Power distribution network investment project optimization method and system

Abstract

The invention relates to the field of investment optimization, and discloses a power distribution network investment project optimization method and system. Comprising the following steps: constructing a contribution degree evaluation index system of a power distribution network investment project, obtaining power distribution network data and dividing data subsets; performing feature parameter estimation on the data subsets in combination with a reverse cloud model and a Bayesian feedback method to obtain estimated feature values, constructing a contribution degree objective function in combination with the estimated feature value of each data subset, obtaining historical data corresponding to the data subsets, and establishing a balance degree objective function; combining the contribution degree objective function and the balance degree objective function to establish a project optimization objective function; and setting a project investment amount constraint, a project necessity construction constraint and a project coupling constraint, constructing a final objective function in combination with the project optimization objective function, and solving by using a grey wolf algorithm to obtain a final investment scheme of the power distribution network investment project. According to the method, factors such as the contribution degree, the balance degree and the cost of the project can be comprehensively considered, and the accuracy and rationality of project optimization are improved.

Description

Technical Field

[0001] The present invention relates to the field of investment optimization, and in particular to a method and system for optimizing distribution network investment projects. Background Art

[0002] With the proposal of the "dual carbon" and new power system construction goals, scientific and reasonable distribution network investment allocation and project optimization methods are of great significance in guiding the planning and construction of distribution networks.

[0003] At present, the investment allocation methods of distribution networks are mainly divided into two scenarios. The first scenario is: in the case of detailed planning reports and equipment unit prices, the investment amount is divided into several parts, and a calculation formula is proposed for each part based on the planning report and equipment unit price, thereby obtaining the investment amount of each part. The second scenario is: making macro investment allocation on the basis of the planning report not yet given.

[0004] Regarding the study of the first type of scenario, the existing technology has a method of dividing the investment amount into rigid investment, economic investment, actual investment, and planned investment based on the relevant project data of the planning report, and then proposing a detailed calculation method for each part of the investment. However, this method requires a detailed planning report to be feasible, has strict requirements on the initial data, and has poor applicability in actual situations.

[0005] The research on the second type of scenario is mainly divided into evaluation method and prediction method. The evaluation method of allocating investment refers to constructing a set of investment evaluation index system and collecting relevant index data. Each index is weighted by selecting a weighting method, and then the relevant evaluation model is used for evaluation. In the prior art, there is a method of studying investment allocation through evaluation method, and then calculating the investment allocation result of the distribution network by constructing an investment allocation index system and assigning weights. The advantage of this method is that it is simple and convenient, but it also has disadvantages, that is, the evaluation method is more difficult to construct the investment allocation index system. In order to solve this shortcoming, the prior art mostly directly uses the distribution network evaluation index system as the investment allocation index system, but this approach lacks scientificity and accuracy, and there is no relevant basis for directly allocating investment according to the evaluation results. The prediction method is different from the evaluation method. It selects the relevant influencing factors of investment allocation and studies the functional relationship between the relevant factors and investment allocation. The prediction method is essentially an improvement of the evaluation method. The evaluation method does not deeply find the relationship between the evaluation results and the investment allocation, while the prediction method further explores the mathematical relationship between the two by fitting functions, thereby improving accuracy. There are methods in the prior art for allocating investment from a macro perspective, which optimize the calculation steps through a particle swarm algorithm, and then calculate the required investment amounts for various locations based on the evaluation results of the distribution network. There are also methods for predicting investment allocation amounts based on investment demand, electricity demand, etc., which use panel data models. This type of method predicts the allocation of investment amounts as a whole, and its allocation results are significantly improved in rationality and accuracy compared to the evaluation method.

[0006] At present, the optimization methods of distribution network projects are mainly divided into two categories: evaluation methods for project optimization and optimization methods. The evaluation method for project optimization is similar to the evaluation method for investment allocation. The basic steps are to establish a distribution network project benefit evaluation index system, evaluate the project benefits by weighting indicators and constructing an evaluation model, and then select according to the evaluation ranking results. In the prior art, when constructing a distribution network investment effectiveness evaluation index system, there are methods of determining the weights of each indicator by the Delphi method and ranking projects by defining the optimal decision value. There are also methods that are different from the traditional weighting method and calculate the weights of each indicator based on the marginal benefit of the indicator. There are also methods that use the combined weighting method to weight and construct an evaluation system based on the improvement of indicators. However, the evaluation method for project optimization has certain limitations, that is, it only considers the project benefits, and treats the investment funds as indicators or denominators, which cannot fully reflect the constraint effect of funds; or directly selects based on the ranking results, which cannot consider the coupling relationship between projects in actual situations, and only selects based on the ranking results, which cannot consider the balance of project investment and other aspects in combination with reality.

[0007] The optimization method builds a project optimization model, constructs optimization goals and solves to obtain a combination of investment plans. Compared with the evaluation method of project optimization, its advantages are: on the one hand, multiple optimization goals can be set for consideration, making the investment plan more applicable; on the other hand, it can fully consider the coupling relationship and capital constraint relationship between projects, and convert the coupling relationship and funds into constraints for processing, which is more scientific. In the prior art, there is a practice of constructing an indicator system to represent project benefits, but this practice only considers the optimal project benefits, and the consideration angle is too single. There is also a practice of introducing risk as the second goal of project optimization on the basis of considering project benefits, taking into account the risk of the investment portfolio through conditional risk value, and taking it together with the benefits as the optimization goal and solving it; there is also a practice of using relatively robust conditional risk value to replace traditional risk value to construct the objective function.

[0008] However, the existing practices have limitations: they do not take into account the ambiguity of some parameters in the project investment process and the balance of project investment. In the actual investment process, since project optimization should be carried out before project investment, the estimation process of project benefits is ambiguous, and the final total investment of the project is also uncertain, resulting in inaccurate and unreasonable final project investment plan. Summary of the invention

[0009] To this end, the technical problem to be solved by the present invention is to overcome the deficiencies in the prior art and provide a method and system for optimizing distribution network investment projects, which can comprehensively consider various factors such as project contribution, balance, cost, etc., and improve the accuracy and rationality of project optimization.

[0010] In order to solve the above technical problems, the present invention provides a method for optimizing distribution network investment projects, comprising:

[0011] Constructing a contribution evaluation index system for distribution network investment projects, acquiring distribution network data and dividing the data into subsets according to the contribution evaluation index system for distribution network investment projects;

[0012] Combining the inverse cloud model and the Bayesian feedback method to estimate characteristic parameters of each of the data subsets, to obtain an estimated characteristic value of each of the data subsets;

[0013] Constructing a contribution objective function in combination with the estimated characteristic values ​​of each of the data subsets, obtaining historical data corresponding to each of the data subsets and establishing a balance objective function, and establishing a project optimization objective function in combination with the contribution objective function and the balance objective function;

[0014] Set the project investment amount constraints, project necessity construction constraints, and project coupling constraints, and build the final objective function in combination with the project optimization objective function. Use the gray wolf algorithm to solve the final objective function to obtain the final investment plan for the distribution network investment project.

