Method and system for determining equipment transformation priority based on improved Schwang optimization algorithm

Abstract

The invention relates to an equipment transformation priority determination method and system based on an improved Sougee optimization algorithm, and the method comprises the steps: firstly constructing an evaluation index system used for reflecting the equipment transformation priority, then calculating a comprehensive weight, and carrying out the optimization based on a multi-objective optimization improved Sougee optimization algorithm to obtain an optimal comprehensive weight; and finally, a TOPSIS method based on DBSCAN-AGNES double clustering redefinition is adopted to calculate and obtain the closeness degree and sort the closeness degree so as to determine the transformation priority of the old production equipment of the power grid. According to the method, subjective and objective weights are calculated in combination with the AHP and the improved CRITIC method, the dynamic crowding speed and the global-local search mechanism are introduced to improve the South goose optimization algorithm, intelligent optimization of the weights is achieved, weight distribution is more accurate and reasonable, meanwhile, based on DBSCAN-AGNES dual clustering, the problem that a traditional TOPSIS method is interfered by abnormal values and noise is avoided, and the method is suitable for large-scale popularization and application. And the robustness and reliability of the evaluation result are improved, so that the equipment transformation priority can be determined more scientifically and accurately.

Description

TECHNICAL FIELD

[0001] The present application relates to the priority determination means of the old equipment reconstruction of power grid, and especially relates to a device reconstruction priority determination method and system based on an improved snow goose optimization algorithm. BACKGROUND

[0002] With the expansion of the scale of the power grid and the running time of the equipment, a large number of power grid equipment gradually ages, and its operation reliability, safety and economy are significantly reduced. Old equipment often becomes a weak link in the operation of the power grid, not only increases the risk of accidents, but also threatens the stability and safety of the power grid. In order to ensure the safe, reliable and efficient operation of the power grid, the technical reconstruction of the old equipment has become an important task of the power enterprise. However, the existing technology has the problems of strong subjectivity and difficulty in comprehensively considering multiple factors when evaluating the priority. Therefore, scientifically and reasonably evaluating the priority of equipment reconstruction has become an important problem to be solved.

[0003] In the prior art, there are documents that propose an evaluation method based on entropy weight TOPSIS, which reduces the subjectivity of the TOPSIS method by adding entropy weight, but the entropy weight method needs to distribute the weight according to the standard deviation, information entropy and other distribution characteristics of the data, which may underestimate or overestimate the importance of key indicators. There are also documents that establish an evaluation index system of regional innovation ecological system based on AHP-CRITIC-TOPSIS method, but when the AHP method and CRITIC method are used to distribute subjective and objective weights, there is a problem that the method depends too much on the experience of experts. In addition, the traditional TOPSIS method uses fixed ideal solution and negative ideal solution, which is easy to be disturbed by data outliers and noise, and the robustness of the evaluation result is insufficient, thereby affecting the accuracy of the result. SUMMARY

[0004] The purpose of the present application is to overcome the above-mentioned defects and problems in the prior art, and to provide a more accurate device reconstruction priority determination method and system based on an improved snow goose optimization algorithm.

[0005] To achieve the above purpose, the technical solution of the present application is: a device reconstruction priority determination method based on an improved snow goose optimization algorithm, comprising:

[0006] Based on the current situation of the reconstruction investment of the old production equipment of the power grid, an evaluation index system reflecting the priority of equipment reconstruction is constructed;

[0007] Based on the evaluation index system, the subjective weight is calculated, and the CRITIC method combining adaptive adjustment K-NN and mutual information is used to calculate the objective weight, and the comprehensive weight is calculated by combining the subjective weight and the objective weight;

[0008] The improved snow goose optimization algorithm based on multi-objective optimization is used to optimize the comprehensive weight, and the optimal comprehensive weight is obtained;

[0009] The improvement of the snow goose optimization algorithm includes updating the snow goose speed in the exploration stage of the snow goose algorithm based on a dynamic congestion speed updating mechanism; and updating the snow goose position in the development stage of the snow goose algorithm based on a global-local search mechanism according to fitness changes;

[0010] Based on the optimal comprehensive weight, the distance between each device and the positive and negative ideal solutions is calculated by using the TOPSIS method redefined based on the DBSCAN-AGNES double clustering, the closeness is obtained and sorted to determine the transformation priority of the old production equipment of the power grid;

[0011] The DBSCAN-AGNES double clustering includes global density clustering based on the DBSCAN clustering algorithm and local subdivision clustering based on the AGNES clustering algorithm.

[0012] The CRITIC method combined with adaptive adjustment of K-NN and mutual information is used to calculate the objective weight, specifically including:

[0013] The different index data samples in the evaluation index system are standardized, and the standard deviation of the index is calculated, and the expression is as follows:

[0014] ;

[0015] Wherein: is the standard deviation of the first index, is the standardized index value of the first sample on the first index, is the mean value of the first index, is the sample number;

[0016] The local density of each sample point is calculated, and the number of neighbors of each sample is adaptively adjusted according to the local density;

[0017] The expression of the local density is as follows:

[0018] ;

[0019] ;

[0020] Wherein: is the local density of the first sample, is the number of neighbors of the current sample, is the distance between the first sample and the first neighbor, , respectively, the first sample and the sample on the th index, is the number of indexes;

[0021] The expression of self-adaptive adjustment of the number of neighbors of each sample is as follows:

[0022] ;

[0023] Wherein: is the number of neighbors of each sample, is the maximum value of the number of neighbors, is the floor function;

[0024] Based on the number of neighbors of each sample, the marginal probability of the th index and the th index is calculated, and the expression is as follows:

[0025] ;

[0026] ;

[0027] Wherein: , is the marginal probability of the th index and the th index, is a Gaussian kernel function, , are all bandwidths; is the value of the th index, is the normalized value of the th sample on the th index;

[0028] The joint probability of the th index and the th index is calculated, and the expression is as follows:

[0029] ;

[0030] Wherein: is the joint probability of the th index and the th index taking a certain value at the same time;

[0031] Based on the marginal probability and the joint probability, the mutual information of the th index and the th index is calculated, and the expression is as follows:

[0032] ;

[0033] wherein: is the mutual information of the first index and the second index, is the logarithm reflecting the difference between the marginal probability and the joint probability;

[0034] Based on the mutual information and the standard deviation, the information of each index is calculated, and the objective weight is calculated;

[0035] The expression of the information is as follows:

[0036] ;

[0037] wherein: is the information of the first index, is the standard deviation of the first index;

[0038] The expression of the objective weight is as follows:

[0039] ;

[0040] wherein: is the objective weight of the first index.

[0041] The improved snow goose optimization algorithm based on multi-objective optimization is used to optimize the comprehensive weight, and the optimal comprehensive weight is obtained, specifically including:

[0042] The position, speed, and global and individual best positions of the snow goose are initialized;

[0043] wherein: the position of the initial snow goose is generated, and the position of each snow goose represents a possible weight coefficient; the speed of each snow goose is initialized to zero or a small random value; the position of each snow goose is set as the individual best position, and the global best position is updated when the comprehensive fitness of the current snow goose is better than the global best comprehensive fitness;

[0044] The comprehensive fitness value of the current position of each snow goose is calculated based on the fitness function, and the expression is as follows:

[0045] ;

[0046] ;

[0047] ;

[0048] ;

[0049] wherein: For the fitness function, , , All are weighting coefficients. , , Both are objective functions. To adjust the penalty strength of the gradient penalty term, This is a gradient penalty term;

[0050] The snow goose speed in the exploration phase of the snow goose algorithm is updated based on the dynamic congestion speed update mechanism, and the snow goose position in the development phase of the snow goose algorithm is updated based on the global-local search mechanism of fitness change, until the number of iterations or the improvement of the comprehensive fitness value is less than the set threshold, then the update stops and the weight coefficient corresponding to the current global best position is returned.

