New energy power generation data double-layer clustering method, system and device and storage medium

Abstract

The invention discloses a new energy power generation data double-layer clustering method, system and device and a storage medium, and relates to the field of new energy power generation power prediction and data clustering analysis, and the method comprises the steps: collecting energy power generation power data, and carrying out the normalization and complement processing to obtain a sample sequence; establishing and training an improved fuzzy adaptive resonance network model; initializing K-means clustering by using the trained model, and preliminarily dividing a sample sequence; performing convergence judgment based on the preliminary result to obtain a final clustering result; the attention mechanism is introduced to improve the feature capture capability; comprehensively calculating and selecting parameters for accurate classification; dynamically updating the model to adapt to data change; the problem that the initial value of the K-mean is sensitive is avoided, and convergence speed and quality are improved; convergence is effectively judged, it is guaranteed that the clustering result is stable and effective, and a basis is provided for subsequent new energy power generation power prediction.

Description

Technical Field

[0001] The present invention relates to the field of new energy power generation power prediction and data cluster analysis, and in particular to a double-layer clustering method, system, device and storage medium for new energy power generation data. Background Art

[0002] Renewable energy generation is an environmentally friendly and renewable energy source, and its installed capacity has been steadily increasing in recent years. However, renewable energy generation is susceptible to various weather factors, such as season and temperature, resulting in significant output uncertainty. With the large-scale integration of renewable energy into the grid, accurate forecasting of renewable energy generation power is essential for the grid to address randomness and intermittency, and is of great significance to grid dispatch and operation.

[0003] Data-driven forecasting methods are currently the primary means of predicting renewable energy power generation. However, the performance of data-driven forecasting models is highly dependent on sample quality. Properly classifying samples and building differentiated forecasting models is an effective approach to improving forecast accuracy. Given the significant variability of renewable energy power curves under different weather types, using only a single renewable energy power forecasting model can result in sudden changes in weather type being interpreted as outliers, impacting neural network training and thus affecting the robustness of forecast results, leading to output smoothness and reduced forecast performance. Therefore, classifying samples by weather type and modeling them separately can improve forecast accuracy. Because a large number of historical renewable energy power samples do not contain complete and accurate weather information, clustering methods are a widely used technique for sample classification. However, currently used clustering methods share common drawbacks: first, they cannot adaptively determine the number of cluster centers; second, randomly initialized cluster centers significantly influence the clustering results. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that in the prediction of renewable energy power generation, due to incomplete and inaccurate sample weather information, traditional clustering methods cannot adaptively determine the number of cluster centers and the random initialization of cluster centers has a great impact.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a two-layer clustering method for new energy power generation data, comprising:

[0008] As a preferred solution for the two-layer clustering method of renewable energy power generation data,

[0009] The method of establishing and training an improved fuzzy adaptive resonant network model based on a new energy power generation power sample sequence includes:

[0010] Set the alert parameters, learning rate, and selection factor of the adaptive resonant network, randomly assign the attention matrix, and assign the first sample to the weight vector of the first neuron.

[0011] As a preferred solution for the two-layer clustering method of renewable energy power generation data,

[0012] The method of establishing and training an improved fuzzy adaptive resonant network model based on a new energy power generation sample sequence further includes:

[0013] The attention mechanism is introduced into the fuzzy adaptive resonant network structure, the learnable weight matrix is ​​updated through unsupervised learning, and the normalized exponential function is used to generate the attention weight vector.

[0014] The beneficial effects of this preferred technical solution are: the introduction of the attention mechanism enables the model to dynamically focus on important features in the data, suppress redundant noise, and update the weight matrix through unsupervised learning, so that the model can automatically adapt to the characteristic distribution of the data, generate a reasonable attention weight vector, and improve the model's ability to capture data features, thereby improving the accuracy of clustering.

[0015] As a preferred solution for the two-layer clustering method of renewable energy power generation data,

[0016] The method of establishing and training an improved fuzzy adaptive resonant network model based on a new energy power generation sample sequence further includes:

[0017] Calculate the selection parameter between the attention weight vector and each neuron in the category representation layer. The selection parameter is calculated based on the weighted average of the fuzzy coefficient and dynamic time warping distance between the attention weight vector and the neuron. The neuron with the largest parameter is selected as the winning neuron to determine the category to which the input sequence belongs.

