described above. Attached Figure Description

[0045] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0046] Figure 1 This is a flowchart illustrating the wind power data clustering method according to an embodiment of the present invention;

[0047] Figure 2 This is a flowchart illustrating another wind power data clustering method according to an embodiment of the present invention;

[0048] Figure 3 This is a schematic diagram of the first historical power generation data set of a wind farm over 30 days according to an embodiment of the present invention.

[0049] Figure 4 This is a flowchart illustrating another wind power data clustering method according to an embodiment of the present invention;

[0050] Figure 5 This is a flowchart illustrating another wind power data clustering method according to an embodiment of the present invention;

[0051] Figure 6 This is a flowchart illustrating the adaptive classification method for wind power based on sample profile coefficients and particle swarm optimization according to an embodiment of the present invention.

[0052] Figure 7 This is a schematic diagram of the clustering results corresponding to the manually specified method according to an embodiment of the present invention;

[0053] Figure 8This is a schematic diagram of the clustering results of the wind power adaptive classification method based on sample profile coefficient and particle swarm algorithm according to an embodiment of the present invention;

[0054] Figure 9 This is a structural block diagram of a wind power data clustering device according to an embodiment of the present invention;

[0055] Figure 10 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] This invention provides a wind power data clustering method that uses a particle swarm optimization algorithm to automatically find the optimal data classification length and number of classifications to achieve better classification results.

[0058] According to an embodiment of the present invention, a method for clustering wind power data is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0059] This embodiment provides a wind power data clustering method for wind farms; Figure 1 This is a flowchart of a wind power data clustering method according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps:

[0060] Step S101: Obtain the wind power data set to be clustered and the historical target power generation data set of the wind farm.

[0061] Among them, the historical target power generation dataset can be power generation data processed within any time period.

[0062] Step S102: Based on the preset period length, the target sample profile coefficients corresponding to the wind power data set to be clustered are obtained through the preset particle swarm algorithm.

[0063] Specifically, by introducing a preset period length into a preset particle swarm optimization algorithm, the contour coefficients of the target samples corresponding to the wind power dataset to be clustered can be obtained.

[0064] Step S103: Based on the preset period length, with the target sample profile coefficient as the optimization target, the preset particle swarm optimization algorithm is used to optimize the wind power dataset to be clustered, so as to obtain the target classification number and target data classification length.

[0065] Specifically, when using the pre-defined particle swarm optimization algorithm to optimize the wind power dataset to be clustered, the improved target sample profile coefficient is used as the optimization target. At the same time, the classification length and number of classifications of the wind power dataset to be clustered are used as optimization variables, which improves the classification effect of the wind power dataset to be clustered and makes the accuracy of the target classification number and target data classification length higher.

[0066] Step S104: Based on the number of target categories and the length of target data categories, a preset clustering algorithm is used to perform clustering processing on the wind power dataset to be clustered, and the target clustering result is obtained.

[0067] The preset clustering algorithm can be the K-means clustering algorithm with improved Euclidean distance.

[0068] Specifically, the optimization results of the preset particle swarm optimization algorithm, namely the number of target categories and the length of target data categories, are used as parameters to input the improved Euclidean distance K-means clustering algorithm. The improved Euclidean distance K-means clustering algorithm is then used to cluster the wind power dataset to be clustered, thereby obtaining the target clustering result of the wind power dataset to be clustered.

[0069] The wind power data clustering method provided in this embodiment uses the particle swarm optimization algorithm to automatically find the optimal data classification length and number of classifications, which has a better classification effect than manually specifying the classification length and number of classifications.

[0070] This embodiment provides a wind power data clustering method for wind farms; Figure 2 This is a flowchart of a wind power data clustering method according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps:

[0071] Step S201: Obtain the wind power data set to be clustered and the historical target power generation data set of the wind farm.

[0072] Specifically, step S201 includes:

[0073] Step S2011: Obtain the wind power data set to be clustered and the first historical power generation data set of the wind farm.

[0074] The first historical power generation dataset consists of initial power generation data for any time period, such as a data interval of 15 minutes and a data length of one year.

[0075] like Figure 3 The image shows the first historical power generation dataset of a certain wind farm over a 30-day period.

[0076] Step S2012: Preprocess the first historical power generation dataset to obtain the second historical power generation dataset.

