A method for analyzing dam service performance based on strong similarity sets

Abstract

The invention belongs to the technical field of dam service performance analysis, and specifically relates to a dam service performance analysis method based on strong similarity sets, comprising: dividing a dam into multiple parts, obtaining multi-measurement point monitoring data of each part; performing strong similarity set mining on the multi-measurement point monitoring data in the same part to obtain a strong similarity set covering all the measurement points; using a dam service performance analysis method within a single strong similarity set range to obtain dam service performance analysis results within the range of each strong similarity set; determining the influence weight of the dam service performance analysis results of each strong similarity set on the upper part, and fusing the analysis results of each strong similarity set upward layer by layer to obtain the service performance analysis results of each part in the part layer and the service performance analysis results of the dam as a whole.

Description

Technical Field

[0001] The invention belongs to the technical field of dam service performance analysis, and in particular relates to a dam service performance analysis method based on a strong similarity set. Background Art

[0002] During the operation of the dam, under the influence of complex and changeable environment and load factors, as well as the invasion of various sudden disasters, its service performance is constantly changing. Therefore, it is necessary to use various safety monitoring methods to grasp the service performance of the dam in real time, discover abnormal changes in time, and take corresponding measures to prevent major accidents or even accidents of the dam, so as to ensure the long-term safe operation of the project. At present, the dam service performance analysis methods used in practical engineering applications are mostly single-point methods, but the monitoring information of a single point cannot fully reflect the service performance changes of the entire project, and the change rules of the monitoring effect quantities of different points are not consistent, which is prone to contradictory analysis results. Therefore, it is necessary to use the monitoring information of multiple points from the perspective of multiple points, and obtain the service performance analysis results of the entire project through fusion. However, most of the existing multi-point service performance analysis methods only use the measurement value information of the measurement point itself, and do not make full use of the similar information between the measurement values ​​of the measurement points, which may cause the loss of key information and misjudgment of performance.

[0003] When using strong similarity sets to conduct real-time analysis of the dam's service performance, the strong similarity set mining results obtained in previous monitoring periods are needed. During the dam's service, due to the continuous changes in its own conditions and external working conditions, its changing properties will change, which may cause changes in the similarity relationship between the measured values ​​of the monitoring effect quantities at each measuring point, thereby causing the original strong similarity set mining results to become invalid at the current moment. If the invalid strong similarity set mining results are used to analyze the service performance, it is easy to cause misjudgment of the service performance. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to address the deficiencies of the prior art and provide a dam service performance analysis method based on strong similarity sets. Through strong similarity set mining and multi-level fusion analysis, it can comprehensively reflect the changes in the service performance of the dam and provide a new analysis method for dam safety monitoring.

[0005] Technical solution: The dam service performance analysis method based on strong similarity sets described in the present invention comprises the following steps:

[0006] S1. Divide the dam into multiple parts and obtain monitoring data of multiple measuring points in each part;

[0007] S2. Conduct strong similarity set mining on the monitoring data of multiple measuring points in the same part to obtain strong similarity sets covering all measuring points. According to the changes in the measurement sequence, adopt the strong similarity set composition status tracking and analysis method to obtain the discrimination results of the strong similarity set composition status at the current moment under the changes in the measurement sequence, and build a hierarchical analysis system for dam service performance based on multiple strong similarity sets;

[0008] S3, using the dam service performance analysis method within the scope of a single strong similarity set, obtaining the dam service performance analysis results within the scope of each strong similarity set;

[0009] S4. Calculate the discrete degree and quantitative characteristics of the service performance analysis results of each element in this layer, construct a judgment matrix based on the calculated discrete degree and quantitative characteristics of the service performance analysis results of each element, and determine the influence weight of the service performance analysis results of the dam of each strong similarity set on the upper layer;

[0010] S5. The analysis results of each strong similarity set are connected and fused upward layer by layer to obtain the service performance analysis results of each part in the part layer and the service performance analysis results of the dam as a whole.