[0015] In one embodiment of the present invention, the contribution evaluation index system of the distribution network investment project includes the following indicators:

[0016] Economic and social benefit data sets, including: data subsets of the reduction in the proportion of light loads on lines, data subsets of the reduction in the proportion of light loads on distribution transformers, and data subsets of the reduction in the comprehensive line loss rate;

[0017] The load supply capacity data set includes: the data subset of the increase in the average distribution transformer capacity per household, the data subset of the decrease in the proportion of heavily overloaded substations, the data subset of the decrease in the proportion of heavily overloaded high-voltage lines, the data subset of the decrease in the proportion of heavily overloaded medium-voltage lines, and the data subset of the decrease in the proportion of heavily overloaded distribution transformers;

[0018] The data set at the network structure level includes: the data subset of the increase in the inter-station contact rate of the line, the data subset of the increase in the transferable power supply rate of the line, the data subset of the increase in the N-1 pass rate of the main transformer, and the data subset of the increase in the N-1 pass rate of the line; N-1 represents the N-1 operation mode, which means that if any independent component of the N components of the power system fails and is removed, no accident should occur;

[0019] The data sets of intelligence and green level include: the data subset of the increase in typical wiring rate, the data subset of the increase in effective coverage rate of distribution automation, the data subset of the increase in distribution communication networking rate, and the data subset of the increase in the proportion of new energy power grid access.

[0020] In one embodiment of the present invention, the inverse cloud model and the Bayesian feedback method are combined to estimate the characteristic parameters of each of the data subsets to obtain the estimated characteristic values ​​of each of the data subsets, specifically:

[0021] Using the reverse cloud model to calculate the characteristic parameters of the data subset, the characteristic parameters include characteristic expectation Ex, entropy En, and super entropy He;

[0022] Calculate the degree of difference of the data subsets and determine whether the characteristic parameters are credible, and use the Bayesian feedback method to correct the characteristic parameters when the characteristic parameters are not credible;

[0023] The calculating the difference degree of the data subsets and judging whether the characteristic parameters are credible, and using the Bayesian feedback method to correct the characteristic parameters when the characteristic parameters are not credible, comprises:

[0024] Define the trusted domain and the rejected domain as:

[0025] When samples in a subset of data When x i Located in the trusted domain, where x iis the i-th sample in the data subset, Z α / 2 is the quantile on both sides of the standard normal distribution, 1-α is the confidence level; when the sample in the data subset When x i is within the rejection region;

[0026] Calculate the deviation σ of the sample mean falling on both sides of the rejection region as:

[0027]

[0028] in, For samples less than The sample mean at time , For samples larger than The sample mean at time d0 is the width of the credible region,

[0029] If the deviation σ is less than or equal to the preset threshold, the calculated characteristic parameters Ex, En, and He are considered to be credible; if the deviation σ is greater than the preset threshold, the characteristic parameters are corrected using the Yes feedback method;

[0030] The method of using the Yesian feedback method to correct the characteristic parameters includes:

[0031] The super entropy is lowered, and Ex, En, and He are corrected through the Bayesian posterior probability distribution to obtain the corrected feature expectation Ex', entropy En', and super entropy He':

[0032]

[0033]

[0034]

[0035] Where N is the total number of samples in the data subset, represents the sample mean in the data subset, c0 = Ex, S 2 is the sample variance,

[0036] The deviation of the sample means on both sides of the rejection region is recalculated for the corrected feature expectation, entropy, and super entropy. If the deviation of the sample means on both sides of the rejection region is greater than the preset threshold, the feature expectation, entropy, and super entropy are re-corrected until the deviation of the sample means on both sides of the rejection region is less than or equal to the preset threshold. The feature expectation, entropy, and super entropy at this time are used as the final estimated feature values ​​of the data subset.

[0037] In one embodiment of the present invention, a contribution objective function is constructed in combination with the estimated characteristic value of each data subset, specifically:

[0038] Assuming that there are m investment projects to be decided in the project library, each investment project to be decided has n indicators under the contribution evaluation index system of the distribution network investment project, and the value of each indicator is the estimated characteristic value of the corresponding data subset. The contribution objective function is constructed as follows:

[0039]

[0040] Among them, F1 represents the contribution objective function, x i ′ indicates whether the i-th investment project to be decided is selected, x i ′=1 means the project is selected for investment, x i ′=0 means that the project is not selected and no investment is made; D ij represents the value of the jth indicator under the i-th investment project to be decided, w j represents the weight of the jth indicator under the i-th investment project to be decided, and n represents the total number of indicators; the optimization goal of the contribution objective function F1 is to maximize the value.

[0041] In one embodiment of the present invention, historical data corresponding to each of the data subsets is obtained and a balance objective function is established, specifically:

[0042] Get the value e of the jth indicator under the i-th investment project in the last round of investment j and the value of the jth indicator under the i-th investment project to be decided in this round of investment, calculate the incremental deviation e of the jth indicator under the i-th investment project to be decided j 'for:

[0043]

[0044] The objective function for calculating the degree of balance is:

[0045]

[0046] Among them, F2 represents the balance objective function, Represents the average value of all incremental deviations. The optimization goal of the balance objective function F2 is to minimize the value.

[0047] In one embodiment of the present invention, a project optimization objective function is established by combining the contribution objective function and the balance objective function, specifically:

[0048] Use single-objective optimization to solve the contribution objective function and obtain the project investment plan F1 when the contribution objective function F1 is maximized.max , solve F1 max Substitute the balance objective function to obtain the project investment plan

[0049] Use single-objective optimization to solve the balance objective function and obtain the project investment plan when the balance objective function F2 is minimized. Will solve Substitute the contribution objective function to obtain the project investment plan F1 (2) ;

[0050] Construct boundary value F1 min and F1 min Meet F1 min ≤F1 (2) , satisfy Construct the target boundary [F1 min , F1 max ]and

[0051] The normalized representation ρ(F1) of the contribution target calculated using the membership function is:

[0052]

[0053] in, d1=(F1-F1 min ) / (F1 max -F1 min ), d2=(F1-F1 max ) / (F1 max -F1 min );

[0054] The normalized representation ρ(F2) of the balance target calculated using the membership function is:

[0055]

[0056] in,

[0057] The project optimization objective function F is established as:

[0058] F=λ1ρ(F1)+λ2ρ(F2),

[0059] Among them, λ1 and λ2 are target allocation weights.

