Distribution network equipment state real-time monitoring and early warning method

Abstract

The invention relates to a distribution network equipment state real-time monitoring and early warning method, which comprises the following steps of: 1, establishing an equipment model by adopting an equipment fault early warning method of a nonlinear state estimation technology, and predicting equipment by utilizing the model; 2, considering from three aspects of voltage quality, frequency quality and power supply reliability, constructing an electric energy evaluation index system, and evaluating the electric energy quality; 3, further constructing a power distribution area voltage quality index evaluation system, and forming a power distribution area power supply quality early warning parameter system model; 4, determining a single weight by adopting a combined weighting method based on an AHP-entropy weight; 5, solving an optimization weight based on a cooperative game; step 6, carrying out comprehensive evaluation on the electric energy quality by adopting an evaluation method based on improved TOPSIS; the system has the advantages of real-time variable data acquisition, automatic analysis of related indexes, all-weather monitoring and timely, effective and real-time early warning.

Description

Technical Field

[0001] The present invention belongs to the technical field of distribution network monitoring and early warning, and in particular relates to a method for real-time monitoring and early warning of distribution network equipment status. Background Art

[0002] At present, "low voltage", "three-phase imbalance" and other power "overvoltage", "heavy load area" and quality indicators are the main indicators of comprehensive evaluation. The quality of these indicators is directly related to the economic interests of the power supply company. Due to the particularity of electricity, the transformer is prone to accidents if it runs at low voltage, overload, heavy load or three-phase imbalance for a long time. At present, the "low voltage", "heavy load", "overload" and "phase imbalance" of the existing distribution area are mainly discovered through the power user electricity consumption information collection system. The system collects the 15-minute load data block of the area master meter yesterday through the wireless public network. Due to the intelligent meter in the system Due to the large number, large amount of data, system architecture design and other reasons, it is impossible to conduct real-time analysis and early warning, and only historical data can be analyzed to find problems. Therefore, there is a lag of 2 days. If the real-time data of the total meter of each combined area is collected manually, manual memory and judgment are required, the task volume is large, the efficiency is low, and it is impossible to achieve 24-hour uninterrupted memory and judgment, which does not meet the actual production needs of the power system; therefore, it is very necessary to provide a real-time monitoring and early warning method for the status of distribution network equipment that can collect variable data in real time, automatically analyze relevant indicators, monitor around the clock, and realize timely and effective real-time early warning of abnormal situations. Summary of the invention

[0003] The purpose of the present invention is to overcome the deficiencies of the prior art and to provide a distribution network equipment status real-time monitoring and early warning method that can collect variable data in real time, automatically analyze relevant indicators, monitor around the clock, and achieve timely and effective real-time early warning.

[0004] The object of the present invention is achieved by: a real-time monitoring and early warning method for distribution network equipment status, the method comprising the following steps:

[0005] Step 1: Use the equipment failure early warning method based on nonlinear state estimation technology to establish an equipment model and use the model to predict the equipment;

[0006] Step 2: Considering the three aspects of voltage quality, frequency quality and power supply reliability, a power evaluation index system is constructed to evaluate the power quality;

[0007] Step 3: Further construct the voltage quality index evaluation system of the distribution substation area and form a power supply quality early warning parameter system model based on the distribution substation area;

[0008] Step 4: Use the combined weighting method based on AHP-entropy weight to determine a single weight;

[0009] Step 5: Use cooperative game-based optimization to obtain the optimal weight;

[0010] Step 6: Use the improved TOPSIS-based evaluation method to comprehensively evaluate the power quality.

[0011] The equipment failure early warning method using nonlinear state estimation technology in step 1 establishes an equipment model and uses the model to predict the equipment, specifically including the following steps:

[0012] Step 1.1: Nonlinear state estimation modeling: Assume that a process or device has n interrelated variables. At a certain time i, the n observed variables are recorded as observation vectors, that is, X(i) = [x1 x2…x n ] T (1) The construction of the process memory matrix is ​​the first step in nonlinear state estimation modeling. During the normal operation period of the process or equipment, m historical observation vectors are collected under different operating conditions to form the process memory matrix: Each column of observation vectors in the process memory matrix represents a normal working state of the equipment. The subspace formed by the m historical observation vectors in the process memory matrix after reasonable selection can represent the entire dynamic process of normal operation of the process or equipment. The input of nonlinear state estimation is the observation vector X of the process or equipment at a certain moment. obs The output of the model is the prediction vector X for the input est , for any input observation vector X obs , the nonlinear state estimation generates an m-dimensional weight vector: W = [w1 w2…w m ] T (3), so that: X est =DW=w1·X(1)+w2·X(2)+...w m X(m)(4), that is, the predicted output of the nonlinear state estimation model is a linear combination of the m observation vectors in the process memory matrix. The weight vector W is determined by the following method: The residual of the input and output prediction vectors of the nonlinear state estimation model is constructed as: ε = X obs -X est (5), minimize the residual and obtain the weight vector W[11, 14] as: In the formula, is a nonlinear operator used to replace the multiplication operation in ordinary matrix operations; Substituting equation (6) into equation (4), the final result of the nonlinear estimation model for the process or equipment prediction is:

[0013] Step 1.2: Select modeling variables for equipment nonlinear estimation, determine the modeling variables in the observation vector, that is, the detection points related to each equipment;

[0014] Step 1.3: Form the similarity curve and similarity baseline of the equipment: The actual value can be used to generate the predicted value of the model through a nonlinear estimation algorithm. The predicted value of the model is compared with the actual value to generate the similarity value of the model. This similarity value is the distance between the predicted value of the model and the actual value. The similarity curve of the equipment is formed by continuously generating similarity values. If the residual error generated is ε within a certain period of time GT =X obs -X est (9), the nonlinear estimation model predicts the residual sequence as follows: GT =[ε1 ε2…ε N ](10), the similarity value is: By calculating the data state of the normal state of the equipment through the normal operation data model, the similarity value of the normal state can be obtained, and the minimum value of the similarity value of the normal state is taken to form the similarity baseline.

