A method and system for constructing a wind power typical scenario set based on morphological feature clustering
By extracting multidimensional morphological features and a comprehensive distance metric matrix of wind power output, and combining K-means clustering algorithm and elbow method, the problem of insufficient morphological feature extraction and similarity measurement in wind power output curve clustering is solved, and more accurate wind power scene construction is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- POWERCHINA FUJIAN ELECTRIC POWER SURVEY & DESIGN INST CO LTD
- Filing Date
- 2026-06-03
- Publication Date
- 2026-07-31
AI Technical Summary
Existing clustering methods for wind power output curves are difficult to effectively extract morphological features, and the similarity measurement is not comprehensive enough, resulting in an inability to accurately describe the uncertainty of wind power output.
By extracting multi-dimensional morphological features from wind power output data in the time domain, frequency domain, and run-length domain, a comprehensive distance metric matrix is constructed. The optimal number of clusters is determined using the K-means clustering algorithm and the elbow method, and typical wind power scenarios are extracted.
It achieves comprehensive and multi-scale characterization of wind power output, ensuring the physical fidelity and statistical feature mapping of clustering results, and avoiding the problems of feature loss and insufficient recognition in traditional methods.
Smart Images
Figure CN122490141A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method and system for constructing a typical wind power scenario set based on morphological feature clustering. It belongs to the field of wind power output scenario generation technology. Background Technology
[0002] As the scale of wind power grid connection continues to expand, the uncertainty of wind power output poses a significant challenge to power system planning and operation. To accurately describe this uncertainty, typical wind power output scenarios are often extracted from historical data as the basis for subsequent scheduling optimization and other work. Existing methods for constructing typical wind power scenarios mainly include statistical methods, machine learning methods, and scenario clustering methods, among which clustering methods are widely used due to their intuitiveness and computational efficiency.
[0003] On the one hand, traditional clustering methods directly perform numerical clustering on time-series data, which is easily dominated by the magnitude of power output and ignores key morphological features such as the fluctuation pattern and climbing characteristics of the curve, making it impossible to effectively distinguish different power output patterns. On the other hand, most existing methods use Euclidean distance as a similarity measure, which is difficult to comprehensively reflect the similarity of curves in multiple dimensions such as amplitude correlation, waveform shape, and statistical characteristics. For example, some methods directly perform distance measurement and clustering on preprocessed time-series data without extracting any morphological features, making it difficult for the clustering results to reflect the deep morphological features of wind power output curves, focusing only on numerical similarity. In addition, some methods use a single distance measurement criterion, which can adapt to nonlinear changes in curves, but cannot comprehensively consider the similarity of curves in multiple dimensions, resulting in insufficient comprehensiveness of similarity measurement.
[0004] Therefore, there is an urgent need for a method to construct typical wind power scenarios that can effectively extract the morphological features of wind power output curves and integrate multi-dimensional similarity measurements, so as to more accurately describe the uncertainty of wind power output. Summary of the Invention
[0005] To address the problems existing in the prior art, this invention proposes a method and system for constructing a typical wind power scenario set based on morphological feature clustering.
[0006] The technical solution of the present invention is as follows: On the one hand, this invention proposes a method for constructing a typical wind power scenario set based on morphological feature clustering, including the following steps: Historical wind power output data of wind farms are acquired and preprocessed to obtain a daily wind power output sample set consisting of multiple single-day samples. Extract multidimensional morphological features from each daily sample and construct a multidimensional morphological feature vector for the daily sample. Calculate the Euclidean distance between the same type of multidimensional morphological feature vectors of each daily sample to obtain the domain distance; at the same time, calculate the Pearson correlation coefficient between each daily sample to obtain the correlation distance; in addition, calculate the Euclidean distance between each daily sample to obtain the shape distance; based on the domain distance, correlation distance and shape distance, construct a comprehensive distance metric matrix. Within a preset range of K values, the optimal number of clusters is determined using the elbow method, and the centroids are iteratively updated using the K-means clustering algorithm with the comprehensive distance metric matrix as the similarity, dividing the sample set into multiple clusters. Using the closest centroid method, the daily sample closest to the centroid in each cluster is extracted as a typical wind power scenario, and its occurrence probability is calculated to construct a typical wind power scenario set.