[0060] In one embodiment of the present invention, the project investment amount constraint is expressed as:

[0061]

[0062] Among them, x i ″ represents the i-th investment project to be decided, m represents the number of investment projects to be decided, C i represents the funds required for the investment of the i-th investment project to be decided, and C represents the total investment limit;

[0063] The project necessity construction constraint is expressed as:

[0064]

[0065] in, represents the minimum improvement requirement of the jth indicator under the i-th investment project to be decided; x ij Indicates whether the i-th investment project to be decided contributes to the j-th indicator. If so, x ij =1, if there is no contribution then x ij =0.

[0066] In one embodiment of the present invention, the project coupling constraints include mutually exclusive constraints and dependency constraints.

[0067] The mutually exclusive constraint is expressed as:

[0068]

[0069] Among them, S1 represents a set of mutually exclusive items;

[0070] The dependency constraints are specifically:

[0071] Construct a set S2 of projects with subordinate relationships, and construct a matrix B with S2-1 rows and S2 columns. S2 represents the number of investment projects to be decided in the set S2. According to all the projects x1″, …, x2 in the set S2, S2 "Construct item vector X = [x1", ..., x S2 ″] T , x1″ represents the first investment project to be decided in set S2, x S2 ″ represents the S2th investment project to be decided in the set S2;

[0072] The dependency constraint is expressed as: B·X=0.

[0073] In one embodiment of the present invention, the final objective function is:

[0074]

[0075] Among them, F represents the project optimization objective function, and max() represents the operation of taking the larger value of the two.

[0076] The present invention also provides a distribution network investment project optimization system, comprising:

[0077] An indicator data acquisition module is used to construct a contribution evaluation indicator system for distribution network investment projects, acquire distribution network data, and divide the data into subsets according to the contribution evaluation indicator system for distribution network investment projects;

[0078] An indicator processing module, used to estimate characteristic parameters of each of the data subsets by combining the inverse cloud model and the Bayesian feedback method, so as to obtain an estimated characteristic value of each of the data subsets;

[0079] A project optimization objective function construction module is used to construct a contribution objective function in combination with the estimated characteristic values ​​of each of the data subsets, obtain the historical data corresponding to each of the data subsets and establish a balance objective function, and establish a project optimization objective function in combination with the contribution objective function and the balance objective function;

[0080] The solution module is used to set the project investment amount constraint, the project necessity construction constraint, and the project coupling constraint, and to construct the final objective function in combination with the project optimization objective function, and to use the gray wolf algorithm to solve the final objective function to obtain the final investment plan of the distribution network investment project.

[0081] The above technical solution of the present invention has the following advantages compared with the prior art:

[0082] The present invention constructs a contribution evaluation index system for distribution network investment projects, combines the inverse cloud model and the Bayesian feedback method to extract the characteristic values ​​of the data, and on this basis constructs an objective function that comprehensively considers contribution, balance, project investment amount constraints, project necessity construction constraints, and project coupling constraints, thereby effectively improving the accuracy and rationality of project optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein:

[0084] Figure 1 It is a flow chart of the present invention.

[0085] Figure 2 It is a structural diagram of the project contribution evaluation index system in the present invention.

[0086] Figure 3 is a schematic diagram of the normalized representation of the contribution target.

[0087] Figure 4 It is a schematic diagram of the normalized representation of the balance objective.

[0088] Figure 5 It is a result diagram of the contribution of the candidate items calculated in the embodiment of the present invention.

[0089] Figure 6 It is a result diagram of the optimization of projects with project numbers 1 to 25 under four investment plans in an embodiment of the present invention.

[0090] Figure 7 It is a result diagram of the optimization of projects with project numbers 26 to 50 under four investment plans in an embodiment of the present invention.

[0091] Figure 8 It is a comparison chart of the estimated improvement value of each indicator of the investment plan in the embodiment of the present invention and the expected target value. DETAILED DESCRIPTION

[0092] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0093] Reference Figure 1 As shown, the present invention discloses a method for optimizing distribution network investment projects, comprising the following steps:

[0094] S1: construct a contribution evaluation index system for distribution network investment projects, obtain distribution network data and divide the data into subsets according to the contribution evaluation index system for distribution network investment projects.

[0095] Contribution refers to the contribution of the project to the local distribution network after investment. The present invention constructs a project contribution evaluation index system to evaluate the contribution of the project. When constructing the contribution evaluation system, the change in the index brought by the project should be considered, that is, the improvement of the positive index and the reverse evaluation index system should be measured by the change in the index. Therefore, based on the distribution network evaluation index system, the present invention constructs a corresponding contribution evaluation index system such as Figure 2 The contribution evaluation index system of the distribution network investment project includes the following indicators:

[0096] Economic and social benefit data sets, including: data subsets of the reduction in the proportion of light loads on lines, data subsets of the reduction in the proportion of light loads on distribution transformers, and data subsets of the reduction in the comprehensive line loss rate;

[0097] The load supply capacity data set includes: the data subset of the increase in the average distribution transformer capacity per household, the data subset of the decrease in the proportion of heavily overloaded substations, the data subset of the decrease in the proportion of heavily overloaded high-voltage lines, the data subset of the decrease in the proportion of heavily overloaded medium-voltage lines, and the data subset of the decrease in the proportion of heavily overloaded distribution transformers;

[0098] The data set at the network structure level includes: the data subset of the increase in the inter-station contact rate of the line, the data subset of the increase in the transferable power supply rate of the line, the data subset of the increase in the N-1 pass rate of the main transformer, and the data subset of the increase in the N-1 pass rate of the line; "N-1" means the N-1 operation mode, which means that after any independent component (generator, transmission line, transformer, etc.) of the N components of the power system fails and is removed, it should not cause power outages to users due to overload tripping of other lines, destroy the stability of the system, and avoid accidents such as voltage collapse;

[0099] The data sets of intelligence and green level include: the data subset of the increase in typical wiring rate, the data subset of the increase in effective coverage rate of distribution automation, the data subset of the increase in distribution communication networking rate, and the data subset of the increase in the proportion of new energy power grid access.

[0100] S2: Combine the reverse cloud model and the Bayesian feedback method to estimate the characteristic parameters of each of the data subsets to obtain the estimated characteristic values ​​of each of the data subsets. By using the reverse cloud model, the qualitative data in the data subset is converted into quantitative specific samples. On this basis, the Bayesian feedback method is used to correct the specific samples to obtain the estimated characteristic parameter values ​​within the allowable error range, and the estimated characteristic parameter values ​​that meet the allowable error range are used as the estimated characteristic values ​​of the data subset.

