A wind power forecasting method based on weighted fuzzy clustering

Through weighted fuzzy clustering and gray correlation screening, combined with backpropagation neural network, the problems of insufficient prediction accuracy of wind power and lack of comprehensive clustering results in the existing technology are solved, and more efficient wind power prediction and Accuracy of clustering results.

CN114219179BActive Publication Date: 2024-11-12INST OF ENERGY HEFEI COMPREHENSIVE NAT SCI CENT (ANHUI ENERGY LAB) +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210005354.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-05
Publication Date
2024-11-12
Estimated Expiration
2042-01-05

AI Technical Summary

Technical Problem

The existing short-term wind power power prediction methods are highly dependent on numerical weather forecast accuracy and a large number of training samples, making it difficult to obtain precise meteorological data in sudden weather conditions or new wind farms, resulting in insufficient prediction accuracy. Traditional clustering methods ignore wind speed and wind direction. Differences, clustering results lack comprehensiveness and accuracy.

Method used

The wind power power prediction method based on weighted fuzzy clustering is adopted to screen key meteorological features through a gray correlation matrix, and a subset of data samples is divided using feature-weighted fuzzy C-mean clustering algorithm, and a prediction model is established in combination with the backpropagation neural network to introduce discrimination. Threshold processing boundary data and combining multiple models to determine wind power output.

Benefits of technology

It improves the prediction accuracy of the wind power model, simplifies the model input dimension, quickly converges, improves the processing speed of the clustering algorithm, accurately recognizes boundary data, and enhances the accuracy and applicability of wind power output prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114219179B_ABST
    Figure CN114219179B_ABST
Patent Text Reader

Abstract

The present invention discloses a wind power prediction method based on weighted fuzzy clustering, and the steps include: 1. Selecting strongly correlated meteorological features based on grey correlation theory; 2. Using feature-weighted fuzzy C-means clustering algorithm to divide the key meteorological feature set into data sample subsets, and adopting back-propagation neural network classification modeling; 3. Calling the model most similar to the current period feature for power prediction, and using membership function to integrate multiple prediction models to confirm its wind power output in order to improve the belonging relationship of boundary data. The present invention strengthens the matching degree between meteorological features and sample subsets from the two perspectives of classification modeling and feature matching, excavates potential connections from randomly fluctuating wind power signals, and improves the prediction accuracy of wind power models.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of short-term wind power prediction, and in particular to a wind power prediction method based on weighted fuzzy clustering. Background Art

[0002] Wind power forecasting is an important auxiliary decision-making means to ensure the timely consumption of wind power and the economic and safe operation of the power grid. Therefore, accurate and effective wind power forecasting methods have become a research hotspot. At present, short-term wind power forecasting methods can be divided into physical methods and statistical methods. Among them: the physical method uses numerical weather forecast data and combines the wind speed power characteristic curve of the wind turbine to predict the output power of the wind farm; the statistical method analyzes the mapping relationship between meteorological factors and wind power output according to historical data, and predicts the output power of the wind farm based on numerical weather forecast data. There are mainly artificial neural networks, support vector machines and other methods. The above methods are highly dependent on the accuracy of numerical weather forecasts and a large number of training samples. For sudden weather conditions or newly commissioned wind farms, it is difficult to obtain accurate meteorological data and sufficient training samples, and the model cannot ensure good prediction accuracy.

[0003] In response to the above problems, some studies have used the spatial correlation of wind farms and integrated the meteorological forecast data of multiple wind farms to deal with sudden weather conditions. The results of each model are combined in the form of weights to form the final output. However, when applying meteorological data in multiple spatial dimensions, the influence of actual topography cannot be ignored. The modeling process must consider the reduction or change of wind speed and wind direction by geographical factors, and the degree of refinement must be improved.

