Comprehensive benefit evaluation method and system for power distribution system

Abstract

The invention discloses a comprehensive benefit evaluation method and system for a power distribution system, and the method comprises the steps: firstly, collecting the multi-dimensional benefit index data of a power distribution system scheme, and constructing a contribution cloud picture through parameter inversion, so as to describe the fuzziness and uncertainty of the data; secondly, fusing subjective and objective weights by adopting a combined weighting method so as to balance expert experience and data laws; then, performing weighted fusion on the cloud model parameters of the plurality of indexes to quantify the overall benefit performance of the scheme; and finally, in combination with an improved grey clustering analysis method, introducing a Logistic curve to fit a benefit level boundary, in combination with a graph neural network, extracting a complex relation structure between the schemes, fusing a chaos theory to improve a fuzzy boundary identification capability, replacing a traditional whitening function, calculating a membership degree of each scheme to a benefit level, and completing sorting. Compared with a traditional method, the method is better in processing nonlinearity, uncertainty and dynamic information fusion, and the scientificity and decision credibility of comprehensive benefit evaluation of the power distribution system are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution system evaluation, and in particular to a method and system for evaluating the comprehensive benefits of a power distribution system. Background Art

[0002] Currently, in the actual evaluation process of distribution systems, efficiency indicators are derived from a wide range of sources and types, including both expert judgment and raw operational data from the acquisition system. Their inherent ambiguity and randomness are significant. This uncertainty poses a severe challenge to traditional evaluation models. Existing methods, such as the fuzzy comprehensive evaluation method, can capture ambiguity to a certain extent, but lack the ability to characterize random fluctuations, resulting in poor model stability and robustness. Weighting mechanisms based on entropy weights or the analytic hierarchy process are also prone to an imbalance between subjectivity and objectivity, making it difficult to fully reflect the actual importance of each efficiency indicator.

[0003] When it comes to assigning weights to multiple indicators, current methods generally adopt a single weighting strategy. Subjective methods, such as the Analytic Hierarchy Process (AHP), rely on expert scoring, which can lead to the accumulation of human bias and difficulty ensuring consistency. Purely objective methods, such as the entropy weight method or the gray entropy weight method, while based on data distribution patterns, often overlook actual business logic and the inherent connections between indicators, failing to reflect the guiding value of expert experience in different application scenarios. Therefore, finding a balanced mechanism that integrates weights between subjectivity and objectivity has become a key path to improving evaluation credibility.

[0004] On the other hand, the grey clustering method, as a common means of multi-level evaluation, is widely used in various evaluation scenarios of power systems because it is suitable for small samples and incomplete information systems. However, traditional grey clustering methods usually rely on linear whitening functions to calculate membership, which cannot accurately characterize the nonlinear transformation relationship between actual benefit levels, resulting in blurred cluster boundaries and insufficient level differentiation. This is especially true when dealing with data with drastic indicator fluctuations or obvious boundary effects. In addition, when dealing with multi-level indicator comprehensive problems, this method lacks detailed modeling of the interaction between indicator layers, which can easily cause information fusion distortion and affect the integrity and reliability of the evaluation results.

[0005] In this context, there is an urgent need for a method that can break through the bottlenecks of traditional models in data uncertainty processing, weight allocation and hierarchical clustering, which can not only reasonably integrate multi-source information, but also comprehensively reflect the system operation characteristics and adapt to the complex interactive relationship between different indicators. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a method and system for evaluating the comprehensive benefits of a power distribution system, which can achieve a more accurate, more robust and more decision-supporting evaluation of the comprehensive benefits of the power distribution system.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is: A comprehensive benefit evaluation method for a power distribution system solution comprises the following steps: Collect multiple benefit indicator data of each distribution system solution, normalize each benefit indicator data, and obtain multiple normalized indicator data; Calculating characteristic parameters of the cloud model corresponding to each benefit indicator data based on the plurality of normalized indicator data, and generating a contribution cloud graph of each benefit indicator data according to the characteristic parameters of the cloud model corresponding to each benefit indicator data; Based on the data of multiple benefit indicators of each distribution system scheme, the hierarchical analysis method is used to determine the subjective weight of each benefit indicator, and the gray entropy weight method is used to determine the objective weight of each benefit indicator. The subjective weight and objective weight of each benefit indicator are integrated to obtain the combined weight of each benefit indicator. According to the combined weights of each benefit indicator, the characteristic parameters of the cloud model corresponding to each benefit indicator are weighted and synthesized to generate a comprehensive cloud model of the distribution system solution; According to the comprehensive cloud model, multiple cloud droplet samples of various benefit indicator data are generated. Based on the multiple cloud droplet samples of the various benefit indicator data and the combined weights of the benefit indicators, a grayscale clustering analysis method that introduces logistic curves, graph neural networks, and chaos theory is combined to calculate the total membership of the distribution system solution to each gray class level, and the gray class level with the largest total membership is used as the comprehensive evaluation level of the distribution system solution.

[0008] In order to solve the above technical problems, another technical solution adopted by the present invention is: A comprehensive benefit evaluation system for a power distribution system comprises a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, each step of the above-mentioned comprehensive benefit evaluation method for a power distribution system is implemented.

[0009] The beneficial effects of the present invention are as follows: first, multi-dimensional benefit indicator data of distribution system solutions is collected, and a contribution cloud map is constructed through parameter inversion to characterize the fuzziness and uncertainty of the data; second, a combined weighting method is used to fuse subjective and objective weights to balance expert experience and data patterns; then, the cloud model parameters of multiple indicators are weighted and fused to quantify the overall benefit performance of the solution; finally, an improved gray clustering analysis method is combined with the introduction of logistic curve fitting benefit level boundaries, combined with graph neural networks to extract the complex relationship structure between solutions, and chaos theory is integrated to enhance the ability to identify fuzzy boundaries, replacing the traditional whitening function, calculating the membership of each solution to the benefit level and completing the ranking. Compared with traditional methods, it performs better in handling nonlinearity, uncertainty and dynamic information fusion, effectively improving the scientific nature and decision-making credibility of the comprehensive benefit evaluation of the distribution system. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 This is a flow chart of a method for evaluating comprehensive benefits of a power distribution system according to an embodiment of the present invention; Figure 2 Schematic diagram of a comprehensive benefit evaluation system for a power distribution system according to an embodiment of the present invention; Figure 3 This is a flowchart of an embodiment of the present invention.

[0011] Description of labels: 1. A comprehensive benefit evaluation system for a power distribution system; 2. Memory; 3. Processor. DETAILED DESCRIPTION

[0012] To illustrate the technical content, achieved objectives and effects of the present invention in detail, the following description is given in conjunction with the embodiments and accompanying drawings.

