CN114066196A - Power grid investment strategy optimization system - Google Patents

Abstract

A power grid investment strategy optimization system comprises an information acquisition and storage module, an information processing module, an optimization calculation module and a human-computer interaction platform; the optimization calculation module comprises a power load measuring and calculating module, a power grid investment amount measuring and calculating module, a power grid equipment text state identification module and an investment strategy optimization module, wherein the investment strategy optimization module is used for optimizing a target function and a constraint condition according to an investment strategy and obtaining an optimal power grid investment strategy combination based on an intuitionistic normal cloud algorithm. The design fully considers the direct relevant information such as investment management related data, power grid equipment data, electric energy load data and the like, processes basic data to obtain a prediction index, and proposes suggestions related to power grid space distribution and structure optimization under the existing investment decision in a targeted manner, so that the design has guiding significance on a power grid investment strategy.

Description

Power grid investment strategy optimization system

Technical Field

The invention relates to the technical field of power grids, in particular to a power grid investment strategy optimization system.

Background

The power grid is one of the foundations of national economic construction, has extremely important significance for supporting industrial production, guaranteeing the life of residents and maintaining stable social development, and has monopoly characteristics, so that the main investment source is capital and investors outside power companies are few; and the construction investment of the power grid is huge, the cost recovery and the profit cycle are long, so that the investment of the power grid has larger uncertainty and risk, and the investment strategy of the power grid needs to be optimized. The existing investment strategy is only based on the prediction of income, the age structure, the health condition and the space distribution condition of various devices of the power grid are neglected, and the influence of coupling factors such as the power supply quality of the power grid and the load levels of various regions is not considered, so that the formulated investment strategy is directly difficult to directly meet the investment requirement of the power grid.

Disclosure of Invention

The invention aims to overcome the defect and the problem of poor applicability of the existing investment strategy in the prior art, and provides a power grid investment strategy optimization system with good applicability.

In order to achieve the above purpose, the technical solution of the invention is as follows: a power grid investment strategy optimization system comprises an information acquisition and storage module, an information processing module, an optimization calculation module and a human-computer interaction platform;

the information acquisition and storage module comprises a data center and a measuring, calculating and temporary storage module;

the information processing module comprises an information screening and classifying management module, a data filling and abnormal correcting module and a data format normalizing module, wherein the information screening and classifying management module is respectively in signal connection with the data center and the measuring, calculating and temporary storing module, and the information screening and classifying management module is in signal connection with the data format normalizing module through the data filling and abnormal correcting module;

the optimization calculation module comprises a power load measuring and calculating module, a power grid investment amount measuring and calculating module, a power grid equipment text state identification module and an investment strategy optimization module, wherein the power load measuring and calculating module, the power grid investment amount measuring and calculating module and the power grid equipment text state identification module are in signal connection with a data format standardization module, a measuring and calculating temporary storage module and an investment strategy optimization module, and the investment strategy optimization module is in signal connection with a human-computer interaction platform;

the data center is used for extracting and integrating data in the power grid system;

the measuring, calculating and temporary storage module is used for storing the processed index information and sending the index information to the information screening and classification management module;

the information screening and classifying management module is used for screening and classifying data input by the data center and performing dimensionality reduction processing on the data by adopting a principal component analysis method;

the data filling and exception correcting module is used for filling and cleaning data;

the data format normalization module is used for normalizing data by setting and evaluating positive and negative contribution indexes;

the power load measuring and calculating module is used for obtaining a predicted value of the power load;

the power grid investment amount measuring and calculating module is used for obtaining a predicted value of the power grid investment amount according to the stored basic data;

the power grid equipment text state identification module is used for quantizing the equipment state text data to establish a power grid equipment state scoring table;

the investment strategy optimization module is used for optimizing a target function and a constraint condition according to an investment strategy and obtaining an optimal power grid investment strategy combination based on an intuitive normal cloud algorithm;

and the human-computer interaction platform is used for displaying the optimal power grid investment strategy combination.

The data padding is realized by the following steps:

and (3) establishing a Kriging interpolation function to estimate the missing data information:

defining a point model to be estimated:

where x, h is the spatial position, t is the time, and λ is i^thAn observation weight of the location;

introducing a variation function 2 gamma (x, h) and a covariance function cov (t)₁，t₂) The observation weights are calculated by the following set of equations:

wherein,

for introduced random noise errors;

three defects exist in the text record of the power grid equipment: the method comprises the steps of establishing an evaluation rule and a matrix by an equipment layering part, a defect description part and a defect grade part, obtaining a new text vector by a dimension reduction method, and promoting a power grid equipment text based on potential Dirichlet distribution according to the defects and the defect grade of the existing text.

The data cleaning is to eliminate error data according to the logic rule of the influence factors.

The power load measuring and calculating module is specifically used for executing the following steps:

s1, sorting all extreme points of the historical power load sequence, and sorting the extreme points at the left end and the right endCarrying out mirror extension, and obtaining an upper envelope line E through cubic spline interpolation_upAnd a lower envelope E_downAnd decomposing the historical power load sequence to obtain a plurality of product functions PF and a residual component u_kThe sum is as follows:

determining a kernel function of the LSSVM:

where σ is the nuclear width, x_i，x_jIs a corresponding historical power load sequence;

s2, establishing an MVO-LSSVM model, taking the average absolute error of each power load sequence decomposition subsequence as an objective function value, optimizing a penalty factor by adopting an improved multivariate cosmic algorithm, and predicting the power load;

and S3, superposing the power load prediction subsequences under different complex components to form a final power load prediction value.