[0007] Preferably, the multidimensional morphological feature vector includes a time-domain feature vector, a frequency-domain feature vector, and a run-length domain feature vector, specifically: The variance, range, and interquartile range of a single-day sample are calculated as characteristic parameters of the fluctuation amplitude of the single-day sample; at the same time, the zero-crossing rate and peak-to-valley density of a single-day sample are calculated as characteristic parameters of the fluctuation frequency of the single-day sample; based on the characteristic parameters of the fluctuation amplitude and fluctuation frequency of the single-day sample, a time-domain feature vector of the single-day sample is constructed. A three-level wavelet packet decomposition is performed on the daily samples to extract the wavelet packet coefficients of each sub-band and calculate the corresponding sub-band energy. The proportion of each sub-band energy to the total energy is arranged in frequency order to construct the frequency domain feature vector of the daily samples. The daily samples are binarized using each threshold in the preset threshold set to generate a binarized sequence corresponding to each threshold. The run test is performed on each binarized sequence to count the total number of runs with consecutive identical elements. The total number of runs corresponding to each threshold is then combined in threshold order to construct the run domain feature vector of the daily samples.
[0008] Preferably, the comprehensive distance metric matrix is obtained according to the following steps: Calculate the Euclidean distance between the time-domain feature vectors, the Euclidean distance between the frequency-domain feature vectors, and the Euclidean distance between the run-length domain feature vectors of any two daily samples; then sum the three Euclidean distances by weight to obtain the domain-specific distances. Calculate the Pearson correlation coefficient between two daily sample output data sequences, and subtract the Pearson correlation coefficient from the number 1 to obtain the correlation distance; Calculate the Euclidean norm of the difference between two daily sample output data sequences to obtain the shape distance; The comprehensive distance between two daily samples is obtained by weighting and summing the regional distance, correlation distance, and shape distance using preset weighting coefficients; wherein the sum of the weighting coefficients is 1. A symmetric matrix is constructed based on the comprehensive distance between two daily samples, which serves as the comprehensive distance metric matrix.
[0009] Preferably, the step of determining the optimal number of clusters using the elbow method includes: Within the preset K value range, each candidate cluster number is traversed sequentially. For each candidate cluster number, the comprehensive distance between each daily sample in each cluster and the corresponding cluster centroid is calculated. The comprehensive distances are squared and summed to obtain the intra-cluster error sum of squares for each cluster. Then, the intra-cluster error sums of squares for all clusters under the current candidate cluster number are accumulated to obtain the total error sum of squares for the current candidate cluster number. Based on the number of each candidate cluster and the corresponding total sum of squared errors, an index trend curve is constructed to show how the sum of squared errors changes with the number of candidate clusters. The inflection point on the trend curve of the indicator is obtained. The turning point where the rate of decrease of the total sum of squared errors changes from sharp to gradual is taken as the elbow point of the curve, and the number of candidate clusters corresponding to the elbow point of the curve is determined as the optimal number of clusters.
[0010] Preferably, the method uses a comprehensive distance metric matrix as the similarity measure and iteratively updates the centroids through the K-means clustering algorithm to divide the sample set into multiple clusters, specifically: Based on the optimal number of clusters, multiple single-day samples are set as initial cluster centers in the daily wind power output sample set; Based on the comprehensive distance metric matrix, the comprehensive distance between each daily sample and each initial cluster center is traversed, and each daily sample is assigned to the cluster to which the initial cluster center with the smallest comprehensive distance belongs, thus constructing multiple initial clusters; For each cluster, retrieve the comprehensive distance between any single-day sample in the cluster and the other single-day samples in the cluster from the comprehensive distance metric matrix, sum them up, and determine the single-day sample with the smallest sum of comprehensive distances as the updated centroid of the cluster. Based on the comprehensive distance metric matrix, the comprehensive distance between each daily sample and each updated centroid is traversed again, and each daily sample is reassigned to the cluster to which the centroid with the smallest comprehensive distance belongs. The centroid update and daily sample reassignment are continued until the daily samples in each cluster no longer change or the preset number of iterations is reached, resulting in multiple clusters after partitioning.