[0077] Specifically, by preprocessing the first historical power generation dataset, abnormal data that does not meet the requirements can be removed.

[0078] Step S2013: Convert the format of the second historical power generation dataset to generate the historical target power generation dataset.

[0079] Specifically, the purpose of the format conversion is to make the second historical power generation dataset meet the requirements of the subsequent preset particle swarm optimization algorithm.

[0080] For example, when running this preset particle swarm algorithm using Python, you need to use the pandas library in Python to convert the data into a dataframe format, which includes date and power columns.

[0081] The dataframe format represents a tabular data structure consisting of a set of data and a pair of indexes (row index and column index).

[0082] In some optional implementations, step S2012 above includes:

[0083] Step a1: Identify null and outlier data in the first historical power generation dataset.

[0084] Step a2 involves using linear interpolation to interpolate the null and outlier data to obtain the second historical power generation dataset.

[0085] Specifically, when the first historical power generation dataset contains non-compliant null values ​​and outliers, linear interpolation can be used for interpolation to obtain the corresponding second historical power generation dataset. This embodiment does not specifically limit the interpolation method, as long as it meets the data processing requirements.

[0086] Step S202: Based on a preset period length, the target sample profile coefficients corresponding to the wind power power dataset to be clustered are obtained through a preset particle swarm optimization algorithm. For details, please refer to [link to relevant documentation]. Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0087] Step S203: Based on the preset period length, and using the target sample profile coefficient as the optimization objective, a preset particle swarm optimization algorithm is used to optimize the wind power dataset to be clustered, obtaining the target number of classifications and the target data classification length. For details, please refer to... Figure 1 Step S103 of the illustrated embodiment will not be described again here.

[0088] Step S204: Based on the target number of categories and the target data category length, a preset clustering algorithm is used to cluster the wind power dataset to be clustered, obtaining the target clustering results. For details, please refer to [link to relevant documentation]. Figure 1 Step S104 of the illustrated embodiment will not be described again here.

[0089] The wind power data clustering method provided in this embodiment improves the accuracy of the data by preprocessing the historical power generation dataset of wind farms and converting it into a historical target power generation dataset that meets the required format. Furthermore, it provides data support for subsequent particle swarm optimization.

[0090] This embodiment provides a wind power data clustering method for wind farms; Figure 4 This is a flowchart of a wind power data clustering method according to an embodiment of the present invention, such as... Figure 4 As shown, the process includes the following steps:

[0091] Step S401: Obtain the wind power data set to be clustered and the historical target power generation data set for the wind farm. For details, please refer to [link to relevant documentation]. Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0092] Step S402: Based on the preset period length, the target sample profile coefficients corresponding to the wind power data set to be clustered are obtained through the preset particle swarm algorithm.

[0093] Specifically, step S402 includes:

[0094] Step S4021: Based on the preset particle swarm algorithm and preset Euclidean distance calculation method, determine the relationship of the first sample contour coefficient.

[0095] The sample silhouette coefficient is a clustering evaluation index used to assess the effectiveness of data clustering (classification). Its value ranges from -1 to 1; a higher value indicates a better clustering result.

[0096] Specifically, by introducing a preset Euclidean distance calculation method into the preset particle swarm algorithm, the first calculation formula for the sample contour coefficients can be determined, namely the first sample contour coefficient formula.

[0097] Step S4022: Based on the preset period length and preset Euclidean distance calculation method, optimize the first sample contour coefficient relationship to obtain the second sample contour coefficient relationship.

[0098] Specifically, by introducing a preset period length into the preset Euclidean distance calculation method, the first sample contour coefficient relationship can be optimized to obtain the first calculation relationship of the sample contour coefficient, which is the second sample contour coefficient relationship.

[0099] Step S4023: Based on the second sample profile coefficient relationship, determine the target sample profile coefficients corresponding to the wind power power dataset to be clustered.

[0100] Specifically, the target sample profile coefficients corresponding to the wind power dataset to be clustered can be calculated using the improved second sample profile coefficient relationship.

[0101] In some optional implementations, step S4021 above includes:

[0102] Step b1: Process the wind power dataset to be clustered based on the preset particle swarm optimization algorithm to generate a first-class population cluster and a second-class population cluster.