[0051] The expression for updating the snow goose velocity during the exploration phase of the snow goose algorithm based on the dynamic congestion velocity update mechanism is as follows:

[0052] ; ;

[0053] in: For the speed of the snow geese, As a weighting factor, As a control factor, The number of objectives in the fitness function. , For the first In the target dimension, the first The fitness values ​​of each individual's neighbors. , The first The maximum and minimum fitness values ​​in each objective dimension. As the initial control factor, This represents the maximum number of iterations.

[0054] The global-local search mechanism based on fitness changes updates the snow goose position during the snow goose algorithm development phase, and its expression is as follows:

[0055] ;

[0056] in: For the first Individuals Location at any given moment As a weighting factor, For the first Individuals Overall adaptability at any time The average overall fitness of the population. This is the step size control factor. the fitness difference of the individual in the jth objective dimension and the fitness difference of the global optimal solution in the jth objective dimension; the comprehensive of the fitness difference of the individual in the jth objective dimension and the fitness difference of the global optimal solution in the jth objective dimension; the fitness value of the jth objective dimension, the optimal solution of the region where the ith individual is located, the fitness value of the jth objective dimension, the optimal solution of the region where the ith individual is located, the fitness value of the jth objective dimension,

[0057] The weight coefficient corresponding to the current global best position is returned, and the subjective weight and the objective weight are optimized respectively to obtain the optimal comprehensive weight, and the expression is as follows:

[0058] ;

[0059] Wherein: the optimal comprehensive weight, the weight coefficient, the subjective weight, the objective weight.

[0060] Based on the optimal comprehensive weight, the distance between each device and the positive and negative ideal solutions is calculated by using the TOPSIS method based on the double clustering redefinition of DBSCAN-AGNES, the closeness is obtained and sorted to determine the priority of the transformation, and the method comprises the following steps:

[0061] Based on the optimal comprehensive weight, the standardized index value is weighted to construct a weighted normalized matrix, and the expression is as follows:

[0062] ; ;

[0062] ; ;

[0063] Wherein: the weighted normalized matrix, the value of the jth device in the jth index in the weighted normalized matrix, the optimal comprehensive weight of the jth index; Based on the DBSCAN clustering algorithm, the global density clustering is carried out, and the effective cluster is extracted, and the expression is as follows:

[0064]

[0065] ;

[0066] Wherein: the effective cluster, the jth set; The data points in the effective cluster

[0067] and and The cosine similarity is converted to distance, and its expression is as follows:

[0068] ;

[0069] in: For data points and Distance based on cosine similarity, , Data points and In the Components in each dimension The total dimension of the data points;

[0070] Based on the cosine distance between data points, the distance between subclusters is calculated using the average distance method, as shown in the following expression:

[0071] ;

[0072] in: The average cosine distance between all pairs of points in the two subclusters. , For the two subclusters to be merged, , Both represent the number of sample points in the cluster;

[0073] Repeat the above steps to gradually merge similar clusters until the minimum distance between clusters exceeds a threshold, ultimately resulting in several sub-clusters, as expressed below:

[0074] ;

[0075] in: For the threshold, For adjustment coefficients, This is the median of the cosine distances between all points within the current global cluster;

[0076] After the merge is complete, each sub-cluster is obtained. The silhouette coefficient method was used to verify and correct the subclusters in AGNES clustering. The quality of the output is used to determine the final partitioning of each sub-cluster. ;

[0077] For each sub-cluster obtained from the final partition The centroid is calculated as the center point, and its expression is as follows:

[0078] ;

[0079] in: For the division of the first Subclusters, The center point of the sub-cluster;

[0080] Based on sub-cluster center point Construct candidate point set Its expression is as follows:

[0081] ;

[0082] Where: each sub-cluster center point corresponds to a multi-dimensional vector. ;

[0083] Determine the positive or negative ideal solution for each index in the candidate point set:

[0084] For a positive ideal solution, the optimal value is selected for each index in the candidate point set, and its expression is as follows:

[0085] ;

[0086] For a negative ideal solution, the worst value is selected for each index in the candidate point set, and its expression is as follows:

[0087] ;

[0088] The distance from each device to the positive and negative ideal solutions is calculated separately, and the expression is as follows:

[0089] ;

[0090] ;

[0091] in: For the first The device reaches the ideal solution distance, For the first The device leads to the negative ideal solution. The distance;

[0092] The proximity is calculated based on the distance of each device to the positive and negative ideal solutions, and its expression is as follows:

[0093] ;

[0094] in: For the first The proximity of each device;

[0095] Iterate through the proximity of all devices and sort them in descending order to obtain the priority of upgrading old power grid production equipment.

[0096] The silhouette coefficient method was used to verify and correct the subclusters in AGNES clustering. The quality specifically includes:

[0097] For each data point , the average distance to all other points in the same cluster is calculated, which is expressed as follows:

[0098] ;

[0099] ;

[0100] wherein: is the intra-cluster compactness of data point to other data points in the same cluster, is the number of points in the sub-cluster , is the Euclidean distance between data point and ; For each data point

[0101] , the average distance to all points in the nearest cluster is calculated, which is expressed as follows:

[0102] ;

[0103] wherein: is the inter-cluster separation of data point to the nearest cluster, is the nearest cluster to the sub-cluster , is the number of points in the nearest cluster , is the Euclidean distance between data point and any data point in the nearest cluster ; The intra-cluster compactness

[0104] is compared with the inter-cluster separation , and the silhouette coefficient is obtained, which is expressed as follows:

[0105] ;

[0106] Based on the silhouette coefficient, the average silhouette coefficient of all sub-clusters is calculated, which is expressed as follows:

[0107] ;

[0108] wherein: is the average silhouette coefficient of all sub-clusters, is the number of total data points,​​​ is the number of sub-clusters;

[0109] all the calculated average profile coefficients of sub-clusters are compared with the profile coefficient threshold value If , it is considered that the clustering effect is not good, and the adjustment coefficient needs to be adjusted a new sub-cluster is generated; if , the final divided each sub-cluster is output .

[0110] An equipment modification priority determination system based on an improved snow goose optimization algorithm, the system comprises:

[0111] An evaluation index system construction unit, configured to construct an evaluation index system reflecting equipment modification priority based on the current situation of power grid old production equipment modification investment;

[0112] A comprehensive weight calculation unit, configured to calculate subjective weights based on the evaluation index system, and calculate objective weights by using a CRITIC method combining adaptive adjustment K-NN and mutual information, and combine the subjective weights and the objective weights to calculate comprehensive weights;

[0113] A comprehensive weight optimization unit, configured to optimize the comprehensive weights based on an improved snow goose optimization algorithm of multi-objective optimization to obtain optimal comprehensive weights;

[0114] The improvement of the snow goose optimization algorithm comprises updating the snow goose speed in the exploration stage of the snow goose algorithm based on a dynamic congestion speed updating mechanism, and updating the snow goose position in the development stage of the snow goose algorithm based on a global-local search mechanism according to the fitness change;

[0115] An equipment modification priority determination unit, configured to calculate the distance between each equipment and the positive and negative ideal solutions by using a TOPSIS method based on DBSCAN-AGNES double clustering redefinition based on the optimal comprehensive weights, obtain closeness and sort to determine the modification priority of the power grid old production equipment;

[0116] Preferably, the evaluation index system construction unit, the comprehensive weight calculation unit, the comprehensive weight optimization unit, and the equipment modification priority determination unit specifically realize the steps of the functions, please refer to the corresponding description in the method part.

[0117] Compared with the prior art, the beneficial effects of the present application are:

[0118] ​The application is a kind of device modification priority determination method and system based on improved snow goose optimization algorithm, and the method first constructs an evaluation index system for reflecting the priority of device modification, then calculates the comprehensive weight and obtains the optimal comprehensive weight based on the improved snow goose optimization algorithm of multi-objective optimization, and finally calculates the closeness degree and sorts it by using the TOPSIS method based on the double clustering redefinition of DBSCAN-AGNES to determine the modification priority of the old production equipment of the power grid; In the application of the design, the subjective and objective weights are calculated by combining AHP and improved CRITIC method, and the dynamic congestion speed and global-local search mechanism are introduced to improve the snow goose optimization algorithm, the intelligent optimization of the weight is realized, the weight distribution is more accurate and reasonable, and based on the double clustering of DBSCAN-AGNES, the problem of interference of abnormal values and noise in the traditional TOPSIS method is avoided, the robustness and reliability of the evaluation result are improved, so that the device modification priority can be determined more scientifically and accurately. BRIEF DESCRIPTION OF DRAWINGS

[0119] Figure 1 The method flowchart of the application.