[0011] Preferably, the method utilizes the closest centroid method to extract the daily sample from each cluster that is closest to the centroid as a typical wind power scenario, and calculates its occurrence probability, specifically: For each cluster after division, retrieve the comprehensive distance between each daily sample in the cluster and the corresponding centroid of the cluster from the comprehensive distance metric matrix; determine the daily sample with the smallest comprehensive distance as the typical wind power scenario corresponding to the cluster; The total number of daily samples contained in each cluster is counted, and the proportion of the total number of daily samples contained in each cluster to the total number of daily samples in the daily wind power output sample set is calculated. This proportion is determined as the probability of occurrence corresponding to a typical wind power scenario.
[0012] On the other hand, this invention also proposes a system for constructing a typical wind power scenario set based on morphological feature clustering, which includes the following modules: The data acquisition and preprocessing module is used to acquire historical power output data of wind farms and preprocess it to obtain a daily wind power output sample set composed of multiple single-day samples. The multidimensional morphological feature extraction module is used to extract multidimensional morphological features from each daily sample and construct a multidimensional morphological feature vector for the daily sample. The comprehensive distance metric matrix construction module is used to calculate the Euclidean distance between the same type of multidimensional morphological feature vectors of each daily sample to obtain the domain distance; at the same time, it calculates the Pearson correlation coefficient between each daily sample to obtain the correlation distance; in addition, it calculates the Euclidean distance between each daily sample to obtain the shape distance; based on the domain distance, correlation distance and shape distance, a comprehensive distance metric matrix is constructed. The sample set clustering module is used to determine the optimal number of clusters within a preset K value range using the elbow method, and to divide the sample set into multiple clusters by iteratively updating the centroids through the K-means clustering algorithm using the comprehensive distance metric matrix as the similarity. A typical wind power scenario set construction model is used to extract the daily sample closest to the centroid in each cluster as a typical wind power scenario using the closest centroid method, and calculate its occurrence probability to construct a typical wind power scenario set.
[0013] In another aspect, the present invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method as described in any embodiment of the present invention.
[0014] In another aspect, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any embodiment of the present invention.
[0015] The present invention has the following beneficial effects: 1. This invention proposes a method and system for constructing a typical wind power scenario set based on morphological feature clustering. By extracting multi-dimensional morphological features in the time domain, frequency domain, and run-length domain from the preprocessed historical daily power output data of wind farms and constructing feature vectors, the morphological evolution law and nonlinear characteristics of wind power output are characterized in a comprehensive and multi-scale manner. This effectively solves the technical problems of feature loss and insufficient recognition in traditional feature extraction methods when facing the strong randomness and spatiotemporal fluctuations of wind power.
[0016] 2. This invention proposes a method and system for constructing a typical wind power scenario set based on morphological feature clustering. By integrating the domain distance, correlation distance and shape distance between each daily sample, a global comprehensive distance metric matrix is constructed. Based on the comprehensive distance metric matrix, the values are directly retrieved and summed in the clustering iteration to dynamically update the centroid, thereby realizing a comprehensive measurement of the amplitude difference, temporal correlation and geometric shape similarity of wind power samples.
[0017] 3. This invention proposes a method and system for constructing a typical wind power scenario set based on morphological feature clustering. The elbow method is used to adaptively determine the optimal number of clusters to eliminate the uncertainty of manual parameter tuning. The closest centroid method is used to retrieve the real daily sample closest to the centroid in each cluster as a typical wind power scenario. The probability of occurrence is calculated by combining the sample proportion of each cluster. This perfectly avoids the defect of traditional mean clustering, which easily generates virtual scenarios that are detached from the actual operation boundary. It ensures that the final constructed typical wind power scenario set has both extremely high physical fidelity and statistical feature mapping degree. Attached Figure Description
[0018] Figure 1 This is a flowchart of the method for constructing a typical wind power scenario set proposed in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the wind power output per-unit standardization result proposed in Embodiment 2 of the present invention; Figure 3 This is the wavelet packet decomposition tree diagram proposed in Embodiment 2 of the present invention; Figure 4 This is the spectrum diagram of a sample after decomposition according to Embodiment 2 of the present invention; Figure 5 This is a schematic diagram of binarization of a sample under a threshold of 0.5, as proposed in Embodiment 2 of the present invention. Figure 6 This is a schematic diagram of the run test results proposed in Embodiment 2 of the present invention; Figure 7 This is a schematic diagram illustrating the determination of the number of clusters using the elbow method as proposed in Embodiment 2 of the present invention; Figure 8 This is a schematic diagram of a typical wind power scenario set proposed in Embodiment 2 of the present invention. Detailed Implementation
[0019] 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, and 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.