The power grid investment amount measuring and calculating module is specifically used for executing the following steps:

a BP neural network optimized by a genetic algorithm is built, a real number coding mode is adopted, an input layer is provided with n neural units, a hidden layer is provided with m neural units, only one unit is output, namely, the required predicted value of the investment amount of the power grid is obtained, and the coding length l is as follows:

l＝nm+2m+1

and the mean square error MSE is used as a fitness function, and the predicted value of the investment amount of the power grid is obtained through continuous intersection, selection and variation processing.

The power grid equipment text state identification module is specifically used for executing the following steps:

s1, defining three quality rating indexes of integrity, accuracy and redundancy aiming at problems existing in an actual defect text, and after a single historical defect text is read in, carrying out word segmentation, stop word removal and part-of-speech tagging preprocessing on the single historical defect text by using a hidden Markov model and a Viterbi algorithm in combination with an electric ontology dictionary;

s2, obtaining the score of the defect text on the redundancy according to the character repetition rate; determining the score of the defect text on the accuracy by combining equipment layering words given by the standard and utilizing a character string matching method; determining the score of the defect text on the integrity by using the regular expression and the part-of-speech tagging result;

s3, calculating the weight of the defect text quality index based on an analytic hierarchy process, and combining each index in a weighted manner through pairwise comparison and judgment;

and S4, calculating the defect text quality score based on the self-adaptive gray correlation analysis method, thereby establishing a power grid equipment state score table.

The power grid equipment state scoring table is as follows:

when the state level of the power grid equipment is healthy, the quality score of the power grid equipment is 0.86-1;

when the state level of the power grid equipment is slight danger, the quality score of the power grid equipment is 0.51-0.86;

when the state level of the power grid equipment is moderate risk, the quality score of the power grid equipment is 0.32-0.51;

when the state level of the power grid equipment is high risk, the quality score of the power grid equipment is 0-0.32.

The investment strategy optimization objective function is as follows:

W＝λ₁·R(x)+λ₂·P(x)+λ₃·NVP(x)

wherein W is the overall efficiency, lambda₁For reliability benefit weighting, λ₁R (x) is reliability benefit, λ₂As a social benefit weight, λ₂P (x) is social benefit, λ₃To weight the economic benefit, λ₃NVP (x) is an economic benefit;

wherein n is the number of items to be selected; x is the number of_iA decision variable of the item i, the value of which is 1 represents that the item is selected, and the value of which is 0 represents that the item is not selected; r is the coefficient of reliability benefit, U_iIs the urgency of item i, c_iThe supply/transmission capacity of item i, c the capacity of the original grid, c_imaxFor maximum cost of power outage, S_ifSales revenue for the ith construction project in the t year, C_itThe operation cost of the ith construction project in the T year, T the life cycle of the project, I the return on investment, and Q_iThe construction cost of the ith construction project.

The constraint conditions are as follows:

(1) and (4) investment capacity constraint:

wherein N is_iThe investment amount of the ith project is N, and the N is the maximum investment amount of the power grid;

(2) power load demand constraints:

wherein D is the power load demand of the region;

(3) and (3) reliability constraint:

wherein alpha is an influence coefficient of the residual capacity of the power grid, beta is an influence coefficient of the insufficient capacity of the power grid, and L is the average load of a main line of the whole power grid system;

(4) and (3) power grid equipment state constraint:

where t is the total number of devices required for item i, m_jAnd M is the minimum value of the equipment score of the normal operation of the regional power grid.

The investment strategy optimization module is specifically configured to perform the following steps:

s1, converting the intuitive language variables into an intuitive normal cloud by combining a golden section method and a cloud model;

s2, quantitatively describing the foreground value by introducing a foreground value function and a probability weight function of income and loss;

the foreground cost function is:

wherein, pi (p)_i) Probability weight coefficient for profit and loss, v (Δ x)_i) As a function of the foreground value, Δ x_iIs the difference between the attribute value and the reference point;

wherein, alpha and beta (alpha is more than 0 and beta is less than 1) are risk sensitivity coefficients under the condition of gain or loss relative to a reference point, and lambda is a loss avoidance coefficient;

wherein gamma is a risk attitude coefficient in a profit state, and delta is a risk attitude coefficient in a loss state;

passing through a cloud moldCompleting conversion from a certain definite concept to quantitative data, determining a reference point and a foreground decision matrix of a target to be selected, and calculating to obtain an attribute weight omega_j：

Wherein, Ed_jDistance entropy, which is an attribute;

s3, calculating the comprehensive foreground value of the object to be selected:

and performing descending arrangement on the items to be selected according to the comprehensive prospect value to obtain the optimal power grid investment strategy combination.

Compared with the prior art, the invention has the beneficial effects that:

in the power grid investment strategy optimization system, direct relevant information such as investment management related data, power grid equipment data and electric energy load data is fully considered, basic data is processed to obtain a prediction index, suggestions related to power grid space distribution and structure optimization under the existing investment decision are provided in a targeted manner, and the power grid investment strategy optimization system has guiding significance.