[0011] To further improve the above technical solution, the method for tracking and analyzing the composition status of a strong similarity set in S2 for changes in the measured value sequence includes:

[0012] S201, normalizing the measurement value sequence of each measurement point in the strong similarity set;

[0013] S202: Determine the length of the sliding period according to the actual situation of the measured values ​​of each measuring point in the strong similarity set. ;

[0014] S203, in Within the sliding period, the TWED distance sequence between each measuring point and the other measuring points in the strong similarity set is calculated in different time periods, and the comprehensive distance sequence of the strong similarity set of each measuring point is constructed respectively;

[0015] S204. Count the maximum value of the comprehensive distance within a fixed length period , as the sample space for calculating the discrimination index of the belonging relationship of the strong similarity set of each measuring point;

[0016] S205. Based on the typical small probability method, the strong similarity set point belonging relationship discrimination index of each measuring point under a certain small probability is calculated respectively. ;

[0017] S206: Determine the time The measured value of the i-th measurement point of the strong similarity set is used to calculate the comprehensive distance of the strong similarity set of the measurement points , and the strong similarity set measurement point belonging relationship discrimination index of the measurement point By comparison, we can get the preliminary judgment result of the belonging relationship of the measurement points in the strong similarity set: , then it is believed that At this moment, the measurement point still belongs to this strong similarity set; if , then it is believed that At this moment, the measurement point no longer belongs to this strong similarity set;

[0018] S207, for the measurement points that are determined not to belong to the strong similarity set in the preliminary identification result of the belonging relationship, the measurement values ​​are processed for gross error identification, and the gross errors are eliminated and then re-analyzed to obtain the supplementary identification result of the belonging relationship;

[0019] S208, the measurement points that have been judged to belong to the original strong similarity set are formed into a strong similarity set subset. For the measurement points that are judged not to belong to the original strong similarity set in the supplementary judgment result of the belonging relationship, they are judged one by one whether they now belong to this strong similarity set subset. If they are judged to belong to this strong similarity set subset, it is considered that At the moment, the measurement point still belongs to the original strong similarity set; if it is judged not to belong to this strong similarity subset, it is considered that At this moment, the measurement point no longer belongs to this strong similarity set;

[0020] S209 , integrating the belonging relationship determination results of all the measurement points of the strong similarity set to obtain the composition status determination result of the strong similarity set at the current moment when the measurement sequence changes.

[0021] Furthermore, the dam service performance hierarchical analysis system divides the entire dam into four hierarchical levels from bottom to top, namely, the measuring point layer, the strong similarity set layer at the same position, the position layer, and the dam overall layer. Each element in the same layer has the same level, and the elements in the upper and lower adjacent layers have a relationship of whole and part.

[0022] Furthermore, for the elements in this layer, an element in the previous layer The corresponding The element set is , in this layer The element corresponds to the next layer The element set is .

[0023] Further, the S3 includes:

[0024] S301: Yes The monitoring effect value of each measuring point in the strong similarity set within the analysis period is normalized;

[0025] S302: For the normalized multidimensional data, input the local linear embedding dimensionality reduction algorithm, adjust the number of nearest neighbors k, and obtain a preliminary dimensionality reduction result;

[0026] S303: adjusting the value range according to the variation range of the average value sequence of the measured values ​​of each measuring point in the strong similarity set, and obtaining the comprehensive effect quantity sequence Y reflecting the variation of the service performance of the dam within the range of the strong similarity set in this period;

[0027] S304: The comprehensive effect size sequence within the time period is input into the reverse cloud generator to obtain three digital features Ex, En and He. Ex, En and He are the expectation, entropy and super entropy of the comprehensive effect size sequence cloud model of the strong similarity set respectively.

[0028] S305: inputting the digital features Ex, En and He into the forward cloud generator to calculate a cloud model reflecting the distribution of the comprehensive effect size;

[0029] S306: Cloud-based "3En" guidelines, proposed in The upper limit of the property discrimination index corresponding to the moment and the lower limit ;

[0030] S307: The comprehensive effect size of the strong similarity set at the moment , and the property discrimination index value as well as By comparison, we can obtain the preliminary analysis results of the dam service performance within the range of this strong similarity set at that moment;

[0031] S308: Utilize formula The preliminary analysis results are quantified to obtain the service performance analysis results of the dam within the range of this strong similarity set at that moment.

[0032] Furthermore, the discrete degree of the service performance analysis results of each element in the lower layer corresponding to the element in this layer is calculated. ;

[0033] Calculate the quantitative characteristics of the elements in this layer corresponding to the elements in the lower layer ;

[0034] according to and Construct judgment matrix D;

[0035] The following formula is used to calculate the influence of the service performance analysis results of the elements of this layer on the upper layer at time t: ,

[0036] ,

[0037] Where: for Elements relative to other elements on the same layer The sum of the relative importance of is the sum of all items in the judgment matrix D, is the number of elements in this layer.