[0020] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.
[0021] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.
[0022] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.
[0023] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.
[0024] Example 1: See Figure 1 This embodiment proposes a method for constructing a typical wind power scenario set based on morphological feature clustering, including the following steps: S100. Obtain the original historical power output data of the wind farm, including: wind power output and total installed capacity of wind turbines at each time point, and calculate the per-unit value of wind power output, expressed by the formula: ; In the formula, Indicates the time index. Indicates the first The per-unit value of wind power output at a given moment. Indicates the first Historical wind power output data at a given moment Indicates the total installed capacity of the wind turbines; By segmenting the per-unit value of wind power output according to a 24-hour cycle, a daily wind power output sample set is obtained, expressed by the formula: ; ; In the formula, Indicates the daily sample index. Indicates the first One daily sample, They represent the first In the first single-day sample The per-unit value of wind power output at a given moment. This represents the daily wind power output sample set. This indicates the total number of samples. They represent the first A single-day sample.
[0025] S200. Based on the daily wind power output sample set, extract time-domain features, frequency-domain features, and run-length domain features from the daily samples to construct a multi-dimensional morphological feature vector. Specifically: S201. Calculate the variance, range, and interquartile range of a single-day sample as characteristic parameters of the daily sample's fluctuation range, expressed by the following formula: ; ; ; In the formula, Indicates the first The variance of a single-day sample This represents the time index of a single day's sample. Indicates the first The first single-day sample was on The per-unit value at each moment. Indicates the first The average of a single day's sample Indicates the first The range of a single-day sample. Indicates the first The maximum value of a single day's sample. Indicates the first The minimum value of a single-day sample. Indicates the first The quartile range of a single-day sample Indicates the first The upper quartiles of a single day's sample in ascending order. Indicates the first The lower quartiles of a single day's sample in ascending order; Simultaneously, the zero-crossing rate and peak-to-valley density of the daily samples are calculated as fluctuation frequency characteristic parameters of the daily samples, expressed by the following formula: ; ; In the formula, This indicates the rate at which the per-unit value of a single-day sample crosses zero. This represents the peak-to-valley density of the per-unit values for a single day's samples. Represents a counting function. Indicates the first The number of times the per-unit value crosses the average per-unit value in a single day's sample. Indicates the first Local maxima of the per-unit value in a single-day sample Indicates the first Local minimum of the per-unit value in a single-day sample Indicates the first The number of local maxima of the per-unit value in a single day's sample. Indicates the first The number of local minima of the per-unit value in a single day's sample; Based on the fluctuation amplitude and fluctuation frequency characteristic parameters of a single day's sample, a time-domain feature vector of the single day's sample is constructed, expressed by the formula: ; In the formula, Indicates the first The temporal feature vector of a single day's sample.
[0026] S202. Further, the daily sample consisting of 24 integer output per-unit values is treated as a discrete time-domain signal. A three-layer wavelet packet decomposition is used to perform progressively orthogonal filtering, decoupling eight independent sub-bands covering different frequency ranges at the third layer. The energy proportion of each sub-band is calculated and arranged in frequency order to construct the frequency domain characteristic direction, expressed by the formula: ; In the formula, Indicates the first Frequency domain morphological feature vector of a single daily sample They represent the first, second, and third parts respectively. The frequency band energy percentage corresponding to the per-unit value in a single day after performing three-level wavelet packet decomposition.
[0027] S203, Further, according to the preset threshold set Perform on daily samples respectively Binarization and run-length validation are performed to generate binarized sequences corresponding to each threshold. Run-length validation is then applied to each binarized sequence to count the total number of runs with consecutive identical elements. The total number of runs corresponding to each threshold is then combined in threshold order to construct a run-length domain feature vector, expressed by the formula: ; In the formula, Indicates the first The run-length domain feature vector of a single-day sample They represent the first The number of runs is obtained by binarizing each daily sample according to a threshold and performing a run test.