[0101] S2-1: The cloud model is a model that combines probability theory and fuzzy mathematics, which can accurately represent the randomness and fuzziness of things. The cloud model is composed of multiple data "cloud drops". The horizontal axis of the cloud droplet represents the value of the element in the domain, and the vertical axis represents the degree of certainty of the element. The cloud model usually uses characteristic expectation Ex, entropy En and super entropy He to describe the uncertain characteristics of the "cloud". The characteristic expectation represents the expected value of the element, which is used to represent the overall characteristics of the element value. The entropy value represents the degree of discreteness of the element, that is, the range fuzziness of the cloud. Super entropy is the entropy value of the entropy value, which represents the degree of discreteness of the entropy. The cloud model can realize the conversion between qualitative and quantitative, and can be divided into forward cloud model and reverse cloud model. The forward cloud model means that when the Ex, En, and He of the characteristics of a certain thing are known, a specific sample that conforms to its distribution law can be generated; the reverse cloud model means that when the specific sample is known, the characteristics of the thing represented by Ex, En, and He are fitted.

[0102] In the present invention, the reverse cloud model is used to calculate the characteristic parameters of the data subset, and the characteristic parameters include characteristic expectation Ex, entropy En, and super entropy He:

[0103]

[0104]

[0105]

[0106] Among them, x i is the i-th sample in the data subset, N is the total number of samples in the data subset; S 2 is the sample variance,

[0107] The traditional inverse cloud model has certain limitations, namely, the cloud droplet sample error generated by using expert opinions will lead to large errors in characteristic parameters, and it is often difficult to determine the cloud map boundary based on limited cloud droplet samples. Therefore, the present invention introduces the Bayesian feedback method for correction.

[0108] S2-2: Calculate the degree of difference of the data subsets and determine whether the characteristic parameters are credible. If the characteristic parameters are not credible, use the Bayesian feedback method to correct the characteristic parameters.

[0109] S2-2-1: Define the credible domain and rejection domain. For each group of cloud droplet samples, there are credible domains and rejection domains related to the confidence level. Sample points located in the credible domain are high-probability events with high credibility. On the contrary, if the sample points are located in the rejection domain, they are called low-probability events, and these samples are usually not credible. The range of the credible domain and the rejection domain is related to the confidence level and can be determined by the central limit theorem. Therefore, the credible domain and the rejection domain are defined by the central limit theorem as:

[0110] When samples in a subset of data When x i Located in the trusted domain, where x i is the i-th sample in the data subset, Z α / 2 is the quantile on both sides of the standard normal distribution, 1-α is the confidence level, and in this embodiment, the value is α=0.05. At this time, z α / 2 =1.645, that is, it is considered that with a probability of 95%, all sample points are located in the credible domain. When x i It is within the rejection region; that is, it is considered that there is only a 5% probability that the sample point is within this range, which is a low-probability event.

[0111] S2-2-2: Calculate the deviation. If in actual situations, there are too many samples falling within the rejection region, it can be considered that there is an error in the parameter estimation, resulting in a lack of accuracy in the estimation of the credible region and the rejection region, and the initial sample needs to be corrected. The present invention calculates the interval width by the sample mean falling on both sides of the rejection region, compares it with the rejection region width, and determines whether the deviation is within a reasonable range, thereby determining the credibility of the parameter estimation. The deviation σ of the sample mean falling on both sides of the rejection region is calculated as:

[0112]

[0113] in, For samples less than The sample mean at time , For samples larger than The sample mean at time d0 is the width of the credible region,

[0114] If the deviation σ is less than or equal to the preset threshold α1, in this embodiment, α1 = 20%, that is, When , it is considered that the difference of sample data is small, and the calculated characteristic parameters Ex, En, and He are credible; if the deviation σ is greater than the preset threshold, it means that the sample difference is large, and there may be inaccurate samples. The Yes feedback method is used to correct the characteristic parameters.

[0115] S2-2-3: Use the Yes feedback method to correct the characteristic parameters:

[0116] S2-2-3-1: lowering the super entropy, that is, He=He-α2, in this embodiment, α2=5%;

[0117] S2-2-3-2: Delete the original samples in the rejection domain, regenerate a certain number of new samples through Ex, En, and He, and delete the samples in the rejection domain and add new samples to form the posterior sample set. Since the cloud model prior distribution Ex satisfies the normal distribution Feedback can be performed through the calculation of the Bayesian posterior probability distribution to correct Ex, En, and He, and the corrected feature expectation Ex', entropy En', and super entropy He' are obtained as follows:

[0118]

[0119]

[0120]

[0121] Where N is the total number of samples in the data subset, represents the sample mean in the data subset, c0 = Ex, S 2 is the sample variance,

[0122] S2-2-4: Recalculate the deviation of the sample means on both sides of the rejection domain for the corrected feature expectation, entropy, and super entropy. If the deviation of the sample means on both sides of the rejection domain is greater than the preset threshold value α1, re-execute S2-2-3 to correct the feature expectation, entropy, and super entropy until the deviation of the sample means on both sides of the rejection domain is less than or equal to the preset threshold value α1. The feature expectation, entropy, and super entropy at this time are used as the final estimated feature values ​​of the data subset.

[0123] The Bayesian posterior probability distribution is expressed as a probability distribution model obtained by correcting the prior probability distribution based on the known previous probability distribution (i.e., prior probability distribution). This can correct the error of the prior probability distribution well, effectively avoid the impact of sample errors on parameter estimation results, and improve the accuracy of cloud model characteristic parameter estimation.

[0124] S3: Construct a contribution objective function based on the estimated characteristic values ​​of each of the data subsets, obtain the historical data corresponding to each of the data subsets and establish a balance objective function, and establish a project optimization objective function based on the contribution objective function and the balance objective function.

[0125] S3-1: Construct a contribution objective function based on the estimated characteristic values ​​of each of the data subsets, specifically:

[0126] Assuming that there are m investment projects to be decided in the project library, each investment project to be decided has n indicators under the contribution evaluation index system of the distribution network investment project, and the value of each indicator is the estimated characteristic value of the corresponding data subset. The contribution objective function is constructed as follows:

[0127]

[0128] Among them, F1 represents the contribution objective function, x i ′ indicates whether the i-th investment project to be decided is selected, x i ′=1 means the project is selected for investment, x i ′=0 means that the project is not selected and no investment is made; D ij represents the value of the jth indicator under the i-th investment project to be decided, w j represents the weight of the jth indicator under the i-th investment project to be decided, w j The value of is set according to the actual situation; n represents the total number of indicators, that is, the total number of data subsets; the optimization goal of the contribution objective function F1 is to maximize the value, that is, the optimization goal is

[0129] S3-2: Obtain the historical data corresponding to each of the data subsets and establish a balance objective function, specifically:

[0130] Balance refers to the degree of fit between the project investment results and the expected plan. In the actual production process, the contribution of project investment should be consistent with the regional planning direction and construction goals, and the investment should take into account all aspects of the project indicators. If the balance goal is not considered, the project optimization results may only select indicators with greater contribution. Since the weight of indicators under load supply capacity is large, the investment results may all be projects for load supply capacity construction, while ignoring the construction and development of other indicators, which is also contrary to the construction goal of "dual carbon". Therefore, the present invention incorporates the project balance into the objective function for consideration.

