Section limit transmission capacity analysis method based on similarity identification and related equipment

Abstract

The invention discloses a section limit transmission capacity analysis method based on similarity identification and related equipment, and relates to the technical field of power system dispatching operation control, and the key points of the technical scheme are as follows: obtaining historical operation data of a power transmission section; screening the historical operation data to obtain a feature set of historical section samples; clustering the feature set by adopting an FCM algorithm to obtain a feature clustering result of each historical section sample; calculating the similarity between the feature clustering result of each historical section sample and a preset feature, and selecting the feature of the historical section sample with the maximum similarity as an analysis value of the section limit transmission capacity of the future section sample; wherein the preset feature refers to the feature quantity of the future section sample corresponding to each historical section sample. According to the method, rapid calculation of the TTC of the future power transmission section can be effectively realized, the manual workload of TTC calculation of the power transmission section in an actual power grid is reduced, and the method has great significance in guaranteeing safe and stable operation of a power system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system dispatching operation control, more particularly, it relates to a section limit transmission capacity analysis method based on similarity recognition and related equipment. BACKGROUND

[0002] Total Transfer Capability (TTC) is an important basic index for evaluating the safe operation state of a power transmission section. In practice, system operators generally monitor the power transmission section as a dimension reduction object of the power grid, and ensure the safety and stability of the power grid by controlling the transmission power of the power transmission section to be lower than the TTC of the power transmission section during power system operation. Therefore, it is of great significance to realize fast and accurate calculation of the TTC of the power transmission section containing complex security constraints for ensuring the safe and stable operation of the power system.

[0003] In the past power grid operation control, the control measures of the power transmission section of the power grid are mainly formulated through a large number of simulation calculations combined with expert experience. Since the complex simulation calculations involved in checking the transient stability constraints will occupy a large amount of computing resources, the TTC calculation is difficult to meet the online requirements. Moreover, various factors such as the dynamic characteristics of the equipment and the coupling characteristics between sections make the TTC calculation considering transient stability more complex. At the same time, with the continuous expansion of the power system scale, the gradual increase of the proportion of renewable energy, and the increasingly complex structure of the power grid, the results obtained by relying on expert experience and traditional calculation methods have gradually failed to meet the actual production and operation requirements, so it is more and more difficult to accurately and efficiently calculate the TTC of the section under different operating modes.

[0004] At present, although the calculation methods provided by the related prior art have realized fast calculation of the TTC of the power transmission section under different operating modes to a certain extent, there are still some problems to be solved: 1) the current calculation method based on successive power flow calculation to explore TTC reduces the calculation complexity of the section TTC and improves the TTC calculation efficiency, but its calculation complexity is high and a large amount of calculation cost is required. Data-driven methods can be used to improve the calculation efficiency. 2) Although the data-driven method can effectively avoid complex physical models and improve the calculation efficiency of the section TTC, the accuracy of the calculation results generally needs to be further verified, and the accuracy of the section TTC calculation results is insufficient. 3) Although the current TTC calculation method based on optimal power flow can effectively solve the maximum section TTC, the solution result is often difficult to reproduce in actual operation, and the section TTC calculation result does not have practical significance. SUMMARY

[0005] The application aims to provide a section limit transmission capacity analysis method based on similarity recognition and a related device, solve the problems in calculating the section limit transmission capacity in the prior art, and consider that the historical operation data of the power system generally contains control information and characteristic data related to the section, if the historical power transmission section similar to the future power transmission section can be effectively recognized, the TTC of the historical power transmission section is taken as the TTC reference value of the future power transmission section, the complex physical modeling and stability constraints in TTC calculation are effectively avoided, the calculation of the section TTC is quickly realized, and the calculation result is ensured to be consistent with the actual operation.

[0006] The above technical purposes of the application are realized by the following technical solutions.

[0007] In a first aspect, the application provides a section limit transmission capacity analysis method based on similarity recognition, which comprises the following steps.

[0008] Obtaining historical operation data of a power transmission section;

[0009] Filtering the historical operation data to obtain a characteristic set of historical section samples;

[0010] Clustering the characteristic set by using a FCM algorithm to obtain a characteristic clustering result of each historical section sample;

[0011] Calculating the similarity between the characteristic clustering result of each historical section sample and a preset characteristic, and selecting the characteristic of the historical section sample with the largest similarity as an analysis value of the section limit transmission capacity of a future section sample; wherein the preset characteristic refers to a characteristic quantity of the future section sample corresponding to each historical section sample.

[0012] In an implementation scheme, filtering the historical operation data to obtain the characteristic set of the historical section samples comprises the following steps.

[0013] Constructing a hierarchical screening model by using a decision tree algorithm;

[0014] Screening the historical operation data by using the hierarchical screening model to obtain the characteristic set of the historical section samples.

[0015] In an implementation scheme, before clustering the characteristic set by using the FCM algorithm, the method further comprises normalizing the characteristic set of the historical section samples.

[0016] In an implementation scheme, clustering the characteristic set by using the FCM algorithm to obtain the characteristic clustering result of each historical section sample comprises the following steps.

[0017] Construct a target function with a distance-weighted square sum of all historical cross-section samples and each cluster center being minimum, and a constraint condition met by each element in a membership matrix; wherein, the element represents a membership degree of the jth historical cross-section sample to the ith category;

[0018] Solve the target function and the constraint condition to obtain an iterative model including a cluster center of each category and a membership degree of the jth historical cross-section sample to the ith category;

[0019] Obtain a category number of the cluster centers, an initial value of each cluster center and an initial membership matrix pre-configured;

[0020] According to the number of the cluster centers and the initial membership matrix, iterative calculation is performed according to the iterative model until a convergence condition is met or a maximum iteration number is reached, and a clustering result is obtained.