[0120] Figure 2 The system structure diagram of the application.

[0121] Figure 3 The device structure diagram of the application.

[0122] In the figure: evaluation index system construction unit 1, comprehensive weight calculation unit 2, comprehensive weight optimization unit 3, device modification priority determination unit 4, processor 5, memory 6, computer program code 61. DETAILED DESCRIPTION

[0123] The application will be further described in detail in combination with the description of the drawings and specific embodiments.

[0124] Embodiment 1:

[0125] Referring to Figure 1 A device modification priority determination method based on improved snow goose optimization algorithm, comprising:

[0126] Based on the current situation of investment in the modification of old production equipment of the power grid, an index evaluation system reflecting the priority of device modification is constructed;

[0127] Further, by analyzing the current situation of investment in the modification of old production equipment of the power grid, an evaluation index system reflecting the investment risk of the modification of old production equipment of the power grid is constructed from the dimensions of operation time, operation state, operation environment, importance, equipment risk and economy, so as to reflect the priority of device modification; The evaluation index system is shown in the following table.

[0128] ;

[0129] Based on the index evaluation system, the subjective weight is calculated, and the CRITIC method combining adaptive adjustment K-NN and mutual information is used to calculate the objective weight, and the comprehensive weight is calculated by combining the subjective weight and the objective weight;

[0130] In this scheme, the operating environment factors are introduced into the quasi-side layer, mainly including the temperature and humidity index and the frequency of extreme weather; among them, the temperature and humidity change directly affects the insulation performance, heat dissipation capacity, corrosion speed and aging degree of the equipment. For example, in a high temperature and humid environment, electrical equipment may be more prone to insulation failure; while in a low temperature environment, some equipment may have problems such as material brittleness and oil solidification. The temperature and humidity index can provide effective information for the performance of equipment under different climate conditions, helping to assess whether the equipment needs to be modified or replaced in advance. The temperature and humidity index can be generated using local meteorological data (temperature, humidity), which can reflect the environmental impact that the equipment may be subjected to.

[0131] Extreme weather events (such as storms, snow, high winds, floods, etc.) have a significant impact on the safety and reliability of power grid equipment. Frequent extreme weather can lead to equipment failure or higher maintenance costs, and even cause equipment to be unable to operate in the short term, increasing the risk of power grid outages. Combining extreme weather events and historical disaster data can help predict the future performance of equipment under extreme climate conditions, identify potential equipment hazards in advance, and prioritize the modification of high-risk equipment. By analyzing historical meteorological data and disaster records in the area where the equipment is located, the frequency of extreme weather events and the impact of each disaster on the equipment can be calculated. For example, analyze the frequency of snowstorms and the impact of strong winds in the past 5-10 years in the region.

[0132] Further, in this scheme, the AHP method is used to calculate the subjective weight; AHP is a decision-making method that combines quantitative and qualitative analysis, which helps decision-makers weigh between multiple alternative solutions and evaluation criteria, so as to make rational choices, the specific steps are as follows:

[0133] Construct a judgment matrix: assume that there are evaluation indexes , compare the importance of each two indexes based on expert scoring, and use a scale to represent the comparison result, the scale table is as follows:

[0134] ;

[0135] For example, assuming that "equipment risk" and "economy" are compared with each other, the expert may think that "equipment risk" is more important, and scores 5.

[0136] Experts make pairwise comparisons by scaling to construct a judgment matrix whose expression is as follows:

[0137]

[0138] The judgment matrix is normalized by column, and the arithmetic mean of each row of the normalized matrix, i.e. the subjective weight of each index, is calculated. Its expression is as follows:

[0139]

[0140] wherein: is the normalized element, is the relative importance of the th index to the th index, obtained by expert judgment, is the subjective weight of the th index.

[0141] After obtaining the subjective weight, the consistency index and the consistency ratio need to be calculated, wherein is the random consistency index; finally, the size relationship between and 0.1 is judged. If , the matrix has satisfactory consistency, otherwise the judgment matrix needs to be adjusted.

[0142] Further, the CRITIC method combining adaptive adjustment K-NN and mutual information is used to calculate the objective weight.

[0143] K-NN (K-Nearest Neighbor) is a simple and intuitive algorithm widely used in classification and regression tasks. Its basic idea is to calculate the distance between the target point and other data points, select the K nearest neighbors, and make predictions according to the categories or values of these neighbors. Applying the idea of K-NN to mutual information calculation can effectively improve the limitations of traditional methods. Traditional mutual information calculation relies on global frequency to estimate marginal probability and joint probability, assuming uniform data distribution, but cannot effectively handle nonlinear relationships and local structures in data. By calculating local neighborhoods through K-NN, local probability distribution can be estimated for each data point, thus more accurately capturing local dependencies and nonlinear relationships in data.

[0144] ​​​​To further optimize the calculation, an adaptive K-NN method is proposed to improve mutual information computation. Adaptive K-NN dynamically adjusts the K value, using a smaller K value in densely populated regions and a larger K value in sparse regions. This method better adapts to the local characteristics of the data and avoids errors caused by a fixed K value. Applying this improved mutual information computation method to the CRITIC method can more accurately measure the correlation between various indicators, thus making the final evaluation result more accurately reflect the true structure of the data. Specifically, this includes:

[0145] Standardize the data samples of different indicators in the evaluation indicator system and calculate the standard deviation of the indicators;

[0146] The expression for the standardization process is as follows:

[0147] ;

[0148] in: For the first The sample at the th Standardized indicator values ​​for each indicator. For the first The sample at the th The original values ​​of each indicator , The first The sample at the th The minimum and maximum values ​​of each indicator; after standardization, each indicator... All are within the interval [0,1].

[0149] The expression for the standard deviation is as follows:

[0150] ;

[0151] in: For the first The standard deviation of each indicator For the first The sample at the th Standardized indicator values ​​for each indicator. For the first The average of the indicators, The number of samples;

[0152] Calculate the local density of each sample point and adaptively adjust the number of neighbors for each sample based on the local density;

[0153] The expression for the local density is as follows:

[0154] ; ;

[0155] wherein: is the local density of the th sample, is the number of neighbors of the current sample, is the distance between the th sample and the th neighbor, , are the values of the th sample and the th sample on the th index, respectively, is the number of indices;

[0156] The expression of the adaptive adjustment of the number of neighbors of each sample is as follows:

[0157] ;

[0158] wherein: is the number of neighbors of each sample, is the maximum value of the number of neighbors, is the floor function;

[0159] Through the above steps, if the local density of the sample is high, the number of neighbors will be smaller; if the local density is lower, the number of neighbors will be larger, thereby achieving adaptive adjustment.

[0160] Based on the number of neighbors of each sample, the marginal probability of the th index and the th index is calculated, and the expression is as follows:

[0161] ;

[0162] ;

[0163] wherein: , are the marginal probabilities of the th index and the th index, respectively, is a Gaussian kernel function, , are both bandwidths; is the value of the th index, is the standardized value of the th sample on the th index;

[0164] The joint probability of the th index and the th index is calculated, and the expression is as follows:​

[0165] ;

[0166] wherein: is the joint probability of the value of the i-th index and the j-th index at the same time; Based on the marginal probability and the joint probability, the mutual information of the i-th index and the j-th index is calculated, and the expression is as follows:

[0167]

[0168] ;

[0169] wherein: is the mutual information of the i-th index and the j-th index; is the logarithm reflecting the difference between the marginal probability and the joint probability, which describes the information sharing amount of the index and the index , if the index and the index are independent, the ratio is 1, and the logarithm is 0. Based on the mutual information and the standard deviation, the information amount of each index is calculated, and the objective weight is obtained, and the expression is as follows:

[0170]

[0171] ; ;

[0172] wherein: is the information amount of the i-th index, is the standard deviation of the i-th index, is the objective weight of the i-th index. Based on the improved snow goose optimization algorithm of multi-objective optimization, the comprehensive weight is optimized, and the optimal comprehensive weight is obtained.