[0028] S300. Based on the time-domain features, frequency-domain features, and run-length-domain feature vectors of a single day's samples, calculate the Euclidean distance between single-day samples, expressed by the formula: ; ; ; In the formula, Indicates different Daily sample index Indicates a single day's sample Compared with single-day samples The time-domain distance between them Indicates a single day's sample The time-domain feature vector, Indicates a single day's sample The time-domain feature vector, Describes the Euclidean norm. Indicates a single day's sample Compared with single-day samples Frequency domain distance between them Indicates a single day's sample The frequency domain eigenvectors, Indicates a single day's sample The frequency domain eigenvectors, Indicates a single day's sample Compared with single-day samples The distance between the travel ranges, Indicates a single day's sample The run-range eigenvectors, Indicates a single day's sample The run-range domain eigenvectors; Based on the preset weighted fusion weights, the domain distance is constructed, expressed by the formula: ; in: Indicates a single day's sample and The distance between the domains, The fusion weights representing the temporal distance are... The fusion weights represent the frequency domain distance. The fusion weights represent the run-range distances; ; ; ; In the formula, Indicates the length of the time-domain feature vector. This represents the length of the frequency domain eigenvector. This represents the length of the feature vector in the run-length domain.
[0029] Furthermore, the correlation distance and shape distance between daily samples are calculated, expressed by the following formula: ; ; In the formula, Indicates a single day's sample Compared with single-day samples The relevant distance between them Indicates a single day's sample , Indicates a single day's sample , Indicates a single day's sample Compared with single-day samples The Pearson correlation coefficient between them Indicates a single day's sample Compared with single-day samples The shape distance between them; The comprehensive distance is calculated by combining the domain distance, relevance distance, and shape distance, and is expressed by the formula: ; In the formula, Indicates a single day's sample Compared with single-day samples The combined distance between them; By traversing all daily sample indices, the comprehensive distance between each pair of daily samples is calculated; the results are arranged in rows and columns into a two-dimensional matrix to obtain the comprehensive distance metric matrix; each element in this matrix corresponds to the comprehensive similarity quantification value between a pair of daily samples, which is used to characterize the overall degree of difference between different wind power output samples.
[0030] S400. Based on the comprehensive distance metric matrix between daily samples, the K-means clustering algorithm is used, with the comprehensive distance in the comprehensive distance metric matrix as the similarity metric, to perform morphological clustering on the daily wind power output sample set, thereby achieving grouping and partitioning of samples with similar output patterns. Specifically: According to the preset The value range is traversed sequentially, and each value within the range is... Value; calculate the sum of squared errors, expressed by the formula: ; In the formula, Represents the sum of squared errors. Represents a cluster index. Represents the number of candidate clusters. This represents the daily sample index within the cluster. Indicates the first The first cluster within the cluster Single-day samples within a cluster, Indicates the first The centroid of a cluster, Indicates the first The first cluster within the cluster Single-day samples within each cluster and the first The combined distance between the centroids of each cluster; Based on the sum of squared errors (SSE), the clustering effect under different numbers of clusters is evaluated, and the sum of squared errors is plotted as a function of the sum of squared errors. For curves showing value changes, select the elbow point corresponding to the curve. The value is used as the optimal cluster number, denoted as . To balance the compactness of clustering with the number of categories, and to avoid the distortion of clustering results caused by too many or too few categories.
[0031] S401. Determine the optimal number of clusters. Then, K-means clustering is performed based on the comprehensive distance metric matrix to divide the sample clusters. Specifically: Randomly selected from all daily samples One sample is used as the initial cluster centroid; Traverse all daily samples, obtain the comprehensive distance between each daily sample and all initial cluster centroids based on the comprehensive distance metric matrix, and assign each daily sample to the cluster corresponding to the centroid with the smallest comprehensive distance. After the allocation is completed, the mean vector of all samples in each cluster is recalculated and used as the new cluster centroid; Repeat the above iterative process of sample allocation and centroid update until the change in cluster centroid is less than the preset threshold or the maximum number of iterations is reached, at which point the algorithm converges.
[0032] At this point, all daily samples were divided into Each cluster consists of a group of samples that are highly similar in terms of multidimensional morphological features, temporal correlation, and waveform morphology. The samples in different clusters have significant differences in output morphology, thus completing the sample grouping based on output morphology.