[0103] Step b2: Calculate the first initial average distance between each wind power data point in the first category of population cluster and other wind power data points in the first category of population cluster using a preset Euclidean distance calculation method.

[0104] Step b3: Calculate the second initial average distance between each wind power data point in the first category of population clusters and the second category of population clusters using a preset Euclidean distance calculation method.

[0105] Step b4: Determine the first sample profile coefficient relationship based on each first initial average distance and each second initial average distance.

[0106] The first initial average distance is used to reflect the cohesion of the data in the wind power data set to be clustered; the second initial average distance is used to reflect the separation of the data in the wind power data set to be clustered.

[0107] First, the wind power dataset to be clustered is processed using a pre-defined particle swarm optimization algorithm. This pre-classifies the dataset into two clusters: a first-category cluster with the same data types and a second-category cluster with different data types. There can be one or more first-category and second-category clusters.

[0108] Secondly, calculate the intra-cluster dissimilarity. For the i-th wind power data in the sample, calculate the first initial average distance from it to its own cluster, that is, to all other wind power data in the first category population cluster.

[0109] Then, the inter-cluster dissimilarity is calculated. For the i-th wind power data in the sample, the second initial average distance from it to other clusters, i.e., the second category population clusters, is calculated.

[0110] Finally, the initial profile coefficients of the wind power dataset to be clustered can be calculated based on the first and second initial average distances.

[0111] The first initial average distance and the second initial average distance are calculated using a preset Euclidean distance calculation method. The calculation formula for the preset Euclidean distance calculation method is shown in the following relation (1):

[0112]

[0113] In the formula: X 1 Sequence 1 represents cluster X; X 2 Let T represent sequence 2 of cluster X; T represents the total periodic length of the sequence. This represents the value of sequence 1 at time t.

[0114] In some alternative implementations, step b4 above includes:

[0115] Step b41: Determine the target first initial average distance corresponding to each wind power data based on each first initial average distance.

[0116] Step b42: Determine the target second initial average distance corresponding to each wind power data based on each second initial average distance.

[0117] Step b43: Based on the first initial average distance and the second initial average distance of each target, determine the third sample profile coefficient relationship.

[0118] Step b44: Determine the first sample profile coefficient relationship based on the third sample profile coefficient relationship.

[0119] Wherein, the first initial average distance of the target is the average of all first initial average distances, denoted as a(i); the second initial average distance of the target is the minimum of all second initial average distances, denoted as b(i).

[0120] Specifically, based on the calculation formula of the preset Euclidean distance calculation method shown in the above relation (1), the first initial average distance a(i) and the second initial average distance b(i) of the target can be calculated. The specific calculation formulas are as follows: relation (2) and (3):

[0121]

[0122]

[0123] In the formula: M represents the total number of objects in cluster X; y, z, and c represent different clusters.

[0124] Furthermore, based on the first initial average distance a(i) and the second initial average distance b(i) of the target calculated according to the above relations (2) and (3), the third sample contour coefficient relation can be obtained, as shown in the following relation (4):

[0125]

[0126] In the formula: S(i) represents the initial profile coefficient S(i) corresponding to each wind power data.

[0127] Finally, based on the above relation (4), the corresponding first sample contour coefficient relation can be determined, as shown in the following relation (5):

[0128]

[0129] In the formula: S represents the initial profile coefficient of the wind power data set to be clustered.

[0130] In some optional implementations, step S4022 above includes:

[0131] Step c1: Introduce a preset period length into the preset Euclidean distance calculation method to obtain the target Euclidean distance calculation method.

[0132] Step c2: Optimize the contour coefficient relationship of the first sample using the target Euclidean distance calculation method to obtain the contour coefficient relationship of the second sample.

[0133] First, a preset period length T is introduced. s Then, the calculation formula for the target Euclidean distance calculation method shown in the above relation (1) becomes the following relation (6):

[0134]

[0135] Furthermore, the average distance a(i) of the first target can be calculated based on the above relation (6) and in combination with the above relations (2) and (3). s The average distance b(i) between the second target and the target s .

[0136] Then, by improving the above relation (4), we can obtain the following relation (7):

[0137]

[0138] In the formula: This represents the target profile coefficient corresponding to each wind power data point.