[0021] In an implementation scheme, the category number of the cluster centers is determined in the following manner:

[0022] Calculate an intra-category distance sum between each cluster center according to the distance;

[0023] According to the target function and the intra-category distance sum after clustering is completed, the category number of the cluster centers is determined in combination with an elbow point method.

[0024] In an implementation scheme, an expression for calculating the distance between each historical cross-section sample and each cluster center is as follows:

[0025] wherein, ω1 represents a weight coefficient of the cosine distance, ω2 represents a weight coefficient of the Euclidean distance, d euc (x, y) represents a Euclidean distance between the jth historical cross-section sample x j and the ith cluster center v i d cos (x, y) represents a cosine distance between the jth historical cross-section sample x j and the ith cluster center v i .

[0026] In an implementation scheme, the initial value of the cluster center is determined in the following manner:

[0027] Calculate the distance between any two historical cross-section samples, and combine multiple distances to obtain a distance matrix with a size of m*m; wherein, m is a total number of historical cross-section samples;

[0028] Select two first historical cross-section samples corresponding to the minimum distance value from the distance matrix, and take a midpoint of the two first historical cross-section samples as the initial value of the cluster center of the first category;

[0029] The distance sum of the rest of the historical cross section samples and two first historical cross section samples is calculated, the second historical cross section sample with the maximum distance sum is selected, the third historical cross section sample closest to the second historical cross section sample is searched, and the midpoint of the second historical cross section sample and the third historical cross section sample is taken as the initial value of the cluster center of the second category; wherein the second historical cross section sample is selected from the rest of the historical cross section samples;

[0030] The calculation of the distance sum and the selection of the maximum distance sum are repeatedly performed until the initial value of the cluster center of the cth category is obtained; wherein c represents the number of categories of the cluster center.

[0031] The second aspect of the present application provides a cross section limit transmission capacity analysis device based on similarity recognition, the device comprising:

[0032] An information acquisition module is configured to acquire historical operation data of a power transmission cross section.

[0033] A feature screening module is configured to screen the historical operation data to obtain a feature set of the historical cross section samples.

[0034] A feature clustering module is configured to cluster the feature set by using a FCM algorithm to obtain a feature clustering result of each historical cross section sample.

[0035] An analysis module is configured to calculate the similarity between the feature clustering result of each historical cross section sample and a preset feature, and select the feature of the historical cross section sample with the maximum similarity as an analysis value of the cross section limit transmission capacity of a future cross section sample; wherein the preset feature refers to a feature quantity of the future cross section sample corresponding to each historical cross section sample.

[0036] The third aspect of the present application provides an electronic device comprising a memory and a processor.

[0037] The memory is configured to store a computer program, and the computer program comprises program instructions.

[0038] The processor is configured to execute the program instructions to enable the electronic device to perform the steps of the method for analyzing the cross section limit transmission capacity based on similarity recognition according to the first aspect of the present application.

[0039] The fourth aspect of the present application provides a computer readable storage medium comprising a computer program, and the computer program, when executed by one or more processors, implements the method for analyzing the cross section limit transmission capacity based on similarity recognition according to the first aspect of the present application.

[0040] Compared with the prior art, the present application has the following beneficial effects:

[0041] 1. The layered decision model is established, the power grid basic characteristic quantity is screened and improved, more comprehensive section characteristic samples are obtained, the effectiveness of section similarity recognition results is improved, and the accuracy of future section TTC calculation results is improved.

[0042] 2. The shape characteristics and numerical characteristics of the section samples are comprehensively considered, the value in the section historical operation data is explored through the improved FCM clustering algorithm, the similarity of the section is recognized, the TTC of the similar historical sample is taken as the reference value of the TTC of the future sample, the complex calculation process of the TTC is effectively avoided, the calculation efficiency is improved, the TTC of the actual historical section is referred to, the obtained result has practical significance, and the result can be reproduced in the actual system. BRIEF DESCRIPTION OF DRAWINGS

[0043] The drawings described herein are used to provide further understanding of the embodiments of the present application, constitute a part of the present application, and do not constitute a limitation of the embodiments of the present application. In the drawings:

[0044] Figure 1 A flowchart of a section limit transmission capacity analysis method based on similarity recognition provided for the embodiments of the present application is shown in the drawings.

[0045] Figure 2 A power system transmission section characteristic quantity screening flowchart provided for the embodiments of the present application is shown in the drawings.

[0046] Figure 3 A schematic diagram of determining the optimal clustering number based on the elbow point method provided for the embodiments of the present application is shown in the drawings.

[0047] Figure 4 A flowchart of clustering based on the improved FCM algorithm provided for the embodiments of the present application is shown in the drawings.

[0048] Figure 5 A 1IEEE 39-node system transmission section division diagram provided for the embodiments of the present application is shown in the drawings.

[0049] Figure 6 A schematic diagram of selecting the optimal clustering number based on the elbow point method provided for the embodiments of the present application is shown in the drawings.

[0050] Figure 7 A clustering number and evaluation index relationship diagram provided for the embodiments of the present application is shown in the drawings.

[0051] Figure 8 A comparison diagram of each feature of a certain future sample and similar and non-similar sample centers provided for the embodiments of the present application is shown in the drawings. DETAILED DESCRIPTION

[0052] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0053] It should be noted that the terms "comprising" or "may include" used in the various embodiments of this application indicate the presence of the claimed function, operation, or element, and do not limit the addition of one or more functions, operations, or elements. Furthermore, as used in the various embodiments of this application, the terms "comprising," "having," and their cognates are intended only to indicate a specific feature, number, step, operation, element, component, or combination of the foregoing, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing, or adding one or more combinations of the foregoing.

[0054] It should be understood that terms such as "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0055] Please refer to Figure 1 This paper presents a method for analyzing the cross-sectional limit transmission capacity based on similarity identification. The method includes:

[0056] S101, Obtain historical operating data of the power transmission section.