[0013] Please refer to Figure 1 The embodiment of the present invention provides a comprehensive benefit evaluation method for a power distribution system solution, comprising the steps of: Collect multiple benefit indicator data of each distribution system solution, normalize each benefit indicator data, and obtain multiple normalized indicator data; Calculating characteristic parameters of the cloud model corresponding to each benefit indicator data based on the plurality of normalized indicator data, and generating a contribution cloud graph of each benefit indicator data according to the characteristic parameters of the cloud model corresponding to each benefit indicator data; Based on the data of multiple benefit indicators of each distribution system scheme, the hierarchical analysis method is used to determine the subjective weight of each benefit indicator, and the gray entropy weight method is used to determine the objective weight of each benefit indicator. The subjective weight and objective weight of each benefit indicator are integrated to obtain the combined weight of each benefit indicator. According to the combined weights of each benefit indicator, the characteristic parameters of the cloud model corresponding to each benefit indicator are weighted and synthesized to generate a comprehensive cloud model of the distribution system solution; According to the comprehensive cloud model, multiple cloud droplet samples of various benefit indicator data are generated. Based on the multiple cloud droplet samples of the various benefit indicator data and the combined weights of the benefit indicators, a grayscale clustering analysis method that introduces logistic curves, graph neural networks, and chaos theory is combined to calculate the total membership of the distribution system solution to each gray class level, and the gray class level with the largest total membership is used as the comprehensive evaluation level of the distribution system solution.

[0014] From the above description, it can be seen that the beneficial effects of the present invention are as follows: first, the multi-dimensional benefit index data of the distribution system scheme is collected, and a contribution cloud map is constructed through parameter inversion to characterize the fuzziness and uncertainty of the data; second, a combined weighting method is used to fuse subjective and objective weights to balance expert experience and data patterns; then, the cloud model parameters of multiple indicators are weighted and fused to quantify the overall benefit performance of the scheme; finally, combined with an improved gray clustering analysis method, the logistic curve is introduced to fit the benefit level boundary, and the graph neural network is combined to extract the complex relationship structure between schemes. Chaos theory is integrated to improve the fuzzy boundary recognition ability, replacing the traditional whitening function, calculating the membership of each scheme to the benefit level and completing the ranking. Compared with traditional methods, it performs better in handling nonlinearity, uncertainty and dynamic information fusion, effectively improving the scientific nature and decision-making credibility of the comprehensive benefit evaluation of the distribution system.

[0015] Furthermore, the various benefit indicator data are normalized to obtain multiple normalized indicator data, including: Normalize each benefit indicator data in turn. When the benefit indicator is a positive indicator, the normalized indicator data x' is:

[0016] When the benefit indicator is a reverse indicator, the normalized indicator data x' is:

[0017] In the formula, x represents the benefit index data, x min Indicates the minimum value of the benefit indicator data, x max Indicates the maximum value of the benefit indicator data.

[0018] From the above description, it can be seen that by normalizing the data of various indicators, the interference of different dimensions on the comprehensive evaluation can be eliminated.

[0019] Furthermore, the characteristic parameters of the cloud model corresponding to each benefit indicator data are calculated based on the multiple normalized indicator data, including: The normalized indicator data of each distribution system solution are regarded as a sample data set, and the characteristic parameters of the cloud model corresponding to each normalized indicator are calculated according to the sample data set of each normalized indicator data. The characteristic parameters include expected value, entropy and super entropy.

[0020] Furthermore, a contribution cloud diagram of each benefit indicator data is generated based on the characteristic parameters of the cloud model corresponding to each benefit indicator data, including: Using a forward cloud generator, cloud droplets are generated according to characteristic parameters of the cloud model corresponding to each normalized indicator, and membership values ​​of the cloud droplets are calculated; The contribution cloud graph of each benefit indicator is formed based on all generated cloud droplets and their membership values.

[0021] From the above description, it can be seen that a large number of cloud droplets are generated by using the forward cloud generator to form a contribution cloud diagram of the indicator, so as to reflect the performance distribution characteristics and uncertainty of the distribution system solution under the indicator.

[0022] Furthermore, the analytic hierarchy process is used to determine the subjective weight of each benefit indicator, including: By constructing a pairwise comparison judgment matrix and performing a consistency test, the subjective weight of each benefit indicator is calculated using the hierarchical analysis method:

[0023] Where, Indicates the j The subjective weight of each benefit indicator, n represents the dimension of the judgment matrix, λ max Represents the maximum eigenvalue of the judgment matrix; The subjective weight of each benefit indicator is modified according to the input importance score of each benefit indicator.

[0024] From the above description, we can see that by combining the direct scores of the importance of each indicator by experts in multiple fields, the subjective weight calculation results of the hierarchical analysis method are revised, thereby improving the multi-dimensional rationality and representativeness of the subjective weight.

[0025] Furthermore, the gray entropy weight method is used to determine the objective weight of each benefit indicator, including: Calculate the grey correlation degree of the power distribution system scheme in terms of benefit indicators:

[0026]

[0027] Where, Indicates the i The power distribution system solution is in j The grey correlation degree of the indicators, ρrepresents the resolution coefficient, x ij Indicates the i The power distribution system solution is in j The gray value of an indicator, Indicates the j The optimal grayscale value of each indicator; The grey information entropy of each benefit indicator is calculated according to the grey correlation degree of each benefit indicator:

[0028]

[0029]

[0030] Where, p ij Indicates the i The power distribution system solution is in j The membership probability distribution of the indicators, n Indicates the total number of solutions, H j Indicates the j Grey information entropy of indicators; The objective weight of each benefit indicator is calculated according to its grey information entropy:

[0031] Where, Indicates the j The objective weight of each indicator, m Indicates the total number of indicators.

[0032] From the above description, we can see that the introduction of grey entropy weight method based on grey system theory for objective weighting can reveal the distribution characteristics and differences of various benefit indicator data.

[0033] Furthermore, the characteristic parameters of the cloud models corresponding to the benefit indicators are weighted and synthesized according to the combined weights of the benefit indicators to generate a comprehensive cloud model of the power distribution system solution, which also includes: Based on the comprehensive benefit evaluation objectives of the power distribution system solution, a multi-level evaluation indicator system is constructed using multiple benefit indicator data of the power distribution system solution. The cloud model of each level is generated based on the indicators under each level. For each level of cloud model, a reverse cloud generator is used to generate cloud droplet samples from multiple benefit indicator data of the distribution system solution, and characteristic parameters of the cloud model of each level are calculated based on the cloud droplet samples; Adaptively adjust the weights of indicators at each level based on the variance and entropy of indicator data at each level; According to the adjusted combined weights of each indicator, the characteristic parameters of the cloud model corresponding to each indicator are weighted and synthesized to obtain the synthetic expected value, synthetic entropy and synthetic super entropy; Based on the synthetic expectation value, the synthetic entropy and the synthetic super entropy, a multidimensional cloud droplet is generated in combination with a covariance matrix, and a comprehensive cloud model is obtained according to the multidimensional cloud droplet.