[0038] Furthermore, the analysis results of each strong similarity set are fused upward layer by layer to obtain the service performance analysis results of each part in the part layer and the service performance analysis results of the dam as a whole, including:

[0039] At the tth moment, the service performance analysis result set of each element in the next layer is obtained as follows: , and fuse them to obtain this layer Analysis results of service properties of elements , further fused to get the previous layer Analysis results of service properties of elements , thus completing the process of handover and fusion layer by layer from bottom to top.

[0040] Furthermore, the analysis results of each strong similarity set are fused upward layer by layer to obtain the service performance analysis results of each part in the part layer and the service performance analysis results of the dam as a whole, including: at the tth moment, according to the elements in the previous layer The service performance analysis results of the corresponding elements in this layer and the influence weights of the elements in this layer on the previous layer are combined to obtain the elements in the previous layer. The service performance analysis results , and its calculation formula is:

[0041] .

[0042] Beneficial effects: Compared with the prior art, the advantages of the present invention are: (1) In order to ensure the real-time effectiveness of the results of strong similarity set mining, the similarity changes between the measured values ​​of the measuring points are considered, and a method for calculating the time deviation editing distance between the measuring point and the other measuring points in the strong similarity set by sliding in different time periods is studied. By using this method, a comprehensive distance that can quantitatively represent the belonging relationship between the measuring point and the strong similarity set is constructed, and the belonging relationship between the measuring point and the strong similarity set is quantitatively represented. Based on the typical small probability method, a strong similarity set measuring point belonging relationship discrimination index is constructed, and a tracking and analysis method for the composition status of the strong similarity set under the change of the measurement value sequence is proposed. The method can track the changes in the composition status of the strong similarity set in real time, and timely discover the changes in the belonging relationship between the measuring point and the strong similarity set, thereby ensuring the real-time effectiveness of the results of strong similarity set mining. (2) From the perspective of information fusion, the dimensionality reduction processing of the multi-point data in the strong similarity set is carried out based on the local linear embedding method, and the performance discrimination index is proposed in combination with the cloud model theory. A method for analyzing the service performance of the dam within a single strong similarity set is proposed. The information of multiple measuring points in the strong similarity set can be integrated to obtain the comprehensive effect quantity reflecting the service performance of the dam in the local range, and the analysis of the service performance of the dam within the single strong similarity set is realized. (3) In order to further analyze the service performance changes of each part and even the whole dam, a hierarchical analysis system of the dam service performance based on multiple strong similarity sets is proposed using the idea of ​​hierarchical analysis. A judgment matrix fusion method based on importance is constructed, and a layer-by-layer fusion analysis method of the dam service performance within the range of multiple strong similarity sets is proposed. It can realize the analysis of the dam service performance from local to overall, obtain the service performance analysis results of each part and the whole dam, and identify abnormal parts and give them special attention.

[0043] The present invention can comprehensively reflect the service performance changes of the dam through strong similarity set mining and multi-level fusion analysis, and provides a new analysis method for dam safety monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a structural schematic diagram of the layer-by-layer fusion analysis method of the dam service performance within the scope of multiple strong similarity sets of the present invention;

[0045] Figure 2 It is a method for tracking and analyzing the composition status of strong similarity sets under changes in the measurement sequence in the present invention;

[0046] Figure 3 It is a hierarchical analysis system of dam service performance based on multiple strong similarity sets in the present invention;

[0047] Figure 4 It is a flow chart of dam service performance analysis within the scope of a single strong similarity set in the present invention. DETAILED DESCRIPTION

[0048] The technical solution of the present invention is described in detail below with reference to the accompanying drawings, but the protection scope of the present invention is not limited to the embodiments.