[0057] In this embodiment, historical operating data of the transmission section is directly collected through the power grid monitoring equipment, and operating data obtained through simulation calculation is also collected. The two are combined to establish a basic feature quantity library, some of which are shown in Table 1:

[0058] Table 1 Basic Feature Library

[0059]

[0060] S102, filter the historical operation data to obtain the feature set of historical cross-section samples.

[0061] In this embodiment, a hierarchical decision screening model is constructed based on a decision tree. The basic features described above are progressively screened and improved to obtain a feature set that meets the requirements. The specific screening steps are as follows:

[0062] According to whether the characteristic quantity can be represented by a numerical value, numerical characteristics and attribute characteristics are distinguished. Text characteristics such as network node information, and attribute variables such as load and unit number are not suitable for clustering algorithms.

[0063] According to whether the characteristic dimension changes with the topology, whether it is a variable dimension characteristic is distinguished. Characteristics such as node connection relationship change in dimension when the network frame changes. Such characteristics are difficult to be directly used in clustering algorithms, but part of them can be used again through improvement.

[0064] According to the distinguishability of the characteristic quantity, characteristics with low distinguishability are removed from the above characteristics. For example, 0-1 variables such as unit start-stop, and node voltage amplitude in direct current flow calculation. Such characteristics may affect the accuracy of the clustering results due to low distinguishability.

[0065] Some characteristic quantities are improved to realize secondary use of the characteristic quantities. For example, the maximum and average values of the characteristic sequence are calculated. Finally, the characteristic quantity set F used for clustering is obtained according to the above steps. The screening and improvement process is shown in FIG. 1. Figure 2 Each characteristic quantity in F can be represented by formula (1): (1), in the formula, f i That is, the i-th component in the characteristic quantity set F is obtained by the screening method h and the improvement method g from the corresponding basic characteristic quantity p in Table 1.

[0066] To avoid the masking phenomenon caused by the numerical difference between different characteristic quantities, normalization processing of the characteristic set of each sample is required before clustering. The specific operation is as follows: (2); in the formula, x is the normalized clustering input sample, which is calculated from the corresponding characteristic quantity in the characteristic quantity set F by formula (2).

[0067] S103, the FCM algorithm is used to cluster the characteristic set to obtain the characteristic clustering result of each historical cross section sample.

[0068] In this embodiment, the FCM algorithm is used to cluster the characteristic set to obtain the characteristic clustering result of each historical cross section sample, including:

[0069] A target function is constructed to minimize the weighted sum of squares of the distances between all historical cross section samples and each clustering center, and a constraint condition that each element in the membership matrix satisfies. Wherein, element represents the membership degree of the j-th historical cross section sample to the i-th class;

[0070] The target function and the constraint condition are solved to obtain an iterative model including the clustering center of each class and the membership degree of the j-th historical cross section sample to the i-th class;

[0071] Obtain the number of cluster centers, the initial value of each cluster center, and the initial membership matrix of the pre-configured cluster centers;

[0072] Based on the number of cluster centers and the initial membership matrix, iterative calculations are performed according to the iterative model until the convergence condition is met or the maximum number of iterations is reached, at which point the clustering result is obtained.

[0073] For a transmission section sample set X=[x1, x2, ..., xm] containing m samples... m ] T Each sample is represented by a set of features as shown in equation (2), and satisfies x j =[x j1 x j2 , ..., x jN x jN+1 The clusters j = 1, 2, ..., m are divided into c classes using the FCM algorithm. During the iterative process of this clustering calculation, the optimization objective is to minimize the weighted sum of squared distances between all samples and the centers of each class. The objective function is as follows: (3), where V = [v1, v2, ..., v] c ] T This represents a matrix consisting of cluster centers for each category, where vector v i =[v i1 v i2 , ..., v iN v iN+1 ], i=1,2,…,c, which are the cluster centers of each class, and d (·) represents the vector x. j With v i Euclidean distance, The ambiguity coefficient is typically set to 2, U = [μ] ij That is, the membership matrix, which represents the membership relationship of each sample to each category, where μ ij Let represent the membership degree of the j-th sample to the i-th class, satisfying the following constraints: (4).

[0074] Solving equations (3) and (4), we obtain the iterative model:

[0075] (5).

[0076] After giving an initial value, iterate according to Equation (5) until the condition shown in Equation (6) is met or the maximum number of iterations is reached. Then the clustering ends and each sample is classified according to the membership value in the final U.

[0077] (6), where the superscript k represents the iteration number k. To preset the precision, a small positive number is used. The greater the value is, the less the iteration number is, but the accuracy of the calculation result is lower.

[0078] The traditional FCM clustering method has the following shortcomings, resulting in poor clustering effect, and the following three improvements are proposed in the embodiment.

[0079] (1) Adaptive selection of optimal cluster number:

[0080] In the iteration process described above, the number of clusters c needs to be specified artificially in advance. In order to reduce the subjectivity of the algorithm, the target function J and the sum of intra-class distances d in after clustering are used as indicators to reflect the clustering degree of the clustering result, and the "elbow point method" is used to determine the optimal cluster number. For different cluster numbers c, the calculation method of the sum of intra-class distances is as follows:

[0081] (7), wherein C k is the set composed of the samples contained in the kth class in the clustering result. As can be seen from formula (7), the smaller d in is, the more concentrated the samples in each class are, and the better the clustering effect is. When c = 1, all samples are in the same class, and at this time, the sample concentration degree is the lowest, d in takes the maximum value, and as c increases, d in shows a downward trend as a whole, and when the number of clusters is equal to the number of samples, at this time, each sample is the clustering center of each class, and d in is 0.