[0034] From the above description, it can be seen that a multi-level evaluation index system is constructed, and the indicators under each level are processed separately through the cloud model to generate their own cloud models. Therefore, the cloud model can accurately reflect the distribution pattern, uncertainty and volatility of the data; automatically adjust the weights when the data changes, and more flexibly and accurately reflect the true contribution of each level to the overall comprehensive benefits, thereby improving the model's adaptability and stability to external disturbances; through the multi-dimensional cloud reconstruction mechanism, a joint distributed comprehensive cloud model is constructed to facilitate the characterization of the interactive relationship between multiple levels.

[0035] Furthermore, multiple cloud droplet samples of various benefit indicator data are generated based on the comprehensive cloud model. Based on the multiple cloud droplet samples of various benefit indicator data and the combined weights of the benefit indicators, a grayscale clustering analysis method that introduces logistic curves, graph neural networks, and chaos theory is combined to calculate the total membership of the distribution system solution to each gray class level, including: Combining Logistic function and chaos theory to construct nonlinear multi-level grey class membership function for each benefit index data; generating a plurality of cloud droplet samples of various benefit indicator data according to the comprehensive cloud model, substituting the cloud droplet samples into the multi-level grey class membership function, and calculating the weight value of each benefit indicator at each grey class level; Based on the graph convolutional neural network, the total membership of the distribution system scheme to each gray class level is calculated by combining the weight value of each benefit indicator at each gray class level and the combined weight of each benefit indicator.

[0036] As can be seen from the above description, the introduction of graph neural networks as a modeling tool for the interdependencies between distribution schemes can capture the complex relationships between them, improving the accuracy and robustness of evaluation. Furthermore, by incorporating chaos theory into chaotic dynamics models such as the logistic map, it provides an effective modeling tool for the dynamic changes in distribution systems, improving the model's adaptability to uncertainty and complex environments, and ensuring efficient and stable evaluation even in the presence of large system fluctuations.

[0037] Furthermore, the Logistic function and chaos theory are combined to construct nonlinear multi-level gray class membership functions for each benefit indicator data, including: The grey clustering hierarchy structure is defined, chaos theory is used to adjust the parameters of the Logistic function mapping, and the adjusted Logistic function is used to establish a nonlinear membership function for each grey clustering level.

[0038] From the above description, we can see that the improved grey clustering analysis is based on the output of the cloud model. By introducing the Logistic curve to replace the traditional linear whitening function, it is more in line with the nonlinear growth trend of the actual data.

[0039] Please refer to Figure 2 Another embodiment of the present invention provides a comprehensive benefit evaluation system for a distribution system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, each step of the above-mentioned comprehensive benefit evaluation method for a distribution system is implemented.

[0040] The above-mentioned method and system for evaluating the comprehensive benefits of a power distribution system of the present invention are used to achieve a more accurate, robust, and decision-supporting evaluation of the comprehensive benefits of the power distribution system by introducing the "fuzzy-random-deterministic" unified expression mechanism of the cloud model, a combined weighting strategy integrating subjective and objective weights, and a clustering method that optimizes nonlinear hierarchical division. The following is an explanation of the specific implementation method: Please refer to Figure 1 and Figure 3 , embodiment 1 of the present invention is: A comprehensive benefit evaluation method for a power distribution system solution comprises the following steps: S1. Collect multiple benefit indicator data of each power distribution system solution, normalize each benefit indicator data, and obtain multiple normalized indicator data.

[0041] S11. Collect multi-dimensional benefit indicator data of each original power distribution system solution.

[0042] S12. Normalize the data of each indicator to eliminate the interference of different dimensions on the comprehensive evaluation. The normalization process is carried out in the following way: For positive indicators, the normalized indicator data x' is:

[0043] When the benefit indicator is a reverse indicator, the normalized indicator data x' is:

[0044] In the formula, x represents the benefit index data, x min Indicates the minimum value of the benefit indicator data, x max Indicates the maximum value of the benefit indicator data.

[0045] S2. Calculate characteristic parameters of the cloud model corresponding to each benefit indicator data based on the plurality of normalized indicator data, and generate a contribution cloud graph of each benefit indicator data according to the characteristic parameters of the cloud model corresponding to each benefit indicator data.

[0046] S21. Consider each normalized indicator data of each distribution system solution as a sample data set, and calculate the characteristic parameters of the cloud model corresponding to each normalized indicator based on the sample data set of each normalized indicator data, wherein the characteristic parameters include expected value, entropy and super entropy.

[0047] Specifically, the normalized indicator data of each distribution system solution is regarded as a sample data set { x 1, x 2, …, x n}, the three characteristic parameters of the cloud model are calculated based on the sample data set as follows: (1) Expected value Ex , represents the mean value of the indicator, reflecting the typical performance of the solution on this indicator. The calculation formula is as follows:

[0048] Where, x i Indicates the i Sample data, n Indicates the sample size.

[0049] (2) Entropy En , which indicates the degree of discreteness of the data and reflects the uncertainty and ambiguity of the indicator. The calculation formula is:

[0050] (3) Super entropy He , which is used to measure the degree of entropy fluctuation, that is, the randomness of the data. It can be estimated accurately or simplified: Accurate estimation method:

[0051] Simplified estimation method:

[0052] S22. Use a forward cloud generator to generate cloud droplets according to the characteristic parameters of the cloud model corresponding to each normalized indicator, calculate the membership value of the cloud droplets, and form a contribution cloud map of each benefit indicator based on all generated cloud droplets and their membership values.

[0053] Specifically, using the forward cloud generator, the parameters obtained in step S21 ( Ex , En , He) is used as input to generate multiple cloud droplets. The cloud droplet generation process includes the following sub-steps: (1) Randomly generate an entropy value from the normal distribution En ', so that it satisfies En '~ N ( En , He 2 ); (2) Ex is the mean, En ' is the standard deviation, generating cloud droplet position yx ,Right now yx ~ N ( Ex , ( En ') 2 ); (3) Calculate the membership value (i.e., certainty) of the cloud droplet:

[0054] Where, μ ( yx )∈[0, 1], which can represent the value yx Corresponding to the typicality of qualitative concepts.

[0055] By repeating the above process, a large number of cloud droplets are generated to form the "contribution score cloud" of the indicator, which is used to reflect the performance distribution characteristics and uncertainty of the solution under the indicator.

[0056] S3. Based on the multiple benefit indicator data of each distribution system scheme, the hierarchical analysis method is used to determine the subjective weight of each benefit indicator, and the gray entropy weight method is used to determine the objective weight of each benefit indicator. The subjective weight and objective weight of each benefit indicator are integrated to obtain the combined weight of each benefit indicator.