[0139] Finally, the relationship of the second sample contour coefficient can be obtained, as shown in the following relationship (8):

[0140]

[0141] In the formula: S new This represents the target profile coefficient of the wind power data set to be clustered.

[0142] Step S403: Based on the preset period length, and using the target sample profile coefficient as the optimization objective, a preset particle swarm optimization algorithm is used to optimize the wind power dataset to be clustered, obtaining the target number of categories and the target data category length. For details, please refer to [link to relevant documentation]. Figure 1 Step S103 of the illustrated embodiment will not be described again here.

[0143] Step S404: Based on the target number of categories and the target data category length, a preset clustering algorithm is used to cluster the wind power dataset to be clustered, obtaining the target clustering results. For details, please refer to [link to relevant documentation]. Figure 1 Step S104 of the illustrated embodiment will not be described again here.

[0144] Step S405: Based on the target clustering results, the first sample profile coefficient is obtained by calculating the second sample profile coefficient relationship.

[0145] Specifically, using the improved second sample silhouette coefficient relation shown in the above relation (8), the target clustering result corresponding to the target clustering result can be calculated.

[0146] Step S406: Use the first sample silhouette coefficient to evaluate the target clustering result, and adjust the target clustering result according to the evaluation result until the target clustering result that meets the conditions is obtained.

[0147] Specifically, the clustering effect of the target clustering result is evaluated using the first sample silhouette coefficient. When the clustering effect does not meet the requirements, the first sample silhouette coefficient is returned to the preset particle swarm algorithm, and the above steps S402 to S404 are repeated until the preset particle swarm algorithm outputs the optimal number of target categories and the target data category length. Under the target number of categories and the target data category length, the K-means clustering algorithm is used to cluster the wind power dataset to be clustered to obtain the best target clustering result, that is, the target clustering result that meets the conditions.

[0148] Step S407: Obtain the preset number of categories and the preset data length.

[0149] The preset number of categories and the preset data length are manually specified.

[0150] Step S408: Based on the preset number of categories and the preset data length, the preset clustering algorithm is used to perform clustering processing on the wind power dataset to be clustered, and the preset clustering results are obtained.

[0151] Specifically, based on the preset number of categories and the preset data length, the K-means clustering algorithm with improved Euclidean distance is used to cluster the wind power dataset to be clustered, and the corresponding preset clustering results can be obtained.

[0152] Step S409: Compare the preset clustering results with the target clustering results that meet the conditions, and determine the clustering effect of the target clustering results that meet the conditions based on the comparison results.

[0153] Specifically, the clustering effect of the optimal target clustering result obtained in step S406 can be determined by comparing the preset clustering result with the target clustering result.

[0154] The greater the difference between the preset clustering result and the above-mentioned optimal target clustering result, the better the clustering effect of the above-mentioned optimal target clustering result.

[0155] The wind power data clustering method provided in this embodiment uses initial profile coefficients to determine the initial number of categories and the initial data category length, providing a basis for subsequent quantitative judgment of the classification results. Simultaneously, it uses the first and second initial average distances to reflect cohesion and separation, allowing the calculated initial profile coefficients to better quantify the classification results. Compared to manually specifying the category length and number of categories, it achieves better classification results.

[0156] This embodiment provides a wind power data clustering method for wind farms; Figure 5 This is a flowchart of a wind power data clustering method according to an embodiment of the present invention, such as... Figure 5 As shown, the process includes the following steps:

[0157] Step S501: Obtain the wind power data set to be clustered and the historical target power generation data set for the wind farm. For details, please refer to [link to relevant documentation]. Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0158] Step S502: Based on a preset period length, the target sample profile coefficients corresponding to the wind power power dataset to be clustered are obtained through a preset particle swarm optimization algorithm. For details, please refer to [link to relevant documentation]. Figure 4 Step S402 of the illustrated embodiment will not be described again here.

[0159] Step S503: Based on the preset period length, with the target sample profile coefficient as the optimization target, the preset particle swarm optimization algorithm is used to optimize the wind power dataset to be clustered, so as to obtain the target classification number and target data classification length.

[0160] Specifically, step S503 includes:

[0161] Step S5031: Obtain the preset range of number of categories and the range of cycle length based on the preset cycle length.