[0082] Using the "elbow point method", first, the value range of the number of clusters needs to be determined, that is, the number of clusters L is set. Secondly, the target function and the sum of intra-class distances of each clustering are calculated, and the curves of J and d in with c are obtained, and the change trend is shown in Figure 3 . When c is less than a certain cluster number c * , the decrease of d in will be very fast, because increasing c will greatly increase the tightness of each class, and when c reaches c * , the degree of tightness growth will gradually decrease, so the decreasing trend of d in will slow down. The relationship between d in and c is similar to the "elbow", and the "elbow point", that is, the c * value corresponding to the maximum change of the slope is the optimal cluster number.

[0083] (2) FCM clustering considering sample shape similarity:

[0084] The cluster sample in the application is essentially a feature sequence, the Euclidean distance as a kind of hard measurement method mainly reflects the numerical difference between samples, and has certain limitations in measuring the similarity between sample sequences of the same kind. The cosine distance judges the similarity between vectors according to the direction of the vector, is not affected by the numerical value of the vector itself, and can better reflect the morphological difference of the sample sequence. Therefore, the cosine distance is introduced in the application, the Euclidean distance and the cosine distance are combined, and the adaptability of the algorithm is ensured. Therefore, the composite distance is defined to replace the Euclidean distance in formula (3) for clustering calculation, and the calculation method is as follows:

[0085] For any vector x=(x1, x2, …, x n ) T and y=(y1, y2, …, y n ) T , the cosine distance d cos (x, y) is calculated as follows:

[0086] (8), wherein the numerator on the right side of the equation represents the inner product of the vector, and the denominator represents the product of the vector modulus, and the composite distance d(x, y) is obtained by weighted summation of the Euclidean distance and the cosine distance:

[0087] (9), wherein represents the weight coefficient of the cosine distance, represents the weight coefficient of the Euclidean distance, d euc (x, y) represents the Euclidean distance between the jth historical cross-section sample x j and the i th clustering center v i , and d cos (x, y) represents the cosine distance between the jth historical cross-section sample x j and the i th clustering center v i .

[0088] The weight coefficient can be regarded as the contribution degree of the cosine distance and the Euclidean distance to the composite distance. In the application, it is considered that both distances are effective in measuring the similarity of vectors, so the contribution degrees of the two distances to the composite distance are equal. In addition, considering that the value ranges of the two distances may be different, but the minimum value is 0, in order to avoid feature masking due to numerical difference, the ratio of the weight coefficients also needs to be adjusted, and the relationship of the weight coefficients is as follows: (10), wherein d euc,max and d cos,max are the maximum value of the sample Euclidean distance and the maximum value of the cosine distance respectively, and takes the value of 0.5.

[0089] (3) The determination method of the initial value of the clustering center is as follows:

[0090] Calculate the distance between any two historical cross-section samples, and combine a plurality of distances to obtain a distance matrix of m*m size; wherein m is the total number of historical cross-section samples;

[0091] Select two first historical cross-section samples corresponding to the minimum distance value from the distance matrix, and take the midpoint of the two first historical cross-section samples as the initial value of the cluster center of the first category;

[0092] Calculate the distance sum of the remaining historical cross-section samples and the two first historical cross-section samples, select the second historical cross-section sample with the maximum distance sum, find the third historical cross-section sample closest to the second historical cross-section sample, and take the midpoint of the second historical cross-section sample and the third historical cross-section sample as the initial value of the cluster center of the second category; wherein the second historical cross-section sample is selected from the remaining historical cross-section samples;

[0093] Repeat the calculation of the distance sum and the selection of the maximum distance sum until the initial value of the cluster center of the cth category is obtained; wherein c represents the number of categories of the cluster center.

[0094] Specifically, the FCM clustering algorithm is sensitive to the selection of the initial value of the cluster center. If the initial value of the membership matrix is not set reasonably, the clustering result is easy to fall into a local optimal condition. In addition, the present application does not focus on whether all the cluster centers in the iteration process fall in the historical samples. Therefore, only the iteration initial value needs to be corrected to obtain an effective clustering result. Considering that the principle of selecting the initial value of the cluster center should make the distance between each initial cluster center large enough but also not too dispersed, so as to avoid the problem that the algorithm falls into a local optimal problem due to the too close distance between the initial cluster centers or too deviating from the region of the sample set. Therefore, the present application selects the iteration initial value based on the following method:

[0095] 1) Calculate the composite distance between any two cluster samples according to formula (9) to form a composite distance matrix ;

[0096] 2) According to the obtained composite distance matrix, find the two first historical cross-section samples closest to each other, take the midpoint of the two first historical cross-section samples as the initial value of the cluster center of the first category, and mark it as v1 (0) ;

[0097] 3) Calculate the distance sum of the remaining historical cross-section samples and all selected first historical cross-section samples, find the second historical cross-section sample with the maximum distance sum, find the third historical cross-section sample closest to the second historical cross-section sample, and take the midpoint of the second historical cross-section sample and the third historical cross-section sample as the initial value of the cluster center of the second category, and mark it as v2 (0) ;

[0098] 4) Repeat the above steps until the initial value V of the cluster center of the c-th category is obtained. (0) .

[0099] Having obtained the initial values ​​of each cluster center according to the above steps, the iterative process shown in equation (5) is slightly modified, no longer using the randomly generated membership matrix U. (0) As the initial value for iteration, the initial value V of each cluster center is obtained. (0) As the initial value for the entire iteration, we obtain Then the calculation continues, thus avoiding the impact of the randomness of the initial value setting of the membership matrix on the clustering algorithm and preventing the algorithm from getting trapped in local optima.