[0131] The various projects in the distribution network investment need to ensure balance, so that investment can be made selectively according to the set goals, and the value of the jth indicator under the i-th investment project in the previous round of investment is obtained. j and the value of the jth indicator under the i-th investment project to be decided in this round of investment, calculate the incremental deviation e of the jth indicator under the i-th investment project to be decided j 'for:

[0132]

[0133] The objective function for calculating the degree of balance is:

[0134]

[0135] Among them, F2 represents the balance objective function, represents the average value of all incremental deviations. The optimization goal of the balance objective function F2 is to minimize the value, that is, the optimization goal is

[0136] S3-3: In the target determination stage of project optimization, the present invention, after comprehensive consideration, selects the contribution and balance of project investment as optimization targets, and constructs multi-target optimization based on this. Considering that there are orders of magnitude differences in this multi-target optimization model, the fuzzy satisfaction theory can be used to process the objective function, and the S-type membership function can be used for normalization. Therefore, the project optimization objective function is established by combining the contribution objective function and the balance objective function, which is specifically:

[0137] S3-3-1: Use single-objective optimization to solve the contribution objective function and obtain the project investment plan F1 when the contribution objective function F1 is maximized max , solve F1 max Substitute the balance objective function to obtain the project investment plan

[0138] Use single-objective optimization to solve the balance objective function and obtain the project investment plan when the balance objective function F2 is minimized. Will solve Substitute the contribution objective function to obtain the project investment plan F1 (2) ; The results correspond to Table 1:

[0139] Table 1 Decision attributes of objective function

[0140]

[0141] S3-3-2: Construct boundary value F1 min and F1 min Meet F1 min ≤F1 (2) , satisfy Construct the target boundary [F1 min , F1 max ]and

[0142] S3-3-3: The normalized representation ρ(F1) of the contribution target calculated using the membership function is:

[0143]

[0144] in, d1=(F1-F1 min ) / (F1 max -F1 min ), d2=(F1-F1 max ) / (F1 max -F1 min );

[0145] The normalized representation ρ(F2) of the balance target calculated using the membership function is:

[0146]

[0147] in,

[0148] By fuzzy processing each objective through normalized representation, the marginal effect can be effectively considered, making the optimization result more accurate. Figure 3 represents the normalized representation of the contribution target, Figure 4 Represents the normalized representation of the balance target.

[0149] S3-3-4: Establish the project optimization objective function F as follows:

[0150] F=λ1ρ(F1)+λ2ρ(F2),

[0151] Among them, λ1 and λ2 are target allocation weights, which are set according to actual conditions.

[0152] S4: Set the project investment amount constraint, project necessity construction constraint, and project coupling constraint, and construct the final objective function in combination with the project optimization objective function. Use the gray wolf algorithm to solve the final objective function to obtain the final investment plan for the distribution network investment project.

[0153] S4-1: Set project investment amount constraints, project necessity construction constraints, and project coupling constraints, specifically:

[0154] The project investment amount constraint means that the total investment amount spent in this round of investment does not exceed the investment limit, which is expressed as:

[0155]

[0156] Among them, x i ″ represents the i-th investment project to be decided, m represents the number of investment projects to be decided, C i represents the funds required for the investment of the i-th investment project to be decided, and C represents the total investment limit.

[0157] In distribution network project investment, there are often minimum requirements for the above indicators. Therefore, the necessary construction constraints of the project are expressed as:

[0158]

[0159] in, represents the minimum improvement requirement of the jth indicator under the i-th investment project to be decided; x ij Indicates whether the i-th investment project to be decided contributes to the j-th indicator. If so, x ij =1, if there is no contribution then x ij =0.

[0160] The project coupling constraints include mutually exclusive constraints and subordinate constraints. The mutually exclusive constraints indicate that there may be coupling relationships between projects and need to be considered in the constraints, which can be expressed as:

[0161]

[0162] Among them, S1 represents a set of mutually exclusive projects, that is, the projects in set S1 cannot be invested in at the same time;

[0163] The dependency constraints are specifically:

[0164] Construct a set S2 of projects with a subordinate relationship. The subordinate relationship constraint is that for projects in the same set S2, if one of the projects is selected for construction, S2 needs to be constructed together; construct a matrix B with S2-1 rows and S2 columns, where S2 represents the number of investment projects to be decided in the set S2. In this embodiment, According to all items x1″, …, x1″, …, x2 in the set S2 S2 "Construct item vector X = [x1", ..., x S2 ″] T , x1″ represents the first investment project to be decided in set S2, x S2 ″ represents the S2th investment project to be decided in the set S2;

[0165] The dependency constraint is expressed as: B·X=0.

[0166] S4-2: The final objective function is constructed by combining the project optimization objective function:

[0167]

[0168] Among them, F represents the project optimization objective function, and max() represents the operation of taking the larger value of the two.

[0169] S4-3: Use the Grey Wolf Algorithm to solve the final objective function and obtain the final investment plan for the distribution network investment project.

[0170] The Gray Wolf Algorithm is usually used to solve continuous optimization problems. However, in this embodiment, the solution is a Boolean variable whose value is limited to 0 and 1. Therefore, the traditional Gray Wolf Algorithm needs to be discretized. Discretization refers to converting the original continuous solution into Through mapping relationship to discrete space, we finally get discrete solution in, represents the continuity solution X c The i-th solution in, m1 represents the continuous solution X c and the discrete solution X d The number of solutions, Denotes the discrete solution X d To this end, the present invention adopts the following encoding conversion method:

[0171]

[0172] Wherein, lb and ub are the upper bound and lower bound of the continuity solution, respectively. In this embodiment, lb=0 and ub=1 are taken, and [] represents a rounding-down operation.