[0162] The preset category range is a pre-defined range of the number of data categories, such as 3-7 categories.

[0163] The period length range is the range of data period lengths with a preset period length as the boundary, such as 48-192 hours.

[0164] Step S5032: Using the preset range of classification quantity and period length as constraints, and the target sample profile coefficient as the optimization target, the preset particle swarm optimization algorithm is used to optimize the wind power dataset to be clustered, so as to obtain the target classification quantity and target data classification length.

[0165] Specifically, when using the preset particle swarm optimization algorithm to optimize the wind power dataset to be clustered, the period length and number of categories of the wind power dataset to be clustered are used as target optimization variables, and the preset range of the number of categories and the range of the period length are used as constraints to constrain the optimization process. When the target profile coefficient is maximized, the preset particle swarm optimization algorithm outputs the optimal number of clusters and the length of the clustered data, that is, the target number of categories and the target data length.

[0166] By using the period length and number of categories of the wind power dataset to be clustered as target optimization variables, we can find better time lengths and number of categories, so that the classification results can better serve the prediction and simulation of wind power.

[0167] For example, if the simulation or prediction time for wind power is 7 days, then 7 predictions are needed if the classification period is 1 day, but only 1 prediction is needed if the classification period is 7 days.

[0168] Step S504: Based on the target number of categories and the target data category length, a preset clustering algorithm is used to cluster the wind power dataset to be clustered, obtaining the target clustering results. For details, please refer to [link to relevant documentation]. Figure 1 Step S104 of the illustrated embodiment will not be described again here.

[0169] The wind power data clustering method provided in this embodiment introduces a preset period length to improve the profile coefficient, enabling the improved target profile coefficient to better quantify and judge the quality of the classification results. Furthermore, by using the preset period length and the initial number of classifications as target variables, a better target number of classifications and target data classification length can be found, resulting in better classification performance.

[0170] In one example, an adaptive classification method for wind power based on sample profile coefficients and particle swarm optimization is provided, such as... Figure 6 As shown, the specific steps are as follows:

[0171] 1. Obtain 30 days of historical power generation data from the wind farm, preprocess the data, and interpolate missing and outlier values ​​using linear interpolation.

[0172] 2. Use the pandas library in Python to convert the data into a dataframe format, containing date and power columns.

[0173] 3. Define the parameters of the particle swarm optimization algorithm, including particle dimension, number of particles, number of iterations, etc.

[0174] 4. Define and set the period length T s Within this period length, using the optimal silhouette coefficient (the larger the value) as the optimization function, the particle swarm optimization algorithm is used to find the optimal number of classifications C. s .

[0175] 5. Define the range for the data length and the number of categories. For example, the data length can be defined as 48-192 hours, and the number of categories can be defined as 3-7 categories.

[0176] 6. Using the improved profile coefficient (the larger the value) as the optimization function, and the period length T and the number of categories C as optimization variables, the optimal solution is found. The data length and number of categories corresponding to the optimal solution (the value of the improved profile coefficient) are the optimal data length and number of categories for clustering.

[0177] 7. Input the results of the particle swarm optimization algorithm (data length and number of classes) as parameters into the improved Euclidean distance K-means clustering algorithm to obtain the clustering results.

[0178] Furthermore, taking the aforementioned wind farm as an example, clustering was performed using both the manually specified length or number of categories and the wind power adaptive classification method based on sample profile coefficient and particle swarm algorithm provided in this example, and the results were compared and analyzed.

[0179] Specifically, in the manually specified method, the specified time length is 24 hours, that is, according to expert experience, clustering is performed using a day as the time unit. Five categories are specified. After clustering using the Euclidean K-means method, 10 curves are selected for each category, and the clustering results are plotted graphically, as shown below. Figure 7 As shown.

[0180] from Figure 7 As can be seen from the data, in the clustering results with a specified time length of 24 hours and 5 categories, the characteristics of each category curve are somewhat different, and the boundaries between categories can be observed. However, the cross-blurring of the classification results is quite serious, and the classification effect is unsatisfactory.

[0181] Furthermore, clustering is performed using the wind power adaptive classification method based on sample profile coefficients and particle swarm optimization provided in this example: a time length T is set. s With 96 points (24 hours), the optimal classification length was found to be 48 points (12 hours), and the optimal number of categories was 4. Ten curves were selected for each category, and the classification results are as follows. Figure 8 As shown.