[0100] Thus, the process of clustering based on the improved FCM algorithm is as follows: Figure 4 As shown, the specific steps are as follows:

[0101] Step 4.1: Use the feature sample set X obtained above as the input sample for the improved FCM clustering algorithm, and set the relevant parameters that need to be set in advance for the improved FCM clustering, including the fuzzy coefficient α and the maximum number of iterations k. max The upper and lower bounds c of the number of clusters min c max And set the iteration count k=1, c=c min ;

[0102] Step 4.2: Calculate the composite distance between any two clustered samples according to equation (9) above, and form the composite distance matrix. And based on the improved calculation of the initial value V for iteration (0) and U (0) ;

[0103] Step 4.3: Iteratively update U according to equation (5). (k) V (k) ;

[0104] Step 4.4: According to equation (6) or k≥k max Determine whether to terminate the iterative calculation. If any of the above conditions are met, record the value of U for that iteration. (k) V (k) Calculation results;

[0105] Step 4.5: If none of the above conditions are met, let k = k + 1, and repeat step 4.3 until the judgment condition mentioned in step 4.4 is met, then terminate the iterative calculation and record the final iterative result;

[0106] Step 4.6: Based on the output calculation results, obtain the classification labels, calculate and record the intra-class distance and d. in Used to select the optimal number of clusters;

[0107] Step 4.7: judge whether c≥c max is not established, let c=c+1 and k=1, then repeat step 4.2-step 4.6 until the inequality is established, then output J and d in The curve of c changes, the optimal clustering number c* is determined by the "elbow method", finally, the U of c=c* is output (k) , V (k) and other calculation results, thus, the program ends.

[0108] In order to effectively evaluate the effectiveness of the clustering results, the silhouette coefficient, Davies-Bouldin index and Calinski-Harabaz index are selected to evaluate the advantages and disadvantages of different clustering results, and the specific calculation is as follows:

[0109] (11), wherein SC is the silhouette coefficient, , The average Euclidean distance of sample j and other samples of the same class and non-same class. The value range of SC is-1~1, and the larger the value is, the better the clustering result is.

[0110] (12), wherein DB is the Davies-Bouldin index, The average Euclidean distance of sample j and other samples of the same class and non-same class. The value range of SC is-1~1, and the larger the value is, the better the clustering result is.

[0111] (13)

[0112] (14), wherein CH is the Calinski-Harabaz index, the index B represents the inter-class dispersion degree, and the index W represents the intra-class compactness, The average value of all samples. The larger the CH is, the better the clustering result is.

[0113] S104, calculate the similarity between the feature clustering result of each historical cross-section sample and the preset feature, and select the feature of the historical cross-section sample with the largest similarity as the analysis value of the future cross-section sample cross-section limit transmission capacity; wherein, the preset feature refers to the characteristic quantity corresponding to each historical cross-section sample.

[0114] Specifically, the FCM clustering algorithm is improved to cluster the similarity identification of the power transmission section sample, and based on the feature clustering result, the difference between the future sample and the historical sample in the load level, unit output, section power and other characteristics is quantitatively calculated, and the similarity of the future sample and the historical sample is compared. And the existing TTC value of the historical section with high similarity to the future section is used as the reference value of the TTC of the future section. Thus, the TTC evaluation of the section based on the similarity identification of the power transmission section is completed, and the TTC of the future section is quickly and accurately obtained. The present application identifies the historical operation mode similar to the future operation mode through the clustering method, and then uses the existing TTC of the similar historical power transmission section as the TTC reference value of the future power transmission section. When the power system is disturbed and the TTC of the section is needed to be quickly obtained for section control, the accurate calculation of the TTC of the future section can be quickly realized based on the existing historical data, and the huge manual calculation cost in practice can be effectively reduced.

[0115] To verify the effectiveness and adaptability of the proposed section limit transmission capacity evaluation method, Example 1 selects the IEEE 39-node system as a test example, and considering that the object concerned by the present application is the power transmission section, the power transmission lines of the 39-node system are divided in advance to obtain four power transmission sections, which are used for subsequent testing. The composition of each power transmission section and the first and last node numbers of the power transmission lines contained are shown in Table 2, and the specific division is shown in Table 3. Figure 5

[0116] Table 2 Section information of IEEE 39-node system

[0117]

[0118] (2) Analysis of the results of Example 1

[0119] The present application takes the typical parameters of the IEEE 39-node system as the basic working condition, the load is set to fluctuate within ±10%, and the power flow calculation is ensured to converge and meet the thermal stability constraint. According to the above setting, 3000 samples are generated, and TTC is obtained in advance through calculation to simulate the historical operation mode with TTC label in practice. Subsequently, 2500 samples are randomly selected as historical power transmission section samples, and 500 samples are randomly selected as future power transmission section samples.

[0120] ​Firstly, the optimal clustering number is selected based on the elbow point method. The range of the clustering number c is set to 2-11, that is, 10 times of clustering are performed; secondly, the historical samples are clustered by the improved FCM clustering algorithm, and the objective function and the intra-class distance sum under different clustering numbers are calculated; finally, the change curves of each clustering index and the clustering number are obtained, and the optimal clustering number is selected by the elbow point method. Since the elbow point method selects the elbow point by the curve shape, and is not sensitive to the numerical value, therefore, in order to better show the relationship between J and d in and the clustering number, the two are divided by the maximum value of each other, and the results are shown in Figure 6 It can be seen that the optimal clustering number c*=6. In addition, the SC, CH and DB three indexes are calculated to evaluate the advantages and disadvantages of the clustering results, and the results are shown in Figure 7 From the relationship between the three indexes SC, CH and DB and the clustering number in Figure 7 , it can be obtained that when the clustering number is 6, the DB takes the minimum value, and the SC and CH also take the maximum value, although when the clustering number is 4, there is a smaller DB value, but the SC and CH values are not large at this time, therefore, when the clustering number is 6, the clustering effect is best, which is consistent with the conclusion of determining the optimal clustering number by the elbow point method. Therefore, the subsequent clustering number c of the application is selected as 6.

[0121] Secondly, in order to reflect the advantages of the algorithm proposed in the application, the following schemes are set for comparison, and the specific scheme setting is shown in Table 3. The clustering number of scheme 1 is also set to 6.