Drawings

Fig. 1 is a block diagram of a power grid investment strategy optimization system according to the present invention.

Fig. 2 is a diagram of a data cluster topology in the present invention.

FIG. 3 is a schematic diagram of the data preprocessing flow of the present invention.

Fig. 4 is a schematic diagram of the calculation process of the investment amount of the power grid in the invention.

FIG. 5 is a schematic diagram of the investment strategy optimization process based on the intuitive normal cloud.

Detailed Description

The present invention will be described in further detail with reference to the following description and embodiments in conjunction with the accompanying drawings.

Referring to fig. 1 to 5, a power grid investment strategy optimization system comprises an information acquisition and storage module, an information processing module, an optimization calculation module and a human-computer interaction platform;

the information acquisition and storage module comprises a data center and a measuring, calculating and temporary storage module;

the information processing module comprises an information screening and classifying management module, a data filling and abnormal correcting module and a data format normalizing module, wherein the information screening and classifying management module is respectively in signal connection with the data center and the measuring, calculating and temporary storing module, and the information screening and classifying management module is in signal connection with the data format normalizing module through the data filling and abnormal correcting module;

the optimization calculation module comprises a power load measuring and calculating module, a power grid investment amount measuring and calculating module, a power grid equipment text state identification module and an investment strategy optimization module, wherein the power load measuring and calculating module, the power grid investment amount measuring and calculating module and the power grid equipment text state identification module are in signal connection with a data format standardization module, a measuring and calculating temporary storage module and an investment strategy optimization module, and the investment strategy optimization module is in signal connection with a human-computer interaction platform;

the data center is used for extracting and integrating data in the power grid system;

the measuring, calculating and temporary storage module is used for storing the processed index information and sending the index information to the information screening and classification management module;

the information screening and classifying management module is used for screening and classifying data input by the data center and performing dimensionality reduction processing on the data by adopting a principal component analysis method;

the data filling and exception correcting module is used for filling and cleaning data;

the data format normalization module is used for normalizing data by setting and evaluating positive and negative contribution indexes;

the power load measuring and calculating module is used for obtaining a predicted value of the power load;

the power grid investment amount measuring and calculating module is used for obtaining a predicted value of the power grid investment amount according to the stored basic data;

the power grid equipment text state identification module is used for quantizing the equipment state text data to establish a power grid equipment state scoring table;

the investment strategy optimization module is used for optimizing a target function and a constraint condition according to an investment strategy and obtaining an optimal power grid investment strategy combination based on an intuitive normal cloud algorithm;

and the human-computer interaction platform is used for displaying the optimal power grid investment strategy combination.

The data padding is realized by the following steps:

and (3) establishing a Kriging interpolation function to estimate the missing data information:

defining a point model to be estimated:

where x, h is the spatial position, t is the time, and λ is i^thAn observation weight of the location;

introducing a variation function 2 gamma (x, h) and a covariance function cov (t)₁，t₂) The observation weights are calculated by the following set of equations:

wherein,

for introduced random noise errors;

three defects exist in the text record of the power grid equipment: the method comprises the steps of establishing an evaluation rule and a matrix by an equipment layering part, a defect description part and a defect grade part, obtaining a new text vector by a dimension reduction method, and promoting a power grid equipment text based on potential Dirichlet distribution according to the defects and the defect grade of the existing text.

The data cleaning is to eliminate error data according to the logic rule of the influence factors.

The power load measuring and calculating module is specifically used for executing the following steps:

s1, sequencing all extreme points of the historical power load sequence, carrying out mirror image continuation on the extreme points at the left end and the right end, and obtaining an upper envelope line E through cubic spline interpolation_upAnd a lower envelope E_downAnd decomposing the historical power load sequence to obtain a plurality of product functions PF and a residual component u_kThe sum is as follows:

determining a kernel function of the LSSVM:

where σ is the nuclear width, x_i，x_jIs a corresponding historical power load sequence;

s2, establishing an MVO-LSSVM model, taking the average absolute error of each power load sequence decomposition subsequence as an objective function value, optimizing a penalty factor by adopting an improved multivariate cosmic algorithm, and predicting the power load;

and S3, superposing the power load prediction subsequences under different complex components to form a final power load prediction value.

The power grid investment amount measuring and calculating module is specifically used for executing the following steps:

a BP neural network optimized by a genetic algorithm is built, a real number coding mode is adopted, an input layer is provided with n neural units, a hidden layer is provided with m neural units, only one unit is output, namely, the required predicted value of the investment amount of the power grid is obtained, and the coding length l is as follows:

l＝nm+2m+1

and the mean square error MSE is used as a fitness function, and the predicted value of the investment amount of the power grid is obtained through continuous intersection, selection and variation processing.

The power grid equipment text state identification module is specifically used for executing the following steps:

s1, defining three quality rating indexes of integrity, accuracy and redundancy aiming at problems existing in an actual defect text, and after a single historical defect text is read in, carrying out word segmentation, stop word removal and part-of-speech tagging preprocessing on the single historical defect text by using a hidden Markov model and a Viterbi algorithm in combination with an electric ontology dictionary;

s2, obtaining the score of the defect text on the redundancy according to the character repetition rate; determining the score of the defect text on the accuracy by combining equipment layering words given by the standard and utilizing a character string matching method; determining the score of the defect text on the integrity by using the regular expression and the part-of-speech tagging result;

s3, calculating the weight of the defect text quality index based on an analytic hierarchy process, and combining each index in a weighted manner through pairwise comparison and judgment;

and S4, calculating the defect text quality score based on the self-adaptive gray correlation analysis method, thereby establishing a power grid equipment state score table.