[0173] Based on the improved snow goose optimization algorithm of multi-objective optimization, the comprehensive weight is optimized, and the optimal comprehensive weight is obtained.

[0174] ​​​​​​​​The traditional method in determining the weight coefficient usually relies on expert experience, which often has a certain subjectivity and limitation, while the Xueyan optimization algorithm can dynamically allocate weights according to data characteristics by automatically adjusting the weights without expert intervention, so as to realize more objective and efficient optimization. In order to further improve the performance of the algorithm, the scheme is improved. Specifically, the scheme designs a multi-objective fitness function to ensure the rationality and effectiveness of weight optimization, and introduces a dynamic congestion speed updating mechanism and a global-local search mechanism based on fitness change to avoid premature convergence, realize more accurate local search, and balance global optimization and local optimization. The specific steps are as follows:

[0175] Firstly, a weight optimization mathematical model, i.e. the optimal comprehensive weight, is constructed, and its expression is as follows:

[0176] ;

[0177] Among them: is the optimal comprehensive weight, is the weight coefficient, is the subjective weight, is the objective weight. The optimization goal of the formula is to find the value of the optimal weight coefficient , so that can not only keep the balance between subjective and objective weights, but also improve the scientificity and reliability of the evaluation results.

[0178] In order to optimize the weight coefficient , it is necessary to define a multi-objective optimization comprehensive fitness function to comprehensively measure the rationality of weight optimization, and the multi-objective optimization comprehensive fitness function is as follows:

[0179] ;

[0180] ;

[0181] ;

[0182] ;

[0183] Among them: is the fitness function, , , are all weight coefficients, , , are all objective functions, is the punishment degree of adjustment gradient punishment item, Gradient penalty term; In this scheme, a gradient penalty term is introduced into the fitness function, aiming to constrain the gradient change of weights to ensure the stability of weight update during optimization, prevent excessive fluctuations, and improve the stability of the optimization process;

[0184] Objective function The deviation between the optimized combination weight and the AHP and CRITIC weight can be measured to ensure the balance of subjective and objective weights; Objective function By evaluating the error between the optimized combination weight and expert experience, the prediction accuracy is ensured; Objective function The balance between AHP and CRITIC weight is ensured, avoiding the bias caused by the excessive weight of one side; The comprehensive weight obtained by expert experience; subscript Represents the index of each data point.

[0185] Snow Geese Algorithm (SGA) is inspired by the migration behavior of snow geese, especially their unique "chevron" and "straight line" flight patterns during migration. This algorithm simulates the flight behavior of snow geese to achieve efficient search and optimization in the solution space. SGA algorithm mainly consists of three stages: initialization stage, exploration stage and development stage.

[0186] In the initialization stage, the individual position needs to be initialized, and the position of snow geese is randomly generated in the search space. The initial position is determined according to the population size, solution space and dimension, and the position update formula is as follows:

[0187] ;

[0188] Where: is the initialized set position, and are the upper and lower bounds of the solution space, is a random number in the range [0, 1].

[0189] In the exploration stage, the "chevron" flight pattern of snow geese is simulated to enhance the search ability of population individuals in the solution space, and the speed update formula is as follows:

[0190] ; ;

[0191] Where: is the velocity, is the iteration speed, is the weight factor, is the acceleration, is the maximum number of iterations;

[0192] The position update formula is as follows:

[0193] ;

[0194] Wherein: is the position of the individual at the moment is the position of the individual at the moment is the position of the individual at the moment is the position of the individual at the moment is the position of the individual at the moment is the position of the individual at the moment is the position of the current optimal individual is the weight coefficient.

[0195] In the development stage, the "straight line" flight mode of the snow goose is simulated, and the concentrated search is used to improve the development ability of the algorithm, and the position update formula is as follows:

[0196] ;

[0197] Wherein: is a random number in the range [0, 1], is an element-by-element multiplication, is Brownian motion;

[0198] Further, in the scheme, the limitations of the snow goose algorithm are improved, as follows:

[0199] In the exploration stage, the congestion distance is introduced into the speed update formula, which helps to maintain diversity in the global search stage and prevent excessive update when approaching the optimal solution; the congestion distance can dynamically adjust the speed, especially when the solution space is relatively dense, which can slow down the individual speed and enhance local optimization. By controlling the direction and pace of speed update, the balance between global exploration and local convergence is achieved, avoiding premature convergence or excessive concentration. When approaching the optimal solution, the introduction of congestion distance can avoid excessive jumping and achieve more accurate local search. At the same time, in order to ensure appropriate diversity in different stages of search, the value of can be dynamically adjusted. In the early stage of exploration, the group is relatively dispersed, so it needs a larger congestion distance influence (i.e. larger ), and as the search gradually converges, the value of can be gradually reduced to enhance the convergence ability. The improved speed update formula in the exploration stage is as follows:

[0200] ; ;

[0201] Wherein: is the snow goose speed, is the weight factor, is the control factor, is the number of fitness function targets, , For the first In the target dimension, the first The fitness values ​​of each individual's neighbors. , The first The maximum and minimum fitness values ​​in each objective dimension. As the initial control factor, This represents the maximum number of iterations.

[0202] During the development phase, the overall fitness value is used to guide the global search, while the inter-target fitness value is used to control the degree of conflict between targets and guide the local search. The improved snow goose position formula for the development phase is as follows:

[0203] ;

[0204] in: For the first Individuals Location at any given moment As a weighting factor, For the first Individuals Overall adaptability at any time The average overall fitness of the population. This is the step size control factor. For individuals in the first The fitness differences in each objective dimension and the global optimal solution in the th objective dimension A synthesis of fitness differences across all target dimensions; For the first Fitness values ​​for each objective dimension For the first The optimal solution for the region where each individual is located;

[0205] With the above improvements, if an individual's overall fitness is high, then the overall fitness can be used to guide the individual toward the global optimum. The algorithm moves and adjusts the step size for a broad search; if the overall fitness of an individual is low, it enters a local search phase, adjusting the step size based on the fitness differences between objectives; larger fitness differences indicate greater conflict between objectives, increasing the step size to promote exploration; smaller differences indicate that objectives are converging, reducing the step size for a more refined search. This improvement makes the optimization process more flexible and adaptive, achieving a better balance between global and local optimization, and enhancing the performance of multi-objective optimization algorithms.

[0206] The position of the snow goose is iterated by the above method until the iteration number or the improvement of the comprehensive fitness value is less than the set threshold, then the updating is stopped and the weight coefficient corresponding to the current global optimal position is returned, and the subjective weight and the objective weight are respectively optimized by returning the weight coefficient corresponding to the current global optimal position, and the optimal comprehensive weight is obtained, and the expression is as follows:

[0207] ;

[0208] Among them: The optimal comprehensive weight is wopt, The weight coefficient is w, The subjective weight is ws, The objective weight is wo.