[0057] Among them, the combined weighting method integrates subjective and objective weights to balance human experience and the inherent laws of data by integrating the analytic hierarchy process (AHP), expert evaluation and gray entropy weight method. Subjective weights reflect the expert's judgment on the importance of indicators, and objective weights use the entropy weight method to explore the correlation between data, and finally generate combined weights through a normalized formula. This method avoids the one-sidedness of single weighting and ensures that the weights can reflect professional experience and conform to the actual data distribution. The combined weights directly affect the calculation of the synthetic cloud and are the key bridge connecting cloud model parameters and comprehensive evaluation results. It provides a scientific basis for weight allocation for the subsequent integration of multi-level indicators. The specific steps are as follows: S31. Use the analytic hierarchy process to determine the subjective weight of each benefit indicator.

[0058] Specifically, an indicator hierarchical model is first constructed. According to the actual needs of the comprehensive benefit evaluation of the distribution system, a three-layer structural model including the target layer, the criterion layer and the indicator layer is constructed. This model is used to systematically summarize and organize the evaluation factors to ensure the logical consistency of subjective judgments. After discussion by experts, each evaluation indicator is classified into the corresponding criterion layer to form a clear evaluation system structure.

[0059] Then the analytic hierarchy process (AHP) is used to determine the relative importance of the indicators: By constructing a pairwise comparison judgment matrix and performing a consistency check, the relative weight coefficients between indicators are obtained. This process is usually performed by experts based on their experience and judgment. Each indicator at the same level is compared to the higher-level goal or criterion. Scoring is performed using the Saaty 1–9 scale and entered into an n×n matrix with diagonal 1s and pairwise reciprocal relationships. This matrix measures whether the i-th indicator is more important than the j-th indicator. The AHP weight calculation process is as follows:

[0060] Where, Indicates the j The subjective weight of each benefit indicator, M represents the dimension of the judgment matrix, λ max Represents the maximum eigenvalue of the judgment matrix.

[0061] After consistency testing, the obtained subjective weights can be used as a quantitative reflection of expert knowledge and subjective judgment.

[0062] Afterwards, combined with the direct scores of the importance of each indicator by experts in multiple fields, the AHP weight calculation results are corrected using fuzzy average or weighted average methods, thereby improving the multivariate rationality and representativeness of the subjective weights.

[0063] S32. Use the gray entropy weight method to determine the objective weight of each benefit indicator.

[0064] Specifically, the original data matrices of multiple benefit indicators of each distribution system scheme are first normalized according to the attributes of positive, negative or neutral indicators to ensure that each benefit indicator is comparable and unified in the interval [0, 1].

[0065] In order to reveal the distribution characteristics and differences of each benefit indicator data, the grey entropy weight method is introduced based on the grey system theory to perform objective weighting. The grey correlation calculation formula is as follows:

[0066]

[0067] Where, Indicates the i The power distribution system solution is in j Grey correlation degree of each indicator; ρ It represents the resolution coefficient, which is usually set to 0.5; x ij Indicates the i The power distribution system solution is in j Gray value of each indicator; Indicates the j The optimal gray value of an indicator, is the absolute difference between the reference sequence and the evaluation sequence; Then, the grey correlation matrix is ​​normalized and entropy value analysis is performed to obtain the grey information entropy of each indicator:

[0068]

[0069]

[0070] Where, p ij Indicates the i The power distribution system solution is in j The membership probability distribution of each indicator; n Indicates the total number of solutions; H j Indicates the j The grey information entropy of the indicators, H j The smaller the value, the greater the information content of the indicator, the greater the variability, and the higher the objective weight should be given; Through normalization processing, the objective weight of each benefit indicator is calculated according to the grey information entropy of each benefit indicator:

[0071] Where, Indicates the j The objective weight of each indicator, m Indicates the total number of indicators.

[0072] S33. The subjective weight of each indicator and the objective weight of each indicator are integrated to obtain the combined weight of each indicator.

[0073] In order to achieve the organic integration of subjective and objective information, the linear weighting method is used to and objective weight Combine to get the final comprehensive weight W j , the specific formula is as follows:

[0074] Where, W j Indicates the j The combined weight of the evaluation indicators; α It represents the adjustment coefficient of subjective weight and objective weight, and its value range is between [0, 1]. It can be assigned a fixed value or determined by adaptive strategy optimization according to the actual scenario. α =0.5, indicating that subjective and objective weights are equally important. If you want to rely more on expert knowledge, you can increase α If you trust the data more, adjust it down. α value.

[0075] The combined weighting method can effectively take into account the subjective judgment of expert knowledge and the objective laws of actual data, overcome the one-sidedness brought by a single weighting method, improve the scientificity, rationality and stability of indicator weight distribution, and enhance the robustness and adaptability of the comprehensive benefit evaluation model of the distribution system.

[0076] S4. Perform weighted synthesis on the characteristic parameters of the cloud model corresponding to each benefit indicator according to the combined weight of each benefit indicator to generate a comprehensive cloud model of the distribution system solution.

[0077] Specifically, the characteristic parameters of the cloud model of each benefit indicator are weighted and synthesized to form a "synthetic cloud" for each distribution system solution.

[0078] The synthesis formula is as follows: (1) Synthetic expected value:

[0079] (2) Synthetic entropy:

[0080] (3) Synthetic super entropy:

[0081] Where, m Indicates the total number of indicators, w j Indicates the j The weight of the indicator, Ex j 、 En j 、 He j are the cloud model parameters corresponding to the indicator respectively.

[0082] S5. Generate multiple cloud droplet samples of various benefit indicator data based on the comprehensive cloud model. Calculate the total membership of the distribution system solution to each gray class level based on the multiple cloud droplet samples of the various benefit indicator data and the combined weights of each benefit indicator, combined with the grayscale clustering analysis method that introduces the Logistic curve, graph neural network, and chaos theory. Take the gray class level with the largest total membership as the comprehensive evaluation level of the distribution system solution.

[0083] Among them, the improved gray cluster analysis, based on the output of the cloud model, introduces a logistic curve instead of the traditional linear whitening function, better aligning with the nonlinear growth trends of actual data. This method combines the Ex, En, and He parameters generated by the cloud model to construct a multi-level gray membership function (such as excellent, good, fair, and poor). It then calculates the clustering coefficient for each development plan and ultimately prioritizes the plan based on the maximum membership. This improvement addresses the traditional method's lack of adaptability to nonlinear relationships and enhances classification accuracy. It integrates the data characteristics of the cloud model in a logical chain, transforming abstract cloud parameters into concrete classification results, which serve as the direct basis for plan ranking.

