CN110837939A

CN110837939A - Power grid multi-target project screening method and system

Info

Publication number: CN110837939A
Application number: CN201911133657.4A
Authority: CN
Inventors: 王晞; 胥威汀; 任志超; 叶强; 王海燕; 程超; 马瑞光; 陈博
Original assignee: Economic and Technological Research Institute of State Grid Sichuan Electric Power Co Ltd
Current assignee: Economic and Technological Research Institute of State Grid Sichuan Electric Power Co Ltd
Priority date: 2019-11-19
Filing date: 2019-11-19
Publication date: 2020-02-25

Abstract

A power grid multi-target project screening method and a system thereof are provided, the method comprises the following steps: establishing a three-level network structure of a power grid project, wherein the three-level network structure comprises a three-level demand layer, a two-level demand layer and a one-level demand layer, and the one-level demand layer comprises economic requirements, safety requirements and policy requirements; constructing a multi-target project screening model; respectively calculating the satisfaction degree of each demand index in the first-level demand layer by using a target function of the screening model for the candidate items in the k candidate areas, and screening out the items meeting the demand; and evaluating the project meeting the requirements by adopting the constraint conditions of the screening model to obtain an optimal project planning scheme. The method and the system respectively consider the requirements of an economic target, a safety target and a technical policy type target c target and consider the constraint conditions such as capital constraint, project constraint and the like to determine project screening, can realize the optimization of all the satisfaction degrees, and have high calculation accuracy and high calculation efficiency.

Description

Power grid multi-target project screening method and system

Technical Field

The invention relates to the technical field of electric power system analysis, in particular to a method and a system for screening multi-target projects of a power grid.

Background

With the entering of the economic development of China into a new normal state, the contradiction between the high-intensity investment development, the increase of the cost rigidity and the gradual increase of the electric quantity speed and the difficulty in the increase of the benefit is increasingly prominent, the difficulty in keeping stable operation and completing the profit target is increased, and higher requirements are put forward on the investment decision of a power grid. Therefore, the development of investment optimization planning of power grid projects has important theoretical and practical guiding significance for ensuring reasonable investment scale, accurate direction, structure optimization, scientific time sequence, strict control on low-efficiency investment and stopping of invalid investment.

No calculations regarding investment direction and structure are seen in the prior art. However, the investment optimization management is the key for the safety development of the power grid and the health development of enterprises. When the power grid project investment planning optimization is carried out, a plurality of candidate projects exist in different regions, and how to select a more optimal investment structure and investment direction from the plurality of candidate projects in the plurality of regions has important significance.

Disclosure of Invention

The invention provides a method and a system for screening multi-target projects of a power grid, which solve the problems of low efficiency and ineffective investment caused by lack of optimization of power grid project investment planning when a more optimal investment structure and investment direction are selected from a plurality of candidate projects of a plurality of regions in the prior art.

A power grid multi-target project screening method comprises the following steps:

establishing a three-level network structure of a power grid project, wherein the three-level network structure comprises a three-level demand layer, a two-level demand layer and a one-level demand layer, and the one-level demand layer comprises economic requirements, safety requirements and policy requirements;

constructing a multi-target project screening model, wherein the screening model has a target function and constraint conditions;

respectively calculating the satisfaction degree of each demand index in a first-level demand layer by using a target function of a screening model for the candidate items in the k candidate areas, and screening the items meeting the demand, wherein at least one candidate item is arranged in each candidate area;

and evaluating the project meeting the requirements by adopting the constraint conditions of the screening model to obtain an optimal project planning scheme.

The screening method disclosed by the technical scheme is suitable for being executed in computing equipment, and is suitable for selecting an optimal planning scheme from l candidate items in k candidate areas, wherein each candidate area is provided with at least one candidate item.

Specifically, the screening method comprises the steps of firstly establishing a three-level network structure of a power grid project and constructing a multi-target project screening model. The three-level network structure comprises a three-level demand layer, a two-level demand layer and a one-level demand layer, wherein the three-level demand layer is a demand index layer. The third-level demand layer comprises at least one third-level demand, the second-level demand layer comprises at least one second-level demand, and the first-level demand layer comprises three first-level demands, namely an economic demand, a safety demand and a policy demand. The multi-target project screening model is provided with an objective function and constraint conditions, wherein the objective function comprises the maximum economic requirement satisfaction degree, the maximum safety requirement satisfaction degree and the maximum policy requirement satisfaction degree.

After the three-level network structure and the screening model are established, the satisfaction degree of each candidate item in each candidate area to three first-level requirements is calculated by adopting an objective function, the screening model is solved by adopting a preset algorithm, and the optimal planning scheme is obtained and comprises a plurality of objective items in the k candidate areas. In other words, after the screening model is built, items meeting the requirements are screened out from the candidate items by using the objective function, and then the items meeting the requirements are ranked and ordered by using the constraint conditions, so that the optimal planning scheme is obtained finally.