The power grid equipment state scoring table is as follows:

when the state level of the power grid equipment is healthy, the quality score of the power grid equipment is 0.86-1;

when the state level of the power grid equipment is slight danger, the quality score of the power grid equipment is 0.51-0.86;

when the state level of the power grid equipment is moderate risk, the quality score of the power grid equipment is 0.32-0.51;

when the state level of the power grid equipment is high risk, the quality score of the power grid equipment is 0-0.32.

The investment strategy optimization objective function is as follows:

W＝λ₁·R(x)+λ₂·P(x)+λ₃·NVP(x)

wherein W is the overall efficiency, lambda₁For reliability benefit weighting, λ₁R (x) is reliability benefit, λ₂As a social benefit weight, λ₂P (x) is social benefit, λ₃To weight the economic benefit, λ₃NVP (x) is an economic benefit;

wherein n is the number of items to be selected; x is the number of_iA decision variable of the item i, the value of which is 1 represents that the item is selected, and the value of which is 0 represents that the item is not selected; r is the coefficient of reliability benefit, U_iIs the urgency of item i, c_iThe supply/transmission capacity of item i, c the capacity of the original grid, c_{i max}For maximum cost of power outage, S_itSales revenue for the ith construction project in the t year, C_itThe operation cost of the ith construction project in the T year, T the life cycle of the project, I the return on investment, and Q_iThe construction cost of the ith construction project.

The constraint conditions are as follows:

(1) and (4) investment capacity constraint:

wherein N is_iThe investment amount of the ith project is N, and the N is the maximum investment amount of the power grid;

(2) power load demand constraints:

wherein D is the power load demand of the region;

(3) and (3) reliability constraint:

wherein alpha is an influence coefficient of the residual capacity of the power grid, beta is an influence coefficient of the insufficient capacity of the power grid, and L is the average load of a main line of the whole power grid system;

(4) and (3) power grid equipment state constraint:

where t is the total number of devices required for item i, m_jAnd M is the minimum value of the equipment score of the normal operation of the regional power grid.

The investment strategy optimization module is specifically configured to perform the following steps:

s1, converting the intuitive language variables into an intuitive normal cloud by combining a golden section method and a cloud model;

s2, quantitatively describing the foreground value by introducing a foreground value function and a probability weight function of income and loss;

the foreground cost function is:

wherein, pi (p)_i) Probability weight coefficient for profit and loss, v (Δ x)_i) As a function of the foreground value, Δ x_iIs the difference between the attribute value and the reference point;

wherein, alpha and beta (alpha is more than 0 and beta is less than 1) are risk sensitivity coefficients under the condition of gain or loss relative to a reference point, and lambda is a loss avoidance coefficient;

wherein gamma is a risk attitude coefficient in a profit state, and delta is a risk attitude coefficient in a loss state;

the conversion from a certain definite concept to quantitative data is completed through a cloud model, a reference point and a target foreground decision matrix to be selected are determined, and an attribute weight omega is obtained through calculation_j：

Wherein, Ed_jDistance entropy, which is an attribute;

s3, calculating the comprehensive foreground value of the object to be selected:

and performing descending arrangement on the items to be selected according to the comprehensive prospect value to obtain the optimal power grid investment strategy combination.

The principle of the invention is illustrated as follows:

as shown in fig. 2, a large data platform at the user side of the power grid is established, wherein a data center data set comprises basic data stored by each system, such as investment risk, equipment use and aging condition texts, power quality, load level, equipment use rate, space distribution condition and the like, and factors influencing the power grid investment strategy are complicated. One PC is used as a main node to be responsible for resource allocation and job scheduling of the whole cluster and management of calling and maintaining of information data, the other PCs are used as data nodes and are mainly responsible for storing data information, the main node divides and stores the data, each data division block is stored in each data node in a redundant mode, and each data division block is stored in 3 data nodes by default. The basic information can be obtained by extracting data stored in existing PMS (equipment management system) systems and SCADA (data acquisition and monitoring control) systems in the power grid, which are collectively called data centers, as basic input information.

The LSSVM is an improvement on an SVM (support vector machine), can convert a problem into a problem of solving a linear equation set, changes inequality constraint into equality constraint and improves convergence speed.

Example (b):