[0209] Based on the optimal comprehensive weight, the distance between each device and the positive and negative ideal solutions is calculated by using the TOPSIS method based on the double clustering redefinition of DBSCAN-AGNES, the closeness is obtained and sorted to determine the priority of the transformation of the old production equipment of the power grid;

[0210] The main deficiency of the traditional TOPSIS method is that it depends on the global maximum and minimum values to define the positive and negative ideal solutions, and does not fully consider the heterogeneity in the data and the similarity between devices, and is sensitive to outliers. In view of the above shortcomings, the scheme proposes a method for determining the positive and negative ideal solutions based on DBSCAN-AGNES double clustering. First, the global density clustering is performed by DBSCAN, and according to the density distribution of the device index data, the devices are divided into different clusters, so as to discover the cluster structure of any shape in the data set. Then, hierarchical clustering is performed by AGNES, which further identifies high-density sub-clusters to more carefully mine the similarity between devices, and devices with higher density and more similar characteristics are classified into the same sub-cluster. Finally, after obtaining the candidate set of sub-cluster centers, the optimal value and the worst value in each sub-cluster are selected to determine the positive and negative ideal solutions. By selecting the optimal value and the worst value in the homogeneous sub-cluster, this method can more accurately reflect the characteristics and differences of the devices. Compared with the traditional TOPSIS method which uniformly finds the positive and negative ideal solutions on the entire data set, the scheme can mine the essential mode through data distribution by introducing DBSCAN-AGNES double clustering, and construct ideal solutions with mathematical optimality and business feasibility, so that the TOPSIS method is more accurate and robust in the priority evaluation of power grid devices. The specific steps are as follows:

[0211] Based on the optimal comprehensive weight, the distance between each device and the positive and negative ideal solutions is calculated by using the TOPSIS method based on the double clustering redefinition of DBSCAN-AGNES, the closeness is obtained and sorted to determine the priority of the transformation of the old production equipment of the power grid;

[0212] ; ;

[0213] Among them: For a weighted normalized matrix, For the weighted normalized matrix, the first... The device in the The value of each indicator, For the first The optimal comprehensive weight of each indicator;

[0214] Global density clustering based on the DBSCAN clustering algorithm is performed to extract effective clusters, specifically including:

[0215] For any sample within the weighted normalization matrix ,That The neighborhood is defined as follows:

[0216] ; ;

[0217] in: For the sample of Neighborhood, For the sample and The Euclidean distance between them For the first The device in the The value of each indicator;

[0218] based on Neighborhood judgment sample Core point, boundary point, and noise point;

[0219] like Then the sample As the core point; Minimum number of samples;

[0220] If the sample Not a core point, but located at a certain core point Within the neighborhood, the sample For boundary points;

[0221] If the sample Points that are neither core points nor boundary points are considered noise points.

[0222] Starting from any unvisited core point, all samples with density reachable from it are grouped into the same cluster, and all samples not grouped into a cluster are labeled as noise (labeled -1).

[0223] Let the set of cluster labels generated after clustering be... ,in The remaining labels are valid tags, while the remaining labels are noise. After further noise removal, the valid clusters are:

[0224] ;

[0225] wherein: is an effective cluster, is the th set;

[0226] The local subdivision clustering is based on the AGNES clustering algorithm, and specifically includes:

[0227] The cosine similarity of data points and in the effective cluster is calculated and converted into a distance, and the expression is as follows:

[0228] ;

[0229] wherein: is the distance based on the cosine similarity of data points and , , are the components of data points and in the th dimension, is the total dimension of the data points;

[0230] Based on the cosine distance between data points, the distance between subclusters is calculated using the average distance method, and the expression is as follows:

[0231] ;

[0232] wherein: is the average cosine distance of all point pairs between two subclusters, , are the two subclusters to be merged, , are the numbers of sample points in the clusters, respectively;

[0233] The above steps are repeated to gradually merge similar clusters until the minimum distance between clusters exceeds the threshold value, and finally a number of subclusters are obtained, and the expression is as follows:

[0234] ;

[0235] wherein: is the threshold value, is the adjustment coefficient, is the median of the cosine distance of all points in the current global cluster;

[0236] After merging, each subcluster is obtained, and the silhouette coefficient method is used to verify and correct the subclusters quality of the final partitioned each sub-cluster is output ;

[0237] Further, the silhouette coefficient method can objectively measure the clustering effect, identify and correct problems in the clustering process, improve the accuracy and robustness of the clustering results, and ensure that the final evaluation is more scientific and reasonable, providing higher precision and reliability for power grid equipment renovation priority evaluation. The steps are as follows:

[0238] For each data point , the average distance from all other points in the same cluster is calculated, and the expression is as follows:

[0239] ;

[0240] ;

[0241] Wherein: is the intra-cluster compactness of data point to other data points in the same cluster, is the number of points in the sub-cluster , is the Euclidean distance between data points and in the sub-cluster ;

[0242] For each data point , the average distance from all points in the nearest cluster is calculated, and the expression is as follows:

[0243] ;

[0244] Wherein: is the inter-cluster separation of data point to the nearest cluster , is the nearest cluster to the sub-cluster , is the number of points in the nearest cluster , is the Euclidean distance between data point and any data point in the nearest cluster ;

[0245] Compare the intra-cluster compactness and the inter-cluster separation to obtain the silhouette coefficient , and the expression is as follows:

[0246] ;

[0247] Based on the profile coefficient, the average profile coefficient of all sub-clusters is calculated, representing the matching degree of data points with their clusters, and its expression is as follows:

[0248] ;

[0249] Wherein: is the average profile coefficient of all sub-clusters, is the number of total data points, is the number of sub-clusters;

[0250] The calculated average profile coefficient of all sub-clusters is compared with the profile coefficient threshold ; if , it is considered that the clustering effect is poor, and the adjustment coefficient needs to be adjusted A new sub-cluster is generated; if , the final divided each sub-cluster is output .

[0251] For each sub-cluster obtained by final division , the centroid is calculated as the center point, and its expression is as follows:

[0252] ;

[0253] Wherein: is the sub-cluster of division, is the sub-cluster center point;

[0254] Based on the sub-cluster center point , the candidate point set is constructed, and its expression is as follows:

[0255] ;

[0256] Wherein: each sub-cluster center point corresponds to a multi-dimensional vector ;

[0257] Determine the positive and negative ideal solutions of each index in the candidate point set:

[0258] For the positive ideal solution, the optimal value of each index in the candidate point set is selected, and its expression is as follows:

[0259] ;

[0260] For the negative ideal solution, the worst value of each index in the candidate point set is selected, and its expression is as follows:

[0261] ;

[0262] The distance of each device to the positive and negative ideal solution is calculated respectively, and the expression is as follows:

[0263]

[0264]

[0265] Wherein: is the distance of the i-th device to the positive ideal solution, is the distance of the i-th device to the negative ideal solution; Based on the distance of each device to the positive and negative ideal solution, the closeness degree is calculated, and the expression is as follows:

[0266]

[0267]

[0268] Wherein: is the closeness degree of the i-th device;

[0269] Traverse the closeness degrees of all devices, and sort all devices in descending order to obtain the transformation priority of the old power grid production device.

[0270] In this embodiment, the transformation priority of the old device based on the method is simulated and analyzed.

[0271]

[0272] The above table is the combination weight optimized by the improved snow goose optimization algorithm based on multi-objective optimization. As can be seen from the table, the optimized weight combines subjectivity (AHP) and objectivity (CRITIC), and through the adjustment of the improved snow goose algorithm, the weight distribution is more scientific and reasonable. For example, the weight of "state evaluation" is optimized from 0.15 of AHP and 0.14 of CRITIC to 0.145, indicating that the index occupies an important position under the two methods, and the weight after optimization is more balanced; and the weight of "device loss risk" is optimized from 0.20 of AHP and 0.18 of CRITIC to 0.19, which reflects the compromise between the index in the subjective and objective weights. Although the weight is slightly lower than AHP, it still maintains a high level, indicating that the importance of the index in the device transformation priority evaluation is reasonably reflected.

[0273]

[0274] ​​​​​​​​​​The above table is five priorities based on closeness degree. As can be seen from the table, there are five levels: 1st level is the highest priority, such devices may have serious safety hazards, poor running state, or have very high importance, which is the key to ensure the stable operation of power grid, and need to be transformed first; 2nd level is higher priority, these devices although the urgency is slightly lower than 1st level, but the transformation demand is higher, should be arranged as soon as possible under the condition of resources; 3rd level is medium priority, such devices current transformation demand is general, but may need to be dynamically tracked and evaluated in the future due to gradual aging; 4th level is lower priority, these devices running state is relatively stable, and has less impact on the overall power grid, resources should be allocated to more urgent devices first; 5th level is the lowest priority, these devices running condition is good, or importance is low, can be temporarily maintained, as part of the long-term transformation plan.