[0173] Aiming at the limitation of the traditional gray wolf algorithm that it is easy to fall into the local optimum, this paper introduces the average fitness and proposes an improved gray wolf algorithm. In order to solve the problem of easy falling into the local optimum and slow convergence speed, a convergence factor a is designed. i for:

[0174]

[0175] In the formula, a i represents the convergence factor of the ith individual, where one individual represents a feasible project investment plan; F i is the value of the project optimization objective function F of the ith individual, F avg is the average value of the project optimization objective function F of the current population. The gray wolf algorithm will randomly generate a series of feasible project investment plans, each of which is an individual. The set of all feasible plans generated in one iteration is called the population; F max , F min is the value of the optimal objective function F for the maximum and minimum items in the current population, e represents the natural logarithm, t represents the current number of iterations, and t max Indicates the preset maximum number of iterations. When the value of the individual project optimization objective function F is less than the average value of the project optimization objective function F, the individual is closer to the prey position, and the exponential function convergence form is used to reduce its search range. If the value of the individual project optimization objective function F is greater than the average value of the project optimization objective function F, it is considered that the individual performs poorly. At this time, the search range is increased to continue looking for the prey position.

[0176] When the final objective function F′ converges or reaches the preset maximum number of iterations, the Grey Wolf Algorithm is solved and the solution at this time is used as the final investment plan for the distribution network investment project.

[0177] The present invention also discloses a distribution network investment project optimization system, comprising:

[0178] An indicator data acquisition module is used to construct a contribution evaluation indicator system for distribution network investment projects, acquire distribution network data, and divide the data into subsets according to the contribution evaluation indicator system for distribution network investment projects;

[0179] An indicator processing module, used to estimate characteristic parameters of each of the data subsets by combining the inverse cloud model and the Bayesian feedback method, so as to obtain an estimated characteristic value of each of the data subsets;

[0180] A project optimization objective function construction module is used to construct a contribution objective function in combination with the estimated characteristic values ​​of each of the data subsets, obtain the historical data corresponding to each of the data subsets and establish a balance objective function, and establish a project optimization objective function in combination with the contribution objective function and the balance objective function;

[0181] The solution module is used to set the project investment amount constraint, the project necessity construction constraint, and the project coupling constraint, and to construct the final objective function in combination with the project optimization objective function, and to use the gray wolf algorithm to solve the final objective function to obtain the final investment plan of the distribution network investment project.

[0182] The present invention also discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for optimizing distribution network investment projects is implemented.

[0183] The present invention also discloses a device, including a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the distribution network investment project optimization method when executing the computer program.

[0184] The present invention is a method for optimizing distribution network investment projects taking into account fuzziness and balance. By processing the fuzziness of various parameters in the actual production process in a targeted manner and constructing an optimization model, the project optimization result is finally obtained. Compared with the prior art, the method has the following beneficial effects:

[0185] (1) Generate characteristic parameters using an improved inverse cloud model based on Bayesian feedback. The inverse cloud model is used to extract the characteristic expectation, entropy and super entropy of the expert evaluation points. In view of the problem that the traditional inverse cloud model is easily affected by error samples and causes poor accuracy, a Bayesian feedback link is introduced to adjust the super entropy to regenerate credible cloud droplet samples to replace the cloud droplet samples that fall into the rejection domain, thereby improving the accuracy of characteristic parameter calculation.

[0186] (2) Considering fuzziness and balance, the objective function of the project optimization model is constructed. The present invention constructs contribution and balance as the objective function of the optimization model, takes the evaluation value of each project under the index system as the project contribution, compares the improvement of the investment plan under each indicator with the expected improvement of each indicator before, and takes the sum of their deviations as the balance. For the multiple solution objectives constructed, the fuzzy satisfaction theory is used to transform each objective. In view of the problem that the traditional linear membership function does not consider the marginal effect, the S-type membership transformation function is used to replace it to obtain the final objective function, which can effectively consider the marginal effect and make the optimization result more accurate.

[0187] (3) Using the discrete gray wolf algorithm to solve the model in multiple scenarios. Since whether a project is invested can be expressed as a Boolean variable, the present invention makes targeted improvements to the traditional gray wolf algorithm, proposes an adaptive convergence strategy, and discretizes it using coding transformation, and finally solves the project investment portfolio schemes under the contribution-focused type, the balance-focused type, and the multi-objective balance type. The solution efficiency is higher and the results are more accurate and reasonable.

[0188] In order to further illustrate the beneficial effects of the present invention, a simulation experiment is performed using the method of the present invention in this embodiment. Specifically:

[0189] (1) Extracting data features using the reverse cloud model based on Bayesian feedback

[0190] Taking a certain city as an example, firstly, the project contribution evaluation index system is evaluated. 20 experts are selected to evaluate the contribution of the project under each indicator, and the expected contribution of the 20 experts to each project and each indicator is obtained. The 20 sets of data of each project under each indicator are substituted into the reverse cloud model based on Bayesian feedback for feature extraction. Taking the evaluation values ​​of Project No. 13 and Project No. 17 under the pass rate of Line N-1 as an example, the expert evaluation values ​​are shown in Tables 2 and 3.

[0191] Table 2 Expert evaluation value of project No. 13 under the pass rate of line N-1

[0192] Expert No. 1 2 3 4 5 6 7 8 9 10 Value / % 0.46 0.27 0.46 0.25 0.24 0.17 0.30 0.22 0.27 0.28 Expert No. 11 12 13 14 15 16 17 18 19 20 Value / % 0.16 0.19 0.29 0.05 0.24 0.16 0.33 0.19 0.32 0.35 . Table 3 Expert evaluation value of project No. 17 under the pass rate of line N-1

[0193] Expert No. 1 2 3 4 5 6 7 8 9 10 Value / % 0.45 0.23 0.14 0.24 0.15 0.01 0.32 0.12 0.33 0.08 Expert No. 11 12 13 14 15 16 17 18 19 20 Value / % 0.23 0.36 0.37 0.07 0.29 0.36 0.23 0.01 0.20 0.31 .

[0194] The characteristic expectation, entropy, super entropy and deviation of the two groups of data for Project 13 and Project 17 are calculated, and the results are shown in Table 4.

[0195] Table 4 Calculation results of characteristic parameters of items No. 13 and No. 17

[0196] project Ex En He σ No. 13 0.2604 0.0912 0.0358i 0.2085 No. 17 0.2252 0.1277 0.0268i 0.0473 .

[0197] As shown above, the deviation of item 13 exceeds the set value, and Ex, En, and He need to be corrected. The corrected results are shown in Table 5.

[0198] Table 5 Parameter results of project 13 after the inverse cloud model was modified based on Bayesian feedback

[0199] Parameter results Ex' En' He' No. 13 0.2422 0.0731 0.0228i .

[0200] It can be seen that improving the feature extraction of the inverse cloud model can effectively reduce the result error caused by some sample errors and improve the accuracy of feature extraction.

[0201] (2) Calculate project contribution

[0202] The expected feature extracted by the cloud model is used as the score of the jth indicator under the i-th project. Calculate the contribution, fi is the contribution of the i-th project, and the contribution evaluation results of the 50 candidate projects are as follows: Figure 5 shown.