referring to fig. 1, a power grid investment strategy optimization system comprises an information acquisition and storage module, an information processing module, an optimization calculation module and a human-computer interaction platform; the information acquisition and storage module comprises a data center and a measuring, calculating and temporary storage module; the information processing module comprises an information screening and classifying management module, a data filling and abnormal correcting module and a data format normalizing module, wherein the information screening and classifying management module is respectively in signal connection with the data center and the measuring, calculating and temporary storing module, and the information screening and classifying management module is in signal connection with the data format normalizing module through the data filling and abnormal correcting module; the optimization calculation module comprises a power load measuring and calculating module, a power grid investment amount measuring and calculating module, a power grid equipment text state identification module and an investment strategy optimization module, wherein the power load measuring and calculating module, the power grid investment amount measuring and calculating module and the power grid equipment text state identification module are in signal connection with a data format standardization module, a measuring and calculating temporary storage module and an investment strategy optimization module, and the investment strategy optimization module is in signal connection with a human-computer interaction platform; the data center is used for extracting and integrating data in the power grid system; the measuring, calculating and temporary storage module is used for storing the processed index information and sending the index information to the information screening and classification management module; the information screening and classifying management module is used for screening and classifying data input by the data center and performing dimensionality reduction processing on the data by adopting a principal component analysis method; the data filling and exception correcting module is used for filling and cleaning data; the data format normalization module is used for normalizing data by setting and evaluating positive and negative contribution indexes; the power load measuring and calculating module is used for obtaining a predicted value of the power load; the power grid investment amount measuring and calculating module is used for obtaining a predicted value of the power grid investment amount according to the stored basic data; the power grid equipment text state identification module is used for quantizing the equipment state text data to establish a power grid equipment state scoring table; the investment strategy optimization module is used for optimizing a target function and a constraint condition according to an investment strategy and obtaining an optimal power grid investment strategy combination based on an intuitive normal cloud algorithm; the human-computer interaction platform comprises an investment candidate item input interface and an investment strategy optimization display interface, the initial investment strategy combination, the basic information input and the expression editing of the optimization model are realized through the human-computer interaction input interface, the investment direction possibly existing in the investment is determined, and the final optimal power grid investment strategy combination is displayed through the investment strategy optimization display interface.

Firstly, data is extracted and integrated, and the integration method adopted by the embodiment is an open-source Sqoop tool, so that data transmission of information among various databases is realized. The client can directly extract the relevant information from the data node, and can also access the main node to extract the packaged overall relevant data. The information is input into the information screening and classifying management module for screening and classifying, and the screening and classifying process is influenced by the intermediate information processed before being stored in the measuring and calculating temporary storage module. Because the data contained in the data center is too complicated, the principal component analysis method is adopted to perform dimensionality reduction processing on the data, and the screened data not only comprises the asset information of the existing power grid, but also comprises the service condition and aging information of various devices in the power grid and the power load and power quality information of each place.

Data padding refers to missing or unrecorded data due to possible missing or mismanagement of data recording. Because factors influencing the power grid investment strategy are not only related to time, such as equipment use condition, power load and other factors, but also related to the geographic position, through data self logic, based on the influence of the related factors such as the geographic position, the time and the like, referring to fig. 3, a kriging interpolation function is established to estimate data information missing per se, and data basis for optimizing the power grid investment strategy is added:

defining a point model to be estimated:

wherein x, h are spacesPosition, t is time, λ is i^thAn observation weight of the location;

introducing a variation function 2 gamma (x, h) and a covariance function cov (t)₁，t₂) The observation weights are calculated by the following set of equations:

wherein,

for introduced random noise errors;

meanwhile, for the particularity of the work of the power grid, text information of the use condition of the equipment is recorded by text methods such as an operation ticket, so that three defects exist in the text recording of the equipment: the method comprises the steps of establishing an evaluation rule and a matrix by an equipment layering part, a defect description part and a defect grade part, obtaining a new text vector by a dimension reduction means, promoting the text of the power grid equipment based on potential Dirichlet distribution according to the defects and the defect grade of the existing text, and increasing the basis of investment strategy optimization. The data cleaning is to remove the error data according to the logic rule of the basic influence factor, for example, the condition of obvious data error such as the load fluctuation level exceeding 100% and the like. The normalization of the data format refers to the process of making the data information consistent in format, length and logic structure, for example, by setting and evaluating positive and negative contribution indexes to normalize the data, the influence of data units is eliminated.

The optimization calculation module can further process the screened final multi-dimensional indexes, and data stored in the data center are basic data, namely the data can be directly collected and recorded through the sensor, so that in order to increase the accuracy of optimization of the power grid investment strategy, the data needs to be processed to obtain intermediate indexes such as power load, power grid investment amount and power grid equipment state.

Referring to fig. 4, the electrical load calculation module is specifically configured to perform the following steps:

s1, historySequencing all extreme points of the power load sequence, carrying out mirror image continuation on the extreme points at the left end and the right end, and obtaining an upper envelope line E through cubic spline interpolation_upAnd a lower envelope E_downAnd decomposing the historical power load sequence to obtain a plurality of product functions PF and a residual component u_kThe sum is as follows:

determining a kernel function of the LSSVM:

where σ is the nuclear width, x_i，x_jIs a corresponding historical power load sequence;

then converting the regression problem into a constraint problem according to a power grid risk minimization principle;

s2, establishing an MVO-LSSVM model, taking the average absolute error of each power load sequence decomposition subsequence as an objective function value, optimizing a penalty factor by adopting an improved multivariate cosmic algorithm, and predicting the power load;

establishing a mathematical model of a universe, a white hole and a black hole, wherein the universe corresponds to a penalty factor object in an LSSVM, and the object with high expansion rate always tends to an object with low expansion rate in the random creation process of the universe, so that the object can be transferred under the action of universal gravitation, and a search space tends to an optimal position;

firstly, changing a search process through establishing an initial universe matrix and giving a penalty factor initial value and through an established white hole model, a black hole model and a wormhole model, and providing a new travel distance TDR as shown in the specification as a limit for each step of search distance, wherein L is the current iteration number, and L is the maximum iteration number;

further changing the expansion rate of the universe, outputting an optimal penalty factor universe matrix when the search times reach a preset upper limit value, and optimizing the prediction process of the LSSVM;

and S3, superposing the power load prediction subsequences under different complex components to form a final power load prediction value.