[0018] The beneficial effects of this preferred technical solution are: comprehensively considering the fuzzy coefficient and dynamic time warping distance to calculate the selection parameters, which can more comprehensively measure the relationship between the attention weight vector and the neuron, accurately determine the category to which the input sequence belongs, and take into account the fuzziness and time series characteristics of the data, so that the clustering results are more consistent with the actual data distribution.

[0019] As a preferred solution for the two-layer clustering method of renewable energy power generation data,

[0020] The method of establishing and training an improved fuzzy adaptive resonant network model based on a new energy power generation sample sequence further includes:

[0021] After determining the category to which the input sequence belongs, the matching parameter is calculated. If the matching parameter is greater than the warning parameter, the attention matrix and neurons of this category are updated using unsupervised gradient descent. If the matching parameter is less than the warning parameter, new neurons that meet the conditions are searched again. When all neurons fail to meet the conditions after performing the warning test, new neurons are generated in the competitive layer using the weight vector extracted from the initial input sample.

[0022] As a preferred solution for the two-layer clustering method of renewable energy power generation data,

[0023] The K-means clustering initialization is performed based on the trained improved fuzzy adaptive resonant network model. The initialized K-means clustering algorithm is used to preliminarily divide the new energy power generation sample sequence. The preliminary classification results include:

[0024] According to the number of cluster centers given by the adaptive resonance network and the cluster centers as the initialization parameters of K-means clustering, the distances from all samples to each center are calculated, and each sample is divided into the cluster represented by the center closest to it.

[0025] The beneficial effects of this preferred technical solution are: using the number of cluster centers obtained by the improved fuzzy adaptive resonant network model and the cluster centers to initialize the K-means clustering algorithm can provide a more reasonable initial state for the K-means clustering, avoid the problem of the K-means clustering algorithm being sensitive to the initial value, and improve the convergence speed and clustering quality of the K-means clustering.

[0026] As a preferred solution for the two-layer clustering method of renewable energy power generation data,

[0027] The clustering convergence judgment based on the preliminary classification results to obtain the final clustering results includes:

[0028] Calculate the mean distance within each sample class and determine whether the mean no longer changes or whether it reaches the upper limit of the number of cycles; if the mean no longer changes or reaches the upper limit of the number of cycles, clustering ends and the final clustering result is obtained; otherwise, continue with preliminary classification.

[0029] In a second aspect, the present invention provides a two-layer clustering system for new energy power generation data, comprising:

[0030] The sample acquisition module is used to collect actual energy generation power data, perform normalization and complement processing, and obtain a new energy generation power sample sequence;

[0031] The fuzzy adaptive resonant network modeling module is used to establish and train an improved fuzzy adaptive resonant network model based on a new energy power generation power sample sequence;

[0032] The preliminary classification module is used to initialize K-means clustering based on the trained improved fuzzy adaptive resonant network model, and to perform preliminary classification on the renewable energy power generation sample sequence using the initialized K-means clustering algorithm to obtain preliminary classification results.

[0033] The clustering convergence judgment module is used to perform clustering convergence judgment based on the preliminary classification results to obtain the final clustering results.

[0034] In a third aspect, the present invention provides an electronic device, comprising:

[0035] memory and processor;

[0036] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the one or more programs are executed by the one or more processors, the one or more processors implement the two-layer clustering method for new energy power generation data as described in the present invention.

[0037] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the two-layer clustering method for new energy power generation data.