[0049] Example 1: Figure 1 The layer-by-layer fusion analysis method for dam service performance within the scope of multiple strong similarity sets is shown in the figure. The specific implementation steps are as follows:

[0050] S1. According to the actual situation of the dam, the dam is divided into E parts, and the multi-point monitoring data of each part is obtained;

[0051] S2. Perform strong similarity set mining on the monitoring data of multiple measuring points in the same part to obtain strong similarity sets covering all measuring points. According to the changes in the measurement sequence, a strong similarity set composition status tracking and analysis method is used to obtain the discrimination results of the strong similarity set composition status at the current moment under the changes in the measurement sequence, thereby constructing a hierarchical analysis system for the service performance of the dam;

[0052] S3, using the dam service performance analysis method within the scope of a single strong similarity set, obtain the dam service performance analysis results s within the scope of each strong similarity set;

[0053] S4. Determine the influence weight of the dam service performance analysis results of each strong similarity set on the upper part based on the normalized data of the monitoring effect measurement values ​​of multiple measuring points in each strong similarity set ;

[0054] S5. Fusion calculation to obtain the service performance analysis results QZ of each part of the dam;

[0055] S6. According to the service performance analysis results of each strong similarity set in each part of the dam, determine the influence weight of the service performance analysis results of each part of the dam on the overall service performance of the dam ;

[0056] S7. The fusion calculation results Z of the overall service performance analysis of the dam are obtained.

[0057] In step S2, the time deviation edit distance similarity analysis method is used to analyze the multi-point monitoring data of the dam to obtain the time deviation edit distance similarity matrix, and the time deviation edit distance similarity matrix is ​​converted into a multi-point graph adjacency matrix, and the multi-point monitoring effect quantity measurement information and the adjacency matrix are jointly represented as a multi-point graph; the greedy module is used to maximize the strong similarity set mining method to perform strong similarity set mining on the multi-point graph to obtain the multi-point graph strong similarity set subgraph partitioning result, and the time deviation edit distance similarity matrix is ​​converted into a multi-point graph adjacency matrix; the multi-point monitoring effect quantity measurement information and the adjacency matrix are jointly represented as a multi-point graph; the greedy module is used to maximize the strong similarity set mining method to perform strong similarity set mining on the multi-point graph to obtain the multi-point graph strong similarity set subgraph partitioning result.

[0058] For the measured value of a certain measuring point in the strong similarity set, when the change of the dam's variable state causes the similarity relationship between the measured values ​​of the monitoring effect quantities of each measuring point to change, it may no longer have a strong similarity with the measured values ​​of the other measuring points in the original strong similarity set, that is, it no longer belongs to the strong similarity set. At this time, the TWED distance between the measured value of the measuring point and the measured values ​​of the other measuring points in the original strong similarity set calculated by the time deviation edit distance method will increase. Using this property, the comprehensive distance between the measured value of the measuring point and the measured values ​​of all other measuring points in the strong similarity set (hereinafter referred to as the comprehensive distance of the strong similarity set of the measuring point) can be constructed, which is used as a quantitative representation of the belonging relationship between the measuring point and the strong similarity set. On this basis, based on the idea of ​​the typical small probability method, a discrimination index for the belonging relationship of the measuring points in the strong similarity set is constructed as the basis for discriminating the belonging relationship between the measuring point and the strong similarity set. The method for tracking and analyzing the composition status of the strong similarity set under the change of the measured value sequence, its basic process is shown in Figure 2 , the specific implementation steps are as follows:

[0059] S201, normalizing the measurement value sequence of each measurement point in the strong similarity set;

[0060] S202: Determine the length of the sliding period according to the actual situation of the measured values ​​of each measuring point in the strong similarity set. ;

[0061] S203, in Within the sliding period, the TWED distance sequence between each measuring point and the other measuring points in the strong similarity set is calculated in different time periods. On this basis, the comprehensive distance sequence of the strong similarity set of each measuring point is constructed respectively.

[0062] S204. Count the maximum value of the comprehensive distance within a fixed length period In the comprehensive distance sequence between the measured value of the measuring point and the measured values ​​of other measuring points in the strong similarity set, the maximum value of the comprehensive distance in each fixed length period is Extract them one by one to form a sample space X for calculating the strong similarity set of each measuring point's belonging relationship discrimination index;

[0063] S205, X belongs to a small sample space, calculate the statistical eigenvalue of the sample space X (sample mean) and (sample mean square error), let is the extreme value of the comprehensive distance, when When , it is considered that the similarity between the measured value of the measurement point and the measured values ​​of other measurement points in the strong similarity set has significantly weakened, and the measurement point no longer belongs to this strong similarity set, with a probability of Based on the typical small probability method, the strong similarity set point belonging relationship discrimination index of each measuring point under small probability is calculated respectively. , , F is determined by statistical test method Distribution function;

[0064] S206: Determine the time The measured value of the i-th measurement point of the strong similarity set is used to calculate the comprehensive distance of the strong similarity set of the measurement points ,

[0065]

[0066] Where: m is the number of measurement points in this strong similarity set; n is the length of the measurement sequence; t is the current calculation time; is the length of the sliding period; is the difference between the measured values ​​of this measuring point and other measuring points at time t and before TWED distance within the sliding period length; is the measurement point at time t and before The comprehensive distance of the strongly similar set of measurement points within the sliding period length.