The power grid investment amount measuring and calculating module is specifically used for executing the following steps:

a BP neural network optimized by a genetic algorithm is built, a real number coding mode is adopted, an input layer is provided with n neural units, a hidden layer is provided with m neural units, only one unit is output, namely, the required predicted value of the investment amount of the power grid is obtained, and the coding length l is as follows:

l＝nm+2m+1

and the mean square error MSE is used as a fitness function, and the predicted value of the investment amount of the power grid is obtained through continuous intersection, selection and variation processing.

The power grid equipment text state identification module is specifically used for executing the following steps:

s1, defining three quality rating indexes of integrity, accuracy and redundancy aiming at problems existing in an actual defect text, and after a single historical defect text is read in, carrying out word segmentation, stop word removal and part-of-speech tagging preprocessing on the single historical defect text by using a hidden Markov model and a Viterbi algorithm in combination with an electric ontology dictionary;

s2, obtaining the score of the defect text on the redundancy according to the character repetition rate; determining the score of the defect text on the accuracy by combining equipment layering words given by the standard and utilizing a character string matching method; determining the score of the defect text on the integrity by using the regular expression and the part-of-speech tagging result;

s3, calculating the weight of the defect text quality index based on an analytic hierarchy process, and combining each index in a weighted manner through pairwise comparison and judgment;

s4, calculating a defect text quality score based on a self-adaptive grey correlation analysis method, determining a resolution coefficient according to the actual quality of the defect text, and enabling the final quality score to clearly represent the quality of the defect text, thereby establishing a power grid equipment state score table.

The power grid equipment state scoring table is as follows:

when the state level of the power grid equipment is healthy, the quality score of the power grid equipment is 0.86-1;

when the state level of the power grid equipment is slight danger, the quality score of the power grid equipment is 0.51-0.86;

when the state level of the power grid equipment is moderate risk, the quality score of the power grid equipment is 0.32-0.51;

when the state level of the power grid equipment is high risk, the quality score of the power grid equipment is 0-0.32.

The investment strategy optimization objective function is as follows:

W＝λ₁·R(x)+λ₂·P(x)+λ₃·NVP(x)

wherein W is the overall efficiency, lambda₁For reliability benefit weighting, λ₁R (x) is reliability benefit, λ₂As a social benefit weight, λ₂P (x) is social benefit, λ₃To weight the economic benefit, λ₃NVP (x) is an economic benefit;

wherein n is the number of items to be selected; x is the number of_iA decision variable of the item i, the value of which is 1 represents that the item is selected, and the value of which is 0 represents that the item is not selected; r is the coefficient of reliability benefit, U_iIs the urgency of item i, c_iThe supply/transmission capacity of item i, c the capacity of the original grid, c_{i max}For maximum cost of power outage, S_itFor the ith construction projectSales revenue of the t year, C_itThe operation cost of the ith construction project in the T year, T the life cycle of the project, I the return on investment, and Q_iThe construction cost of the ith construction project.

The constraint conditions are as follows:

(1) and (4) investment capacity constraint:

wherein N is_iThe investment amount of the ith project is N, and the N is the maximum investment amount of the power grid;

(2) power load demand constraints:

wherein D is the power load demand of the region;

(3) and (3) reliability constraint:

the reliability of investment is reflected by calling the current power grid capacity-load ratio stored in the SCADA, and the capacity-load ratio is selected to be in the range of [1.5, 1.8] and [1.8, 2.0] as the constraint condition of the reliability, as follows:

wherein alpha is an influence coefficient of the residual capacity of the power grid, beta is an influence coefficient of the insufficient capacity of the power grid, and L is the average load of a main line of the whole power grid system;

(4) and (3) power grid equipment state constraint:

where t is the total number of devices required for item i, m_jThe state score of the equipment is obtained, and M is the minimum value of the equipment score of the normal work of the regional power grid;

(5) investment project constraint:

assuming that the grid has n items in total, there are several relationship constraints between these items:

if the items are independent of each other, then there is x₁+x₂+…x_n≤n；

If the items are in a mutually exclusive relationship, then there is x₁+x₂+…x_n≤l；

If the items are interdependent, that is, only if item 1 is selected, item 2 may be selected; on the contrary, if item 1 is not selected, item 2 is not likely to be selected, and x is present₁-x₂≥0；

If there is a close dependency relationship between items, that is: if the two items must be selected or not selected at the same time, then x is present₁-x₂＝0；

If the items are in a complementary relationship, it is shown that item 1 and item 2 can be used as complementary item x₁₂Selected at the same time, but item 1 and complementary item x₁₂Item 2 and complementary item x₁₂Cannot be selected at the same time, then x is₁+x₂+x₁₂≤1；

Because the complexity of the power grid investment decision problem can affect uncertain factors such as behaviors, psychology, cognition and the like of decision makers, and if the risk acceptance degrees of various decision makers to power grid investment projects are different, the investment strategies cannot be accurately optimized only by setting an investment strategy optimization objective function and constraint conditions, and a proper power grid investment project combination is selected. Therefore, the investment strategy is further optimized by introducing the intuitive normal cloud into the investment strategy optimization module, and the subjective uncertainty of a decision maker is considered from the aspect of combining the qualitative and quantitative aspects.