[0015] The combined weighting method based on AHP-entropy weight is adopted in step 4 to determine a single weight, which specifically includes the following steps:

[0016] Step 4.1: Determine the subjective weight based on the improved AHP method;

[0017] Step 4.2: Determine the objective weight using the entropy weight method: This is an indirect weighting method. According to the column vector in the sample, the information entropy of the indicator is calculated to indirectly determine the indicator weight. The larger the information entropy value, the higher the indicator weight, and vice versa.

[0018] Step 4.3: Determine the comprehensive weight based on the minimum information entropy increase: After obtaining the subjective and objective weights of the indicators respectively, the weight calculation model is established with the minimum sum of the comprehensive weight and the information entropy increase of the two weights, that is, the minimum information entropy increase: In the formula, u j 、v j 、w j are the subjective, objective and comprehensive weights of indicator j respectively; n is the total number of indicators, and the weight calculation formula is:

[0019]

[0020] Determining the subjective weight based on the improved AHP method in step 4.1 specifically includes the following steps:

[0021] Step 4.11: Establish a comparison matrix X: Each element of the comparison matrix can reflect the relative importance of the two indicators, with x ij For example, its value reflects the importance of indicator i relative to indicator j: It should be noted that each element satisfies That is, the importance of index i relative to index j is the reciprocal of the importance of index j relative to index i; in addition, the importance of any index relative to itself is 1, so the diagonal elements are 1; therefore, to construct an m-order square matrix, m(m-1) / 2 comparison operations are required;

[0022] Step 4.12: Consistency check: Due to cognitive differences among evaluation experts and calculation errors, the comparison matrix may violate the importance transitivity principle, such as x ij >1 and x jk >1, and x ijkj <1, so the consistency check is performed on the comparison matrix, and the consistency index C is defined respectively I and consistency factor C R for: In the formula, λ max is the maximum characteristic root of the comparison matrix; m is the matrix order; R I is the random consistency coefficient;

[0023] Step 4.13: Calculate the indicator weight: Use the eigenvector method to determine the indicator weight. For the weight ω of indicator i i , the weight based on the feature vector is:

[0024] The entropy weight method in step 4.2 specifically includes the following steps:

[0025] Step 4.21: Construct the initial comparison matrix Y: If the number of indicators and the number of evaluation objects are m and n respectively, the initial evaluation matrix is ​​an m×n matrix, and the element y ij is the jth index value of object i, and the decision matrix is:

[0026]

[0027] Step 4.22: Dimensionless processing: Considering that the units of each indicator are different, the entropy values ​​obtained are not comparable, so the initial comparison matrix is ​​dimensionless and normalized. ij Normalize and get the normalized comparison matrix P, then the element P in P ij The calculation formula is: Where Y j is the column vector consisting of the jth column of the matrix; max(Y j )、min(Y j ) are the maximum and minimum values ​​of the elements in the jth column respectively;

[0028] Step 4.23: Calculate the weight of each indicator: First calculate the entropy value of each indicator and determine the weight according to the entropy value: In the formula, E j is the information entropy of the jth column; wj is the weight of the indicator corresponding to the jth column.

[0029] The optimization weights obtained based on cooperative game in step 5 are specifically as follows: Assume that in the power quality evaluation system, the power levels of n monitoring points need to be evaluated and ranked, and there are m single evaluation methods that can be selected. The set of evaluation methods M = {1, 2, ..., m}, M is regarded as the set of objects in the game theory, and its subsets are 2 m And any subset constitutes one of the alliances in the evaluation system, and the alliance is evaluated as a whole; according to the objectivity of the data itself, the consistency correlation coefficient is used to optimize the combination of multiple evaluation methods, and the p weights W(1), W(2), ..., W(p) obtained by different calculation methods are used as input to form a whole, and the optimized weight W is obtained through processing, which is the output.

[0030] The step 5 of obtaining the optimization weight specifically includes the following steps:

[0031] Step 5.1: Calculate the correlation coefficient: Use W(i) to represent the weight obtained by the i-th weight calculation method, and W(p_i) to represent the combined weight of the other p-1 weights other than the i-th weight; then the consistency correlation coefficient between the two can be calculated using the following formula and represented by L(i): In the formula, q represents the number of indicators in the evaluation system; — represents the average value of the elements;

[0032] Step 5.2: Calculate the combined weight W′:

[0033] Step 5.3: Calculate the final comprehensive weight W: Normalize W′ to get the final comprehensive weight W. Recursively call the method. Each time the calculation is performed, the number of weight types is reduced by one until the number of types is 2. The calculation is terminated when there are only two weights. The calculation is as follows:

[0034]

[0035] The step 6 adopts the evaluation method based on improved TOPSIS to comprehensively evaluate the power quality, which specifically includes the following steps:

[0036] Step 6.1: Construct the original evaluation matrix: Assume that there are m monitoring points in the system, select n evaluation indicators, and use x ij represents the jth (j=1,2,...,n) index value of monitoring point i (i=1,2,...,m). All index values ​​of each monitoring point to be evaluated can be regarded as a sequence curve. By combining all the data sets of m monitoring points, the original evaluation matrix X=(xij ) m×n ;

[0037] Step 6.2: Construct a weighted matrix: For each element x of the initial matrix ij , use the following formula to calculate its weighted value e ij , thus obtaining the weighted matrix E,e ij =x ij ·w j , (i=1,2,...,n; j=1,2,...,m)(25);

[0038] Step 6.3: Standardization When conducting quantitative evaluation, the original evaluation matrix needs to be standardized. Different calculation methods are used according to the differences in indicator categories in the evaluation system: The calculation method for income-based indicators is: The calculation method of cost-based indicators is: Each element in the original evaluation matrix becomes the standard matrix R = (u ij ) m×n ,u ij ∈[0,1], the weighted standard matrix is ​​calculated based on the optimized weights: R * =(r ij ) m×n =(u ij ω j ) m×n (28), where ω j is the weight of the jth indicator;

[0039] Step 6.4: Obtain the extreme value vector from the weighted evaluation matrix E using the following formula: Among them, x b 、x w is the best and worst vector; best(E j ) is the optimal value of the jth column of matrix E; worst(E j ) is the worst value;

[0040] Step 6.5: Determine the optimal and worst-case scenarios: Use the formula Get the optimal and worst values ​​corresponding to each indicator in the weighted standard matrix respectively, and then get the maximum value sequence and the minimum sequence They are the virtual optimal solution and the worst solution respectively;