[0033] S403. Using the closest centroid method, select the daily sample with the smallest comprehensive distance from the centroid of each cluster as the typical wind power scenario of the current cluster; at the same time, calculate the occurrence probability of each typical wind power scenario, which is the ratio of the number of daily samples in the corresponding cluster to the total number of samples in the daily wind power output sample set. Based on typical wind power scenarios and their occurrence probabilities, a set of typical wind power scenarios is constructed, expressed by the formula: ; In the formula, This represents a typical wind power scenario set. Represents the optimal cluster number. This represents a typical wind power output scenario sample corresponding to the first cluster. This represents the probability of occurrence of a typical wind power output scenario sample corresponding to the first cluster. This represents a typical wind power output scenario sample corresponding to the second cluster. This represents the probability of occurrence of a typical wind power output scenario sample corresponding to the second cluster. Indicates the first Typical wind power output scenario samples corresponding to each cluster. Indicates the first The probability of occurrence of a typical wind power output scenario sample corresponding to a cluster.
[0034] Example 2: See Figure 2 In this embodiment, taking the annual wind power output data of a wind farm in a certain region as an example, the original historical output data is normalized to per unit value; the per-unit data is segmented to obtain a sample set consisting of several samples; wherein, the sample data is represented in a table as follows: Table 1 Sample Data Example
[0035] Furthermore, time-domain feature vectors are extracted from the samples. An example of the time-domain vector extraction results is shown in the table below: Table 2 Examples of sample time-domain vector extraction
[0036] The frequency domain feature vector is extracted by performing three-level wavelet packet decomposition on the sample. The three-level wavelet decomposition tree is as follows: Figure 3 As shown, the decomposed spectrum of a certain sample is as follows: Figure 4 As shown, an example of the sample frequency domain vector extraction results is presented in a table: Table 3. Examples of sample frequency domain vector extraction results
[0037] The samples were binarized to 0-1 values using different thresholds [0.1, 0.2, ..., 0.9], and the run-length detection was performed to extract the run-length domain feature vector. For example, using a threshold of 0.5, the binarization result and run-length detection result for a certain sample are shown below. Figure 5-6 As shown in the figure, the number of runs is 6 when the threshold is 0.5. Similarly, the extraction results of the sample's run-domain feature vectors are presented in a table as follows: Table 4. Examples of sample run-length domain vector extraction results
[0038] Then, the domain distance, feature distance, correlation coefficient distance, and shape distance between samples are calculated, and a comprehensive distance between samples is constructed. Taking the above samples as an example, the calculation results are shown in the table below: Table 5. Examples of Comprehensive Distance Calculation Results
[0039] The elbow rule is used to determine the number of clusters. Taking the preset K value range [2, 18] as an example, the elbow diagram is as follows: Figure 7 As shown. Based on the rate of change of the curve, it can be divided into three segments, with the middle segment being [4, 8]. Therefore, the elbow point can be determined as K=6.
[0040] K-means clustering algorithm was used for clustering, and a typical wind power scenario set was calculated and output. The results are as follows: Figure 8 As shown.
[0041] To further illustrate the beneficial effects of the embodiments, the clustering effects of the present invention and traditional scene clustering methods are compared in terms of cluster density, variance, correlation coefficient, and outlier ratio, and are presented in a table as follows: Table 6 Comparison of Clustering Results
[0042] Based on the above results, the method proposed in this invention achieves similar cluster mean variance to traditional clustering methods, but with improvements in cluster density, correlation coefficient, and outlier ratio. Therefore, the scene construction method proposed in this invention has better clustering performance.
[0043] Example 3: This embodiment proposes a system for constructing a typical wind power scenario set based on morphological feature clustering, which includes the following modules: The data acquisition and preprocessing module is used to acquire historical power output data of wind farms and preprocess it to obtain a daily wind power output sample set composed of multiple single-day samples. The multidimensional morphological feature extraction module is used to extract multidimensional morphological features from each daily sample and construct a multidimensional morphological feature vector for the daily sample. The comprehensive distance metric matrix construction module is used to calculate the Euclidean distance between the same type of multidimensional morphological feature vectors of each daily sample to obtain the domain distance; at the same time, it calculates the Pearson correlation coefficient between each daily sample to obtain the correlation distance; in addition, it calculates the Euclidean distance between each daily sample to obtain the shape distance; based on the domain distance, correlation distance and shape distance, a comprehensive distance metric matrix is constructed. The sample set clustering module is used to determine the optimal number of clusters within a preset K value range using the elbow method, and to divide the sample set into multiple clusters by iteratively updating the centroids through the K-means clustering algorithm using the comprehensive distance metric matrix as the similarity. A typical wind power scenario set construction model is used to extract the daily sample closest to the centroid in each cluster as a typical wind power scenario using the closest centroid method, and calculate its occurrence probability to construct a typical wind power scenario set.