[0084] This embodiment introduces a graph neural network (GNN) as a modeling tool for the interdependence between distribution schemes. Through graph convolution operations, GNN can capture the complex correlations between distribution schemes and improve the accuracy and robustness of the evaluation. Unlike independent evaluations in traditional methods, GNN considers the mutual influence between schemes by processing graph structure data between distribution schemes, making the evaluation results more comprehensive and accurate. At the same time, this embodiment considers introducing chaos theory by introducing chaotic dynamic models such as Logistic mapping to provide an effective modeling method for the dynamic changes of the distribution system. Chaos theory can simulate highly nonlinear and complex dynamic changes in the distribution system, improve the model's adaptability to uncertainty and complex environments, and ensure that efficient and stable evaluation can be maintained even when the system fluctuates greatly. The specific steps are as follows: S51. Define the grey clustering level structure, and use the Logistic function to establish a nonlinear membership function for each grey clustering level.

[0085] Specifically, a grey clustering hierarchy is defined, dividing the evaluation levels into five categories: extremely poor, poor, fair, good, and excellent, representing the different quality levels of the power distribution scheme in the comprehensive benefit evaluation. Each level corresponds to a grey class, which is used to express the affiliation of the indicator values ​​within that level.

[0086] In order to improve the nonlinear expression ability of cluster evaluation, the traditional linear gray whitening function is replaced by the Logistic curve function. Its function form has an "S" shape characteristic, which can more realistically reflect the gradual change trend of indicator values ​​and the fuzzy boundary state in reality. Taking the "excellent" class as an example, its whitening function is defined as follows:

[0087] Where, μ s ( x ) indicates an indicator x Belong to s The degree of membership of the class ranges from [0, 1]. x ' represents the normalized index, x 0 represents the center position of membership change, k represents the slope coefficient of the curve; Where, Indicates the indicator value x The membership degree belongs to the “excellent” class, and its value range is [0, 1]; x 'Indicates the normalized index value; x 0 represents the center position of the membership change (i.e., the inflection point), which represents the typical value of the indicator value approaching "excellent"; k Represents the curve slope coefficient, which is used to control the steepness of the function.

[0088] Similarly, for the "extreme difference" class, the function form is reversed, that is:

[0089] This function is used to represent the affiliation when the indicator value tends to be worse. Its meaning and symbol are consistent with the above formula. Other intermediate grades (poor, medium, good) are also modeled using the Logistic function. The form is the same but the parameters are x 0 and k Different to achieve smoother level transition and interval control.

[0090] S52. Generate multiple cloud droplet samples of various benefit indicator data according to the comprehensive cloud model, substitute the cloud droplet samples into the multi-level gray class membership function, and calculate the weight value of each benefit indicator at each gray class level.

[0091] Specifically, according to the cloud model parameters (i.e. expected values) corresponding to each solution under each indicator Ex ,entropy En and super entropy He ), using the inverse cloud generator to generate a large number of cloud droplets x i , each cloud droplet is treated as a simulation sample, which contains the uncertainty characteristics of the indicator.

[0092] Substitute the cloud droplet samples into the Logistic function to calculate their membership in each level. Specifically, each cloud droplet sample generated x i Substitute into the five types of whitening functions in turn to obtain their membership at each levelμ k ( x i ), the average weight of the index under a certain level can be obtained by averaging the membership of all cloud droplets. The calculation formula is as follows:

[0093] Where, Indicates that an evaluation index is k The average membership degree under the gray level, N is the total number of cloud droplets generated, μ k ( x i ) indicates the i A cloud droplet k The average membership of all levels is combined to form the gray class clustering weight vector of the indicator.

[0094] S53. Calculate the clustering coefficient matrix Integrate the gray class membership of all indicators and construct the clustering coefficient matrix of the scheme. i The plan in j On the first indicator k The membership degree of the gray class is r ijk , then the clustering coefficient matrix of the scheme is R i It can be expressed as:

[0095] Where, R i For the i The plan in m Cluster membership structure under each indicator; r ijk For the j The indicator in k The weight value under the gray class level.

[0096] Combined with the indicator weights, the comprehensive cluster membership of the scheme to each gray class level is calculated. The cluster membership of each scheme to a certain gray class level is the weighted sum of the membership of each indicator. The specific formula is as follows:

[0097] Where, S ik Indicates the i The plan for k The total membership of the gray class level, w j For the jThe weight of each benefit indicator, m is the total number of indicators.

[0098] S54. Introducing Graph Neural Networks (GNNs) and Chaos Theory In order to improve the accuracy and robustness of the comprehensive benefit evaluation of the distribution system, considering the complex dependencies and nonlinear characteristics between the distribution schemes in the system, graph neural networks (GNN) and chaos theory were introduced into this framework to further optimize the scheme evaluation process.

[0099] Each solution in the power distribution system can be viewed as a node in a graph, and the relationships between solutions (such as similarity and interdependence) are represented by edges. Therefore, the solutions in the power distribution system form a complex graph structure. The nodes in the graph represent the individual distribution solutions, and the edges indicate the similarities or connections between the solutions. The edge weights reflect the strength of these similarities or dependencies.

[0100] By constructing such a graph structure, graph neural networks can capture the complex interactions and potential correlations between schemes, which helps to consider the nonlinear relationship between distribution schemes in the comprehensive benefit evaluation process.

[0101] Graph neural networks can effectively process dependencies between nodes (distribution solutions) through graph convolution operations. During solution evaluation, each solution not only relies on its own indicator weight but is also influenced by its similarity to other solutions. The basic formula for graph convolution is as follows:

[0102] Where, For the k Node feature representation after layer convolution, N( i ) is a node i The set of adjacent nodes of c ij is the normalized coefficient of the adjacent nodes, W (k) and b (k) They are k The layer's weight matrix and bias terms, σ is the activation function.

[0103] Through graph convolution operations, distribution plans can update their evaluation results based on information from their neighboring nodes, better accounting for the mutual influence between different plans. In this way, GNN can help optimize the ranking of distribution plans, making the final comprehensive benefit evaluation results more accurate.

[0104] The operation of power distribution systems is influenced by a variety of factors, including emergencies, environmental changes, and internal system uncertainties. These factors often cause the system to exhibit strong nonlinear dynamic characteristics. To model these complex nonlinear changes, chaos theory has been introduced into the framework.

[0105] The logistic map in chaos theory is a classic nonlinear dynamic model that can simulate dramatic fluctuations caused by small changes in a system. In scheme evaluation, by adjusting the parameters of the logistic map, the nonlinear perturbations of different schemes in the distribution system can be simulated. The equation is as follows: x n+1 = rx n (1- x n ) Where, x n Indicates the n The state value of the iteration, r is the control parameter, usually 3.57≤ r When ≤ 4, the system will exhibit chaotic behavior. By introducing the disturbance term of chaos theory, the evaluation results of the power distribution scheme can be dynamically adjusted, thereby improving the robustness of the evaluation system in dealing with complex and uncertain environments.

[0106] S55. Determine comprehensive membership and program ranking The gray class level with the largest membership degree in each scheme is selected as the final classification result: according to the membership degree of each scheme under the five categories, the level corresponding to the maximum value is extracted as the comprehensive evaluation level of the scheme.