[0041] Step 6.6: Calculate the Euclidean distance and closeness between each object in the object set and the extreme value vector: in, are the Euclidean distances between object i and the extreme value vector, is the j-th dimension component of the extreme and extreme vector; define the closeness S to ii for:

[0042] Step 6.7: Calculate the degree of association: Use the following formula to calculate the Euclidean distance between each solution to be evaluated and the best and worst solutions, Calculate the correlation between each scheme to be evaluated and the virtual optimal and worst respectively. The relative difference generated on the jth indicator of the evaluation system is called the correlation coefficient. and The calculation formula is: In the formula, ρ represents the resolution coefficient; the correlation coefficient ε ij The larger the value of is, the higher the correlation between the two sequences on index j is; the correlation between the i-th scheme and the virtual optimal and the worst is respectively and The calculation formula is:

[0043]

[0044] Step 6.8: Criterion Fitting: When a solution to be evaluated and When the values ​​of are relatively large, it indicates that the scheme is relatively good; similarly, D i + and E i - When the value of is large, it indicates that the scheme is relatively poor; accordingly, the four variables can be fitted into two new feature quantities: In the formula, the coefficients α1 and α2 represent the evaluator's preference for the two factors of spatial position and graphic proximity, where α1+α2=1; The larger the value, the better the solution. On the contrary, The larger the value, the worse the solution.

[0045] Step 6.9: Determine the relative closeness: Using the new feature quantity obtained by fitting in the previous step, the relative closeness of each scheme to be evaluated can be calculated as: G i The larger it is, the better the power quality is.

[0046] A real-time monitoring and early warning system for the status of distribution network equipment, the system includes an electricity user electricity consumption information collection system and a short message gateway. The real-time monitoring and early warning system for the status of distribution network equipment is combined with the electricity user electricity consumption information collection system to form a real-time early warning system for power supply quality based on process automation technology. The power supply quality early warning parameter model based on the distribution station area established by the above-mentioned real-time monitoring and early warning method for the status of distribution network equipment is used to automate the collection, judgment, early warning, notification and other processes by utilizing the existing wireless public network collection channel, so as to realize "one-click" sending of power supply quality abnormalities in all areas of the county or power supply station to relevant responsible persons in real time. After receiving the abnormal situation notification, the relevant responsible persons take avoidance measures in time, and ultimately avoid more serious accidents.

[0047] The system utilizes WEB browser embedding technology, process automation technology, and wireless public network communication technology, and takes the power supply quality warning parameter system model based on the distribution station area established by the above-mentioned distribution network equipment status real-time monitoring and early warning method as the calculation basis, collects the power quality variable data of the distribution transformer area master meter in real time, and automatically analyzes the abnormal conditions of power quality related indicators; and sends the abnormal conditions to the relevant responsible persons in real time through the short message gateway, while realizing 24-hour all-weather monitoring, and finally achieving real-time and efficient power quality warning of the distribution station area.

[0048] The beneficial effects of the present invention are as follows: the present invention provides a real-time monitoring and early warning method for the status of distribution network equipment. During use, the present invention proposes an early warning mode that combines "management, control, and technology". It is based on the premise of timely obtaining abnormal data of power quality indicators in the substation area and ensuring the safe operation of power supply equipment. It is fundamentally based on improving the safety operation efficiency of the power grid, and aims to ensure that the power quality indicators of the combined area are lower than the assessment standards. It uses a nonlinear state estimation model, a cooperative game, and an improved TOPSIS model to make a systematic and comprehensive judgment on the power quality indicators of the substation area, so as to achieve timely and effective real-time early warning, and constructs a power supply quality early warning parameter system model based on the distribution station area, which can detect abnormal conditions of the power quality indicators of the distribution station area in real time. The present invention has the advantages of real-time acquisition of variable data, automatic analysis of relevant indicators, all-weather monitoring, and timely and effective real-time early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the electric energy evaluation index system of the present invention.

[0050] Figure 2 It is a schematic diagram of the voltage quality evaluation index system of the distribution station area of ​​the present invention.

[0051] Figure 3 It is a simplified flow chart of the present invention.

[0052] Figure 4This is a flow chart of the method for determining extreme value vectors based on K-Means clustering of the present invention.

[0053] Figure 5 The present invention is a flow chart of the method for evaluating voltage quality in a distribution station area based on TOPSIS. DETAILED DESCRIPTION

[0054] The present invention will be further described below in conjunction with the accompanying drawings.

[0055] Example 1

[0056] like Figure 1-5 As shown, a method for real-time monitoring and early warning of distribution network equipment status includes the following steps:

[0057] Step 1: Use the equipment failure early warning method based on nonlinear state estimation technology to establish an equipment model and use the model to predict the equipment;

[0058] Step 2: Considering the three aspects of voltage quality, frequency quality and power supply reliability, a power evaluation index system is constructed to evaluate the power quality;

[0059] Step 3: Further construct the voltage quality index evaluation system of the distribution substation area and form a power supply quality early warning parameter system model based on the distribution substation area;

[0060] Step 4: Use the combined weighting method based on AHP-entropy weight to determine a single weight;

[0061] Step 5: Use cooperative game-based optimization to obtain the optimal weight;

[0062] Step 6: Use the improved TOPSIS-based evaluation method to comprehensively evaluate the power quality.