[0004] As a typical method of data mining, clustering algorithm can analyze the internal connection between samples based on existing historical data and effectively identify scenes with similar features. However, the traditional prediction method based on weather type clustering ignores the differences between physical quantities such as wind speed and wind direction, and lacks clustering results for multi-feature data samples. In addition, the boundary data between cluster subsets is simply divided into a single category, which lacks comprehensiveness and obviously insufficient prediction accuracy. Therefore, improving clustering methods and improving the matching degree of sample data has become a new research idea. Summary of the invention

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the present invention proposes a wind power prediction method based on weighted fuzzy clustering, in order to enhance the matching degree between meteorological characteristics and sample subsets from the two perspectives of classification modeling and feature matching, and to explore potential connections from randomly fluctuating wind power signals, thereby improving the prediction accuracy of the wind power model.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] The wind power prediction method based on weighted fuzzy clustering of the present invention is characterized by being performed in the following steps:

[0008] Step 1: Select a sample set consisting of l historical samples, each sample has n meteorological characteristics, and normalize the sample set to obtain a meteorological characteristic matrix with a dimension of n×l Among them, x ij represents the i-th meteorological feature after normalization of the j-th historical sample, i = 1, 2, ..., n, j = 1, 2, ..., l;

[0009] Step 2: Collect the power data of l historical samples corresponding to the meteorological characteristics and perform normalization to obtain the power sequence P = [p1, p2, ..., p j ,…,p l ], p j represents the normalized power of the jth historical sample;

[0010] Step 3: Use formula (1) to calculate the i-th meteorological characteristic x of the j-th historical sample: ij and the power p of the jth historical sample j Grey correlation coefficient ξ ij , thus constructing the grey incidence matrix

[0011]

[0012] In formula (1), ρ is the resolution coefficient;

[0013] Step 4: Calculate the i-th meteorological characteristic x of l historical samples according to formula (2): i Correlation degree γ with power sequence P i :

[0014]

[0015] Step 5: Sort the meteorological features in descending order according to the degree of relevance, and select the meteorological feature matrix X n×l Select m key meteorological features that dominate wind power output as the screening feature set;

[0016] Step 6: Use the feature-weighted fuzzy C-means clustering algorithm to divide the selected feature set into data sample subsets, and use the back propagation neural network to classify and model the data sample subsets:

[0017] Step 6.1, determine the optimization criterion parameter as α, the number of cluster categories as c, and the termination threshold of the iterative calculation as ε, define and initialize the current iteration number t=0, where c≥2, α≥1;

[0018] Step 6.2: Initialize the weight corresponding to the pth selected key meteorological feature of the tth iteration The membership degree between the qth historical sample and the hth sample subset in the feature set selected at the tth iteration The fuzzy cluster center element of the pth row and hth column at the tth iteration Wherein, h = 1, ..., c, p = 1, ..., m, q = 1, ..., l, m represents the number of key meteorological features screened, and l represents the number of historical samples;

[0019] Step 6.3: Use formula (3) to calculate the qth historical sample r in the feature set selected at the tth iteration: q =[r 1q ,…,r pq ,…,r mq ] and the hth fuzzy cluster center vector The weighted Euclidean distance Among them, r pq is the pth key meteorological feature of the qth historical sample in the selected feature set, is the pth element value of the hth fuzzy cluster center vector at the tth iteration:

[0020]

[0021] Step 6.4: Update the weight of the tth iteration using equations (4), (5), and (6) respectively. Membership Fuzzy cluster center element Thus, the weight of the t+1th iteration is obtained Membership Fuzzy cluster center element

[0022]

[0023]

[0024]

[0025] In formula (4), formula (5) and formula (6), r kq is the kth meteorological feature of the qth historical sample in the filtered feature set, is the kth element value of the hth fuzzy cluster center vector at the tth iteration, is the p-th element value of the k-th fuzzy cluster center vector at the t-th iteration;

[0026] Step 6.5: If satisfied The iteration ends, and the weight of the t+1th iteration is obtained Membership Fuzzy cluster center element The weight ω as the final iterative output p , membership degree u hq , fuzzy cluster center element v ph , and is used to divide the selected feature set to obtain a sample subset with c categories. Otherwise, assign t+1 to t and go to step 6.3 to execute sequentially;

[0027] Step 6.6, using c sample subsets to train the back propagation neural network respectively, and establishing a prediction model between meteorological data and fluctuating power in each sample subset;

[0028] Step 7: Call the prediction model that is most similar to the meteorological characteristics of the current forecast period to perform power forecasting:

[0029] Step 7.1: According to the weight ω output by the final iteration p , fuzzy cluster center element v ph , using formula (7) to calculate the membership degree u between the normalized vth meteorological characteristic sample and the hth sample subset in the current forecast period hv , thus obtaining the membership matrix between the normalized meteorological characteristic samples and each sample subset of the current forecast period Let u v =[u 1v ,…,u cv ] T represents the membership vector of the vth meteorological feature sample in the current forecast period, and

[0030]

[0031] In formula (7), r′ pv is the pth key meteorological feature in the vth sample of the current period to be predicted; h = 1, ..., c, p = 1, ..., m, v = 1, ..., z, z represents the number of meteorological feature samples in the current period to be predicted;

[0032] Step 7.2, set the discrimination threshold to σ, define variables a, b∈{1,2,…,c} and a≠b;

[0033] Step 7.3: The membership vector u of the vth meteorological characteristic sample in the current forecast period v , if |u av -u bv |<σ, then the vth meteorological characteristic sample is identified as a boundary sample, and its wind power output is confirmed by integrating multiple prediction models using formula (8); otherwise, execute step 7.4;

[0034] pv =[p 1v ,…,p cv ]·[u 1v ,…,u cv ] T (8)

[0035] In formula (8): p v is the final wind power output of the boundary sample, p cv is the prediction result of the vth meteorological characteristic sample in the current prediction period under the cth model;

[0036] Step 7.4: Match the membership vector u of the vth meteorological feature sample in the current forecast period v =[u 1v ,…,u cv ] T The prediction model corresponding to the largest element in predicts the wind power of the vth meteorological characteristic sample;

[0037] Step 7.5: After assigning v+1 to v, determine whether v>z is established. If so, complete the wind power forecast for the current forecast period. Otherwise, return to step 7.3 and execute sequentially.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] 1 The present invention targets various meteorological factors related to wind power fluctuations, uses grey correlation theory to screen out key meteorological features, simplifies the input dimension of the model, ensures rapid convergence of the model while achieving optimal prediction results, and avoids the increase in model training difficulty caused by too many input features. In addition, simplifying the feature dimension can improve the processing speed of the clustering algorithm to obtain an accurate subset of wind power meteorological features.

[0040] 2 The present invention aims at the traditional clustering method that ignores the differences between different input quantities in the process of multi-feature sample clustering. It adopts a feature weighting method to divide more accurate clustering results, classifies and constructs a prediction model, and calls the prediction model that is most similar to the characteristics of the current time period during the prediction process, thereby improving the overall wind power prediction accuracy.

[0041] 3 Based on weighted fuzzy clustering, the present invention introduces a judgment threshold σ to search for boundary data that is difficult to define, and determines the final wind power by integrating multiple sample subsets through the membership matrix, thereby improving the belonging relationship of the boundary data, replacing the single division result in traditional clustering, and taking into account the wind power output prediction results in each time period. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the process of short-term wind power prediction of the present invention;

[0043] Figure 2 This is a comparison chart of prediction results under different clustering algorithms. DETAILED DESCRIPTION

[0044] In this embodiment, a wind power prediction method based on weighted fuzzy clustering is to improve the prediction accuracy by optimizing a combination of multiple links, taking into account the weights of key meteorological factors in the multi-feature sample clustering process, comprehensively weighing multiple sample subsets to determine the boundary data wind power output, and establishing a more accurate classification prediction model to improve the accuracy and applicability of the prediction method. Specifically, Figure 1 As shown, the prediction method is carried out in the following steps:

[0045] Step 1: Select a sample set consisting of l historical samples, each of which has n meteorological characteristics, and normalize the sample set to obtain a meteorological feature matrix with a dimension of n×l. Among them, x ij represents the i-th meteorological feature after normalization of the j-th historical sample, i = 1, 2, ..., n, j = 1, 2, ..., l;

[0046] Step 2: Collect the power data of l historical samples corresponding to the meteorological characteristics and perform normalization to obtain the power sequence P = [p1, p2, ..., p j ,…,p l ], p j represents the normalized power of the jth historical sample;

[0047] Step 3: Use formula (1) to calculate the i-th meteorological characteristic x of the j-th historical sample: ij and the power p of the jth historical sample j Grey correlation coefficient ξ ij , thus constructing the grey incidence matrix