[0038] The present invention has the following beneficial effects: the present invention reasonably sets and assigns key parameters such as the warning parameter and learning rate of the adaptive resonant network, provides stable starting conditions for model training, and makes the training process repeatable and controllable; introduces an attention mechanism into the network structure, generates an attention weight vector through unsupervised learning, can dynamically focus on important data features, suppress redundant noise, and improve the model's ability to capture the characteristics of renewable energy power generation data; comprehensively considers the fuzzy coefficient and dynamic time warping distance when calculating and selecting parameters, can comprehensively measure the relationship with neurons, and accurately determine the category to which the input sequence belongs by combining the time series characteristics and fuzziness of renewable energy power generation data; through warning testing and matching parameter judgment, the model parameters and structure can be dynamically updated, can adapt to changes in renewable energy power generation data with factors such as seasons and weather, and ensure clustering accuracy; uses the cluster center and number of the improved fuzzy adaptive resonant network to initialize K-means clustering, avoids the problem of K-means being sensitive to initial values, and improves convergence speed and clustering quality; calculates the mean of the intra-class distance and sets the convergence condition to determine the end of clustering, avoids unnecessary iterations, improves efficiency, ensures stable and effective clustering results, accurately reflects the internal structure of renewable energy power generation data, and provides a reliable basis for subsequent analysis and decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0040] Figure 1 This is the overall flow chart of the double-layer clustering method for new energy power generation data provided by the present invention;

[0041] Figure 2 This is the photovoltaic clustering visualization result in the simulation example of the double-layer clustering method for new energy power generation data provided by the present invention. DETAILED DESCRIPTION

[0042] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0043] Example 1, with reference to Figure 1 , which is the first embodiment of the present invention, provides a two-layer clustering method for new energy power generation data, including:

[0044] S1: Collect actual energy generation power data, perform normalization and complement processing, and obtain a new energy generation power sample sequence;

[0045] S2: Based on the renewable energy power generation power sample sequence, an improved fuzzy adaptive resonant network model is established and trained;

[0046] S3: Based on the trained improved fuzzy adaptive resonant network model, K-means clustering is initialized. The initialized K-means clustering algorithm is used to preliminarily divide the renewable energy power generation sample sequence to obtain preliminary classification results.

[0047] S4: Based on the preliminary classification results, cluster convergence judgment is performed to obtain the final clustering results.

[0048] It should be noted that through steps S1-S4, the number of cluster centers can be automatically determined, and the cluster center of each category can be given. After the training is deployed, it can continue to adaptively determine whether new cluster centers should be generated based on the new data input.

[0049] Example 2, reference Figure 1, which is an embodiment of the present invention, provides a two-layer clustering method for new energy power generation data based on the previous embodiment, including:

[0050] In this embodiment, in step S1, actual energy generation power data is collected, normalized and complemented, and the new energy generation power sample sequence obtained includes:

[0051] Collect actual energy generation power data, and denote the input data as an n-dimensional sequence P, expressed as:

[0052] P=(P1,P2,...,P n )

[0053] Normalize the data:

[0054]

[0055] Among them, p i represents the normalized sequence, p max 、p min Represent the maximum and minimum values ​​in the sequence respectively.

[0056] Then do the complement processing to generate the input vector p′ after complement processing:

[0057] p′=[p,1-p]

[0058] In this embodiment, in the above step S2, establishing and training the improved fuzzy adaptive resonant network model based on the new energy power generation sample sequence includes:

[0059] The improved fuzzy adaptive resonant network model introduces an attention mechanism into the network structure. A weighted average calculation method considering the DTW distance (Dynamic Time Warping Distance) is designed in the calculation of sample category selection parameters. When the alert test fails, the weight vector extracted from the attention matrix is ​​used as the new neuron weight.

[0060] Set the alert parameter ρ of the adaptive resonance network, the learning rate β of the network, the selection factor δ, and randomly assign the attention matrix W∈r 2N×2N , and assign the first sample to the weight vector of the first neuron.

[0061] Specifically, the attention mechanism layer is used to dynamically learn feature weights and suppress redundant noise;

[0062] The attention mechanism layer adds a learnable weight matrix W∈R 2N×2N , can be updated through unsupervised learning, using softmax to generate the attention weight vector α, and the calculation formula is as follows:

[0063] α=softmax(p′·W)

[0064] Softmax(·) is a normalized exponential function, which is calculated as follows:

[0065]

[0066] In another possible implementation, a multi-head attention mechanism can be used. The input vector p′ is subjected to multiple linear transformations to obtain multiple query, key, and value matrices. The attention weights are then calculated for each of them. Finally, the results of multiple attention heads are concatenated and passed through a linear layer to obtain the final attention weight vector α.