[0063] The equipment failure early warning method using nonlinear state estimation technology in step 1 establishes an equipment model and uses the model to predict the equipment, specifically including the following steps:

[0064] Step 1.1: Nonlinear state estimation modeling: Assume that a process or device has n interrelated variables. At a certain time i, the n observed variables are recorded as observation vectors, that is, X(i) = [x1 x2…x n ] T (1) The construction of the process memory matrix is ​​the first step in nonlinear state estimation modeling. During the normal operation period of the process or equipment, m historical observation vectors are collected under different operating conditions (such as low load, high load, etc.), and the process memory matrix is: Each column of observation vectors in the process memory matrix represents a normal working state of the equipment. The subspace (represented by D) spanned by the m historical observation vectors in the process memory matrix after reasonable selection can represent the entire dynamic process of normal operation of the process or equipment. Therefore, the construction of the process memory matrix is ​​essentially the learning and memory process of the normal operation characteristics of the process or equipment. The input of the nonlinear state estimation is the observation vector X of the process or equipment at a certain moment. obs The output of the model is the prediction vector X for the input est , for any input observation vector X obs , the nonlinear state estimation generates an m-dimensional weight vector: W = [w1 w2…w m ] T (3), so that: X est =DW=w1·X(1)+w2·X(2)+...w m X(m)(4), that is, the predicted output of the nonlinear state estimation model is a linear combination of the m observation vectors in the process memory matrix. The weight vector W is determined by the following method: The residual of the input and output prediction vectors of the nonlinear state estimation model is constructed as: ε = X obs -X est (5), minimize the residual and obtain the weight vector W[11, 14] as: In the formula, is a nonlinear operator used to replace the multiplication operation in ordinary matrix operations; the nonlinear operator is selected as the Euclidean distance between two vectors, that is: The nonlinear operator has an intuitive physical meaning. When the two vectors are the same or similar, the distance is 0 or close to 0. The greater the difference between the two vectors, the greater the result of the nonlinear operation. The weight vector W reflects the similarity between the input observation vector of the nonlinear estimation model and the vectors in the process memory matrix. Substituting equation (6) into equation (4), the final result of the nonlinear estimation model for the process or equipment prediction is: When the process or equipment works normally, the new input observation vector of the nonlinear estimation is located in the normal working space represented by the process memory matrix, and is close to some historical observation vectors in the D matrix. The corresponding predicted value of the nonlinear estimation is X est It has high accuracy. When the working state of the process or equipment changes and potential faults occur, the input observation vector will deviate from the normal working space due to the change of dynamic characteristics. The corresponding accurate prediction value cannot be constructed by combining the historical observation vectors in the D matrix, resulting in a decrease in prediction accuracy and an increase in residual error.

[0065] Step 1.2: Select modeling variables for equipment nonlinear estimation, determine the modeling variables in the observation vector, that is, the detection points related to each equipment;

[0066] Step 1.3: Form the similarity curve and similarity baseline of the equipment: The actual value can be used to generate the predicted value of the model through a nonlinear estimation algorithm. The predicted value of the model is compared with the actual value to generate the similarity value of the model. This similarity value is the distance between the predicted value of the model and the actual value. The similarity curve of the equipment is formed by continuously generating similarity values. If the residual error generated is ε within a certain period of time GT =X obs -X est (9), the nonlinear estimation model predicts the residual sequence as follows: GT =[ε1 ε2…ε N ](10), the similarity value is: By calculating the data state of the normal state of the equipment through the normal operation data model, the similarity value of the normal state can be obtained, and the minimum value of the similarity value of the normal state is taken to form the similarity baseline.

[0067] The combined weighting method based on AHP-entropy weight is adopted in step 4 to determine a single weight, which specifically includes the following steps:

[0068] Step 4.1: Determine the subjective weight based on the improved AHP method;

[0069] Step 4.2: Determine the objective weight using the entropy weight method: This is an indirect weighting method. According to the column vector in the sample, the information entropy of the indicator is calculated to indirectly determine the indicator weight. The larger the information entropy value, the higher the indicator weight, and vice versa.

[0070] Step 4.3: Determine the comprehensive weight based on the minimum information entropy increase: After obtaining the subjective and objective weights of the indicators respectively, the weight calculation model is established with the minimum sum of the comprehensive weight and the information entropy increase of the two weights, that is, the minimum information entropy increase: In the formula, u j 、v j 、w j are the subjective, objective and comprehensive weights of indicator j respectively; n is the total number of indicators, and the weight calculation formula is:

[0071]

[0072] Determining the subjective weight based on the improved AHP method in step 4.1 specifically includes the following steps:

[0073] Step 4.11: Establish a comparison matrix X: Each element of the comparison matrix can reflect the relative importance of the two indicators, with x ij For example, its value reflects the importance of indicator i relative to indicator j: It should be noted that each element satisfies That is, the importance of index i relative to index j is the reciprocal of the importance of index j relative to index i; in addition, the importance of any index relative to itself is 1, so the diagonal elements are 1; therefore, to construct an m-order square matrix, m(m-1) / 2 comparison operations are required;

[0074] Step 4.12: Consistency check: Due to cognitive differences among evaluation experts and calculation errors, the comparison matrix may violate the importance transitivity principle, such as x ij >1 and x jk >1, and x ijkj <1, so the consistency check is performed on the comparison matrix, and the consistency index C is defined respectively I and consistency factor C R for: In the formula, λ max is the maximum characteristic root of the comparison matrix; m is the matrix order; R I is the random consistency coefficient; when the consistency factor is less than 0.1, the comparison matrix is ​​considered valid; in order to avoid repeated scoring due to failure to pass the consistency test, an improvement is made here. First, the indicators are sorted separately. Assuming that there are m indicators in total, the indicator with the highest importance has an importance score of m. In order of importance, the scores decrease in turn, and the indicator with the lowest importance has a score of 1; each scoring result has an equal weight, and the average score of each indicator is calculated to determine the order of indicator importance, thereby establishing an expert consensus mechanism; as long as the expert scoring does not violate the consensus on the indicator weight, the consistency requirement can be met;

[0075] Step 4.13: Calculate the indicator weight: Use the eigenvector method to determine the indicator weight. For the weight ω of indicator i i , the weight based on the feature vector is:

[0076] The entropy weight method in step 4.2 specifically includes the following steps:

[0077] Step 4.21: Construct the initial comparison matrix Y: If the number of indicators and the number of evaluation objects are m and n respectively, the initial evaluation matrix is ​​an m×n matrix, and the element y ij is the jth index value of object i, and the decision matrix is:

[0078]

[0079] Step 4.22: Dimensionless processing: Considering that the units of each indicator are different, the entropy values ​​obtained are not comparable, so the initial comparison matrix is ​​dimensionless and normalized. ij Normalize and get the normalized comparison matrix P, then the element P in P ij The calculation formula is: Where Yj is the column vector consisting of the jth column of the matrix; max(Y j )、min(Y j ) are the maximum and minimum values ​​of the elements in the jth column respectively;

[0080] Step 4.23: Calculate the weight of each indicator: First calculate the entropy value of each indicator and determine the weight according to the entropy value: In the formula, E j is the information entropy of the jth column; w j is the weight of the indicator corresponding to the jth column.

