Wind and light output scene clustering method based on information entropy and smoothness

Abstract

The invention provides a wind and light output scene clustering method based on information entropy and smoothness, and relates to the technical field of wind and light operation, and the method comprises the steps: carrying out the minimum-maximum standardization processing of original wind and light output data, and limiting the data range between 0 and 1; after standardization processing, analyzing information entropy and smoothness of wind power photovoltaic output, and extracting characteristics of wind and light output data; determining the number of clustering centers of the wind power photovoltaic output scene by combining the information entropy of the wind and light output and the smoothness analysis result; and in the process of iteratively calculating the clustering result, calculating a difference value between the clustering center Euclidean distances of the nth iteration and the (n-1) th iteration, comparing the difference value with a set threshold value, and when the difference value is smaller than the threshold value, indicating that the clustering result is converged. According to the method, the information entropy and the smoothness of the historical wind and light output are analyzed, the key features of the wind and light output are extracted, the number of the clustering centers of the wind and light output scene is determined, and the representative wind and light output clustering scene can be generated.

Description

Technical Field

[0001] The present invention relates to the field of wind and solar operation technology, and in particular to a wind and solar output scene clustering method based on information entropy and smoothness. Background Art

[0002] In the management and optimization of modern complex systems, clustering wind and solar output scenarios has important application value, especially in the fields of power systems, energy management, climate modeling, etc. These systems usually have high uncertainty and dynamics, and are affected by a variety of internal and external factors, resulting in the complexity and unpredictability of their operating states. Therefore, generating representative wind and solar output scenarios based on actual needs can help system designers and decision makers better understand the operating laws of the system, thereby improving the accuracy of scheduling, optimization and risk management.

[0003] In complex systems, it is necessary not only to generate scenarios that can truly reflect the operating status of the system, but also to ensure that these scenarios have sufficient generalization capabilities to adapt to different operating conditions that may arise in the future. However, existing clustering algorithms and optimization methods often cannot balance the representativeness and generalization capabilities of wind and solar output scenarios well, resulting in the generated scenarios being unable to cope with extreme situations or unpredictable changes in the future, and having limitations. Summary of the invention

[0004] The present invention proposes a wind and solar power output scenario clustering method based on information entropy and smoothness. Through an optimized iterative clustering method, it can effectively process a large amount of complex data and generate representative typical scenarios, thereby providing data support for subsequent decision optimization, system scheduling and risk assessment to solve the problems existing in the background technology.

[0005] A wind and solar power output scene clustering method based on information entropy and smoothness includes the following steps:

[0006] S100, firstly, the acquired original wind and solar power output data is processed according to the minimum-maximum standardization, and the data range is limited to between 0 and 1;

[0007] S200, analyzing the information entropy and smoothness of wind and solar power output, and extracting wind and solar power output data features;

[0008] S300, combining the information entropy and smoothness analysis results of wind power and photovoltaic output, determining the number of cluster centers of wind power and photovoltaic output scenarios;

[0009] S400, in the process of iteratively calculating the clustering results, the difference between the Euclidean distances of the cluster centers of the nth iteration and the n-1th iteration is calculated, and the difference is compared with a set threshold value, until the difference between the two is less than the threshold value, indicating that the clustering results converge;

[0010] S500: Generate a typical scenario based on the converged clustering result.

[0011] Further, based on the above, in step S100, the original wind and solar power output data is standardized:

[0012]

[0013] In the formula, x z Represents a value of attribute z; max z 、min z Respectively represent the maximum and minimum values ​​of attribute z, x′ z Represents the data after the normalization of a certain value of attribute z. By using the interval normalization operation, all wind and solar power output values ​​are normalized within the range of [0,1].