[0107] When multiple solutions are classified into the same level, they can be classified according to their maximum membership value. S ik The numerical values ​​of are further sorted, and the higher the membership degree, the better the scheme is, and finally a complete ranking result of the comprehensive benefits of the distribution system is formed.

[0108] By incorporating graph neural networks (GNNs) and chaos theory, the evaluation process can fully account for the complex interrelationships between distribution schemes and the dynamics of the system. Compared to traditional methods, the combination of GNNs and chaos theory with gray clustering analysis achieves breakthroughs and improvements in several key technical aspects. Traditional gray clustering methods typically rely on linear whitening functions and static indicator inputs, ignoring the underlying structural connections between schemes and the dynamic changes during system evolution. This leads to problems such as blurred cluster boundaries and unstable ranking results when dealing with high-dimensional, multi-coupled, and highly uncertain data. In contrast, GNNs effectively extract high-order dependency features by modeling the graph structural relationships between schemes, addressing the shortcomings of traditional methods in relational modeling. Chaos theory, through nonlinear mapping (such as the logistic mapping), simulates the system response of schemes under different initial conditions, enhancing the model's ability to perceive dynamic evolution trends. When these two methods are combined in gray clustering analysis, they not only enrich the representation dimensionality of input features but also introduce a dynamic perturbation control mechanism for the clustering process, significantly improving the sensitivity of cluster boundary identification and the robustness of scheme ranking. Furthermore, leveraging the cloud model's strengths in handling ambiguity and randomness, this method maintains high assessment accuracy and reliability even under uncertain data conditions. Overall, this method outperforms traditional methods in structural modeling, dynamic adaptability, and uncertainty handling, making it more suitable for the complex demands of modern intelligent distribution systems.

[0109] By introducing the Logistic curve to replace the linear whitening function in traditional grey cluster analysis and combining it with the cloud model's ability to process uncertain data, the sensitivity of cluster boundary identification and the credibility of comprehensive evaluation results are significantly improved under the condition of coexistence of fuzziness and randomness. It is particularly suitable for the comprehensive benefit evaluation of distribution system solutions and multi-solution ranking decision-making.

[0110] The second embodiment of the present invention is: On the basis of the first embodiment, the generation process of the comprehensive cloud model in step S4 is further defined, specifically: S41. Based on the comprehensive benefit evaluation target of the power distribution system solution, a multi-level evaluation index system is constructed using multiple benefit index data of the power distribution system solution, and a cloud model of each level is generated based on the indicators under each level.

[0111] Specifically, a multi-level evaluation indicator system is constructed: First, based on the target requirements of the comprehensive benefit evaluation of the distribution system, a multi-level evaluation indicator system is defined and constructed. This system usually includes multiple levels, such as technical indicators, economic indicators, environmental indicators, social benefit indicators, etc. After the acquisition in step S1, each level contains several specific benefit indicators, which facilitates the comprehensive and systematic evaluation of the distribution plan. In practical applications, indicators at different levels have different weights and influences. The technical indicator layer mainly focuses on the technical performance and reliability of the system, the economic indicator layer focuses on cost-effectiveness and return on investment, and the environmental indicator layer considers environmental factors such as sustainability and carbon emissions. Therefore, the indicators at each level will have different roles and contributions in the synthetic cloud.

[0112] The indicators at each level are processed separately through the cloud model to generate their own cloud model. The cloud model can accurately reflect the distribution pattern, uncertainty and volatility of data, and is an important tool in comprehensive benefit evaluation. The cloud model at each level will contain three basic parameters: expected value Ex ,entropy En and super entropy He , these parameters reflect the core characteristics, uncertainty measures, and volatility of the data at this level.

[0113] S42. For each level of the cloud model, a reverse cloud generator is used to generate cloud droplet samples from the benefit indicators, and characteristic parameters of the cloud model at each level are calculated based on the cloud droplet samples.

[0114] Specifically, for each layer, a reverse cloud generator is used to generate a large number of cloud droplet samples from the benefit indicator data collected in step S1. The generated cloud droplets can truly reflect the distribution characteristics of the data. These cloud droplet samples not only retain the expected value information of the data, but also reflect the ambiguity and randomness of the data.

[0115] For each cloud model level, the three basic parameters of the cloud model at that level are calculated through statistical analysis of cloud droplet samples: expected value Ex k , entropy En k and super entropy He k Expected value Ex k Indicates the central tendency or average level of the data at this level, entropy En k Indicates the uncertainty of data, super entropy He k It reflects the degree of uncertainty of entropy. The larger it is, the more significant the volatility of the data at that level is.

[0116] The processed cloud model parameters will be used for subsequent weighted synthesis to provide basic data for the generation of synthetic clouds.

[0117] S43. Adaptively adjust the weights of indicators at each level based on the variance and entropy of indicator data at each level.

[0118] Specifically, to reduce subjectivity in weight setting and enhance the synthetic cloud's responsiveness to real-world data characteristics, this embodiment introduces a "hierarchical adaptive dynamic weight adjustment mechanism." Although some models in the field of multi-level data processing and intelligent decision-making have attempted to incorporate dynamic weight mechanisms to improve system adaptability, within the cloud model framework, weight setting methods are still primarily static configurations or based on a single metric (such as entropy). This is partly due to the cloud model's inherent design philosophy of "uncertainty modeling" and its emphasis on the integration of qualitative and quantitative methods. This has led existing research to favor static weight allocation strategies with simple structures and stable computations. In contrast, the "hierarchical adaptive dynamic weight adjustment mechanism" proposed in this embodiment, while maintaining the cloud model's flexible expression advantages, incorporates synergistic information between data variance and entropy. By adjusting parameters, it achieves a dynamic balance between structural volatility and the impact of uncertainty, significantly enhancing the model's adaptability and robustness to complex data structures.

[0119] This mechanism adaptively adjusts the weights based on the variance and entropy information of each level of data. The specific expression is as follows:

[0120]

[0121] Where, Indicates the k The comprehensive adjustment value of each indicator is determined by its data volatility (variance) and uncertainty (entropy) in each scheme; Indicates the j The adjustment value corresponding to each indicator is used for normalization calculation; Indicates the k The data vector of the benefit index under all distribution system schemes, Var( X k ) represents its variance; represents the adjusted weight, En k Its entropy value, α and β It is an adjustment parameter used to balance the impact of structural volatility and uncertainty on weights.

[0122] Compared with static weighting, this mechanism can automatically adjust weights when data changes, more flexibly and accurately reflecting the true contribution of each level to the overall comprehensive benefits, and improving the model's adaptability and stability to external disturbances.

[0123] S44. Perform weighted synthesis on the characteristic parameters of the cloud model corresponding to each indicator according to the adjusted combined weights of each indicator to obtain a synthetic expected value, a synthetic entropy, and a synthetic super entropy.