[0182] from Figure 8 As can be seen, compared with the manually specified classification results, the wind power adaptive classification method based on sample silhouette coefficient and particle swarm optimization algorithm provided in this example has better classification features. Clear boundaries can be observed between classes, and the cross-classification ambiguity is significantly reduced, resulting in satisfactory classification performance. Therefore, the classification results of the two schemes were measured using the improved silhouette coefficient. The result for the manually specified scheme was 0.22, while the result for the proposed scheme was 0.51, demonstrating a significant improvement in the improved sample silhouette coefficient.

[0183] This embodiment also provides a wind power data clustering device, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0184] This embodiment provides a wind power data clustering device for wind farms; such as Figure 9 As shown, it includes:

[0185] The acquisition module 901 is used to acquire the wind power data set to be clustered and the historical target power data set of wind farms.

[0186] The first processing module 902 is used to obtain the target sample profile coefficients corresponding to the wind power data set to be clustered by processing it with a preset particle swarm algorithm based on a preset period length.

[0187] The second processing module 903 is used to optimize the wind power dataset to be clustered based on a preset period length and with the target sample profile coefficient as the optimization target, using a preset particle swarm optimization algorithm to obtain the target classification number and target data classification length.

[0188] The third processing module 904 is used to perform clustering processing on the wind power dataset to be clustered based on the number of target categories and the length of target data categories, using a preset clustering algorithm to obtain the target clustering result.

[0189] In some optional implementations, the acquisition module 901 includes:

[0190] The first acquisition submodule is used to acquire the wind power data set to be clustered and the first historical power generation data set of the wind farm.

[0191] The preprocessing submodule is used to preprocess the first historical power generation dataset to obtain the second historical power generation dataset.

[0192] The format conversion submodule is used to convert the format of the second historical power generation dataset to generate the historical target power generation dataset.

[0193] In some alternative implementations, the first processing module 902 includes:

[0194] The first determination submodule is used to determine the relationship of the first sample contour coefficients based on a preset particle swarm algorithm and a preset Euclidean distance calculation method.

[0195] The optimization submodule is used to optimize the contour coefficient relationship of the first sample based on the preset period length and the preset Euclidean distance calculation method to obtain the contour coefficient relationship of the second sample.

[0196] The second determination submodule is used to determine the target sample profile coefficients corresponding to the wind power power dataset to be clustered based on the second sample profile coefficient relationship.

[0197] In some alternative implementations, the first determining submodule includes:

[0198] The first processing unit is used to process the wind power dataset to be clustered based on a preset particle swarm optimization algorithm to generate a first-class population cluster and a second-class population cluster.

[0199] The first calculation unit is used to calculate the first initial average distance between each wind power data in the first category population cluster and other wind power data in the first category population cluster using a preset Euclidean distance calculation method.

[0200] The second calculation unit is used to calculate the second initial average distance between each wind power data in the first category population cluster and the second category population cluster using a preset Euclidean distance calculation method.

[0201] The first determining unit is used to determine the first sample profile coefficient relationship based on each first initial average distance and each second initial average distance.

[0202] The second determining unit is used to determine the first sample profile coefficient relationship based on each first initial average distance and each second initial average distance.

[0203] In some optional implementations, the second determining unit includes:

[0204] The first determining subunit is used to determine the target first initial average distance corresponding to each wind power data based on each first initial average distance.

[0205] The second determining subunit is used to determine the target second initial average distance corresponding to each wind power data based on each second initial average distance.

[0206] The third determining sub-unit is used to determine the third sample profile coefficient relationship based on the first initial average distance and the second initial average distance of each target.

[0207] The fourth determining sub-unit is used to determine the first sample profile coefficient relationship based on the third sample profile coefficient relationship.

[0208] In some alternative implementations, the optimization submodule includes:

[0209] An introduction unit is used to introduce a preset period length into a preset Euclidean distance calculation method to obtain the target Euclidean distance calculation method.

[0210] The optimization unit is used to optimize the contour coefficient relationship of the first sample using the target Euclidean distance calculation method to obtain the contour coefficient relationship of the second sample.