[0081] The optimization weights obtained based on cooperative game in step 5 are specifically as follows: Assume that in the power quality evaluation system, the power levels of n monitoring points need to be evaluated and ranked, and there are m single evaluation methods that can be selected. The set of evaluation methods M = {1, 2, ..., m}, M is regarded as the set of objects in the game theory, and its subsets are 2 m And any subset constitutes one of the alliances in the evaluation system. Comprehensively evaluating the alliance as a whole can reduce the error. According to the objectivity of the data itself, the consistency correlation coefficient is used to optimize the combination of multiple evaluation methods to make the final result more reasonable. The p weights W(1), W(2), ..., W(p) obtained by different calculation methods are used as input to form a whole. The optimized weight W is obtained through processing, which is the output.

[0082] The step 5 of obtaining the optimization weight specifically includes the following steps:

[0083] Step 5.1: Calculate the correlation coefficient: Use W(i) to represent the weight obtained by the i-th weight calculation method, and W(p_i) to represent the combined weight of the other p-1 weights other than the i-th weight; then the consistency correlation coefficient between the two can be calculated using the following formula and represented by L(i): In the formula, q represents the number of indicators in the evaluation system; — represents the average value of the elements;

[0084] Step 5.2: Calculate the combined weight W′:

[0085] Step 5.3: Calculate the final comprehensive weight W: After normalizing W', the final comprehensive weight W is obtained. It should be pointed out that according to the meaning of W(p_i), its calculation method is the same as that of W'; using a recursive call, the number of weight types is reduced by one each time the calculation is performed, until the number of types is 2, the calculation is terminated, for the calculation with only two weights, as shown in the following formula:

[0086] The evaluation method based on improved TOPSIS in step 6, that is, the TOPSIS ranking method based on K-Means clustering, is used to comprehensively evaluate the power quality, which specifically includes the following steps:

[0087] Step 6.1: Construct the original evaluation matrix: Assume that there are m monitoring points in the system, select n evaluation indicators, and use x ij represents the jth (j=1,2,...,n) index value of monitoring point i (i=1,2,...,m). All index values ​​of each monitoring point to be evaluated can be regarded as a sequence curve. By combining all the data sets of m monitoring points, the original evaluation matrix X=(x ij ) m×n ;

[0088] Step 6.2: Construct a weighted matrix: For each element x of the initial matrix ij , use the following formula to calculate its weighted value e ij , thus obtaining the weighted matrix E,e ij =x ij ·w j , (i=1,2,...,n; j=1,2,...,m)(25);

[0089] Step 6.3: Standardization When conducting quantitative evaluation, the original evaluation matrix needs to be standardized. Different calculation methods are used according to the differences in indicator categories in the evaluation system: The calculation method for income-based indicators is: The calculation method of cost-based indicators is: Each element in the original evaluation matrix becomes the standard matrix R = (u ij ) m×n ,u ij ∈[0,1], the weighted standard matrix is ​​calculated based on the optimized weights: R * =(r ij ) m×n =(u ij ω j ) m×n (28), where ω j is the weight of the jth indicator;

[0090] Step 6.4: Obtain the extreme value vector from the weighted evaluation matrix E using the following formula: Among them, x b 、x w is the best and worst vector; best(E j ) is the optimal value of the jth column of matrix E; worst(E j ) is the worst value;

[0091] Step 6.5: Determine the optimal and worst-case scenarios: Use the formula Get the optimal and worst values ​​corresponding to each indicator in the weighted standard matrix respectively, and then get the maximum value sequence and the minimum sequence They are the virtual optimal solution and the worst solution respectively;

[0092] Step 6.6: Calculate the Euclidean distance and closeness between each object in the object set and the extreme value vector: in, are the Euclidean distances between object i and the extreme value vector, is the j-th dimension component of the extreme and extreme vector; define the closeness S to i i for: Closeness reflects the closeness of the object to the optimal vector and the distance from the extreme phase. Closeness can be regarded as the evaluation score of the object. The closer the object is to the optimal phase and the farther it is from the extreme phase, the higher its score.

[0093] Step 6.7: Calculate the degree of association: Use the following formula to calculate the Euclidean distance between each solution to be evaluated and the best and worst solutions, Calculate the correlation between each scheme to be evaluated and the virtual optimal and worst respectively. The relative difference generated on the jth indicator of the evaluation system is called the correlation coefficient. and The calculation formula is: In the formula, ρ represents the resolution coefficient, which ranges from 0 to 1 and is usually taken as 0.5; the correlation coefficient ε ij The larger the value of is, the higher the correlation between the two sequences on index j is; the correlation between the i-th scheme and the virtual optimal and the worst is respectively and The calculation formula is:

[0094] Step 6.8: Criterion Fitting: When a solution to be evaluated and When the values ​​of are relatively large, it indicates that the scheme is relatively good; similarly, and When the value of is large, it indicates that the scheme is relatively poor; accordingly, the four variables can be fitted into two new feature quantities: In the formula, the coefficients α1 and α2 represent the evaluator's preference for the two factors of spatial position and graphic closeness, where α1+α2=1, and usually both are taken as 0.5; The larger the value, the better the solution. On the contrary, S i - The larger the value, the worse the solution.

[0095] Step 6.9: Determine the relative closeness: Using the new feature quantity obtained by fitting in the previous step, the relative closeness of each scheme to be evaluated can be calculated as: G i The larger it is, the better the power quality is.

[0096] In this embodiment, the present invention realizes real-time detection of abnormal conditions of power quality indicators in distribution stations, including low voltage, overload, phase loss, overvoltage, overtemperature, power theft, clock deviation, etc., based on the power supply quality warning parameter system model of distribution stations; the present invention adopts cooperative game theory to combine various types of weights to obtain more reasonable optimization weights, avoiding the shortcomings of a single weight determination method; the improved TOPSIS method is used to conduct a comprehensive quantitative evaluation of power quality, and the result is more accurate. According to the quantitative evaluation results, the power level can be simply and clearly understood.