[0067] In the process of constructing the comprehensive distance of the strong similarity set of measuring points, it is first necessary to calculate the TWED distance between the measured value of the measuring point and the measured values ​​of other measuring points in the strong similarity set by sliding in different time periods, and perform mean square processing on them, so as to obtain the comprehensive distance between the measured value of the measuring point and the measured values ​​of other measuring points in the strong similarity set and the discrimination index of the belonging relationship of the measuring point in the strong similarity set of the measuring point By comparison, we can get the preliminary judgment result of the belonging relationship of the measurement points in the strong similarity set: , then it is believed that At this moment, the measurement point still belongs to this strong similarity set; if , then it is believed that At this moment, the measurement point no longer belongs to this strong similarity set;

[0068] S207, for the measurement points that are determined not to belong to the strong similarity set in the preliminary identification result of the belonging relationship, the measurement values ​​are processed for gross error identification, and the gross errors are eliminated and then re-analyzed to obtain the supplementary identification result of the belonging relationship;

[0069] S208, the measurement points that have been judged to belong to the original strong similarity set are formed into a strong similarity set subset. For the measurement points that are judged not to belong to the original strong similarity set in the supplementary judgment result of the belonging relationship, it is necessary to judge whether they now belong to this strong similarity set subset. If they are judged to belong to this strong similarity set subset, it is considered that At the moment, the measurement point still belongs to the original strong similarity set; if it is judged not to belong to this strong similarity subset, it is considered that At this moment, the measurement point no longer belongs to this strong similarity set;

[0070] S209 , integrating the belonging relationship determination results of all the measurement points of the strong similarity set to obtain the composition status determination result of the strong similarity set at the current moment when the measurement sequence changes.

[0071] Multiple measuring points in a strong similarity set have similar change properties, which reflect the changes in the service performance of the dam in a certain part or a certain range. The local linear embedding method based on the dimensionality reduction method is used to reduce the dimensionality of the monitoring effect quantity monitoring data of multiple measuring points in the strong similarity set and fuse them into a one-dimensional comprehensive effect quantity that can reflect the changes in the service performance of the dam within the range, thereby achieving the purpose of information fusion analysis.

[0072] like Figure 4 As shown in the figure, the basic flow chart of the dam service performance analysis method within the scope of a single strong similarity set is as follows:

[0073] S301: Yes The monitoring effect value of each measuring point in the strong similarity set within the analysis period is normalized;

[0074] S302: For the normalized multidimensional data, input the local linear embedding dimensionality reduction algorithm, adjust the number of nearest neighbors k, and obtain a preliminary dimensionality reduction result;

[0075] S303: adjusting the value range according to the variation range of the average value sequence of the measured values ​​of each measuring point in the strong similarity set, and obtaining the comprehensive effect quantity sequence Y reflecting the variation of the service performance of the dam within the range of the strong similarity set in this period;

[0076] S304: The cloud generator includes a forward cloud generator and a reverse cloud generator. The comprehensive effect quantity sequence within the time period is input into the reverse cloud generator to obtain three digital features Ex, En and He. The expectation Ex is the expected value of the distribution of cloud droplets in the domain. The closer the cloud droplet distribution is to the expectation, the more concentrated it is, indicating that the concept cognition is consistent. The farther the cloud droplet distribution is from the expectation, the more dispersed it is, indicating that the difference in concept cognition is greater. Entropy En is used to measure the fuzziness of qualitative concepts. The larger the entropy, the more vague the qualitative concept is, the wider the quantitative value range corresponding to the qualitative concept in the domain, and the greater the discreteness of the cloud map. Super entropy He measures the uncertainty of entropy and can better reflect the stability of the indicator, that is, it reflects the degree of condensation of cloud droplets in the cloud map. Therefore, super entropy is also called entropy of entropy, which is determined by the fuzziness and randomness of entropy.