As a preferred embodiment of the objective function in the present invention, the objective function includes:

wherein, maxf₁、maxf₂、maxf₃Respectively means that the economic demand satisfaction degree is maximum, the safety demand satisfaction degree is maximum, the policy demand satisfaction degree is maximum, and Z_1,j、Z_2,j、Z_3,jRespectively the satisfaction degrees of the ith project on economic requirements, safety requirements and policy requirements; u shape_iThe decision variable, which takes the value 0 or 1, represents the decision whether to select the ith item.

As a preferred embodiment of the constraints of the present invention, the constraints comprise one or more of capital constraints, project constraints, regional constraints, and minimum requirements constraints.

The capital constraint is that the sum of investment capital of all projects does not exceed the total capital upper limit I_maxThe calculation formula is as follows:

wherein, I_iTo invest the funds required for the ith project.

The project constraint is that the total number of investment projects is within a predetermined range, and the calculation formula is as follows:

wherein N is_minAs a lower limit on the total number of investment items, N_maxAn upper limit on the total number of investment items.

The region constraint is that each region has items to be selected, and each candidate region is set as M_j(j ═ 1,2, …, k), then the area constraint is calculated as:

the minimum requirement constraint is that the satisfaction degree of each first-level requirement is not lower than a lower limit value, and the calculation formula is as follows:

wherein A is_min、B_min、C_minAnd respectively setting the lower limit of the meeting degree of all target projects on the total economic requirement, the total safety requirement and the total policy requirement.

Further, the predetermined algorithm is an improved particle swarm algorithm, and the multi-target project screening model is decided according to the improved particle swarm algorithm.

Further, the step of calculating the satisfaction degree of each candidate item to each demand index in the first-level demand layer by using the objective function of the screening model specifically includes:

for the three-level network structure, an ANP network analysis method is adopted, the upper-level network layer is used as a criterion, each element set in the current-level network layer is used as a secondary criterion, and a first weight matrix of all three-level requirements in the three-level requirement layer and a second weight matrix of all two-level requirements in the two-level requirement layer are respectively calculated;

acquiring a satisfaction degree matrix of each three-level requirement of the project, and establishing a first mapping relation from the three-level requirement to the second-level requirement and a second mapping relation from the second-level requirement to the first-level requirement;

and calculating the satisfaction degree of each item in each candidate area to the primary requirement based on the established first mapping relation and the second mapping relation.

In some embodiments, the first weight matrix is

The second weight matrix isThe first mapping relationship is Y ═ X ° a, the second mapping relationship is Z ═ Y ° B, wherein,

is the weight value of the i-th index,

the value is the weighted value of the r-th requirement, X is a satisfaction degree matrix of each requirement index, Y and Z represent the satisfaction degree of the item on the second-level requirement and the first-level requirement respectively, and omicron is a generalized model synthesis operator.

Further, the second-level requirements in the second-level requirement layer comprise one or more of newly increased load requirements, conventional power supply sending-out, power supply access, marketing propaganda, grid structure optimization, grid reliability improvement, heavy overload solving, potential safety hazard elimination, coal-to-electricity, rural power grid transformation, remote power-on engineering and new energy source sending-out engineering.

Further, the three-level requirements in the three-level requirement layer comprise one or more of N-1 passing rate, outage rate, capacity-load ratio, old equipment ratio, equipment availability factor, low-voltage household number, available power generation power, frontier defense area, electric vehicle occupancy, intelligent electric meter household number, number of per capita office equipment and information security frequency.

Further, the method also comprises the step of determining three-level requirements of each item, namely a requirement index:

determining a candidate index set of each item of a power grid company, wherein the candidate index set comprises one or more candidate indexes;

collecting historical index values and current index values of the candidate indexes, and calculating standard value intervals of the candidate indexes according to the historical index values by adopting a machine learning algorithm;

for any candidate index, if the current index value is outside the standard value interval of the index, determining the candidate index as a three-level requirement;

and calculating the deviation between the current index value of the three-level requirement and the standard value interval as the requirement degree of the project on the three-level requirement.

In some embodiments, the machine learning algorithm used in the step of determining the demand indicator is a long-short term memory neural network LSTM model algorithm.