[0014] Further, based on the above, in step S200, the smoothness s and information entropy w of the wind and solar power output are analyzed;

[0015] Smoothness s is used to describe the fluctuation intensity of the output distribution curve. The processing formula is as follows:

[0016]

[0017] In the formula, s(x) represents the smoothness, which reflects the fluctuation degree of wind and solar power output in the time series; N represents the total number of data; Indicates the wind and solar power output value of the attribute at the i-th time point;

[0018] Information entropy w is used to quantify the uncertainty and randomness of output distribution:

[0019]

[0020] In the formula, w(x) represents the information entropy of the random variable x, which measures the uncertainty or information content of the variable; p(x) represents the probability of the random variable x taking a value; log b Represents logarithmic operation, the base b can be any positive number;

[0021] The numerical judgment basis of smoothness s and information entropy w is as follows:

[0022] When s = 0, the distribution curve is consistent with the ideal smooth curve, and the fluctuation is gentle;

[0023] When s>0, the right side of the distribution fluctuates more violently, and the right tail has a clear protrusion;

[0024] When s<0, the left side of the distribution fluctuates more violently, and the left tail has an obvious protrusion.

[0025] w=w 0(consistent with the ideal entropy value), it means that the degree of disorder of the output force distribution is the same as that of the standard state;

[0026] w>w 0 When , it indicates that the output distribution is more dispersed and the uncertainty of the system increases;

[0027] w <w 0 , it indicates that the output force is more concentrated and the certainty of the system increases.

[0028] Further, based on the above, in step S300, the number of cluster centers of wind power and photovoltaic output scenarios is determined:

[0029]

[0030] In the formula, R i is the kth cluster, t is R k The sample point, y k YesR k BG(m) is the centroid of all samples, and the m value with which BG(m) decreases significantly is selected as the number of cluster centers.

[0031] Further, based on the above, in step S400, during the iterative calculation of the clustering results, the difference between the Euclidean distances of the cluster centers of the nth iteration and the n-1th iteration is calculated, and compared with the set threshold value, until the difference between the two is less than the threshold value, indicating that the clustering results converge:

[0032] The threshold is 0.0001, and the Euclidean distance calculation formula is:

[0033]

[0034] In the formula, a h is the coordinate of the representative point a in the hth dimension, b h is the coordinate of the representative point b in the i-th dimension, and d is the distance between the data points;

[0035] Δd=|d n -d n-1 |

[0036] Where, d n is the distance between data points at the nth iteration, d n-1 is the distance between data points at the n-1th iteration, and Δd is the absolute value of the distance difference between data points.

[0037] The present invention has outstanding substantive features and significant progress compared to the prior art, specifically:

[0038] 1. The present invention extracts representative wind and solar power output scenarios through effective algorithms, so that the multi-dimensional complex data in the system can be classified and merged, thereby simplifying the complexity of the system and revealing its potential operating rules.

[0039] 2. Based on a large amount of historical operating data and changeable meteorological data, the present invention obtains convergence values ​​through a clustering algorithm, and then generates typical scenarios. Based on the typical scenarios, the stability of power dispatching is effectively improved, the energy allocation plan is optimized, and the risks caused by uncontrollable factors are reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic diagram of the process of the present invention;

[0041] Figure 2 It is a curve chart of historical photovoltaic output data;

[0042] Figure 3 It is the photovoltaic output clustering curve;

[0043] Figure 4 Historical wind power output data curve chart;

[0044] Figure 5 It is a clustering curve diagram of wind power output. DETAILED DESCRIPTION

[0045] The following will be combined with the accompanying drawings in the embodiments of the present invention to provide a detailed and comprehensive description of the technical solution of the present invention. Obviously, the embodiments are only examples of some of the present invention and do not cover all implementation methods. Based on these embodiments, various other variations or extensions obtained by those skilled in the art without creative work should be deemed to belong to the protection scope of the present invention.

[0046] like Figure 1-5 As shown in the figure, a wind and solar power output scene clustering method based on information entropy and smoothness is proposed, focusing on determining the optimal and generating efficient and reliable clustering scenes. The specific implementation steps are as follows:

[0047] S100, firstly, the acquired original wind and solar power output data is processed according to the minimum-maximum standardization, and the data range is limited to between 0 and 1;

[0048]

[0049] In the formula, x z Represents a value of attribute z; max z 、min z Respectively represent the maximum and minimum values ​​of attribute z, x′ z It represents the data after the attribute z value is standardized, and all wind and solar power output values ​​are standardized within the range of [0,1].