[0067] The competition layer is used to calculate the membership between the output of the attention mechanism layer and the neurons in the category representation layer, and to perform a warning test on the data. The attention weight vector α obtained by the attention mechanism layer is applied to the competition layer to calculate the selection parameter T between the vector α and each neuron in the category representation layer. i , select the neuron with the largest parameter as the winning neuron, which is the category to which the input sequence P belongs.

[0068] Specifically, calculate the vector α and the neuron w i The fuzzy coefficient ε between i :

[0069]

[0070] The || operator represents the sum of the elements of the vector, δ is the selection factor, which is generally a constant slightly greater than 0 and can be regarded as a regularization parameter that penalizes large weights, α∧w i =min(α,w i ).

[0071] Calculate the vector α and neuron w i The DTW distance between i , select parameter T i The calculation is as follows:

[0072]

[0073] In another possible implementation, when calculating the selection parameter Ti, in addition to considering the fuzzy coefficient ε i and DTW distance d i , and also introduce the time feature weight t of the sample i First, the time feature weight t is calculated based on the timestamp of the sample iFor example, different weights can be assigned according to different time periods of the day (such as peak and trough periods). Then, the calculation formula for the selection parameter is updated as

[0074] After determining the category to which the vector α belongs, a warning test is performed to calculate the matching parameter M i , the calculation formula is as follows:

[0075]

[0076] Let the warning parameter be ρ∈[0,1], if M i If it is greater than ρ, then the attention matrix of this class is updated using unsupervised gradient descent, and the neuron w of this class is updated at the same time. i , the calculation formula is as follows:

[0077] w i (new)=(1-β)w i +β(α∧w i )

[0078] Among them, β∈[0,1] represents the learning rate of the model.

[0079] If M i If it is less than ρ, then search for new neurons that meet the conditions again.

[0080] The results of the test are reflected in the category representation layer; when all neurons do not meet the conditions after performing the alert test, a new neuron w is generated in the competition layer using the weight vector extracted from the initial input sample j , that is, set w j =α.

[0081] New samples are input one by one, and the operations of the attention mechanism layer, competition layer and category representation layer are repeated until the labels of all samples are given or the upper limit of the number of cycles is reached. Finally, the number of cluster centers given by the improved fuzzy adaptive resonance network and the weight vector of each cluster center are obtained.

[0082] In this embodiment, in the above step S3, K-means clustering initialization is performed based on the trained improved fuzzy adaptive resonant network model, and the new energy power generation sample sequence is preliminarily divided by the initialized K-means clustering algorithm, and the preliminary classification results obtained include:

[0083] The number of cluster centers and the cluster centers given by the adaptive resonance network are used as the initialization parameters of K-means clustering.

[0084] The Euclidean distance method is used to calculate the distance between all samples and each center, and each sample is divided into the cluster represented by the center closest to it.

[0085] In this embodiment, the clustering convergence judgment is performed based on the preliminary classification results in the above step S4, and the final clustering result includes:

[0086] Calculate the mean distance within each sample class and determine whether the mean is stable or has reached the upper limit of the number of cycles. If the mean is stable or has reached the upper limit of the number of cycles, clustering ends and the final clustering result is obtained, which can be used for subsequent applications such as renewable energy power generation prediction. Otherwise, return to S3 to continue preliminary classification.

[0087] In another possible implementation, in addition to calculating the mean intra-class distance, the variance within the class can also be calculated. For each cluster, the variance of its sample vector relative to the cluster center is calculated, and the variance and mean are comprehensively considered to determine whether the clustering has converged. If the mean and variance of the intra-class distance change less than a certain threshold (such as 0.01) in several consecutive iterations (such as 5 times), or the upper limit of the number of cycles is reached, the clustering ends and the final clustering result is obtained; otherwise, return to step S3 to continue the preliminary classification.

[0088] Example 3. The above is a schematic scheme of the two-layer clustering method for renewable energy power generation data of this embodiment. It should be noted that the technical scheme of the two-layer clustering system for renewable energy power generation data and the technical scheme of the two-layer clustering method for renewable energy power generation data described above are based on the same concept. For details not described in detail in the technical scheme of the two-layer clustering system for renewable energy power generation data in this embodiment, please refer to the description of the technical scheme of the two-layer clustering method for renewable energy power generation data described above.