[0122] Table 3 Different scheme scene setting

[0123]

[0124] The screened samples are used as the input of the above schemes. Considering that the initial value of the traditional FCM clustering is randomly generated, the clustering results of each time may not be the same, in order to more objectively evaluate, each scheme is run for 10 times, the average value of each index is taken, the clustering results are obtained and the corresponding evaluation indexes are calculated, and the results are shown in Table 4.

[0125] Table 4 Comparison of evaluation indexes of different clustering results

[0126]

[0127] From Table 4, compared with schemes 1 and 2, the method (scheme 4) of the present application has improved performance in various indicators, and the improvement degree is about 12% and 3% respectively, and does not significantly increase the computational complexity. The indicators of scheme 3 have about 1.2% improvement on the basis of scheme 2, so it can be obtained that the introduction of cosine distance can improve the performance of the clustering algorithm. Although using the compound distance instead of the Euclidean distance will reduce the computational efficiency and increase the calculation time, considering that the FCM clustering algorithm can be completed within seconds even when dealing with tens of thousands of sample inputs. Therefore, considering the compound distance for sample clustering can still meet the demand of practical application. In addition, due to the improvement of the initial value of iteration, compared with scheme 3, the indicators of scheme 4 are improved, and the calculation speed is equivalent.

[0128] Then, the future samples are mixed with the historical samples and clustered again, and optionally a future sample is compared with the center of the same sample and the center of the different sample in the characteristics of the future sample, unit output, load distribution, and cross-section transmission power flow. The results are shown in Figure 8 , and to enhance the visualization effect of the legend, the ordinate is logarithmically processed.

[0129] From Figure 8 , it can be obtained that the same sample and the future sample we are interested in have a very high similarity, and their load distribution, unit output, and cross-section TTC are almost the same, that is, the operation of the future sample and its similar historical sample is basically the same, and the cross-section TTC is also basically the same. For different samples, their characteristics are quite different. Considering that the actual dispatching will at most reserve 10% control margin, and the control limit is often rounded, therefore, when the error of the two is less than 5%, the cross-section TTC calculated by the two can be considered the same; when the error is greater than or equal to 10%, obviously, the cross-section TTC of the two is quite different and cannot be considered the same; when the error is between 5% and 10%, it needs to be further defined by manual. Therefore, the error range of the subsequent error statistics of the present application is distinguished by 5% and 10%. And based on Figure 8 , the error of the centers of 500 future samples, same samples and different samples in the cross-section TTC is quantitatively counted, and the results are shown in Table 5, where δ is the percentage error of the sample, and the calculation objects are the centers of the future samples and the same samples and the centers of the different samples. In addition, in order to ensure that the cross-section similarity recognition result can meet the demand of practical application, considering that the error interval which needs to be manually defined is also set in this paper, it is considered that the effective recognition quantity needs to reach more than 80% to have the value of practical application, and the higher the better.

[0130] Table 5 Sample-error statistical results of IEEE 39-node system

[0131]

[0132] As can be seen from Table 5, about 84% of the future samples with high similarity to the historical samples TTC can be identified by the method adopted by the application; when distinguishing from non-similar samples, the effective identification quantity can also reach more than 85%, and only about 2% of the samples have classification errors. In addition, Table 5 also quantitatively counts the total time consumption from feature preprocessing to the output of the sample sequence error, that is, the total time consumption of a single identification. It can be seen that scheme 1 has the highest solving efficiency, but the effective identification quantity is less than 80%, the number of future operation modes that need to be manually defined or manually adjusted increases greatly, the method adopted by the application has the longest time consumption, about 10.67s, but can effectively reduce the workload of manual calculation. Considering that the time consumption can be completed in seconds, therefore, the method adopted by the application can meet the daily application requirements of system operation.

[0133] Subsequently, the load level of the IEEE 39-node system is set to vary in the range of 70% to 130% of the standard load, and the output upper limit of all units is appropriately increased in the same proportion to balance the load growth and ensure power flow convergence, and the rest is the same as described above. A total of 1000 samples are generated, of which 800 are historical samples and 200 are future samples. The above future samples are identified one by one, and the transmission section identification error statistical results are shown in Table 6. As can be seen from Table 6, the effective identification quantity of the method adopted by the application in the similarity of similar samples and the difference from other samples is about 90%, and compared with other methods, the TTC similarity of scheme 4 is the highest, that is, when the load changes greatly, the method adopted by the application still has effectiveness.

[0134] Table 6 Sample-error statistical results when the load variation range is increased

[0135]

[0136] Finally, node 5 is disconnected from the system due to planned maintenance and other conditions as a new scenario 1, a new generator node 40 is added and connected to node 7 and node 11 as a new scenario 2, and the load is set to fluctuate in the range of ±10%, and the rest is the same as described above. Each generates 500 groups of samples, of which 80% are historical samples and 20% are future samples, and the identification results are shown in Table 7. From it, for the two types of new scenarios, the method adopted by the application can obtain future sections with high similarity to historical sections, and the similarity of TTC is also the highest, more than 80% of the future samples can be accurately identified, that is, it can meet the requirement of taking the historical section TTC as the reference value of the future section TTC when the power grid topology changes to a certain extent.

[0137] Table 7 Comparison of sample-error statistical results when the topology changes

[0138]

[0139] (3)Example 2 introduces

[0140] Example 2 is to verify the effectiveness and adaptability of the method adopted by the present application in the actual power grid application, and a 220kV and above actual power grid in a certain place in southwest China is selected as the test target. Based on experience, four power transmission sections are selected as the research objects, a total of 1000 samples with TTC labels are collected, and it is ensured that the load fluctuation range can basically cover the upper and lower limits of the grid supply load, and 800 samples are randomly selected as historical samples and 200 samples as future samples.