[0050] Step S200, analyzing the smoothness s and information entropy w of wind and solar power output, and extracting wind and solar power output data features;

[0051] Smoothness s is used to describe the fluctuation intensity of the output distribution curve. The processing formula is as follows:

[0052]

[0053] In the formula, s(x) represents the smoothness, which reflects the degree of fluctuation of wind and solar power output in the time series; G represents the total number of data; Indicates the wind and solar power output value of the attribute at the i-th time point;

[0054] Information entropy w is used to quantify the uncertainty and randomness of output distribution. The processing formula is as follows:

[0055]

[0056] In the formula, w(x) represents the information entropy of the random variable x, which measures the uncertainty or information content of the variable; p(x) represents the probability of the random variable x taking a value; log b Represents logarithmic operation. The base b can be any positive number.

[0057] The numerical judgment basis of smoothness s and information entropy w is as follows:

[0058] When s = 0, the distribution curve is consistent with the ideal smooth curve, and the fluctuation is gentle;

[0059] When s>0, the right side of the distribution fluctuates more violently, and the right tail has a clear protrusion;

[0060] When s<0, the left side of the distribution fluctuates more violently, and the left tail has an obvious protrusion.

[0061] w=w 0 When , it indicates that the chaotic degree of output force distribution is the same as that of the standard state;

[0062] w>w 0 When , it indicates that the output distribution is more dispersed and the uncertainty of the system increases;

[0063] w <w 0 , it indicates that the output force is more concentrated and the certainty of the system increases.

[0064] Step S300, determining the number of cluster centers of wind power and photovoltaic output scenarios:

[0065]

[0066] In the formula, R i is the kth cluster, t is R k The sample point, y kYesR k BG(m) is the centroid of all samples, and the m value with which BG(m) decreases significantly is selected as the number of cluster centers.

[0067] Step S400, in the process of iteratively calculating the clustering results, the difference between the Euclidean distances of the cluster centers of the nth iteration and the n-1th iteration is calculated, and the difference is compared with the set threshold value until the difference between the two is less than the threshold value, indicating that the clustering results converge. The specific calculation process is as follows:

[0068] The threshold is 0.0001, and the Euclidean distance calculation formula is:

[0069]

[0070] In the formula, a h is the coordinate of the representative point a in the hth dimension, b h is the coordinate of the representative point b in the i-th dimension, and d is the distance between the data points;

[0071] Δd=|d n -d n-1 |

[0072] Where, d n is the distance between data points at the nth iteration, d n-1 is the distance between data points at the n-1th iteration, and Δd is the absolute value of the distance difference between data points.

[0073] The following is described with reference to a specific embodiment:

[0074] Based on the simulation examples and parameter settings in Table 1, simulation verification and analysis are carried out.

[0075] Table 1 Simulation parameters and variable settings

[0076]

[0077]

[0078] Based on historical photovoltaic and wind power output data, representative photovoltaic and wind power output curves are generated.

[0079] On the one hand, based on the historical photovoltaic output data, this clustering algorithm generates two cluster centers, that is, two representative clustering curves. By calculating the change in the Euclidean distance between each curve and setting a threshold, the forward and backward reduction method is used to eliminate similar curves, and finally the following is formed: Figure 2 , Figure 3 The two cluster intervals and typical curves shown.

[0080] On the other hand, based on historical wind power output data, this clustering algorithm generates 4 cluster centers, that is, 4 representative clustering curves. By calculating the change in the Euclidean distance between each curve and setting a threshold, the forward and backward reduction method is used to eliminate similar curves, and finally the following is formed. Figure 4 , Figure 5 The four cluster intervals and typical curves shown.

[0081] For those skilled in the art, the present invention is not limited to the specific details of the aforementioned exemplary embodiments. Under the premise of maintaining the core ideas and basic features of the present invention, other forms of implementation are also feasible. Therefore, these embodiments should be regarded as examples rather than restrictive descriptions. The scope of protection should be clearly defined by the accompanying claims and should not be limited to the foregoing. All changes or modifications that fall within the scope of the claims and their equivalent features should be included in the scope of protection of the present invention. In addition, any figure marks in the claims should not be regarded as limitations on the relevant claims.