[0077] S305: inputting the digital features Ex, En and He into the forward cloud generator to calculate a cloud model reflecting the distribution of the comprehensive effect size;

[0078] S306: Cloud-based “3En” guidelines, proposed in The upper limit of the property discrimination index corresponding to the moment and the lower limit ;

[0079] S307: The comprehensive effect size of the strong similarity set at the moment , and the property discrimination index value as well as By comparison, we can obtain the preliminary analysis results of the dam service performance within the range of this strong similarity set at that moment;

[0080] S308: Utilize formula Quantify the preliminary analysis results. for The comprehensive effect value of the strong similarity set at the moment is used to obtain the analysis result s of the dam service performance within the scope of this strong similarity set at that moment. The value range of s is greater than or equal to 0. If 0≤s≤1, it indicates that there is no abnormality in the service performance of the dam within the scope of this strong similarity set; if s>1, it indicates that the service performance of the dam within the scope of this strong similarity set may be abnormal.

[0081] A hierarchical analysis system for dam service performance based on multiple strong similarity sets is proposed. The system divides the entire dam into four hierarchical levels from bottom to top. Each part in the same layer (hereinafter referred to as elements) has the same level. The elements of the upper and lower adjacent layers have a whole-part relationship. The main content of each layer and the relationship between adjacent layers are shown in Figure 3 , including: (1) Measurement point layer: represents the monitoring information of the single measurement point monitoring effect quantity of each measurement point, which is the basic layer of the entire multi-strong similarity set dam service performance hierarchical analysis system; (2) Strong similarity set layer in the same part: by mining the strong similarity set of the measurement values ​​of each monitoring effect quantity measurement point within the same part, each measurement point is divided into a certain strong similarity set. The strong similarity in the strong similarity set can be used to obtain a more accurate analysis result of the dam service performance within the range of the single strong similarity set. For example, the measurement points near the dam top in the riverbed dam section have a strong change similarity, and a single strong similarity set is used for them. The dam service performance analysis can obtain the service performance information reflected by the dam top part in the riverbed dam section; (3) Part layer: that is, the main parts of the dam that can be specifically distinguished, and can also be further subdivided into key parts and sections according to actual conditions. In this layer, the service performance analysis results of each strong similarity set in the strong similarity set layer can be obtained by hierarchical fusion. (4) Dam overall layer: The object of the overall layer is the service performance of the dam as a whole. The service performance analysis results of the dam as a whole are obtained by hierarchical fusion of the service performance analysis results of the layers below it.

[0082] The recursive fusion process of the dam service performance analysis results is analyzed: for an element in a certain layer, an element in the previous layer The corresponding The element set is , and in this layer The element corresponds to the next layer The element set is At the tth moment, the service performance analysis result set of each element in the next layer is obtained as , and fuse them to obtain this layer Analysis results of service properties of elements , and further integration is performed in the same way to obtain the previous layer Analysis results of service properties of elements , thus completing the process of handover and fusion layer by layer from bottom to top.

[0083] Assume that the judgment matrix between the service performance analysis results of each element in the same layer is D, and its expression is:

[0084] ,

[0085] Where: Represents an element within a layer The service performance analysis results of the dam are relatively The importance of the service performance analysis results of each element in this layer is the basis for determining the weights of hierarchical fusion of the service performance analysis results of each element in this layer to a higher layer.

[0086] If at time t, The closer the service performance analysis results of each element in the next layer corresponding to the element are, that is, the smaller the overall dispersion is, the more reliable the service performance analysis results are and the higher the credibility is. The present invention uses the standard deviation method for measurement, then at time t The discrete degree of the service performance analysis results of all the lower m elements corresponding to the element The calculation formula is:

[0087] ,

[0088] In the formula, For the next layer The service performance analysis results of the element at time t; It is the mathematical expectation of the service performance analysis results of each element in the next layer at the tth moment.

[0089] Based on this, we can get that in the period 1~T, The average dispersion value of the service performance analysis results of the lower-level elements corresponding to the element for:

[0090] .

[0091] like The smaller the value, the The smaller the discreteness of the service performance analysis results of the lower-level elements corresponding to the element, the smaller the discreteness of the service performance analysis results of the lower-level elements corresponding to the element. Therefore, when performing hierarchical fusion on the upper level, elements are given greater weight.