[0141] The historical samples collected from the actual system are clustered according to the method, and the error of the future samples and the same sample and non-same sample center on the TTC of the section is quantitatively counted, and the results are shown in Table 8. As shown in Table 8, in the actual power grid, the present application can identify more than 85% of the future samples with high similarity to the historical sample TTC. Since the complexity of the actual power grid is significantly increased compared to the 39-node system, the time consumption of the single identification of the method adopted by the present application is significantly increased, but its calculation efficiency can still meet the demand of daily application.

[0142] Table 8 Sample-error statistical results of an actual power grid in a certain place

[0143]

[0144] Since the above samples have been selected to ensure that their load levels can basically cover the range of grid supply load changes, the method adopted by the present application also has adaptability within the load change range of the actual power grid. In addition, considering that the number of nodes of the actual power grid is generally increased in units of years, and the operating personnel often pay special attention to the annual operation mode in actual control, it is more necessary to study the changes of the topology structure caused by daily maintenance or N-1 temporary operation mode for the actual power grid. Therefore, 500 operation samples of the actual power grid in the planned maintenance mode are collected, 400 of which are taken as historical samples and 100 as future samples, and the error statistical results of the identification of their TTC and each feature are shown in Table 9.

[0145] Table 9 Sample-error statistical results of an actual power grid topology change

[0146]

[0147] As shown in Table 9, in the actual power grid, the method adopted by the present application also has good performance when the topology changes, and can effectively identify more than 85% of the future samples with high similarity to the historical sample TTC. Therefore, it also has adaptability to the topology changes in the daily operation mode of the actual power grid.

[0148] From the simulation examples described above, the application proposes a transmission section limit transmission capacity analysis method based on transmission section similarity recognition. Under this framework, first, in order to obtain more comprehensive section feature samples, a hierarchical decision model is established to screen and improve the basic feature quantities of the power grid. Secondly, the improved FCM clustering is used for section similarity recognition, the shape distance is introduced into the clustering algorithm, the numerical and shape characteristics of the samples are more comprehensively considered, and the multi-index dynamic evaluation clustering result is used to adaptively select the optimal clustering number, improve the reliability and objectivity of the clustering algorithm. Finally, taking the IEEE 39-node system and a certain local power grid in Southwest China as examples, the effectiveness of the proposed framework and method is verified. The proposed method can effectively realize the recognition of the future transmission section and the similar historical section, and take the TTC of the historical section as the reference value of the TTC of the future section, so as to realize the fast and accurate acquisition of the TTC of the future transmission section. Therefore, the application has the following beneficial effects:

[0149] (1) More comprehensive consideration of section feature samples, through feature screening and improvement, the performance of section division results can be effectively improved, and the effectiveness of transmission section similarity recognition can be improved.

[0150] (2) Through the improved FCM clustering algorithm embedded with multi-scale key features, historical samples with high similarity to future samples can be effectively recognized and obtained, and historical samples with large differences can be distinguished, so as to realize the fast and accurate acquisition of the TTC of the transmission section, and the calculation time can also meet the requirements of actual application.

[0151] (3) When the load fluctuation and topology change to a certain extent, the method can still effectively recognize the historical samples similar to the future samples and distinguish the other historical samples with large differences, so as to realize the fast and accurate acquisition of the TTC of the transmission section, and can meet the requirements of the daily load fluctuation and maintenance mode of the power grid.

[0152] The embodiment of the application also provides a section limit transmission capacity analysis device based on similarity recognition, and the device comprises:

[0153] An information acquisition module is configured to acquire historical operation data of a transmission section.

[0154] A feature screening module is configured to screen the historical operation data to obtain a feature set of historical section samples.

[0155] A feature clustering module is configured to use an FCM algorithm to cluster the feature set to obtain a feature clustering result of each historical section sample.

[0156] The analysis module is configured to calculate the similarity between the feature clustering result of each historical cross-section sample and a preset feature, and select the feature of the historical cross-section sample with the maximum similarity as the analysis value of the cross-section limit transmission capacity of the future cross-section sample.

[0157] The analysis device for cross-section limit transmission capacity based on similarity recognition in the embodiments of the present application is a technical solution based on the same inventive concept as the analysis method for cross-section limit transmission capacity based on similarity recognition. Figure 1 The analysis device for cross-section limit transmission capacity based on similarity recognition in the embodiments of the present application is a technical solution based on the same inventive concept as the analysis method for cross-section limit transmission capacity based on similarity recognition.

[0158] The present application comprehensively considers the morphological features and numerical features of cross-section samples, explores the value in the historical operation data of cross-sections through the improved FCM clustering algorithm, realizes the similarity recognition of cross-sections, and takes the TTC of the similar historical sample as the reference value of the TTC of the future sample, effectively avoids the complex calculation process of obtaining the TTC, improves the calculation efficiency, and simultaneously refers to the TTC of the actual historical cross-section to ensure that the obtained result has practical significance and can be reproduced in the actual system.

[0159] The embodiments of the present application further provide an electronic device. The electronic device includes a processor, a memory, a communication interface, and at least one communication bus for connecting the processor, the memory, and the communication interface. The memory includes, but is not limited to, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (PROM), or a portable compact disc read-only memory (CD-ROM), and is used for storing related instructions and data.

[0160] The communication interface is configured to receive and send data. The processor can be one or more CPUs, and in the case of one CPU, the CPU can be a single-core CPU or a multi-core CPU. The processor in the electronic device is configured to read one or more programs stored in the memory and perform the following operations: obtaining historical operation data of a power transmission cross-section; filtering the historical operation data to obtain a feature set of historical cross-section samples; clustering the feature set using an FCM algorithm to obtain a feature clustering result of each historical cross-section sample; calculating the similarity between the feature clustering result of each historical cross-section sample and a preset feature, and selecting the feature of the historical cross-section sample with the maximum similarity as the analysis value of the cross-section limit transmission capacity of the future cross-section sample; wherein the preset feature refers to the feature quantity of the future cross-section sample corresponding to each historical cross-section sample.