[0048]

[0049] In formula (1), ρ is the resolution coefficient;

[0050] Step 4: Calculate the i-th meteorological characteristic x of l historical samples according to formula (2): i Correlation degree γ with power sequence P i :

[0051]

[0052] Step 5: Sort the meteorological features in descending order according to the degree of relevance, and select the meteorological feature matrix X n×l Select m key meteorological features that dominate wind power output as the screening feature set;

[0053] Wind power fluctuations involve a variety of meteorological factors, but too many input features will increase the difficulty of model training. Therefore, for the prediction model, it is necessary to select key features from many meteorological factors, establish a mapping relationship between meteorological data and fluctuating power, and ensure that the model converges quickly while achieving the optimal prediction effect. In addition, simplifying the feature dimension can effectively reduce the time of sample data clustering and obtain reasonable and efficient clustering results. The gray correlation method can judge the correlation between the historical power series and the meteorological feature series, and select strongly correlated feature quantities based on the correlation.

[0054] Step 6: Use the feature-weighted fuzzy C-means clustering algorithm to divide the selected feature set into data sample subsets, and use the back propagation neural network to classify and model the data sample subsets:

[0055] Step 6.1, determine the optimization criterion parameter as α, the number of cluster categories as c, and the termination threshold of the iterative calculation as ε, define and initialize the current iteration number t=0, where c≥2, α≥1;

[0056] Step 6.2: Initialize the weight corresponding to the pth selected key meteorological feature of the tth iteration The membership degree between the qth historical sample and the hth sample subset in the feature set selected at the tth iteration The fuzzy cluster center element of the pth row and hth column at the tth iteration Wherein, h = 1, ..., c, p = 1, ..., m, q = 1, ..., l, m represents the number of key meteorological features screened, and l represents the number of historical samples;

[0057] Step 6.3: Use formula (3) to calculate the qth historical sample r in the feature set selected at the tth iteration: q =[r 1q ,…,r pq ,…,r mq ] and the hth fuzzy cluster center vector The weighted Euclidean distance Among them, r pq is the pth key meteorological feature of the qth historical sample in the selected feature set, is the pth element value of the hth fuzzy cluster center vector at the tth iteration:

[0058]

[0059] Step 6.4: Update the weight of the tth iteration using equations (4), (5), and (6) respectively. Membership Fuzzy cluster center element Thus, the weight of the t+1th iteration is obtained Membership Fuzzy cluster center element

[0060]

[0061]

[0062]

[0063] In formula (4), formula (5) and formula (6), r kq is the kth meteorological feature of the qth historical sample in the filtered feature set, is the kth element value of the hth fuzzy cluster center vector at the tth iteration, is the p-th element value of the k-th fuzzy cluster center vector at the t-th iteration;

[0064] Step 6.5: If satisfied The iteration ends, and the weight of the t+1th iteration is obtained Membership Fuzzy cluster center element The weight ω as the final iterative output p , membership degree u hq , fuzzy cluster center element v ph , and is used to divide the selected feature set to obtain a sample subset with c categories. Otherwise, assign t+1 to t and go to step 6.3 to execute sequentially;

[0065] Step 6.6, using c sample subsets to train the back propagation neural network respectively, and establishing a prediction model between meteorological data and fluctuating power in each sample subset;

[0066] The feature-weighted fuzzy clustering method is used to replace K-means, DBSCAN and other clustering methods. The fuzzy clustering method provides the membership degree of each observation sample to each category, reflecting the uncertainty of sample attribution, and replaces the single division result in general clustering. In addition, the traditional clustering algorithm believes that the features have the same weight, but the actual wind speed, wind direction and other meteorological features have different contributions to power fluctuations, so the traditional clustering method cannot divide reasonable clustering results. The fuzzy clustering algorithm based on feature weighting takes into account the influence of meteorological features, determines the weight vector by iteration, and applies greater weight to important meteorological features, so that the clustering results are more in line with the actual rules.