[0092] The more elements there are, the more monitoring information they have, and the better they can reflect the changes in the service performance of the dam. Therefore, the quantitative characteristics of the measured elements can be used as the basis for determining the weights. When the number of elements in the next layer corresponding to the element is m, calculate the number characteristics of the measured points in each strong similarity set ,

[0093] .

[0094] like The larger the value, the The more elements in the lower layer the element corresponds to, the more elements in the upper layer should be considered. elements are given greater weight.

[0095] According to the discrete degree of the service performance analysis results of each element in the next layer and the quantitative characteristics of the elements in the next layer, the judgment matrix at the tth moment can be obtained: In this layer Element relative to Relative importance of the dam service performance analysis results of the elements , and its calculation formula is:

[0096] .

[0097] On this basis, we can further obtain the The influence weight of the service performance analysis results of the element on the upper layer , and its calculation formula is:

[0098] ,

[0099] In the formula, for Elements relative to other elements on the same layer The sum of the relative importance of is the judgment matrix The sum of the terms in is the number of elements in this layer.

[0100] At the tth moment, according to the elements in the previous layer The service performance analysis results of the corresponding elements in this layer and the influence weights of the elements in this layer on the previous layer can be combined to obtain the elements in the previous layer. The service performance analysis results , and its calculation formula is: , and then the service performance analysis result QZ of each part in the part layer is obtained, and finally the service performance analysis result Z of the dam as a whole is obtained.

[0101] Assume that the dam is divided into E parts, and the service performance analysis results of each part at time t are Sorting from largest to smallest, the results are as follows:

[0102] .

[0103] when >1, indicating that the service performance analysis results of the corresponding parts are abnormal, especially for several parts with relatively large service performance analysis results in the sorting results, which should be given special attention.

[0104] As described above, although the present invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the present invention itself. Various changes in form and details may be made without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. A dam service performance analysis method based on strong similarity sets, characterized in that: The following steps are involved: S1. Divide the dam into multiple parts and obtain monitoring data of multiple measuring points in each part; S2. Conduct strong similarity set mining on the monitoring data of multiple measuring points in the same part to obtain strong similarity sets covering all measuring points. According to the changes in the measurement sequence, adopt the strong similarity set composition status tracking and analysis method to obtain the discrimination results of the strong similarity set composition status at the current moment under the changes in the measurement sequence, and build a hierarchical analysis system for dam service performance based on multiple strong similarity sets; The dam service performance hierarchical analysis system divides the entire dam into four hierarchical levels from bottom to top, namely, the measurement point layer, the strong similarity set layer at the same part, the part layer, and the dam overall layer. Each element in the same layer has the same level, and the elements of the upper and lower adjacent layers have the relationship of whole and part. For an element in this layer, an element in the upper layer The corresponding The element set is , in this layer The element corresponds to the next layer The element set is ; S3, using the dam service performance analysis method within the scope of a single strong similarity set, obtaining the dam service performance analysis results within the scope of each strong similarity set; S4. Calculate the discrete degree and quantitative characteristics of the service performance analysis results of each element in this layer, construct a judgment matrix based on the calculated discrete degree and quantitative characteristics of the service performance analysis results of each element, and determine the influence weight of the service performance analysis results of the dam of each strong similarity set on the upper layer; S5. The analysis results of each strong similarity set are connected and fused upward layer by layer to obtain the service performance analysis results of each part in the part layer and the service performance analysis results of the dam as a whole.