[0044] Example 4: This embodiment proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the method described in any embodiment of the present invention.
[0045] Example 5: This embodiment proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in any embodiment of the present invention.
[0046] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.
[0047] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0048] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0049] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0050] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.
Claims
1. A method for constructing a typical wind power scenario set based on morphological feature clustering, characterized in that, Includes the following steps: Historical wind power output data of wind farms are acquired and preprocessed to obtain a daily wind power output sample set consisting of multiple single-day samples. Extract multidimensional morphological features from each daily sample and construct a multidimensional morphological feature vector for the daily sample. Calculate the Euclidean distance between the same type of multidimensional morphological feature vectors of each daily sample to obtain the domain distance; at the same time, calculate the Pearson correlation coefficient between each daily sample to obtain the correlation distance; in addition, calculate the Euclidean distance between each daily sample to obtain the shape distance; based on the domain distance, correlation distance and shape distance, construct a comprehensive distance metric matrix. Within a preset range of K values, the optimal number of clusters is determined using the elbow method, and the centroids are iteratively updated using the K-means clustering algorithm with the comprehensive distance metric matrix as the similarity, dividing the sample set into multiple clusters. Using the closest centroid method, the daily sample closest to the centroid in each cluster is extracted as a typical wind power scenario, and its occurrence probability is calculated to construct a typical wind power scenario set.
2. The method for constructing a typical wind power scenario set based on morphological feature clustering according to claim 1, characterized in that, The multidimensional morphological feature vector includes time-domain feature vectors, frequency-domain feature vectors, and run-length domain feature vectors, specifically: The variance, range, and interquartile range of a single-day sample are calculated as characteristic parameters of the fluctuation amplitude of the single-day sample; at the same time, the zero-crossing rate and peak-to-valley density of a single-day sample are calculated as characteristic parameters of the fluctuation frequency of the single-day sample; based on the characteristic parameters of the fluctuation amplitude and fluctuation frequency of the single-day sample, a time-domain feature vector of the single-day sample is constructed. A three-level wavelet packet decomposition is performed on the daily samples to extract the wavelet packet coefficients of each sub-band and calculate the corresponding sub-band energy. The proportion of each sub-band energy to the total energy is arranged in frequency order to construct the frequency domain feature vector of the daily samples. The daily samples are binarized using each threshold in the preset threshold set to generate a binarized sequence corresponding to each threshold. The run test is performed on each binarized sequence to count the total number of runs with consecutive identical elements. The total number of runs corresponding to each threshold is then combined in threshold order to construct the run domain feature vector of the daily samples.
3. The method for constructing a typical wind power scenario set based on morphological feature clustering according to claim 1, characterized in that, The comprehensive distance metric matrix is obtained according to the following steps: Calculate the Euclidean distance between the time-domain feature vectors, the Euclidean distance between the frequency-domain feature vectors, and the Euclidean distance between the run-length domain feature vectors of any two daily samples; then sum the three Euclidean distances by weight to obtain the domain-specific distances. Calculate the Pearson correlation coefficient between two daily sample output data sequences, and subtract the Pearson correlation coefficient from the number 1 to obtain the correlation distance; Calculate the Euclidean norm of the difference between two daily sample output data sequences to obtain the shape distance; The comprehensive distance between two daily samples is obtained by weighting and summing the regional distance, correlation distance, and shape distance using preset weighting coefficients; wherein the sum of the weighting coefficients is 1. A symmetric matrix is constructed based on the comprehensive distance between two daily samples, which serves as the comprehensive distance metric matrix.