[0211] In some alternative implementations, the second processing module 903 includes:

[0212] The third acquisition submodule is used to acquire a preset range of categories and a range of cycle lengths based on a preset cycle length.

[0213] The optimization processing submodule is used to optimize the wind power dataset to be clustered by using a preset particle swarm optimization algorithm, with the preset range of the number of categories and the range of the period length as constraints and the target sample profile coefficient as the optimization target, to obtain the target number of categories and the target data category length.

[0214] In some alternative implementations, the preprocessing submodule includes:

[0215] The determination unit is used to identify null and abnormal data in the first historical power generation dataset.

[0216] The second processing unit is used to perform interpolation processing on null and outlier data using linear interpolation methods to obtain the second historical power generation dataset.

[0217] In some alternative implementations, the wind power data clustering device further includes:

[0218] The calculation module is used to calculate the first sample profile coefficient based on the target clustering results and the second sample profile coefficient relationship.

[0219] The adjustment module is used to evaluate the target clustering results using the silhouette coefficient of the first sample, and adjust the target clustering results according to the evaluation results until the target clustering results that meet the conditions are obtained.

[0220] In some alternative implementations, the wind power data clustering device further includes:

[0221] The first acquisition module is used to acquire the preset number of categories and the preset data length.

[0222] The fourth processing module is used to perform clustering processing on the wind power dataset to be clustered based on the preset number of categories and the preset data length, and to obtain the preset clustering results.

[0223] The comparison and determination module is used to compare the preset clustering results with the target clustering results that meet the conditions, and determine the clustering effect of the target clustering results that meet the conditions based on the comparison results.

[0224] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0225] In this embodiment, the wind power data clustering device is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0226] This invention also provides a computer device having the above-described features. Figure 9 The wind power data clustering device shown is shown.

[0227] Please see Figure 10 , Figure 10 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 10 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 10 Take a processor 10 as an example.

[0228] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0229] The memory 20 stores instructions executable by at least one processor 10 to cause at least one processor 10 to perform the method shown in the above embodiments.

[0230] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0231] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0232] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0233] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0234] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A wind power data clustering method for wind farms; characterized in that, The method includes: Obtain the wind power data set to be clustered and the historical target power generation data set of the wind farm station; Based on a preset period length, the target sample profile coefficients corresponding to the wind power power dataset to be clustered are obtained after processing by a preset particle swarm algorithm. Based on the preset period length, with the target sample contour coefficient as the optimization target, the preset particle swarm optimization algorithm is used to optimize the wind power dataset to be clustered, so as to obtain the target number of classifications and the target data classification length. Based on the target number of categories and the target data category length, the wind power dataset to be clustered is clustered using a preset clustering algorithm to obtain the target clustering result; Among them, based on a preset period length, and after processing with a preset particle swarm optimization algorithm, the target sample contour coefficients corresponding to the wind power data set to be clustered are obtained, including: The wind power dataset to be clustered is processed based on the preset particle swarm optimization algorithm to generate a first category population cluster and a second category population cluster. The first initial average distance between each wind power data point in the first category of population cluster and other wind power data points in the first category of population cluster is calculated using a preset Euclidean distance calculation method. The second initial average distance between each wind power data point in the first category of population cluster and the second category of population cluster is calculated using the preset Euclidean distance calculation method. Based on each of the first initial average distances and each of the second initial average distances, the first sample contour coefficient relationship is determined; By incorporating the preset period length into the preset Euclidean distance calculation method, a target Euclidean distance calculation method is obtained; Using the target Euclidean distance calculation method, the first sample contour coefficient relationship is optimized to obtain the second sample contour coefficient relationship; Based on the second sample profile coefficient relationship, the target sample profile coefficients corresponding to the wind power power dataset to be clustered are determined.

2. The method according to claim 1, characterized in that, Obtain the wind power data set to be clustered and the historical target power generation data set of the wind farm, including: Obtain the wind power data set to be clustered and the first historical power generation data set of the wind farm station; The first historical power generation dataset is preprocessed to obtain the second historical power generation dataset. The second historical power generation dataset is converted to a new format to generate the historical target power generation dataset.