[0067] Step 7: Call the prediction model that is most similar to the meteorological characteristics of the current forecast period to perform power forecasting:

[0068] Step 7.1: According to the weight ω output by the final iteration p, fuzzy cluster center element v ph , using formula (7) to calculate the membership degree u between the normalized vth meteorological characteristic sample and the hth sample subset in the current forecast period hv , thus obtaining the membership matrix between the normalized meteorological characteristic samples and each sample subset of the current forecast period Let u v =[u 1v ,…,u cv ] T represents the membership vector of the vth meteorological feature sample in the current forecast period, and

[0069]

[0070] In formula (7), r′ pv is the pth key meteorological feature in the vth sample of the current period to be predicted; h = 1, ..., c, p = 1, ..., m, v = 1, ..., z, z represents the number of meteorological feature samples in the current period to be predicted;

[0071] Step 7.2, set the discrimination threshold to σ, define variables a, b∈{1,2,…,c} and a≠b;

[0072] Step 7.3: The membership vector u of the vth meteorological characteristic sample in the current forecast period v , if |u av -u bv |<σ, then the vth meteorological characteristic sample is identified as a boundary sample, and its wind power output is confirmed by integrating multiple prediction models using formula (8); otherwise, execute step 7.4;

[0073] p v =[p 1v ,…,p cv ]·[u 1v ,…,u cv ] T (8)

[0074] In formula (8): p v is the final wind power output of the boundary sample, p cv is the prediction result of the vth meteorological characteristic sample in the current prediction period under the cth model;

[0075] Step 7.4: Match the membership vector u of the vth meteorological feature sample in the current forecast period v =[u 1v ,…,u cv ] T The prediction model corresponding to the largest element in predicts the wind power of the vth meteorological characteristic sample;

[0076] Step 7.5: After assigning v+1 to v, determine whether v>z is established. If so, complete the wind power forecast for the current forecast period. Otherwise, return to step 7.3 and execute sequentially.

[0077] Traditional clustering algorithms divide the boundary data between sample subsets into specific categories, but the boundary data itself is related to adjacent subsets. Once divided into a certain subset, the prediction results of matching the corresponding prediction model are often difficult to guarantee accuracy. With the help of the membership matrix of fuzzy clustering, the probability between the current sample and each subset is determined, and a discrimination threshold is introduced. When the probability difference is lower than the discrimination threshold, that is, the sample data has an obvious connection with multiple subsets, the data is identified as boundary data, and no longer matched with a single model for prediction. Instead, it is determined by combining multiple classification models according to the membership matrix to determine its power size, taking into account the accuracy of each data sample. In order to demonstrate the improvement effect of this method, it is now compared with the K-means method, such as Figure 2 As shown in the figure, the prediction effect of this method is better than that of the K-means method as a whole and is more in line with the actual power curve.

[0078] In summary, the present invention screens key meteorological features through the grey correlation theory, simplifies the input dimension, and improves the convergence speed of clustering algorithms and prediction models; and replaces the traditional clustering algorithm with a feature-weighted fuzzy clustering algorithm, considers the contribution of different meteorological features to power fluctuations, divides more accurate clustering results, and uses the membership matrix to indicate the probability of samples belonging to each category, reflecting the uncertainty of sample attribution; at the same time, for boundary data that is difficult to define, the membership matrix is ​​used for identification, and multiple prediction models are combined to determine its wind power output, taking into account the prediction accuracy of each data sample.