[0203] (3) Objective function combination based on fuzzy satisfaction theory

[0204] The present invention takes the coefficient η before the penalty function as 1000, and first obtains two target values ​​in two single-objective optimization situations, as shown in Table 6.

[0205] Table 6 Solution results of fuzzy satisfaction theory

[0206]

[0207] By extending the above objective function values, the boundaries of the two objective function values ​​can be determined, as shown in Table 7.

[0208] Table 7 Objective function boundary determination

[0209] Objective Function <![CDATA[F1]]> <![CDATA[F2]]> Maximum 1.4246 1.4342 Minimum 0.2685 0.2724 The extended border (0.1，1.5) (0.1，1.5) .

[0210] The determined boundaries are used to perform fuzzy processing on each objective, and the project optimization objective function F = λ1ρ(F1) + λ2ρ(F2) is taken as the optimization objective. According to the expert opinions, weight combinations with different emphases are set as shown in Table 8. Setting weight combinations with different emphases can provide a variety of project investment options and select appropriate project optimization plans based on specific circumstances.

[0211] Table 8 Weight settings of each objective function for different investment plans

[0212]

[0213] As shown in Table 8, Scheme ① focuses on contribution, which is usually the primary consideration for distribution network investment in various regions. The weight of contribution is set to 0.7 and the weight of balance is set to 0.3. Scheme ② focuses on balance. Similar to Scheme ①, the weight of balance is set to 0.7 and the weight of contribution is set to 0.3. Schemes ③ and ④ are multi-objective balanced schemes with similar weights for the two objectives: Scheme ③ focuses on contribution, with the weight of contribution set to 0.6 and the weight of balance set to 0.4; Scheme ④ focuses on balance, with the weight of contribution set to 0.4 and the weight of balance set to 0.6.

[0214] (4) Solve the objective function using the Gray Wolf Algorithm

[0215] According to the above four schemes, the final objective function F′ is constructed and solved by discrete grey wolf algorithm. The number of wolves is set to 100 and the maximum number of iterations is 1000. The solution is obtained as follows: Figure 6 As shown in the figure, the horizontal axis represents the serial number of 1 to 50 groups of projects, and the vertical axis is the value of 0 and 1 of the Boolean variable. When the value is 1, it means that the project with the same number will be invested, otherwise, it means that no investment will be made. Figure 6 The results for item numbers 1 to 25. Figure 7 These are the results for item numbers 26 to 50.

[0216] Depend on Figure 6 , Figure 7 It can be seen that Plan ① invests in 20 projects, Plan ② invests in 7 projects, Plan ③ invests in 21 projects, and Plan ④ invests in 17 projects. Among them, projects No. 4, No. 6, No. 7, and No. 8 are not invested in any plan, indicating that the feasibility of such projects is low. On the contrary, projects No. 26 and No. 40 are selected under multiple plans, and the feasibility of such projects is relatively high.

[0217] Based on the above schemes, the target values ​​and investment amounts corresponding to each scheme can be obtained as shown in Table 9.

[0218] Table 9 The objective function values ​​obtained by solving

[0219]

[0220] As shown in Table 9, the investment amounts of Scheme ① and Scheme ③ are 240.30 million yuan and 243.58 million yuan respectively, which exceed the set constraint of 226.15 million yuan. However, for the fuzzy constraint set by the present invention, it is allowed that the investment amount exceeds the set constraint with a certain probability. The application of fuzzy constraint makes the actual investment process more reasonable. Scheme ② focuses on risk investment, so the optimization goal is mainly to minimize the risk, so the investment result is more conservative.

[0221] Based on the four types of investment plans obtained, the expected change values ​​of each indicator can be calculated as follows: Figure 8 As shown in the figure, the indicator numbers that are expected to increase or decrease in this construction are given, where the broken line represents the expected target value set before investment. It can be seen that under the goal of considering investment balance in the present invention, the four types of investment plans given can effectively take into account investment balance on the basis of ensuring contribution, and obtain more reasonable investment results.

[0222] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0223] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0224] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0225] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0226] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from these are still within the protection scope of the invention.

Claims

1. A method for optimizing distribution network investment projects, characterized in that: include: Constructing a contribution evaluation index system for distribution network investment projects, acquiring distribution network data and dividing the data into subsets according to the contribution evaluation index system for distribution network investment projects; Combining the inverse cloud model and the Bayesian feedback method to estimate characteristic parameters of each of the data subsets, to obtain an estimated characteristic value of each of the data subsets; Constructing a contribution objective function in combination with the estimated characteristic values ​​of each of the data subsets, obtaining historical data corresponding to each of the data subsets and establishing a balance objective function, and establishing a project optimization objective function in combination with the contribution objective function and the balance objective function; Set the project investment amount constraints, project necessity construction constraints, and project coupling constraints, and build the final objective function in combination with the project optimization objective function. Use the gray wolf algorithm to solve the final objective function to obtain the final investment plan for the distribution network investment project.

2. The method for optimizing distribution network investment projects according to claim 1, characterized in that: The contribution evaluation index system of the distribution network investment project includes the following indicators: Economic and social benefit data sets, including: data subsets of the reduction in the proportion of light loads on lines, data subsets of the reduction in the proportion of light loads on distribution transformers, and data subsets of the reduction in the comprehensive line loss rate; The load supply capacity data set includes: the data subset of the increase in the average distribution transformer capacity per household, the data subset of the decrease in the proportion of heavily overloaded substations, the data subset of the decrease in the proportion of heavily overloaded high-voltage lines, the data subset of the decrease in the proportion of heavily overloaded medium-voltage lines, and the data subset of the decrease in the proportion of heavily overloaded distribution transformers; The data set at the network structure level includes: the data subset of the increase in the inter-station contact rate of the line, the data subset of the increase in the transferable power supply rate of the line, the data subset of the increase in the N-1 pass rate of the main transformer, and the data subset of the increase in the N-1 pass rate of the line; N-1 represents the N-1 operation mode, which means that if any independent component of the N components of the power system fails and is removed, no accident should occur; The data sets of intelligence and green level include: the data subset of the increase in typical wiring rate, the data subset of the increase in effective coverage rate of distribution automation, the data subset of the increase in distribution communication networking rate, and the data subset of the increase in the proportion of new energy power grid access.