[0089] This embodiment also provides a two-layer clustering system for new energy power generation data, including:

[0090] The sample acquisition module is used to collect actual energy generation power data, perform normalization and complement processing, and obtain a new energy generation power sample sequence;

[0091] The fuzzy adaptive resonant network modeling module is used to establish and train an improved fuzzy adaptive resonant network model based on a new energy power generation power sample sequence;

[0092] The preliminary classification module is used to initialize K-means clustering based on the trained improved fuzzy adaptive resonant network model, and to perform preliminary classification on the renewable energy power generation sample sequence using the initialized K-means clustering algorithm to obtain preliminary classification results.

[0093] The clustering convergence judgment module is used to perform clustering convergence judgment based on the preliminary classification results to obtain the final clustering results.

[0094] This embodiment further provides an electronic device applicable to a two-layer clustering method for new energy power generation data, including:

[0095] Memory and processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the two-layer clustering method for new energy power generation data proposed in the above embodiment.

[0096] This embodiment further provides a storage medium storing a computer program, which, when executed by a processor, implements the two-layer clustering method for renewable energy power generation data proposed in the above embodiment.

[0097] The storage medium proposed in this embodiment and the two-layer clustering method for new energy power generation data proposed in the above embodiment belong to the same inventive concept. For technical details not described in detail in this embodiment, please refer to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0098] Example 4, with reference to Figure 2 , is an embodiment of the present invention, which provides a two-layer clustering method for new energy power generation data. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0099] Taking the clustering of photovoltaic curves as an example, we used 90 days of measured data from a photovoltaic power station in a certain region in December, January, and February to organize the training samples. The data was sampled at a 5-minute interval, resulting in 288 points per day. Since photovoltaic output is almost zero at night, and considering the local sunrise and sunset times, only the 144 points recorded between 7:00 AM and 7:00 PM were used for subsequent analysis and calculations.

[0100] The 144 PV power points per day are clustered using the process in Section 3.2, with the warning coefficient ρ set to 0.5, the adaptive resonant network learning rate β set to 0.65, and the selection factor δ set to 0.01. The specific process is as follows.

[0101] (1) Randomly initialize the attention matrix W 288×288

[0102] (2) Randomly select a sample as the first center point w1 for initialization. Assume that the first input data is P1. After data normalization and complement processing, the sequence p1 is formed:

[0103] p1=[0.0041,0.0052,…,0.9959] 1×288

[0104] (3) The sequence p1 is multiplied by the attention matrix W through the attention mechanism layer to obtain the weight vector α:

[0105]

[0106] =[0.005,…,0.000018]

[0107] (4) First calculate the fuzzy coefficient ε between vector α and neuron w1 i :

[0108] ε1=0.006

[0109] Then calculate the DTW distance d1 between the vector α and the neuron w1:

[0110] d1=8.18

[0111] (5) Since there is only one center point at this time, the input sequence P1 belongs to category 1. According to formula (9), the alert test is performed and the calculation results are as follows:

[0112] M1=0.4465<0.5

[0113] The input sample P1 fails the alert test, and the weight vector α is used as the new neuron w2:

[0114] w2=[0.00006,…,0.00003]

[0115] The above process is repeated until all samples are traversed and three cluster centers are finally formed.

[0116] (6) The number of clusters of the K-means clustering algorithm is set to 3, and the weight vectors of the three neurons obtained from the improved fuzzy adaptive resonant network model are passed to the K-means algorithm as the initial values ​​of the cluster centers to carry out cluster analysis.

[0117] Finally, the visualization results of daily power curves of different categories are as follows Figure 2 shown.

[0118] The clustering results show that the sample classification is reasonable and samples of the same type have significant similarities. The classification results are strongly correlated with the weather type. The three types correspond to sunny mode, rainy mode and cloudy mode, respectively, which verifies the effectiveness of this method.