[0161] It should be noted that the specific implementation of each operation can be described above with respect to the method embodiments shown Figure 1 According to the description of the method embodiments shown above, the electronic device can be used to perform a section limit transmission capacity analysis method based on similarity recognition, which will not be described in detail here.

[0162] The embodiments of the present application also provide a computer readable storage medium, which is a memory device in a computer device and is used to store programs and data. It can be understood that the computer readable storage medium herein can include an internal storage medium in the computer device, and of course can also include an extended storage medium supported by the computer device. The computer readable storage medium provides a storage space, which stores an operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and the instructions can be one or more computer programs (including program codes). It should be noted that the computer readable storage medium herein can be a high-speed RAM memory or a non-volatile memory such as at least one disk memory. One or more instructions stored in the computer readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the section limit transmission capacity analysis method based on similarity recognition in the above embodiments. Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0163] The embodiments of the present application also provide a computer program product containing program instructions. The computer program product can be software or a program product containing program instructions, which can run on a computing device or be stored in any available medium. When the computer program product runs on at least one electronic device, it makes the at least one electronic device execute a section limit transmission capacity analysis method based on similarity recognition.

[0164] The above specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for analyzing the sectional limit transmission capacity based on similarity identification, characterized in that the method... include: Obtain historical operating data of power transmission sections; By filtering historical operational data, a feature set of historical cross-sectional samples is obtained; The FCM algorithm is used to cluster the feature set to obtain the feature clustering results for each historical cross-section sample; Calculate the similarity between the feature clustering results of each historical cross-section sample and the preset features, and select the feature of the historical cross-section sample with the highest similarity as the analysis value of the cross-sectional limit transmission capacity of the future cross-section sample; where the preset features refer to the feature quantity of the future cross-section sample corresponding to each historical cross-section sample.

2. The method according to claim 1, characterized in that, By filtering historical operational data, a feature set of historical cross-sectional samples is obtained, including: A hierarchical screening model is constructed using the decision tree algorithm; By using a hierarchical screening model to filter historical operational data, a feature set of historical cross-sectional samples is obtained.

3. The method according to claim 1, characterized in that, Before clustering the feature set using the FCM algorithm, the method further includes: normalizing the feature set of historical cross-section samples.

4. The method according to claim 1, characterized in that, The FCM algorithm is used to cluster the feature set to obtain the feature clustering results for each historical cross-section sample, including: Construct an objective function that minimizes the weighted sum of squared distances between all historical cross-section samples and each cluster center, and define the constraints that each element in the membership matrix must satisfy; where each element represents the membership degree of the j-th historical cross-section sample to the i-th category. Solving the objective function and constraints yields an iterative model that includes the cluster centers of each category and the membership of the j-th historical cross-section sample to the i-th category. Obtain the number of cluster centers, the initial value of each cluster center, and the initial membership matrix of the pre-configured cluster centers; Based on the number of cluster centers and the initial membership matrix, iterative calculations are performed according to the iterative model until the convergence condition is met or the maximum number of iterations is reached, at which point the clustering result is obtained.

5. The method according to claim 4, characterized in that, The number of cluster centers is determined as follows: Calculate the sum of intra-cluster distances between each cluster center based on the distance; Based on the objective function and intra-cluster distances after clustering, the number of cluster centers is determined by combining the elbow method.

6. The method according to claim 4, characterized in that, The expression for calculating the distance between each historical cross-section sample and each cluster center is as follows: Where ω1 represents the weighting coefficient of the cosine distance, ω2 represents the weighting coefficient of the Euclidean distance, and d euc (x, y) represents the j-th historical cross-section sample x. j With the i-th cluster center v i The Euclidean distance between them, d cos (x, y) represents the j-th historical cross-section sample x. j With the i-th cluster center v i The cosine distance between them.

7. The method according to claim 4, characterized in that, The initial values ​​of cluster centers are determined as follows: Calculate the distance between any two historical cross-sectional samples, and combine multiple distances to obtain a distance matrix of size m*m; where m is the total number of historical cross-sectional samples. Select the two first historical section samples with the smallest distance value from the distance matrix, and take the midpoint of the two first historical section samples as the initial value of the cluster center of the first category; Calculate the sum of distances between the remaining unselected historical section samples and the two first historical section samples, select the second historical section sample with the largest distance sum, find the third historical section sample that is closest to the second historical section sample, and use the midpoint between the second historical section sample and the third historical section sample as the initial value of the cluster center of the second category; wherein, the second historical section sample is selected from the remaining historical section samples. Repeat the calculation of the distance sum and the selection of the maximum distance sum until the initial value of the cluster center of the c-th category is obtained; where c represents the number of cluster centers.

8. A cross-sectional limit transmission capacity analysis device based on similarity recognition, characterized in that, The device includes: The information acquisition module is used to acquire historical operating data of the power transmission section; The feature filtering module is used to filter historical operational data to obtain a feature set of historical cross-sectional samples; The feature clustering module is used to cluster the feature set using the FCM algorithm to obtain the feature clustering results for each historical cross-section sample; The analysis module is used to calculate the similarity between the feature clustering results of each historical cross-section sample and the preset features, and selects the feature of the historical cross-section sample with the highest similarity as the analysis value of the cross-sectional limit transmission capacity of the future cross-section sample; where preset features refer to the feature quantity of the future cross-section sample corresponding to each historical cross-section sample.

9. An electronic device, characterized in that, Including memory and processor; A memory for storing computer programs, the computer programs including program instructions; A processor is configured to execute the program instructions to cause the electronic device to perform the steps of a cross-sectional limit transmission capacity analysis method based on similarity identification as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a computer program that, when executed by one or more processors, implements a method for cross-sectional limit transmission capacity analysis based on similarity identification as described in any one of claims 1 to 7.

Images