[0275] ;

[0276] The above table is the evaluation result of device priority. As can be seen from the table, the closeness degree of device 1 is 0.81, which is closest to the ideal solution, and the transformation priority is the highest, which should be arranged first; the closeness degrees of device 2 and device 3 are 0.73 and 0.69 respectively, and the transformation demand is higher, it is recommended to transform as soon as possible; the closeness degree of device 4 is 0.57, the transformation demand is medium, the transformation plan can be appropriately delayed, and at the same time, the high priority devices should be focused on; the closeness degree of device 5 is 0.48, the transformation demand is low, and the current running state is relatively stable, which can be temporarily maintained; the closeness degree of device 6 is 0.34, which is farthest from the ideal solution, and the transformation demand is the lowest, which can be temporarily maintained, and the transformation resources should be prioritized for other more urgent devices.

[0277] Example 2:

[0278] Referring to Figure 2 A device transformation priority determination system based on improved snow goose optimization algorithm, the system comprises:

[0279] An evaluation index system construction unit 1 for constructing an evaluation index system reflecting the transformation priority of the device based on the transformation investment status of the old production equipment of the power grid;

[0280] A comprehensive weight calculation unit 2 for calculating subjective weight based on the evaluation index system, and calculating objective weight by using CRITIC method combined with adaptive adjustment K-NN and mutual information, and combining subjective weight and objective weight to calculate comprehensive weight;

[0281] A comprehensive weight optimization unit 3 for optimizing the comprehensive weight based on the improved snow goose optimization algorithm of multi-objective optimization to obtain the optimal comprehensive weight;

[0282] The improvement of the snow goose optimization algorithm includes updating the snow goose speed in the exploration stage of the snow goose algorithm based on a dynamic congestion speed updating mechanism; and updating the snow goose position in the development stage of the snow goose algorithm based on a global-local search mechanism according to the fitness change;

[0283] The equipment modification priority determination unit 4 is configured to calculate the distance between each equipment and the positive and negative ideal solutions based on the optimal comprehensive weight, and obtain the closeness degree and sort the closeness degree by using the TOPSIS method redefined based on the DBSCAN-AGNES double clustering, so as to determine the modification priority of the old production equipment in the power grid.

[0284] The DBSCAN-AGNES double clustering includes global density clustering based on the DBSCAN clustering algorithm and local subdivision clustering based on the AGNES clustering algorithm.

[0285] Further, the steps of realizing the functions of the evaluation index system construction unit 1, the comprehensive weight calculation unit 2, the comprehensive weight optimization unit 3, and the equipment modification priority determination unit 4 are described in the corresponding description of the method in Embodiment 1.

[0286] Embodiment 3

[0287] Referring to Figure 3 An equipment for determining the modification priority of equipment based on the improved snow goose optimization algorithm, the equipment comprising a processor 5 and a memory 6.

[0288] The memory 6 is configured to store computer program code 61 and transmit the computer program code 61 to the processor 5.

[0289] The processor 5 is configured to execute the method for determining the modification priority of equipment based on the improved snow goose optimization algorithm according to the instructions in the computer program code 61.

[0290] In this embodiment, a computer readable storage medium is also included, and the computer readable storage medium stores computer executable instructions, when the computer executable instructions are executed on a computer, the method for determining the modification priority of equipment based on the improved snow goose optimization algorithm is realized.

[0291] Generally, the computer instructions for realizing the method of the present application can be carried by any combination of one or more computer readable storage media. The non-transitory computer readable storage medium can include any computer readable medium except the signal itself temporarily propagating.

[0292] Computer readable storage media can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer readable storage media include: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present disclosure, computer readable storage media can be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0293] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages, and specifically Python language and platform frameworks based on TensorFlow, PyTorch, etc. suitable for neural network computing. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0294] The above-mentioned device and non-transitory computer readable storage medium can refer to the specific description of the device modification priority determination method based on the improved snow goose optimization algorithm and its beneficial effects, which will not be described here.

[0295] Although the embodiments of the present application have been shown and described above, it should be understood by those skilled in the art that the above-mentioned embodiments are exemplary and cannot be interpreted as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above-mentioned embodiments within the scope of the present application.

Claims

1. A device modification priority determination method based on an improved snow goose optimization algorithm, characterized in that, The method comprises the following steps: Based on the current situation of the transformation investment of old power grid production equipment, an evaluation index system reflecting the priority of equipment transformation is constructed; Based on the evaluation index system, the subjective weight is calculated, and the CRITIC method combining adaptive adjustment K-NN and mutual information is used to calculate the objective weight, and the comprehensive weight is calculated by combining the subjective weight and the objective weight; The improved snow goose optimization algorithm based on multi-objective optimization is used to optimize the comprehensive weight to obtain the optimal comprehensive weight; The improvement of the snow goose optimization algorithm includes the following aspects: based on the dynamic congestion speed updating mechanism, the snow goose speed in the exploration stage of the snow goose algorithm is updated; based on the global-local search mechanism of fitness change, the snow goose position in the development stage of the snow goose algorithm is updated; Based on the optimal comprehensive weight, the TOPSIS method based on the double clustering redefinition of DBSCAN-AGNES is used to calculate the distance between each equipment and the positive and negative ideal solutions, obtain the closeness and sort, and determine the transformation priority of the old power grid production equipment; The double clustering based on DBSCAN-AGNES includes the following steps: global density clustering based on the DBSCAN clustering algorithm; Local subdivision clustering based on the AGNES clustering algorithm.

2. The equipment transformation priority determination method based on the improved snow goose optimization algorithm according to claim 1, wherein: The CRITIC method combining adaptive adjustment K-NN and mutual information is used to calculate the objective weight, which specifically includes the following steps: The different index data samples in the evaluation index system are standardized, and the standard deviation of the index is calculated, and the expression is as follows: ； wherein: is the standard deviation of the first index, is the normalized index value of the first sample on the first index, is the mean of the first index, is the number of samples; The local density of each sample point is calculated, and the number of neighbors of each sample is adaptively adjusted according to the local density; The expression of the local density is as follows: ； ； wherein: is the local density of the th sample, is the number of neighbors of the current sample, is the distance between the th sample and the th neighbor, , are the values of the th sample and the th sample on the th index, respectively, is the number of indices. The expression of the adaptive adjustment of the number of neighbors of each sample is as follows: ； wherein: is the number of neighbors for each sample, is the maximum value of the number of neighbors, is the floor function; Based on the number of neighbors of each sample, the edge probability of the first index and the second index is calculated, which is expressed as follows: ； ； wherein: , are the marginal probabilities of the th index and the th index, respectively, is a Gaussian kernel function, , are bandwidths; is the value of the th index, is the normalized value of the th index for the th sample. Calculate the first The first indicator and the first The joint probability of the indicators is expressed as follows: ； in: For the first The first indicator and the first The joint probability of several indicators taking a certain value at the same time; Based on the marginal probability and the joint probability, mutual information of the first index and the second index is calculated, and its expression is as follows: ； wherein: is the mutual information of the first index and the second index, is the logarithm reflecting the difference between the marginal probability and the joint probability; Based on the mutual information and the standard deviation, the information of each index is calculated, and the objective weight is obtained; The expression of the information is as follows: ； wherein: is the information content of the thindex, is the standard deviation of the thindex; The expression of the objective weight is as follows: ； wherein: is the objective weight of the th index.