[0119] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A two-layer clustering method for renewable energy power generation data, characterized in that: include: Collect actual energy generation power data, perform normalization and complement processing, and obtain a new energy generation power sample sequence; Based on the new energy power generation power sample sequence, an improved fuzzy adaptive resonant network model is established and trained; Based on the trained improved fuzzy adaptive resonant network model, K-means clustering initialization is performed. The initialized K-means clustering algorithm is used to preliminarily divide the renewable energy power generation sample sequence and obtain preliminary classification results. Based on the preliminary classification results, clustering convergence judgment is performed to obtain the final clustering results.

2. A two-layer clustering method for new energy power generation data according to claim 1, characterized in that: The method of establishing and training an improved fuzzy adaptive resonant network model based on a new energy power generation power sample sequence includes: Set the alert parameters, learning rate, and selection factor of the adaptive resonant network, randomly assign the attention matrix, and assign the first sample to the weight vector of the first neuron.

3. A two-layer clustering method for new energy power generation data according to claim 2, characterized in that: The method of establishing and training an improved fuzzy adaptive resonant network model based on a new energy power generation sample sequence further includes: The attention mechanism is introduced into the fuzzy adaptive resonant network structure, the learnable weight matrix is ​​updated through unsupervised learning, and the normalized exponential function is used to generate the attention weight vector.

4. A two-layer clustering method for new energy power generation data according to claim 3, characterized in that: The method of establishing and training an improved fuzzy adaptive resonant network model based on a new energy power generation sample sequence further includes: Calculate the selection parameter between the attention weight vector and each neuron in the category representation layer. The selection parameter is calculated based on the weighted average of the fuzzy coefficient and dynamic time warping distance between the attention weight vector and the neuron. The neuron with the largest parameter is selected as the winning neuron to determine the category to which the input sequence belongs.

5. A two-layer clustering method for new energy power generation data according to claim 4, characterized in that: The method of establishing and training an improved fuzzy adaptive resonant network model based on a new energy power generation sample sequence further includes: After determining the category to which the input sequence belongs, the matching parameter is calculated. If the matching parameter is greater than the warning parameter, the attention matrix and neurons of this category are updated using unsupervised gradient descent. If the matching parameter is less than the warning parameter, new neurons that meet the conditions are searched again. When all neurons fail to meet the conditions after performing the warning test, new neurons are generated in the competitive layer using the weight vector extracted from the initial input sample.

6. A two-layer clustering method for renewable energy power generation data according to claim 5, characterized in that: The K-means clustering initialization is performed based on the trained improved fuzzy adaptive resonant network model. The initialized K-means clustering algorithm is used to preliminarily divide the new energy power generation sample sequence. The preliminary classification results include: According to the number of cluster centers given by the adaptive resonance network and the cluster centers as the initialization parameters of K-means clustering, the distances from all samples to each center are calculated, and each sample is divided into the cluster represented by the center closest to it.

7. A two-layer clustering method for renewable energy power generation data according to claim 6, characterized in that: The clustering convergence judgment based on the preliminary classification results to obtain the final clustering results includes: Calculate the mean distance within each sample class and determine whether the mean no longer changes or whether it reaches the upper limit of the number of cycles; if the mean no longer changes or reaches the upper limit of the number of cycles, clustering ends and the final clustering result is obtained; otherwise, continue with preliminary classification.

8. A two-layer clustering system for new energy power generation data, applying the method according to any one of claims 1 to 7, characterized in that: include: The sample acquisition module is used to collect actual energy generation power data, perform normalization and complement processing, and obtain a new energy generation power sample sequence; The fuzzy adaptive resonant network modeling module is used to establish and train an improved fuzzy adaptive resonant network model based on a new energy power generation power sample sequence; The preliminary classification module is used to initialize K-means clustering based on the trained improved fuzzy adaptive resonant network model, and to perform preliminary classification on the renewable energy power generation sample sequence using the initialized K-means clustering algorithm to obtain preliminary classification results. The clustering convergence judgment module is used to perform clustering convergence judgment based on the preliminary classification results to obtain the final clustering results.

9. An electronic device, characterized in that: include: memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that It stores computer-executable instructions, which implement the steps of the method according to any one of claims 1 to 7 when executed by a processor.

Images