[0097] The present invention is a real-time monitoring and early warning method for the status of distribution network equipment. During use, the present invention proposes an early warning mode combining "management, control and technology". It is based on the premise of timely obtaining abnormal data of power quality indicators in the substation area and ensuring the safe operation of power supply equipment, and is fundamentally based on improving the safety operation efficiency of the power grid. The goal is that the power quality indicators of the combined area are lower than the assessment standard. It uses a nonlinear state estimation model, a cooperative game and an improved TOPSIS model to make a systematic and comprehensive judgment on the power quality indicators of the substation area, so as to achieve timely and effective real-time early warning, and constructs a power supply quality early warning parameter system model based on the distribution station area, which can detect the abnormal situation of the power quality indicators of the distribution station area in real time. The present invention has the advantages of real-time acquisition of variable data, automatic analysis of relevant indicators, all-weather monitoring, and timely and effective real-time early warning.

[0098] Example 2

[0099] like Figure 1-5 As shown, a distribution network equipment status real-time monitoring and early warning system, the system includes a power user electricity consumption information collection system and a short message gateway. The distribution network equipment status real-time monitoring and early warning system is combined with the power user electricity consumption information collection system to form a power supply quality real-time early warning system based on process automation technology. Through the power supply quality early warning parameter model based on the distribution station area established by the above-mentioned distribution network equipment status real-time monitoring and early warning method, the existing wireless public network collection channel is used to automate the collection, judgment, early warning, notification and other processes, so that the power supply quality abnormalities of all substations in the county or power supply station can be sent to the relevant responsible persons in real time with one click. After receiving the abnormal situation notification, the relevant responsible persons can take avoidance measures in time, and finally avoid more serious accidents.

[0100] The system utilizes WEB browser embedding technology, process automation technology, and wireless public network communication technology, and takes the power supply quality warning parameter system model based on the distribution station area established by the above-mentioned distribution network equipment status real-time monitoring and early warning method as the calculation basis, collects the power quality variable data of the distribution transformer area master meter in real time, and automatically analyzes the abnormal conditions of power quality related indicators; and sends the abnormal conditions to the relevant responsible persons in real time through the short message gateway, while realizing 24-hour all-weather monitoring, and finally achieving real-time and efficient power quality warning of the distribution station area.

[0101] In this embodiment, the system of the present invention adopts WEB browser embedding technology, process automation technology, and wireless public network communication technology. Since the system of the present invention is based on the power acquisition system and is realized through the existing established wireless public network communication, it is necessary to integrate the system as a WEB application into the application program, control the power acquisition system through javascript scripts, and use the direct copy communication of the power acquisition system to collect the relevant data of the substation master table in real time, analyze whether the returned data is abnormal according to the distribution area warning parameter model, record and warn when abnormal, and notify the relevant responsible person of the abnormal information through the short message gateway device, so as to achieve the purpose of real-time warning; the whole process involves file establishment, data preparation, collection, judgment, warning, notification, sub-delivery and other processes. In order to avoid process errors or errors, these processes need to be integrated and automated to achieve "one-key" completion and 24-hour uninterrupted monitoring and warning;

[0102] The system of the present invention has good economic and social benefits: ① Economic benefits: The power supply quality real-time early warning system based on process automation technology proposed by the present invention makes full use of existing new technologies and power grid infrastructure, is low-cost, and has strong operability. It conforms to the actual situation of existing production technology of the power system and can effectively solve the problem of difficult monitoring of power supply quality indicators in the substation area; The real-time monitoring and early warning tool for the status of the distribution network equipment can realize 7*24 hours of uninterrupted full-automatic monitoring of the operating status of the distribution network equipment. No manual operation is required during the operation of the system, which greatly reduces the labor intensity, and adopts a variety of data algorithm models to timely discover abnormal equipment and issue early warnings in time, so as to reserve sufficient time for troubleshooting equipment failures and restoring the normal operating status of the equipment, further improve the power supply stability and user satisfaction with electricity consumption, and it is estimated that the cost can be saved by 85,000 yuan per year; ② Social benefits: At present, the company lacks the monitoring means for heavy overload, three imbalances and low voltage in the substation area, resulting in the above indicators being assessed, and the abnormal substation area cannot be discovered in time, affecting the normal electricity consumption of users; The system of the present invention can discover and push the abnormal equipment information to the relevant person in charge at the first time, and reserve sufficient time to solve the abnormal distribution network equipment, thereby improving the equipment operation stability and user satisfaction.

[0103] In this embodiment, the power supply quality warning parameter system model based on the distribution station area is slightly modified or expanded and combined, and can also be used to alarm abnormal conditions in substations, such as overload, phase loss, low voltage, overvoltage, overtemperature, power theft, clock deviation, etc.

[0104] The present invention is a real-time monitoring and early warning method for the status of distribution network equipment. During use, the present invention proposes an early warning mode combining "management, control and technology". It is based on the premise of timely obtaining abnormal data of power quality indicators in the substation area and ensuring the safe operation of power supply equipment, and is fundamentally based on improving the safety operation efficiency of the power grid. The goal is that the power quality indicators of the combined area are lower than the assessment standard. It uses a nonlinear state estimation model, a cooperative game and an improved TOPSIS model to make a systematic and comprehensive judgment on the power quality indicators of the substation area, so as to achieve timely and effective real-time early warning, and constructs a power supply quality early warning parameter system model based on the distribution station area, which can detect the abnormal situation of the power quality indicators of the distribution station area in real time. The present invention has the advantages of real-time acquisition of variable data, automatic analysis of relevant indicators, all-weather monitoring, and timely and effective real-time early warning.

Claims

1. A method for real-time monitoring and early warning of distribution network equipment status, characterized in that: The method comprises the following steps: Step 1: Use the equipment failure early warning method based on nonlinear state estimation technology to establish an equipment model and use the model to predict the equipment; Step 2: Considering the three aspects of voltage quality, frequency quality and power supply reliability, a power evaluation index system is constructed to evaluate the power quality; Step 3: Further construct the voltage quality index evaluation system of the distribution substation area and form a power supply quality early warning parameter system model based on the distribution substation area; Step 4: Use the combined weighting method based on AHP-entropy weight to determine a single weight; Step 5: Use cooperative game-based optimization to obtain the optimal weight; Step 6: Use the improved TOPSIS-based evaluation method to comprehensively evaluate the power quality.