4. The method for constructing a typical wind power scenario set based on morphological feature clustering according to claim 1, characterized in that, The method for determining the optimal number of clusters using the elbow method includes the following steps: Within the preset K value range, each candidate cluster number is traversed sequentially. For each candidate cluster number, the comprehensive distance between each daily sample in each cluster and the corresponding cluster centroid is calculated. The comprehensive distances are squared and summed to obtain the intra-cluster error sum of squares for each cluster. Then, the intra-cluster error sums of squares for all clusters under the current candidate cluster number are accumulated to obtain the total error sum of squares for the current candidate cluster number. Based on the number of each candidate cluster and the corresponding total sum of squared errors, an index trend curve is constructed to show how the sum of squared errors changes with the number of candidate clusters. The inflection point on the trend curve of the indicator is obtained. The turning point where the rate of decrease of the total sum of squared errors changes from sharp to gradual is taken as the elbow point of the curve, and the number of candidate clusters corresponding to the elbow point of the curve is determined as the optimal number of clusters.
5. The method for constructing a typical wind power scenario set based on morphological feature clustering according to claim 4, characterized in that, The method uses a comprehensive distance metric matrix as the similarity measure and iteratively updates the centroids through the K-means clustering algorithm to divide the sample set into multiple clusters. Specifically: Based on the optimal number of clusters, multiple single-day samples are set as initial cluster centers in the daily wind power output sample set; Based on the comprehensive distance metric matrix, the comprehensive distance between each daily sample and each initial cluster center is traversed, and each daily sample is assigned to the cluster to which the initial cluster center with the smallest comprehensive distance belongs, thus constructing multiple initial clusters; For each cluster, retrieve the comprehensive distance between any single-day sample in the cluster and the other single-day samples in the cluster from the comprehensive distance metric matrix, sum them up, and determine the single-day sample with the smallest sum of comprehensive distances as the updated centroid of the cluster. Based on the comprehensive distance metric matrix, the comprehensive distance between each daily sample and each updated centroid is traversed, and each daily sample is reassigned to the cluster to which the centroid with the smallest comprehensive distance belongs; The centroid is continuously updated and the daily samples are redistributed until the daily samples in each cluster no longer change or the preset number of iterations is reached, resulting in multiple clusters after partitioning.
6. The method for constructing a typical wind power scenario set based on morphological feature clustering according to claim 1, characterized in that, The method utilizes the closest centroid method to extract the daily sample from each cluster that is closest to the centroid as a typical wind power scenario, and calculates its occurrence probability, specifically: For each cluster after division, retrieve the comprehensive distance between each daily sample in the cluster and the corresponding centroid of the cluster from the comprehensive distance metric matrix; determine the daily sample with the smallest comprehensive distance as the typical wind power scenario corresponding to the cluster; The total number of daily samples contained in each cluster is counted, and the proportion of the total number of daily samples contained in each cluster to the total number of daily samples in the daily wind power output sample set is calculated. This proportion is determined as the probability of occurrence corresponding to a typical wind power scenario.
7. A system for constructing a typical wind power scenario set based on morphological feature clustering, characterized in that, Includes the following modules: The data acquisition and preprocessing module is used to acquire historical power output data of wind farms and preprocess it to obtain a daily wind power output sample set composed of multiple single-day samples. The multidimensional morphological feature extraction module is used to extract multidimensional morphological features from each daily sample and construct a multidimensional morphological feature vector for the daily sample. The comprehensive distance metric matrix construction module is used to calculate the Euclidean distance between the same type of multidimensional morphological feature vectors of each daily sample to obtain the domain distance; at the same time, it calculates the Pearson correlation coefficient between each daily sample to obtain the correlation distance; in addition, it calculates the Euclidean distance between each daily sample to obtain the shape distance; based on the domain distance, correlation distance and shape distance, a comprehensive distance metric matrix is constructed. The sample set clustering module is used to determine the optimal number of clusters within a preset K value range using the elbow method, and to divide the sample set into multiple clusters by iteratively updating the centroids through the K-means clustering algorithm using the comprehensive distance metric matrix as the similarity. A typical wind power scenario set construction model is used to extract the daily sample closest to the centroid in each cluster as a typical wind power scenario using the closest centroid method, and calculate its occurrence probability to construct a typical wind power scenario set.
8. An electronic 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 program, it implements the method as described in any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 6.