Referring to fig. 5, the investment strategy optimization module is specifically configured to perform the following steps:

s1, converting the intuitive language variable into an intuitive normal cloud by combining a golden section method and a cloud model, and fully considering the ambiguity and the randomness of decision information;

s2, quantitatively describing the foreground value by introducing a foreground value function and a probability weight function of income and loss;

the foreground cost function is:

wherein, pi (p)_i) Probability weight coefficient for profit and loss, v (Δ x)_i) As a function of the foreground value, Δ x_iIs the difference between the attribute value and the reference point;

wherein, alpha and beta (alpha is more than 0 and beta is less than 1) are risk sensitivity coefficients under the condition of gain or loss relative to a reference point, and lambda is a loss avoidance coefficient;

wherein gamma is a risk attitude coefficient in a profit state, and delta is a risk attitude coefficient in a loss state;

the conversion from a certain definite concept to quantitative data is completed through a cloud model, a reference point and a target foreground decision matrix to be selected are determined, and an attribute weight omega is obtained through calculation_j：

Wherein, Ed_jDistance entropy, which is an attribute;

s3, calculating the comprehensive foreground value of the object to be selected:

and performing descending arrangement on the items to be selected according to the comprehensive prospect value to obtain the optimal power grid investment strategy combination.

Finally, the optimal investment strategy combination and the specific scores of the investment strategies under different risk conditions are displayed through a human-computer interaction platform, the relevant information such as investment management relevant data, power grid equipment data and power quality data is comprehensively considered, suggestions about power grid space distribution and structure optimization under the existing investment decision are provided in a targeted manner, and the method has guiding significance on the power grid investment strategy combination.

Claims (10)

1. A power grid investment strategy optimization system is characterized by comprising an information acquisition and storage module, an information processing module, an optimization and calculation module and a man-machine interaction platform;

the information acquisition and storage module comprises a data center and a measuring, calculating and temporary storage module;

the information processing module comprises an information screening and classifying management module, a data filling and abnormal correcting module and a data format normalizing module, wherein the information screening and classifying management module is respectively in signal connection with the data center and the measuring, calculating and temporary storing module, and the information screening and classifying management module is in signal connection with the data format normalizing module through the data filling and abnormal correcting module;

the optimization calculation module comprises a power load measuring and calculating module, a power grid investment amount measuring and calculating module, a power grid equipment text state identification module and an investment strategy optimization module, wherein the power load measuring and calculating module, the power grid investment amount measuring and calculating module and the power grid equipment text state identification module are in signal connection with a data format standardization module, a measuring and calculating temporary storage module and an investment strategy optimization module, and the investment strategy optimization module is in signal connection with a human-computer interaction platform;

the data center is used for extracting and integrating data in the power grid system;

the measuring, calculating and temporary storage module is used for storing the processed index information and sending the index information to the information screening and classification management module;

the information screening and classifying management module is used for screening and classifying data input by the data center and performing dimensionality reduction processing on the data by adopting a principal component analysis method;

the data filling and exception correcting module is used for filling and cleaning data;

the data format normalization module is used for normalizing data by setting and evaluating positive and negative contribution indexes;

the power load measuring and calculating module is used for obtaining a predicted value of the power load;

the power grid investment amount measuring and calculating module is used for obtaining a predicted value of the power grid investment amount according to the stored basic data;

the power grid equipment text state identification module is used for quantizing the equipment state text data to establish a power grid equipment state scoring table;

the investment strategy optimization module is used for optimizing a target function and a constraint condition according to an investment strategy and obtaining an optimal power grid investment strategy combination based on an intuitive normal cloud algorithm;

and the human-computer interaction platform is used for displaying the optimal power grid investment strategy combination.

2. The system according to claim 1, wherein the system comprises: the data padding is realized by the following steps:

and (3) establishing a Kriging interpolation function to estimate the missing data information:

defining a point model to be estimated:

where x, h is the spatial position, t is the time, and λ is i^thAn observation weight of the location;

introducing a variation function 2 gamma (x, h) and a covariance function cov (t)₁，t₂) The observation weights are calculated by the following set of equations:

wherein,

for introduced random noise errors;

three defects exist in the text record of the power grid equipment: the method comprises the steps of establishing an evaluation rule and a matrix by an equipment layering part, a defect description part and a defect grade part, obtaining a new text vector by a dimension reduction method, and promoting a power grid equipment text based on potential Dirichlet distribution according to the defects and the defect grade of the existing text.

3. The system according to claim 1, wherein the system comprises: the data cleaning is to eliminate error data according to the logic rule of the influence factors.

4. The system according to claim 1, wherein the system comprises: the power load measuring and calculating module is specifically used for executing the following steps:

s1, sequencing all extreme points of the historical power load sequence, carrying out mirror image continuation on the extreme points at the left end and the right end, and obtaining an upper envelope line E through cubic spline interpolation_upAnd a lower envelope E_downAnd decomposing the historical power load sequence to obtain a plurality of product functions PF and a residual component u_kThe sum is as follows:

determining a kernel function of the LSSVM:

where σ is the nuclear width, x_i，x_jIs a corresponding historical power load sequence;

s2, establishing an MVO-LSSVM model, taking the average absolute error of each power load sequence decomposition subsequence as an objective function value, optimizing a penalty factor by adopting an improved multivariate cosmic algorithm, and predicting the power load;

and S3, superposing the power load prediction subsequences under different complex components to form a final power load prediction value.