3. The method according to claim 1, characterized in that, Based on each of the first initial average distances and each of the second initial average distances, the relationship of the first sample contour coefficients is determined, including: Determine the target first initial average distance corresponding to each wind power data based on each first initial average distance; Determine the target second initial average distance corresponding to each wind power data based on each second initial average distance; Based on the first initial average distance and the second initial average distance of each target, the third sample contour coefficient relationship is determined; Based on the third sample contour coefficient relationship, the first sample contour coefficient relationship is determined.

4. The method according to claim 1, characterized in that, Based on the preset period length, and using the target sample profile coefficient as the optimization objective, the preset particle swarm optimization algorithm is used to optimize the wind power dataset to be clustered, resulting in the target number of classifications and the target data classification length, including: Obtain a preset range of categories and a range of cycle lengths based on the preset cycle length; Using the preset range of the number of classifications and the range of the period length as constraints, and the target sample profile coefficient as the optimization target, the preset particle swarm optimization algorithm is used to optimize the wind power dataset to be clustered, so as to obtain the target number of classifications and the target data classification length.

5. The method according to claim 2, characterized in that, The first historical power generation dataset is preprocessed to obtain the second historical power generation dataset, which includes: Identify null and outlier data in the first historical power generation dataset; The null data and the abnormal data are interpolated using a linear interpolation method to obtain the second historical power generation dataset.

6. The method according to claim 1, characterized in that, The method further includes: Based on the target clustering results, the first sample contour coefficient is obtained by calculating the second sample contour coefficient relationship. The target clustering result is evaluated using the first sample silhouette coefficient, and the target clustering result is adjusted according to the evaluation result until the target clustering result that meets the conditions is obtained.

7. The method according to claim 6, characterized in that, The method further includes: Get the preset number of categories and the preset data length; Based on the preset number of categories and the preset data length, the preset clustering algorithm is used to cluster the wind power dataset to be clustered, and the preset clustering result is obtained. The preset clustering result is compared with the target clustering result that meets the conditions, and the clustering effect of the target clustering result that meets the conditions is determined based on the comparison result.

8. A wind power data clustering device for wind farms; characterized in that, The device includes: The acquisition module is used to acquire the wind power data set to be clustered and the historical target power generation data set of the wind farm. The first processing module is used to obtain the target sample profile coefficients corresponding to the wind power power dataset to be clustered by processing it with a preset particle swarm algorithm based on a preset period length. The second processing module is used to optimize the wind power dataset to be clustered based on the preset period length, with the target sample profile coefficient as the optimization target, using the preset particle swarm optimization algorithm to obtain the target classification number and target data classification length. The third processing module is used to perform clustering processing on the wind power dataset to be clustered based on the target number of categories and the target data category length using a preset clustering algorithm to obtain the target clustering result; The first processing module includes: The first processing unit is used to process the wind power dataset to be clustered based on the preset particle swarm algorithm to generate a first category population cluster and a second category population cluster. The first calculation unit is used to calculate the first initial average distance between each wind power data in the first category population cluster and other wind power data in the first category population cluster using a preset Euclidean distance calculation method; The second calculation unit is used to calculate the second initial average distance between each wind power data in the first category population cluster and the second category population cluster using the preset Euclidean distance calculation method. The first determining unit is configured to determine the first sample contour coefficient relationship based on each of the first initial average distances and each of the second initial average distances; An introduction unit is used to introduce the preset period length into the preset Euclidean distance calculation method to obtain the target Euclidean distance calculation method; The optimization unit is used to optimize the first sample contour coefficient relationship using the target Euclidean distance calculation method to obtain the second sample contour coefficient relationship. The second determining submodule is used to determine the target sample profile coefficients corresponding to the wind power power dataset to be clustered based on the second sample profile coefficient relationship.

9. The apparatus according to claim 8, characterized in that, The acquisition module includes: The first acquisition submodule is used to acquire the wind power data set to be clustered and the first historical power generation data set of the wind farm station; The preprocessing submodule is used to preprocess the first historical power generation dataset to obtain the second historical power generation dataset. The format conversion submodule is used to convert the format of the second historical power generation dataset to generate the historical target power generation dataset.

10. A computer device, characterized in that, include: A memory and a processor are interconnected, the memory stores computer instructions, and the processor executes the wind power data clustering method according to any one of claims 1 to 7 by executing the computer instructions.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the wind power data clustering method according to any one of claims 1 to 7.

Images