2. The method for real-time monitoring and early warning of distribution network equipment status according to claim 1, characterized in that: The equipment failure early warning method using nonlinear state estimation technology in step 1 establishes an equipment model and uses the model to predict the equipment, specifically including the following steps: Step 1.1: Nonlinear state estimation modeling: Assume that a process or device has n interrelated variables. At a certain time i, the n observed variables are recorded as the observation vector, that is, X(i) = [x1 x2 … x n ] T (1) The construction of the process memory matrix is ​​the first step in nonlinear state estimation modeling. During the normal operation period of the process or equipment, m historical observation vectors are collected under different operating conditions to form the process memory matrix: Each column of observation vectors in the process memory matrix represents a normal working state of the equipment. The subspace formed by the m historical observation vectors in the process memory matrix after reasonable selection can represent the entire dynamic process of normal operation of the process or equipment. The input of nonlinear state estimation is the observation vector X of the process or equipment at a certain moment. obs The output of the model is the prediction vector X for the input est , for any input observation vector X obs , the nonlinear state estimation generates an m-dimensional weight vector: W = [w1 w2 … w m ] T (3), so that: X est =DW=w1·X(1)+w2·X(2)+...w m ·X(m) (4), that is, the predicted output of the nonlinear state estimation model is a linear combination of the m observation vectors in the process memory matrix. The weight vector W is determined by the following method: The residual of the input and output prediction vectors of the nonlinear state estimation model is constructed as: ε = X obs -X est (5), minimize the residual and obtain the weight vector W[11, 14] as: In the formula, is a nonlinear operator used to replace the multiplication operation in ordinary matrix operations; Substituting equation (6) into equation (4), the final result of the nonlinear estimation model for the process or equipment prediction is: Step 1.2: Select modeling variables for equipment nonlinear estimation, determine the modeling variables in the observation vector, that is, the detection points related to each equipment; Step 1.3: Form the similarity curve and similarity baseline of the equipment: The actual value can be used to generate the predicted value of the model through a nonlinear estimation algorithm. The predicted value of the model is compared with the actual value to generate the similarity value of the model. This similarity value is the distance between the predicted value of the model and the actual value. The similarity curve of the equipment is formed by continuously generating similarity values. If the residual error generated is ε within a certain period of time GT =X obs -X est (9), the nonlinear estimation model predicts the residual sequence as follows: GT =[ε1 ε2 … ε N ] (10), the similarity value is: By calculating the data state of the normal state of the equipment through the normal operation data model, the similarity value of the normal state can be obtained, and the minimum value of the similarity value of the normal state is taken to form the similarity baseline.

3. The method for real-time monitoring and early warning of distribution network equipment status according to claim 1, characterized in that: The combined weighting method based on AHP-entropy weight is adopted in step 4 to determine a single weight, which specifically includes the following steps: Step 4.1: Determine the subjective weight based on the improved AHP method; Step 4.2: Determine the objective weight using the entropy weight method: This is an indirect weighting method. According to the column vector in the sample, the information entropy of the indicator is calculated to indirectly determine the indicator weight. The larger the information entropy value, the higher the indicator weight, and vice versa. Step 4.3: Determine the comprehensive weight based on the minimum information entropy increase: After obtaining the subjective and objective weights of the indicators respectively, the weight calculation model is established with the minimum sum of the comprehensive weight and the information entropy increase of the two weights, that is, the minimum information entropy increase: In the formula, u j 、v j 、w j are the subjective, objective and comprehensive weights of indicator j respectively; n is the total number of indicators, and the weight calculation formula is:

4. The method for real-time monitoring and early warning of distribution network equipment status according to claim 3, characterized in that: Determining the subjective weight based on the improved AHP method in step 4.1 specifically includes the following steps: Step 4.11: Establish a comparison matrix X: Each element of the comparison matrix can reflect the relative importance of the two indicators, with x ij For example, its value reflects the importance of indicator i relative to indicator j: It should be noted that each element satisfies That is, the importance of index i relative to index j is the reciprocal of the importance of index j relative to index i; in addition, the importance of any index relative to itself is 1, so the diagonal elements are 1; therefore, to construct an m-order square matrix, m(m-1) / 2 comparison operations are required; Step 4.12: Consistency check: Due to cognitive differences among evaluation experts and calculation errors, the comparison matrix may violate the importance transitivity principle, such as x ij >1 and x jk >1, and x ijkj <1, so the consistency check is performed on the comparison matrix, and the consistency index C is defined respectively I and consistency factor C R for: In the formula, λ max is the maximum characteristic root of the comparison matrix; m is the matrix order; R I is the random consistency coefficient; Step 4.13: Calculate the indicator weight: Use the eigenvector method to determine the indicator weight. For the weight ω of indicator i i , the weight based on the feature vector is:

5. The method for real-time monitoring and early warning of distribution network equipment status according to claim 3, characterized in that: The entropy weight method in step 4.2 specifically includes the following steps: Step 4.21: Construct the initial comparison matrix Y: If the number of indicators and the number of evaluation objects are m and n respectively, the initial evaluation matrix is ​​an m×n matrix, and the element y ij is the jth index value of object i, and the decision matrix is: Step 4.22: Dimensionless processing: Considering that the units of each indicator are different, the entropy values ​​obtained are not comparable, so the initial comparison matrix is ​​dimensionless and normalized. ij Normalize and get the normalized comparison matrix P, then the element P in P ij The calculation formula is: Where Y j is the column vector consisting of the jth column of the matrix; max(Y j )、min(Y j ) are the maximum and minimum values ​​of the elements in the jth column respectively; Step 4.23: Calculate the weight of each indicator: First calculate the entropy value of each indicator and determine the weight according to the entropy value: In the formula, E j is the information entropy of the jth column; w j is the weight of the indicator corresponding to the jth column.