[0124] Specifically, based on the dynamically adjusted weight coefficient , the parameters of the comprehensive cloud model are obtained through weighted calculation:

[0125]

[0126]

[0127] Where, g represents the total number of layers. This formula ensures that the proportion of each layer parameter in the synthetic cloud matches its real-time status.

[0128] S45. Based on the synthetic expectation value, the synthetic entropy and the synthetic super entropy, a multidimensional cloud droplet is generated in combination with a covariance matrix, and a comprehensive cloud model is obtained according to the multidimensional cloud droplet.

[0129] Specifically, using weighted synthesis Ex 总 、 En 总 、 He 总 , synthetic cloud droplet samples are generated through the forward cloud generator.

[0130] Traditional methods usually generate a one-dimensional cloud droplet sequence, which makes it difficult to depict the interactive relationship between multiple levels. To this end, this embodiment introduces a "multi-dimensional cloud reconstruction mechanism" to build a joint distributed multi-dimensional cloud model.

[0131] Although modeling the synergistic relationships between variables is relatively mature in modeling frameworks such as multivariate statistical analysis, fuzzy mathematics, and Bayesian networks, with methods such as covariance modeling, conditional dependency structures, and joint probability modeling widely adopted, cloud models have distinct theoretical starting points and application objectives, resulting in inherent limitations in handling multivariate coupling relationships. Specifically, cloud models emphasize the unified expression of "fuzziness" and "randomness," and their core mechanism relies on a cloud droplet generation rule composed of three parameters: expectation, entropy, and superentropy. This mechanism was originally designed for modeling single-variable uncertainty and is highly intuitive and interpretable. However, when the model is extended to multiple dimensions, the original forward cloud generator does not consider the synergistic structure between dimensions. As a result, cloud droplets in each dimension are often generated separately and then spliced ​​together to form multidimensional cloud droplets. This lacks a unified joint distribution foundation and cannot naturally embed the synergistic characteristics between variables.

[0132] Furthermore, current cloud model research focuses primarily on preserving marginal features and qualitative identification, lacking a mechanism or standardized process for unified modeling of statistical correlations and structural uncertainty. Without a universal formula or supporting tools, implementing cloud droplet generation based on joint distributions is complex, making it difficult to develop a widely accepted or standardized methodology. Consequently, despite extensive research in other fields on variable collaborative modeling, a universally recognized, effective, and clearly structured multivariate coupling modeling mechanism has yet to be established in cloud models.

[0133] Therefore, this embodiment expands the ability of cloud models in high-dimensional uncertainty modeling by introducing the covariance matrix and multidimensional normal distribution mechanism without violating the basic idea of ​​the cloud model, filling the gap in the current cloud model in variable coupling modeling.

[0134] When generating cloud droplets, the model samples multivariate cloud droplets from a high-dimensional normal distribution based on the covariance matrix Σ:

[0135] Where, Indicates each level Ex The vector of parameters, Σ, reflects the uncertainty synergy between each level.

[0136] The generation of multidimensional cloud droplets enables the model to characterize the coupled fluctuation characteristics between indicators, introduce more variable correlation information for subsequent grey clustering analysis, and improve the discrimination and robustness of the scheme.

[0137] In this embodiment, the synthetic cloud can also be used to perform comprehensive benefit evaluation, specifically: (1) Calculation of comprehensive benefit score Each distribution system solution corresponds to a set of synthetic cloud droplet samples. By evaluating the solution's membership in the synthetic cloud (i.e., the similarity of its cloud droplets to the ideal cloud center), its comprehensive benefit score is calculated. A higher score indicates a solution that is closer to the optimal synthetic cloud morphology and exhibits better comprehensive benefit performance.

[0138] (2) Scheme ranking and decision-making Sorting the options from high to low scores, selecting the optimal option with the highest score. If multiple high-quality options exist, further adjustments can be made based on other constraints to achieve multi-objective, differentiated decision support.

[0139] Therefore, by introducing the "hierarchical adaptive dynamic weight mechanism" and the "multi-dimensional cloud reconstruction mechanism", the synthetic cloud method is significantly enhanced in handling uncertainty and multi-source coupling information, which not only improves the flexibility, scientificity and adaptability of the model, but also provides a more robust and reliable decision-making basis for the comprehensive benefit analysis of the distribution system.

[0140] For multiple sets of distribution system operating indicators and data, the indicator attributes for each scenario are converted into cloud model eigenvalues ​​(expected value Ex, entropy En, and excess entropy He). By converting various distribution system operating indicators into cloud model parameters, the ambiguity and uncertainty of the performance under each scenario can be effectively characterized. The standard estimate cloud and the comprehensive score cloud for each scenario are clearly distinguishable in terms of morphology, demonstrating the method's sensitive response to the comprehensive benefits of different scenarios. The overall evaluation results are highly discriminative and stable, verifying the effectiveness of the proposed method in multi-source data fusion and classification.

[0141] Please refer to Figure 2 , the third embodiment of the present invention is: A distribution system comprehensive benefit evaluation system 1 includes a memory 2, a processor 3, and a computer program stored in the memory 2 and executable on the processor 3. When the processor 3 executes the computer program, each step of a distribution system comprehensive benefit evaluation method of embodiment 1 or embodiment 2 is implemented.

[0142] In summary, the present invention provides a comprehensive benefit evaluation method and system for distribution systems, aiming to improve the accuracy of benefit evaluation and intelligent decision-making capabilities in a multi-source, heterogeneous data environment. The method comprises the following steps: first, collecting multidimensional benefit indicators for distribution system solutions, standardizing the raw data using a cloud model, and constructing a contribution cloud map through parameter inversion to characterize the data's fuzziness and uncertainty. Second, a combined weighting method (analytic hierarchy process and grey entropy weighting) is employed to integrate subjective and objective weights, balancing expert experience with data patterns. Third, a multidimensional cloud synthesis method is employed to weightedly integrate the cloud model parameters for multiple indicators to construct a comprehensive cloud model and quantify the overall benefit performance of the solutions. Finally, an improved grey clustering analysis method is employed, along with the introduction of a logistic curve to fit benefit level boundaries. A graph neural network is then used to extract the complex relationship structure between solutions. Chaos theory is then integrated to enhance the ability to identify fuzzy boundaries, replacing the traditional whitening function. The membership of each solution to the benefit level is calculated and ranked. Compared to traditional methods, the present invention excels in handling nonlinearity, uncertainty, and dynamic information fusion, effectively enhancing the scientific nature and decision-making credibility of comprehensive benefit evaluation for distribution systems.