3. The equipment transformation priority determination method based on the improved snow goose optimization algorithm according to claim 2, wherein: The improved snow goose optimization algorithm based on multi-objective optimization is used to optimize the comprehensive weight to obtain the optimal comprehensive weight, which specifically includes the following steps: The position, speed and global and individual best position of the snow goose are initialized; Wherein: the position of the initial snow goose is generated, the position of each snow goose represents a possible weight coefficient; the speed of each snow goose is initialized to zero or a small random value; the position of each snow goose is set as the individual best position, and the global best position is updated when the comprehensive fitness of the current snow goose is better than the global best comprehensive fitness; The comprehensive fitness value of the current position of each snow goose is calculated based on the fitness function, and the expression is as follows: ； ； ； ； wherein: is a fitness function, , , are weight coefficients, , , are objective functions, is a penalty strength of the adjustment gradient penalty term, is a gradient penalty term; The snow goose speed in the exploration stage of the snow goose algorithm is updated based on the dynamic congestion speed updating mechanism, and the snow goose position in the development stage of the snow goose algorithm is updated based on the global-local search mechanism of fitness change, until the iteration number or the improvement of the comprehensive fitness value is less than the set threshold, then the updating is stopped and the weight coefficient corresponding to the current global best position is returned. The snow goose speed in the exploration stage of the snow goose algorithm is updated based on the dynamic congestion speed updating mechanism, and the expression is as follows: ； ； wherein: is the snow goose velocity, is the weight factor, is the control factor, is the number of fitness function objectives, , is the fitness value of the individual adjacent to the th individual in the th objective dimension, , are the maximum and minimum fitness values in the th objective dimension, respectively, is the initial control factor, is the maximum number of iterations; The snow goose position in the development stage of the snow goose algorithm is updated based on the global-local search mechanism based on fitness changes, and the expression is as follows: ； in: For the first Individuals Location at any given moment As a weighting factor, For the first Individuals Overall adaptability at any time The average overall fitness of the population. This is the step size control factor. For individuals in the first The fitness differences in each objective dimension and the global optimal solution in the th objective dimension A synthesis of fitness differences across all target dimensions; For the first Fitness values ​​for each objective dimension For the first The optimal solution for the region where each individual is located; The weight coefficient corresponding to the current global optimal position is returned to optimize the subjective weight and the objective weight respectively, and the optimal comprehensive weight is obtained, and the expression is as follows: ； wherein: is the optimal integrated weight, is the weight coefficient, is the subjective weight, is the objective weight.

4. The device renovation priority determination method based on the improved snow goose optimization algorithm according to claim 3, characterized in that: Based on the optimal comprehensive weight, the distance between each device and the positive and negative ideal solutions is calculated by using the TOPSIS method based on the DBSCAN-AGNES double clustering redefinition, the closeness is obtained and sorted to determine the renovation priority, and specifically comprising: Based on the optimal comprehensive weight, the standardized index value is weighted to construct a weighted normalized matrix, and the expression is as follows: ； ； in: For a weighted normalized matrix, For the weighted normalized matrix, the first... The device in the The value of each indicator, For the first The optimal comprehensive weight of each indicator; Based on the DBSCAN clustering algorithm, global density clustering is performed, and effective clusters are extracted, and the expression is as follows: ； wherein: is an effective cluster, is a first set; Computing cosine similarity of data points in an effective cluster and and converting to distance, expressed as follows: ； wherein: is a data point and a distance based on cosine similarity, , are data points and components of the data points in the th dimension, is the total dimension of the data points; Based on the cosine distance between data points, the average distance method is used to calculate the distance between subclusters, and the expression is as follows: ； wherein: is the average of the cosine distances of all pairs of points between the two sub-clusters, , are the two sub-clusters to be merged, , are the number of sample points in the clusters, respectively. Repeat the above steps to gradually merge similar clusters until the minimum distance between clusters exceeds the threshold value, and finally divide to obtain several subclusters, and the expression is as follows: ； wherein: is a threshold value, is a regulation coefficient, is the median of the cosine distances of all points in the current global cluster. After the merge is complete, each sub-cluster is obtained. The silhouette coefficient method was used to verify and correct the subclusters in AGNES clustering. The quality of the output is used to determine the final partitioning of each sub-cluster. ; For each sub-cluster obtained by the final division The centroid is calculated as the center point, and its expression is as follows: ； wherein: is the first subcluster, is the subcluster center point; Based on sub-cluster center points , constructing candidate point set whose expression is as follows: ； wherein: each sub-cluster center point corresponds to a multi-dimensional vector ; Determine the positive and negative ideal solutions of each index in the candidate point set: For the positive ideal solution, the optimal value of each index is selected in the candidate point set, and the expression is as follows: ； For the negative ideal solution, the worst value of each index is selected in the candidate point set, and the expression is as follows: ； The distance of each device to the positive and negative ideal solutions is calculated, and the expression is as follows: ； ； wherein: is the distance of the i-th device to the positive ideal solution, is the distance of the i-th device to the negative ideal solution. is the distance of the i-th device to the positive ideal solution, is the distance of the i-th device to the negative ideal solution. is the distance of the i-th device to the positive ideal solution,​ Based on the distance of each device to the positive and negative ideal solutions, the closeness is calculated, and the expression is as follows: ； wherein: is the proximity of the first device to the second device; Traverse the closeness of all devices, and sort all devices in descending order to obtain the renovation priority of the old production equipment in the power grid.

5. The device renovation priority determination method based on the improved snow goose optimization algorithm according to claim 4, characterized in that: The silhouette coefficient method was used to verify and correct the subclusters in AGNES clustering. The quality specifically includes: For each data point , the average distance to all other points in the same cluster is calculated, which is expressed as follows: ； ； wherein: is the intra-cluster density of data points to other data points within the same cluster, is the intra-cluster density of data points is the number of sub-clusters of the midpoint, is the Euclidean distance of data points in the sub-cluster and in the sub-cluster. For each data point , the average distance to all points in the nearest cluster is computed, which is expressed as follows: ； wherein: is the distance between data points to the nearest cluster the inter-cluster separation, is the distance to the nearest cluster from the data point, is the number of data points in the nearest cluster is the Euclidean distance between data points and any data point in the nearest cluster . Comparing intra-cluster compactness With inter-cluster separation Obtaining profile coefficients The expression of which is as follows: ； Based on the contour coefficient, the average contour coefficient of all subclusters is calculated, and the expression is as follows: ； wherein: is the average profile coefficient for all sub-clusters, is the number of total data points, is the number of sub-clusters; all the calculated average profile coefficients of all sub-clusters are compared with the profile coefficient threshold value ; if , it is considered that the clustering effect is not good and the adjustment coefficient needs to be adjusted , a new sub-cluster is generated; if , the final divided each sub-cluster is output .

6. A system for determining equipment modification priority based on an improved snow goose optimization algorithm, characterized in that, The system comprises: An evaluation index system construction unit (1) for constructing an evaluation index system reflecting the device renovation priority based on the renovation investment status of the old production equipment in the power grid; A comprehensive weight calculation unit (2) for calculating the subjective weight based on the evaluation index system, and calculating the objective weight by using the CRITIC method combined with adaptive adjustment K-NN and mutual information, and calculating the comprehensive weight combined with the subjective weight and the objective weight; A comprehensive weight optimization unit (3) for optimizing the comprehensive weight based on the improved snow goose optimization algorithm of multi-objective optimization to obtain the optimal comprehensive weight; The improvement of the snow goose optimization algorithm includes updating the snow goose speed in the exploration stage of the snow goose algorithm based on the dynamic congestion speed updating mechanism, and updating the snow goose position in the development stage of the snow goose algorithm based on the global-local search mechanism based on fitness changes. The device renovation priority determination unit (4) is configured to calculate distances between each device and positive and negative ideal solutions based on the DBSCAN-AGNES double clustering redefined TOPSIS method based on the optimal comprehensive weight, obtain closeness and sort to determine the renovation priority of the old production device of the power grid. The DBSCAN-AGNES double clustering includes global density clustering based on the DBSCAN clustering algorithm and local subdivision clustering based on the AGNES clustering algorithm.