6. The method for real-time monitoring and early warning of distribution network equipment status according to claim 1, characterized in that: The optimization weights obtained based on cooperative game in step 5 are specifically as follows: Assume that in the power quality evaluation system, the power levels of n monitoring points need to be evaluated and ranked, and there are m single evaluation methods that can be selected. The set of evaluation methods M = {1, 2, ..., m}, M is regarded as the set of objects in the game theory, and its subsets are 2 m And any subset constitutes one of the alliances in the evaluation system, and the alliance is evaluated as a whole; according to the objectivity of the data itself, the consistency correlation coefficient is used to optimize the combination of multiple evaluation methods, and the p weights W(1), W(2), ..., W(p) obtained by different calculation methods are used as input to form a whole, and the optimized weight W is obtained through processing, which is the output.

7. The method for real-time monitoring and early warning of distribution network equipment status according to claim 6, characterized in that: The step 5 of obtaining the optimization weight specifically includes the following steps: Step 5.1: Calculate the correlation coefficient: Use W(i) to represent the weight obtained by the i-th weight calculation method, and W(p_i) to represent the combined weight of the other p-1 weights other than the i-th weight; then the consistency correlation coefficient between the two can be calculated using the following formula and represented by L(i): In the formula, q represents the number of indicators in the evaluation system; — represents the average value of the elements; Step 5.2: Calculate the combined weight W′: Step 5.3: Calculate the final comprehensive weight W: Normalize W′ to get the final comprehensive weight W. Recursively call the method. Each time the calculation is performed, the number of weight types is reduced by one until the number of types is 2. The calculation is terminated when there are only two weights. The calculation is as follows:

8. The method for real-time monitoring and early warning of distribution network equipment status according to claim 1, characterized in that: The step 6 adopts the evaluation method based on improved TOPSIS to comprehensively evaluate the power quality, which specifically includes the following steps: Step 6.1: Construct the original evaluation matrix: Assume that there are m monitoring points in the system, select n evaluation indicators, and use x ij represents the jth (j=1,2,...,n) index value of monitoring point i (i=1,2,...,m). All index values ​​of each monitoring point to be evaluated can be regarded as a sequence curve. By combining all the data sets of m monitoring points, the original evaluation matrix X=(x ij ) m×n ; Step 6.2: Construct a weighted matrix: For each element x of the initial matrix ij , use the following formula to calculate its weighted value e ij , thus obtaining the weighted matrix E,e ij =x ij ·w j , (i=1,2,...,n; j=1,2,...,m)(25); Step 6.3: Standardization When conducting quantitative evaluation, the original evaluation matrix needs to be standardized. Different calculation methods are used according to the differences in indicator categories in the evaluation system: The calculation method for income-based indicators is: The calculation method of cost-based indicators is: Each element in the original evaluation matrix becomes the standard matrix R = (u ij ) m×n ,u ij ∈[0,1], the weighted standard matrix is ​​calculated based on the optimized weights: R * =(r ij ) m×n =(u ij ω j ) m×n (28), where ω j is the weight of the jth indicator; Step 6.4: Obtain the extreme value vector from the weighted evaluation matrix E using the following formula: Among them, x b 、x w is the best and worst vector; best(E j ) is the optimal value of the jth column of matrix E; worst(E j ) is the worst value; Step 6.5: Determine the optimal and worst-case scenarios: Use the formula Get the optimal and worst values ​​corresponding to each indicator in the weighted standard matrix respectively, and then get the maximum value sequence and the minimum sequence They are the virtual optimal solution and the worst solution respectively; Step 6.6: Calculate the Euclidean distance and closeness between each object in the object set and the extreme value vector: in, are the Euclidean distances between object i and the extreme value vector, is the j-th dimension component of the extreme and extreme vector; define the closeness S to i i for: Step 6.7: Calculate the degree of association: Use the following formula to calculate the Euclidean distance between each solution to be evaluated and the best and worst solutions, Calculate the correlation between each scheme to be evaluated and the virtual optimal and worst respectively. The relative difference generated on the jth indicator of the evaluation system is called the correlation coefficient. and The calculation formula is: In the formula, ρ represents the resolution coefficient; the correlation coefficient ε ij The larger the value of is, the higher the correlation between the two sequences on index j is; the correlation between the i-th scheme and the virtual optimal and the worst is respectively and The calculation formula is: Step 6.8: Criterion Fitting: When a solution to be evaluated and When the values ​​of are relatively large, it indicates that the scheme is relatively good; similarly, and When the value of is large, it indicates that the scheme is relatively poor; accordingly, the four variables can be fitted into two new feature quantities: In the formula, the coefficients α1 and α2 represent the evaluator's preference for the two factors of spatial position and graphic proximity, where α1+α2=1; The larger the value, the better the solution. On the contrary, The larger the value, the worse the solution. Step 6.9: Determine the relative closeness: Using the new feature quantity obtained by fitting in the previous step, the relative closeness of each scheme to be evaluated can be calculated as: G i The larger it is, the better the power quality is.

9. The real-time monitoring and early warning system for distribution network equipment status is characterized by: The system includes an electricity user electricity consumption information collection system and a short message gateway. The distribution network equipment status real-time monitoring and early warning system is combined with the power user electricity consumption information collection system to form a power supply quality real-time early warning system based on process automation technology. The power supply quality early warning parameter model based on the distribution station area is established by the distribution network equipment status real-time monitoring and early warning method described in any one of claims 1 to 8, and the existing wireless public network collection channel is used to automate the collection, judgment, early warning, notification and other processes, so that the power supply quality abnormalities of all substations in the county or power supply station can be sent to the relevant responsible persons in real time with one click. After receiving the abnormal situation notification, the relevant responsible persons can take avoidance measures in time, and finally avoid more serious accidents.

10. The real-time monitoring and early warning system for distribution network equipment status according to claim 9, characterized in that: The system utilizes WEB browser embedding technology, process automation technology, and wireless public network communication technology, and takes the distribution substation area warning parameter model established by the real-time monitoring and early warning method for distribution network equipment status described in any one of claims 1 to 8 as the calculation basis, collects the power quality variable data of the distribution substation area master meter in real time, and automatically analyzes the abnormal conditions of power quality related indicators; and sends the abnormal conditions to the relevant responsible persons in real time through the short message gateway, while realizing 24-hour all-weather monitoring, and finally achieving real-time and efficient power quality early warning of the distribution substation area.

Images