5. The system according to claim 1, wherein the system comprises: the power grid investment amount measuring and calculating module is specifically used for executing the following steps:

a BP neural network optimized by a genetic algorithm is built, a real number coding mode is adopted, an input layer is provided with n neural units, a hidden layer is provided with m neural units, only one unit is output, namely, the required predicted value of the investment amount of the power grid is obtained, and the coding length l is as follows:

l＝nm+2m+1

and the mean square error MSE is used as a fitness function, and the predicted value of the investment amount of the power grid is obtained through continuous intersection, selection and variation processing.

6. The system according to claim 1, wherein the system comprises: the power grid equipment text state identification module is specifically used for executing the following steps:

s1, defining three quality rating indexes of integrity, accuracy and redundancy aiming at problems existing in an actual defect text, and after a single historical defect text is read in, carrying out word segmentation, stop word removal and part-of-speech tagging preprocessing on the single historical defect text by using a hidden Markov model and a Viterbi algorithm in combination with an electric ontology dictionary;

s2, obtaining the score of the defect text on the redundancy according to the character repetition rate; determining the score of the defect text on the accuracy by combining equipment layering words given by the standard and utilizing a character string matching method; determining the score of the defect text on the integrity by using the regular expression and the part-of-speech tagging result;

s3, calculating the weight of the defect text quality index based on an analytic hierarchy process, and combining each index in a weighted manner through pairwise comparison and judgment;

and S4, calculating the defect text quality score based on the self-adaptive gray correlation analysis method, thereby establishing a power grid equipment state score table.

7. The power grid investment strategy optimization system of claim 6, wherein: the power grid equipment state scoring table is as follows:

when the state level of the power grid equipment is healthy, the quality score of the power grid equipment is 0.86-1;

when the state level of the power grid equipment is slight danger, the quality score of the power grid equipment is 0.51-0.86;

when the state level of the power grid equipment is moderate risk, the quality score of the power grid equipment is 0.32-0.51;

when the state level of the power grid equipment is high risk, the quality score of the power grid equipment is 0-0.32.

8. The system according to claim 1, wherein the system comprises:

the investment strategy optimization objective function is as follows:

W＝λ₁·R(x)+λ₂·P(x)+λ₃·NVP(x)

wherein W is the overall efficiency, lambda₁For reliability benefit weighting, λ₁R (x) is reliability benefit, λ₂As a social benefit weight, λ₂P (x) is social benefit, λ₃To weight the economic benefit, λ₃NVP (x) is an economic benefit;

wherein n is the number of items to be selected; x is the number of_iA decision variable of the item i, the value of which is 1 represents that the item is selected, and the value of which is 0 represents that the item is not selected; r is the coefficient of reliability benefit, U_iIs the urgency of item i, c_iThe supply/transmission capacity of item i, c the capacity of the original grid, c_imaxFor maximum cost of power outage, S_itSales revenue for the ith construction project in the t year, C_itThe operation cost of the ith construction project in the T year, T the life cycle of the project, I the return on investment, and Q_iThe construction cost of the ith construction project.

9. The system according to claim 8, wherein the power grid investment strategy optimization system comprises:

the constraint conditions are as follows:

(1) and (4) investment capacity constraint:

wherein N is_iThe investment amount of the ith project is N, and the N is the maximum investment amount of the power grid;

(2) power load demand constraints:

wherein D is the power load demand of the region;

(3) and (3) reliability constraint:

wherein alpha is an influence coefficient of the residual capacity of the power grid, beta is an influence coefficient of the insufficient capacity of the power grid, and L is the average load of a main line of the whole power grid system;

(4) and (3) power grid equipment state constraint:

where t is the total number of devices required for item i, m_jAnd M is the minimum value of the equipment score of the normal operation of the regional power grid.

10. The system according to claim 1, wherein the system comprises: the investment strategy optimization module is specifically configured to perform the following steps:

s1, converting the intuitive language variables into an intuitive normal cloud by combining a golden section method and a cloud model;

s2, quantitatively describing the foreground value by introducing a foreground value function and a probability weight function of income and loss;

the foreground cost function is:

wherein, pi (p)_i) Probability weight coefficient for profit and loss, v (Δ x)_i) As a function of the foreground value, Δ x_iIs the difference between the attribute value and the reference point;

wherein, alpha and beta (alpha is more than 0 and beta is less than 1) are risk sensitivity coefficients under the condition of gain or loss relative to a reference point, and lambda is a loss avoidance coefficient;

wherein gamma is a risk attitude coefficient in a profit state, and delta is a risk attitude coefficient in a loss state;

the conversion from a certain definite concept to quantitative data is completed through a cloud model, a reference point and a target foreground decision matrix to be selected are determined, and an attribute weight omega is obtained through calculation_j：

Wherein, Ed_jDistance entropy, which is an attribute;

s3, calculating the comprehensive foreground value of the object to be selected:

and performing descending arrangement on the items to be selected according to the comprehensive prospect value to obtain the optimal power grid investment strategy combination.

Images