7. The device renovation priority determination system based on the improved snow goose optimization algorithm according to claim 6, characterized in that: The comprehensive weight calculation unit (2) is configured to calculate the objective weight according to the following method: The different index data samples in the evaluation index system are standardized, and the standard deviation of the index is calculated, and the expression is as follows: ； wherein: is the standard deviation of the first index, is the normalized index value of the first sample on the first index, is the mean of the first index, is the number of samples; The local density of each sample point is calculated, and the number of neighbors of each sample is adaptively adjusted according to the local density; The expression of the local density is as follows: ； ； wherein: is the local density of the th sample, is the number of neighbors of the current sample, is the distance between the th sample and the th neighbor, , are the values of the th sample and the th sample on the th index, respectively, is the number of indices. The expression of the adaptive adjustment of the number of neighbors of each sample is as follows: ； wherein: is the number of neighbors for each sample, is the maximum value of the number of neighbors, is the floor function; Based on the number of neighbors of each sample, the edge probability of the first index and the second index is calculated, which is expressed as follows: ； ； where: , are the marginal probabilities of the th index and the th index, respectively, is a Gaussian kernel function, , are bandwidths; is the value of the th index, is the normalized value of the th sample on the th index. Calculate the first The first indicator and the first The joint probability of the indicators is expressed as follows: ； in: For the first The first indicator and the first The joint probability of several indicators taking a certain value at the same time; Based on the marginal probability and the joint probability, the mutual information of the first index and the second index is calculated, and its expression is as follows: ； wherein: is the mutual information of the first index and the second index, is the logarithm reflecting the difference between the marginal probability and the joint probability; Based on the mutual information and the standard deviation, the information of each index is calculated, and the objective weight is obtained; The expression of the information is as follows: ； wherein: is the information content of the th index, is the standard deviation of the th index; The expression of the objective weight is as follows: ； wherein: is the objective weight of the th indicator.

8. The device renovation priority determination system based on the improved snow goose optimization algorithm according to claim 7, characterized in that: The comprehensive weight optimization unit (3) is configured to optimize the comprehensive weight according to the following method: The position, speed and global and individual best position of the snow goose are initialized; Wherein: the position of the initial snow goose is generated, the position of each snow goose represents a possible weight coefficient; the speed of each snow goose is initialized to zero or a small random value; the position of each snow goose is set as the individual best position, and when the comprehensive fitness of the current snow goose is better than the global best comprehensive fitness, the global best position is updated; The comprehensive fitness value of the current position of each snow goose is calculated based on the fitness function, and the expression is as follows: ； ； ； ； wherein: is a fitness function, , , are weight coefficients, , , are objective functions, is a penalty strength of the gradient penalty term, is a gradient penalty term; The snow goose speed in the exploration stage of the snow goose algorithm is updated based on the dynamic congestion speed update mechanism, and the snow goose position in the development stage of the snow goose algorithm is updated based on the global-local search mechanism of fitness change, until the number of iterations or the improvement of the comprehensive fitness value is less than the set threshold, then stop updating and return the weight coefficient corresponding to the current global best position; The snow goose speed in the exploration stage of the snow goose algorithm is updated based on the dynamic congestion speed update mechanism, and the expression is as follows: ； ； in: For the speed of the snow geese, As a weighting factor, As a control factor, The number of objectives in the fitness function. , For the first In the target dimension, the first The fitness values ​​of each individual's neighbors. , The first The maximum and minimum fitness values ​​in each objective dimension. As the initial control factor, This represents the maximum number of iterations. The snow goose position in the development stage of the snow goose algorithm is updated based on the global-local search mechanism of fitness change, and the expression is as follows: ； in: For the first Individuals Location at any given moment As a weighting factor, For the first Individuals Overall adaptability at any time The average overall fitness of the population. This is the step size control factor. For individuals in the first The fitness differences in each objective dimension and the global optimal solution in the th objective dimension A synthesis of fitness differences across all target dimensions; For the first Fitness values ​​for each objective dimension For the first The optimal solution for the region where each individual is located; The weight coefficient corresponding to the current global best position is returned to optimize the subjective weight and the objective weight respectively, and the optimal comprehensive weight is obtained, and the expression is as follows: ； wherein: is the optimal integrated weight, is the weight coefficient, is the subjective weight, is the objective weight.

9. The device renovation priority determination system based on the improved snow goose optimization algorithm according to claim 8, characterized in that: The device renovation priority determination unit (4) is configured to determine the renovation priority of the device according to the following method: The standardized index value is weighted based on the optimal comprehensive weight to construct a weighted normalized matrix, and the expression is as follows: ； ； in: For a weighted normalized matrix, For the weighted normalized matrix, the first... The device in the The value of each indicator, For the first The optimal comprehensive weight of each indicator; Based on the DBSCAN clustering algorithm, global density clustering is performed, and effective clusters are extracted, and the expression is as follows: ； wherein: is an effective cluster, is a first set; Computing cosine similarity of data points in an effective cluster and and converting to distance, expressed as follows: ； wherein: is a data point and a distance based on cosine similarity, , are data points and components in the th dimension, is the total dimension of the data point; Based on the cosine distance between data points, the average distance method is used to calculate the distance between sub-clusters, and the expression is as follows: ； wherein: is the average cosine distance of all pairs of points between the two sub-clusters, , are the two sub-clusters to be merged, , are the number of sample points in the clusters, respectively. Repeat the above steps to gradually merge similar clusters until the minimum distance between clusters exceeds the threshold value, and finally divide to obtain several sub-clusters, and the expression is as follows: ； wherein: is a threshold value, is a regulation coefficient, is the median of the cosine distances of all points in the current global cluster. After the merge is complete, each sub-cluster is obtained. The silhouette coefficient method was used to verify and correct the subclusters in AGNES clustering. The quality of the output is used to determine the final partitioning of each sub-cluster. ; For each sub-cluster obtained by final division , the centroid is calculated as the center point, and its expression is as follows: ； wherein: is the first subcluster, is the subcluster center point; Based on sub-cluster center points , constructing candidate point set whose expression is as follows: ； wherein: each sub-cluster center point corresponds to a multi-dimensional vector ; Determine the positive and negative ideal solutions of each index in the candidate point set: For the positive ideal solution, select the optimal value for each index in the candidate point set, and the expression is as follows: ； For the negative ideal solution, select the worst value for each index in the candidate point set, and the expression is as follows: ； Calculate the distance from each device to the positive and negative ideal solutions, and the expression is as follows: ； ； wherein: is the distance of the i-th device to the positive ideal solution, is the distance of the i-th device to the negative ideal solution. ​​​​ Based on the distance from each device to the positive and negative ideal solutions, calculate the closeness, and the expression is as follows: ； wherein: is the proximity of the first device to the second device; Traverse the closeness of all devices and sort all devices in descending order to obtain the priority of the old power grid production equipment.

10. The device renovation priority determination system based on the improved snow goose optimization algorithm according to claim 9, characterized in that: The equipment modification priority determination unit (4) verifies and corrects the sub-clusters in AGNES clustering according to the following steps. The quality specifically includes: For each data point , the average distance to all other points in the same cluster is calculated, which is expressed as follows: ； ； wherein: is the intra-cluster density of data points to other data points within the same cluster, is the intra-cluster density of data points is the number of points in the sub-cluster is the Euclidean distance of data points and in the sub-cluster and in the sub-cluster. For each data point , the average distance to all points in the nearest cluster is computed, which is expressed as follows: ； wherein: is the distance between data points to the nearest cluster the inter-cluster separation, is the distance to the nearest cluster from the data point, is the number of data points in the nearest cluster is the Euclidean distance between data points and any data point in the nearest cluster . Comparing intra-cluster compactness With inter-cluster separation Obtaining profile coefficients The expression of which is as follows: ； Based on the contour coefficient, calculate the average contour coefficient of all sub-clusters, and the expression is as follows: ； wherein: is the average silhouette coefficient for all sub-clusters, is the number of total data points, is the number of sub-clusters; all the calculated average profile coefficients of all sub-clusters are compared with the profile coefficient threshold value ; if , it is considered that the clustering effect is not good and the adjustment coefficient needs to be adjusted , a new sub-cluster is generated; if , the final divided each sub-cluster is output .

Images