[0143] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent transformations made using the contents of the present invention's description and drawings, or directly or indirectly applied in related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A comprehensive benefit evaluation method for a power distribution system solution, characterized in that: Including steps: Collect multiple benefit indicator data of each distribution system solution, normalize each benefit indicator data, and obtain multiple normalized indicator data; Calculating characteristic parameters of the cloud model corresponding to each benefit indicator data based on the plurality of normalized indicator data, and generating a contribution cloud graph of each benefit indicator data according to the characteristic parameters of the cloud model corresponding to each benefit indicator data; Based on the data of multiple benefit indicators of each distribution system scheme, the hierarchical analysis method is used to determine the subjective weight of each benefit indicator, and the gray entropy weight method is used to determine the objective weight of each benefit indicator. The subjective weight and objective weight of each benefit indicator are integrated to obtain the combined weight of each benefit indicator. According to the combined weights of each benefit indicator, the characteristic parameters of the cloud model corresponding to each benefit indicator are weighted and synthesized to generate a comprehensive cloud model of the distribution system solution; According to the comprehensive cloud model, multiple cloud droplet samples of various benefit indicator data are generated. Based on the multiple cloud droplet samples of the various benefit indicator data and the combined weights of the benefit indicators, a grayscale clustering analysis method that introduces logistic curves, graph neural networks, and chaos theory is combined to calculate the total membership of the distribution system solution to each gray class level, and the gray class level with the largest total membership is used as the comprehensive evaluation level of the distribution system solution.

2. The comprehensive benefit evaluation method of a power distribution system solution according to claim 1, characterized in that: The various benefit indicator data are normalized to obtain multiple normalized indicator data, including: Normalize each benefit indicator data in turn. When the benefit indicator is a positive indicator, the normalized indicator data x' is: When the benefit indicator is a reverse indicator, the normalized indicator data x' is: In the formula, x represents the benefit index data, x min Indicates the minimum value of the benefit indicator data, x max Indicates the maximum value of the benefit indicator data.

3. The comprehensive benefit evaluation method of a power distribution system solution according to claim 1, characterized in that: Calculating characteristic parameters of the cloud model corresponding to each benefit indicator data based on the multiple normalized indicator data includes: The normalized indicator data of each distribution system solution are regarded as a sample data set, and the characteristic parameters of the cloud model corresponding to each normalized indicator are calculated according to the sample data set of each normalized indicator data. The characteristic parameters include expected value, entropy and super entropy.

4. The comprehensive benefit evaluation method of a power distribution system solution according to claim 1, characterized in that: Generate contribution cloud graphs of various benefit indicator data based on the characteristic parameters of the cloud model corresponding to each benefit indicator data, including: Using a forward cloud generator, cloud droplets are generated according to characteristic parameters of the cloud model corresponding to each normalized indicator, and membership values ​​of the cloud droplets are calculated; The contribution cloud graph of each benefit indicator is formed based on all generated cloud droplets and their membership values.

5. The comprehensive benefit evaluation method of a power distribution system solution according to claim 3, characterized in that: The analytic hierarchy process is used to determine the subjective weight of each benefit indicator, including: By constructing a pairwise comparison judgment matrix and performing a consistency test, the subjective weight of each benefit indicator is calculated using the hierarchical analysis method: Where, Indicates the j The subjective weight of each benefit indicator, n represents the dimension of the judgment matrix, λ max Represents the maximum eigenvalue of the judgment matrix; The subjective weight of each benefit indicator is modified according to the input importance score of each benefit indicator.

6. The comprehensive benefit evaluation method of a power distribution system solution according to claim 3, characterized in that: The gray entropy weight method is used to determine the objective weight of each benefit indicator, including: Calculate the grey correlation degree of the power distribution system scheme in terms of benefit indicators: Where, Indicates the i The power distribution system solution is in j The grey correlation degree of the indicators, ρ represents the resolution coefficient, x ij Indicates the i The power distribution system solution is in j The gray value of an indicator, Indicates the j The optimal gray value of each indicator; The grey information entropy of each benefit indicator is calculated based on the grey correlation degree of each benefit indicator: Where, p ij Indicates the i The power distribution system solution is in j The membership probability distribution of the indicators, n Indicates the total number of solutions, H j Indicates the j Grey information entropy of indicators; The objective weight of each benefit indicator is calculated according to its grey information entropy: Where, Indicates the j The objective weight of each indicator, m Indicates the total number of indicators.

7. The comprehensive benefit evaluation method for a power distribution system solution according to claim 3, characterized in that: The characteristic parameters of the cloud models corresponding to the benefit indicators are weighted and synthesized according to the combined weights of the benefit indicators to generate a comprehensive cloud model of the power distribution system solution, which also includes: Based on the comprehensive benefit evaluation objectives of the power distribution system solution, a multi-level evaluation indicator system is constructed using multiple benefit indicator data of the power distribution system solution. The cloud model of each level is generated based on the indicators under each level. For each level of cloud model, a reverse cloud generator is used to generate cloud droplet samples from multiple benefit indicator data of the distribution system solution, and characteristic parameters of the cloud model of each level are calculated based on the cloud droplet samples; Adaptively adjust the weights of indicators at each level based on the variance and entropy of indicator data at each level; According to the adjusted combined weights of each indicator, the characteristic parameters of the cloud model corresponding to each indicator are weighted and synthesized to obtain the synthetic expected value, synthetic entropy and synthetic super entropy; Based on the synthetic expectation value, the synthetic entropy and the synthetic super entropy, a multidimensional cloud droplet is generated in combination with a covariance matrix, and a comprehensive cloud model is obtained according to the multidimensional cloud droplet.

8. The comprehensive benefit evaluation method for a power distribution system solution according to claim 1, characterized in that: Multiple cloud droplet samples of various benefit indicator data are generated based on the comprehensive cloud model. Based on the multiple cloud droplet samples of various benefit indicator data and the combined weights of the benefit indicators, a grayscale clustering analysis method that introduces logistic curves, graph neural networks, and chaos theory is combined to calculate the total membership of the distribution system solution to each gray class level, including: Combining Logistic function and chaos theory to construct nonlinear multi-level grey class membership function for each benefit index data; generating a plurality of cloud droplet samples of various benefit indicator data according to the comprehensive cloud model, substituting the cloud droplet samples into the multi-level grey class membership function, and calculating the weight value of each benefit indicator at each grey class level; Based on the graph convolutional neural network, the total membership of the distribution system scheme to each gray class level is calculated by combining the weight value of each benefit indicator at each gray class level and the combined weight of each benefit indicator.

9. The comprehensive benefit evaluation method of a power distribution system solution according to claim 8, characterized in that: Combining the Logistic function and chaos theory, a nonlinear multi-level grey class membership function is constructed for each benefit indicator data, including: The grey clustering hierarchy structure is defined, chaos theory is used to adjust the parameters of the Logistic function mapping, and the adjusted Logistic function is used to establish a nonlinear membership function for each grey clustering level.

10. A comprehensive benefit evaluation system for a power distribution system solution, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, each step of the comprehensive benefit evaluation method for a power distribution system solution according to any one of claims 1 to 9 is implemented.

Images