3. The method for optimizing distribution network investment projects according to claim 1, characterized in that: The inverse cloud model and the Bayesian feedback method are combined to estimate the characteristic parameters of each of the data subsets to obtain the estimated characteristic values ​​of each of the data subsets, specifically: Using the reverse cloud model to calculate the characteristic parameters of the data subset, the characteristic parameters include characteristic expectation Ex, entropy En, and super entropy He; Calculate the degree of difference of the data subsets and determine whether the characteristic parameters are credible, and use the Bayesian feedback method to correct the characteristic parameters when the characteristic parameters are not credible; The calculating the difference degree of the data subsets and judging whether the characteristic parameters are credible, and using the Bayesian feedback method to correct the characteristic parameters when the characteristic parameters are not credible, comprises: Define the trusted domain and the rejected domain as: When samples in a subset of data When x i Located in the trusted domain, where x i is the i-th sample in the data subset, Z α / 2 is the quantile on both sides of the standard normal distribution, 1-α is the confidence level; when the sample in the data subset When x i is within the rejection region; Calculate the deviation σ of the sample mean falling on both sides of the rejection region as: in, For samples less than The sample mean at time , For samples larger than The sample mean at time d0 is the width of the credible region, If the deviation σ is less than or equal to the preset threshold, the calculated characteristic parameters Ex, En, and He are considered to be credible; if the deviation σ is greater than the preset threshold, the characteristic parameters are corrected using the Yes feedback method; The method of using the Yesian feedback method to correct the characteristic parameters includes: The super entropy is lowered, and Ex, En, and He are corrected through the Bayesian posterior probability distribution to obtain the corrected feature expectation Ex', entropy En', and super entropy He': Where N is the total number of samples in the data subset, represents the sample mean in the data subset, c0 = Ex, S 2 is the sample variance, The deviation of the sample means on both sides of the rejection region is recalculated for the corrected feature expectation, entropy, and super entropy. If the deviation of the sample means on both sides of the rejection region is greater than the preset threshold, the feature expectation, entropy, and super entropy are re-corrected until the deviation of the sample means on both sides of the rejection region is less than or equal to the preset threshold. The feature expectation, entropy, and super entropy at this time are used as the final estimated feature values ​​of the data subset.

4. The method for optimizing distribution network investment projects according to claim 1, characterized in that: The contribution objective function is constructed by combining the estimated characteristic values ​​of each of the data subsets, specifically: Assuming that there are m investment projects to be decided in the project library, each investment project to be decided has n indicators under the contribution evaluation index system of the distribution network investment project, and the value of each indicator is the estimated characteristic value of the corresponding data subset. The contribution objective function is constructed as follows: Among them, F1 represents the contribution objective function, x i ′ indicates whether the i-th investment project to be decided is selected, x i′ =1 means the project is selected for investment, x i ′=0 means that the project is not selected and no investment is made; D ij represents the value of the jth indicator under the i-th investment project to be decided, w j represents the weight of the jth indicator under the i-th investment project to be decided, and n represents the total number of indicators; the optimization goal of the contribution objective function F1 is to maximize the value.

5. The method for optimizing distribution network investment projects according to claim 4, characterized in that: Obtain the historical data corresponding to each of the data subsets and establish a balance objective function, specifically: Get the value e of the jth indicator under the i-th investment project in the last round of investment j and the value of the jth indicator under the i-th investment project to be decided in this round of investment, calculate the incremental deviation e of the jth indicator under the i-th investment project to be decided j 'for: The objective function for calculating the degree of balance is: Among them, F2 represents the balance objective function, Represents the average value of all incremental deviations. The optimization goal of the balance objective function F2 is to minimize the value.

6. The method for optimizing distribution network investment projects according to claim 5, characterized in that: The project optimization objective function is established by combining the contribution objective function and the balance objective function, which is specifically: Use single-objective optimization to solve the contribution objective function and obtain the project investment plan F1 when the contribution objective function F1 is maximized. max , solve F1 max Substitute the balance objective function to obtain the project investment plan Use single-objective optimization to solve the balance objective function and obtain the project investment plan when the balance objective function F2 is minimized. Will solve Substitute the contribution objective function to obtain the project investment plan F1 (2) ; Construct boundary value F1 min and F1 min Meet F1 min ≤F1 (2) , satisfy Construct the target boundary [F1 min , F1 max ] and [F2 min , F2 max ]; The normalized representation ρ(F1) of the contribution target calculated using the membership function is: Among them, d1(F1-F1 min ) / (F1 max -F1 min ), d2(F1-F1 max ) / (F1 max -F1 min ) ; The normalized representation ρ(F2) of the balance target calculated using the membership function is: in, The project optimization objective function F is established as: F=λ1ρ(F1)+λ2ρ(F2), Among them, λ1 and λ2 are target allocation weights.

7. The method for optimizing distribution network investment projects according to claim 1, characterized in that: The project investment amount constraint is expressed as: Among them, x ″ i represents the i-th investment project to be decided, m represents the number of investment projects to be decided, C i represents the funds required for the investment of the i-th investment project to be decided, and C represents the total investment limit; The project necessity construction constraint is expressed as: in, represents the minimum improvement requirement of the jth indicator under the i-th investment project to be decided; x ij Indicates whether the i-th investment project to be decided contributes to the j-th indicator. If so, x ij =1, if there is no contribution then x ij =0.

8. The method for optimizing distribution network investment projects according to claim 7, characterized in that: The project coupling constraints include mutually exclusive constraints and subordinate constraints. The mutually exclusive constraint is expressed as: Among them, S1 represents a set of mutually exclusive items; The dependency constraints are specifically: Construct a set S2 of projects with subordinate relationships, and construct a matrix B with S2-1 rows and S2 columns. S2 represents the number of investment projects to be decided in the set S2. According to all the projects x1″, …, x2 in the set S2, S2 "Construct the project vector X = [x1", ..., x s2 ″] T , x1″ represents the first investment project to be decided in set S2, x S2 ″ represents the S2th investment project to be decided in the set S2; The dependency constraint is expressed as: B·X=0.

9. The method for optimizing distribution network investment projects according to claim 8, characterized in that: The final objective function is: Among them, F represents the project optimization objective function, and max() represents the operation of taking the larger value of the two.

10. A distribution network investment project optimization system, characterized in that: include: An indicator data acquisition module is used to construct a contribution evaluation indicator system for distribution network investment projects, acquire distribution network data, and divide the data into subsets according to the contribution evaluation indicator system for distribution network investment projects; An indicator processing module, used to estimate characteristic parameters of each of the data subsets by combining the inverse cloud model and the Bayesian feedback method to obtain an estimated characteristic value of each of the data subsets; A project optimization objective function construction module is used to construct a contribution objective function in combination with the estimated characteristic values ​​of each of the data subsets, obtain the historical data corresponding to each of the data subsets and establish a balance objective function, and establish a project optimization objective function in combination with the contribution objective function and the balance objective function; The solution module is used to set the project investment amount constraint, the project necessity construction constraint, and the project coupling constraint, and to construct the final objective function in combination with the project optimization objective function, and to use the gray wolf algorithm to solve the final objective function to obtain the final investment plan of the distribution network investment project.

Images