Claims

1. A wind power prediction method based on weighted fuzzy clustering, characterized by Proceed as follows: Step 1: Select a sample set consisting of l historical samples, each sample has n meteorological characteristics, and normalize the sample set to obtain a meteorological characteristic matrix with a dimension of n×l Among them, x ij represents the i-th meteorological feature after normalization of the j-th historical sample, i = 1, 2, ..., n, j = 1, 2, ..., l; Step 2: Collect the power data of l historical samples corresponding to the meteorological characteristics and perform normalization to obtain the power sequence P = [p1, p2, ..., p j ,…,p l ], p j represents the normalized power of the jth historical sample; Step 3: Use formula (1) to calculate the i-th meteorological characteristic x of the j-th historical sample: ij and the power p of the jth historical sample j Grey correlation coefficient ξ ij , thus constructing the grey incidence matrix In formula (1), ρ is the resolution coefficient; Step 4: Calculate the i-th meteorological characteristic x of l historical samples according to formula (2): i Correlation degree γ with power sequence P i : Step 5: Sort the meteorological features in descending order according to the degree of relevance, and select the meteorological feature matrix X n×l Select m key meteorological features that dominate wind power output as the screening feature set; Step 6: Use the feature-weighted fuzzy C-means clustering algorithm to divide the selected feature set into data sample subsets, and use the back propagation neural network to classify and model the data sample subsets: Step 6.1, determine the optimization criterion parameter as α, the number of cluster categories as c, and the termination threshold of the iterative calculation as ε, define and initialize the current iteration number t=0, where c≥2, α≥1; Step 6.2: Initialize the weight corresponding to the pth selected key meteorological feature of the tth iteration The membership degree between the qth historical sample and the hth sample subset in the feature set selected at the tth iteration The fuzzy cluster center element of the pth row and hth column at the tth iteration Wherein, h = 1, ..., c, p = 1, ..., m, q = 1, ..., l, m represents the number of key meteorological features screened, and l represents the number of historical samples; Step 6.3: Use formula (3) to calculate the qth historical sample r in the feature set selected at the tth iteration: q =[r 1q ,…,r pq ,…,r mq ] and the hth fuzzy cluster center vector The weighted Euclidean distance Among them, r pq is the pth key meteorological feature of the qth historical sample in the selected feature set, is the pth element value of the hth fuzzy cluster center vector at the tth iteration: Step 6.4: Update the weight of the tth iteration using equations (4), (5), and (6) respectively. Membership Fuzzy cluster center element Thus, the weight of the t+1th iteration is obtained Membership Fuzzy cluster center element In formula (4), formula (5) and formula (6), r kq is the kth meteorological feature of the qth historical sample in the filtered feature set, is the kth element value of the hth fuzzy cluster center vector at the tth iteration, is the p-th element value of the k-th fuzzy cluster center vector at the t-th iteration; Step 6.5: If satisfied The iteration ends, and the weight of the t+1th iteration is obtained Membership Fuzzy cluster center element The weight ω as the final iterative output p , membership degree u hq , fuzzy cluster center element v ph , and is used to divide the selected feature set to obtain a sample subset with c categories. Otherwise, after assigning t+1 to t, go to step 6.3 and execute sequentially; Step 6.6, using c sample subsets to train the back propagation neural network respectively, and establishing a prediction model between meteorological data and fluctuating power in each sample subset; Step 7: Call the prediction model that is most similar to the meteorological characteristics of the current forecast period to perform power forecasting: Step 7.1: According to the weight ω output by the final iteration p , fuzzy cluster center element v ph , using formula (7) to calculate the membership degree u between the normalized vth meteorological characteristic sample and the hth sample subset in the current forecast period hv , thus obtaining the membership matrix between the normalized meteorological characteristic samples and each sample subset of the current forecast period Let u v =[u 1v ,…,u cv ] T represents the membership vector of the vth meteorological feature sample in the current forecast period, and In formula (7), r′ pv is the pth key meteorological feature in the vth sample of the current period to be predicted; h = 1, ..., c, p = 1, ..., m, v = 1, ..., z, z represents the number of meteorological feature samples in the current period to be predicted; Step 7.2, set the discrimination threshold to σ, define variables a, b∈{1,2,…,c} and a≠b; Step 7.3: The membership vector u of the vth meteorological characteristic sample in the current forecast period v , if |u av -u bv |<σ, then the vth meteorological characteristic sample is identified as a boundary sample, and its wind power output is confirmed by integrating multiple prediction models using formula (8); otherwise, execute step 7.4; pp v =[p 1v ,…,p cv ]·[u 1v ,…,u cv ] T (8) In formula (8): p v is the final wind power output of the boundary sample, p cv is the prediction result of the vth meteorological characteristic sample in the current prediction period under the cth model; Step 7.4: Match the membership vector u of the vth meteorological feature sample in the current forecast period v =[u 1v ,…,u cv ] T The prediction model corresponding to the largest element in predicts the wind power of the vth meteorological characteristic sample; Step 7.5: After assigning v+1 to v, determine whether v>z is established. If so, complete the wind power forecast for the current forecast period. Otherwise, return to step 7.3 and execute sequentially.