Claims

1. A wind and solar power output scene clustering method based on information entropy and smoothness, characterized in that: Including the following steps: S100. First, the obtained original wind-solar power output data is processed by min-max normalization, and the data range is limited between 0 and 1; S200. Analyze the information entropy and smoothness of the wind-solar power output, and extract the characteristics of the wind-solar power output data; S300. Combine the analysis results of the information entropy and smoothness of the wind power and photovoltaic power output to determine the number of clustering centers for the wind power and photovoltaic power output scenarios; S400. During the iterative calculation of the clustering results, calculate the difference between the Euclidean distances of the clustering centers in the nth iteration and the (n - 1)th iteration, and compare it with the set threshold. When the difference between the two is less than the threshold, it indicates that the clustering results converge; S500. Generate typical scenarios based on the converged clustering results.

2. According to claim 1, a wind and solar output scene clustering method based on information entropy and smoothness is characterized by: In step S100, the original wind-solar power output data is normalized as follows: In the formula, x z Represents a value of attribute z; max z 、min z Respectively represent the maximum and minimum values ​​of attribute z, x′ z Represents the data after the attribute z value is standardized.

3. A method for clustering wind-solar power output scenarios based on information entropy and smoothness according to claim 1, characterized in that: In step S200, analyze the smoothness s and information entropy w of the wind-solar power output, and extract the characteristics of the wind-solar power output data; The smoothness s is used to describe the degree of fluctuation of the output distribution curve, and the processing formula is as follows: In the formula, s(x) represents the smoothness, which reflects the degree of fluctuation of wind and solar power output in the time series; G represents the total number of data; Indicates the wind and solar power output value of the attribute at the i-th time point; The information entropy w is used to quantify the uncertainty and randomness of the output distribution, and the processing formula is as follows: In the formula, w(x) represents the information entropy of the random variable x, which measures the uncertainty or information content of the variable; p(x) represents the probability of the random variable x taking a value; log b Represents logarithmic operation, the base b can be any positive number; The numerical judgment basis for the smoothness s and the information entropy w is as follows: When s = 0, the distribution curve is consistent with the ideal smooth curve, and the fluctuation is gentle; When s > 0, the right side of the distribution fluctuates more violently, and there is an obvious protrusion at the right tail; When s < 0, the left side of the distribution fluctuates more violently, and there is an obvious protrusion at the left tail; When w = w0 (consistent with the ideal entropy value), it means that the degree of chaos of the output distribution is the same as the standard state; When w > w0, it means that the output distribution is more dispersed, and the uncertainty of the system increases; When w < w0, it means that the output is more concentrated, and the certainty of the system increases.

4. A method for clustering wind-solar power output scenarios based on information entropy and smoothness according to claim 1, characterized in that: In step S300, combine the analysis results of the information entropy and smoothness of the wind power and photovoltaic power output to clarify the number of clustering centers for the wind power and photovoltaic power output scenarios: In the formula, R i is the kth cluster, t is R k The sample point y k YesR k BG(m) is the centroid of all samples, and the m value with which BG(m) decreases significantly is selected as the number of cluster centers.

5. A method for clustering wind-solar power output scenarios based on information entropy and smoothness according to claim 1, characterized in that: In step S400, during the iterative calculation of the clustering results, calculate the difference between the Euclidean distances of the clustering centers in the nth iteration and the (n - 1)th iteration, and compare it with the set threshold. When the difference between the two is less than the threshold, it indicates that the clustering results converge. The specific calculation process is as follows: Take the threshold as 0.0001, and its Euclidean distance calculation formula is: In the formula, a h is the coordinate of the representative point a in the hth dimension, b h is the coordinate of the representative point b in the i-th dimension, and d is the distance between the data points; Δd=|d n -d n-1 Where, d n is the distance between data points at the nth iteration, d n-1 is the distance between data points at the n-1th iteration, and Δd is the absolute value of the distance difference between data points.

Images