2. The dam service performance analysis method based on strong similarity sets according to claim 1 is characterized in that: In S2, the tracking and analysis method of the composition status of the strong similarity set is adopted for the change of the measured value sequence, including: S201, normalizing the measurement value sequence of each measurement point in the strong similarity set; S202: Determine the length of the sliding period according to the actual situation of the measured values ​​of each measuring point in the strong similarity set. ; S203, in Within the sliding period, the TWED distance sequence between each measuring point and the other measuring points in the strong similarity set is calculated in different time periods, and the comprehensive distance sequence of the strong similarity set of each measuring point is constructed respectively; S204. Count the maximum value of the comprehensive distance within a fixed length period , as the sample space for calculating the discrimination index of the belonging relationship of the strong similarity set of each measuring point; S205. Based on the typical small probability method, the strong similarity set measurement point belonging relationship discrimination index of each measurement point under a certain small probability is calculated respectively. ; S206: Determine the time The measured value of the i-th measurement point of the strong similarity set is used to calculate the comprehensive distance of the strong similarity set of the measurement points , and the strong similarity set measurement point belonging relationship discrimination index of the measurement point By comparison, we can get the preliminary judgment result of the belonging relationship of the measurement points in the strong similarity set: , then it is believed that At this moment, the measurement point still belongs to this strong similarity set; if , then it is believed that At this moment, the measurement point no longer belongs to this strong similarity set; S207, for the measurement points that are determined not to belong to the strong similarity set in the preliminary identification result of the belonging relationship, the measurement values ​​are processed for gross error identification, and the gross errors are eliminated and then re-analyzed to obtain the supplementary identification result of the belonging relationship; S208, the measurement points that have been judged to belong to the original strong similarity set are formed into a strong similarity set subset. For the measurement points that are judged not to belong to the original strong similarity set in the supplementary judgment result of the belonging relationship, they are judged one by one whether they now belong to this strong similarity set subset. If they are judged to belong to this strong similarity set subset, it is considered that At the moment, the measurement point still belongs to the original strong similarity set; if it is judged not to belong to this strong similarity subset, it is considered that At this moment, the measurement point no longer belongs to this strong similarity set; S209 , integrating the belonging relationship determination results of all the measurement points of the strong similarity set to obtain the composition status determination result of the strong similarity set at the current moment when the measurement sequence changes.

3. The dam service performance analysis method based on strong similarity sets according to claim 1 is characterized in that: The S3 includes: S301: Yes The monitoring effect value of each measuring point in the strong similarity set within the analysis period is normalized; S302: For the normalized multidimensional data, input the local linear embedding dimensionality reduction algorithm, adjust the number of nearest neighbors k, and obtain a preliminary dimensionality reduction result; S303: adjusting the value range according to the variation range of the average value sequence of the measured values ​​of each measuring point in the strong similarity set, and obtaining the comprehensive effect quantity sequence Y reflecting the variation of the service performance of the dam within the range of the strong similarity set in this period; S304: The comprehensive effect size sequence within the time period is input into the reverse cloud generator to obtain three digital features Ex, En and He. Ex, En and He are the expectation, entropy and super entropy of the comprehensive effect size sequence cloud model of the strong similarity set respectively. S305: inputting the digital features Ex, En and He into the forward cloud generator to calculate a cloud model reflecting the distribution of the comprehensive effect size; S306: Cloud-based "3En" guidelines, proposed in The upper limit of the property discrimination index corresponding to the moment and the lower limit ; S307: The comprehensive effect size of the strong similarity set at the moment , and the property discrimination index value as well as By comparison, we can obtain the preliminary analysis results of the dam service performance within the range of this strong similarity set at that moment; S308: Utilize formula The preliminary analysis results are quantified to obtain the service performance analysis results of the dam within the range of this strong similarity set at that moment.

4. The dam service performance analysis method based on strong similarity sets according to claim 3 is characterized in that: Calculate the discrete degree of the service performance analysis results of each element in the lower layer corresponding to the element in this layer ; Calculate the quantitative characteristics of the elements in this layer corresponding to the elements in the lower layer ; according to and Construct judgment matrix D; The following formula is used to calculate the influence weight of the service performance analysis results of the elements in this layer on the upper layer at time t: , ， Where: for Elements relative to other elements on the same layer The sum of the relative importance of is the sum of all items in the judgment matrix D, is the number of elements in this layer.

5. The dam service performance analysis method based on strong similarity sets according to claim 4 is characterized in that: The analysis results of each strong similarity set are fused upward layer by layer to obtain the service performance analysis results of each part in the part layer and the service performance analysis results of the dam as a whole, including: In the At this moment, the service performance analysis result set of each element in the next layer is obtained as , and fuse them to obtain this layer Analysis results of service properties of elements , further fused to get the previous layer Analysis results of service properties of elements , thus completing the process of handover and fusion layer by layer from bottom to top.

6. The dam service performance analysis method based on strong similarity sets according to claim 5 is characterized in that: At the tth moment, according to the elements in the previous layer The service performance analysis results of the corresponding elements in this layer and the influence weights of the elements in this layer on the previous layer are combined to obtain the elements in the previous layer. The service performance analysis results , and its calculation formula is: 。

Images