Specifically, the step of calculating the standard value interval of each candidate index according to the historical index value by adopting a machine learning algorithm comprises the following steps: regarding any candidate index, taking the historical index value as a first training set { (x)₁,y₁),(x₂,y₂),…,(x_N,y_N) The LSTM model is trained for the first time; calculating model pre-estimation values by using models after first training

And calculates the actual value y_gAnd an estimate

Estimated error of

Predicting an output value y at time t according to the model after the first training and an input value containing known historical information_t(ii) a Based onPredictive value and predictive error in a training set to create a second training setObtaining the model after the second training by adopting a second training set, and calculating a new estimated value, an estimated error and an output value y at the time t_t(ii) a Continuing to generate other training sets according to the steps, and acquiring the model trained by the other training sets, the estimated value, the estimated error and the output value y at the t moment under the model_tUntil the model estimation error reaches the minimum value; a plurality of y obtained_tAnd sorting is carried out, and the numerical value in the preset interval is obtained as the standard value interval of the candidate index.

Further, the air conditioner is provided with a fan,

g＝1,2,……,N，

η_gto follow an equal distribution with a mean of 0 and a variance of 1,

is from

Is obtained by randomly selecting a numerical value with the probability of 1/N,

is derived from the estimated error

In eliminating systematic error

The latter corrected error.

Further, in the method according to the present invention, if the sorted y is to be sorted_tIs recorded as Q₍₁₎,Q₍₂₎,…Q_(c)Then the predetermined interval is Q_(cα2)<y_t<Q_(c-cα2))And α is a coefficient.

The invention provides a power grid multi-target project screening system, which comprises:

the system comprises a relation construction module, a relation calculation module and a relation calculation module, wherein the relation construction module is used for establishing a three-level network structure of a power grid project, the three-level network structure comprises a three-level demand layer, a two-level demand layer and a one-level demand layer, and the one-level demand layer comprises economic requirements, safety requirements and policy requirements;

the model construction module is used for constructing a multi-target project screening model, and the screening model is provided with a target function and constraint conditions;

the module calculation module is used for calculating the satisfaction degree of each demand index in the primary demand layer of each candidate item in the k candidate areas by adopting an objective function of the screening model and screening out items meeting the demand, wherein at least one candidate item is arranged in each candidate area; and evaluating the project meeting the requirement by adopting the constraint conditions of the screening model to obtain the optimal project planning scheme.

When the system is used, the relationship building module and the model building module respectively build a three-level network structure and a multi-target project screening model, then the model calculating module firstly screens out projects meeting requirements from candidate projects by using a target function, and then grades and sorts the projects meeting the requirements by using constraint conditions, so that an optimal planning scheme is finally obtained.

The present invention provides a computing device comprising one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs when executed by the processors implement any of the multi-target project screening methods described above.

The present invention provides a readable storage medium storing program instructions that, when read and executed by a computing device, cause the computing device to perform any of the above-described multi-project screening methods.

One or more technical solutions provided by the present application have at least the following technical effects or advantages:

1. the method screens certain types of items based on a multi-objective decision algorithm, so that the optimal item decision is determined to meet the company requirements, and in order to optimize the meeting degree of the item requirements, the requirements of an economic target, a safety target and a technical policy type target are respectively considered, a multi-objective optimization model is established, and the item screening is determined by considering the constraint conditions such as capital constraint, item constraint and the like;

2. the invention adopts the improved particle swarm optimization to decide the multi-objective optimization problem, and the method improves the PSO of the particle swarm optimization by applying methods of cross operators, dynamic inertia factors, learning factor parameter improvement, chaos mutation operation addition and the like, thereby overcoming the problems of local extremum and premature convergence of the PSO.

Drawings

The accompanying drawings, which are included to provide a further understanding of the embodiments of the invention and are incorporated in and constitute a part of this application, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention;

FIG. 1 is a flow diagram of a method 200 for multi-project screening in accordance with an embodiment of the present invention;

FIG. 2 is a schematic diagram of a three-level network structure of project requirement indicators in an embodiment of the present invention;

FIG. 3 is a diagram illustrating a process for calculating the level of satisfaction of the first level requirements of a project in accordance with an embodiment of the present invention;

FIG. 4 is a flow chart of a method 500 for determining a project requirement index in an embodiment of the present invention;

FIG. 5 is a schematic diagram of a multi-project screening apparatus 600 according to an embodiment of the present invention;

FIG. 6 is a block diagram of computing device 100 in an embodiment of the invention.

Detailed Description

In order that the above objects, features and advantages of the present invention can be more clearly understood, a more particular description of the invention will be rendered by reference to the appended drawings. It should be noted that the embodiments and features of the embodiments of the present application may be combined with each other without conflicting with each other.

In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those specifically described and thus the scope of the present invention is not limited by the specific embodiments disclosed below.

Example 1:

fig. 1 shows a schematic flow diagram of a multi-target item screening method 200 of a power grid, which is suitable for being executed in a computing device, and the model is suitable for selecting an optimal planning scheme from a total of l candidate items in k candidate areas, wherein at least one candidate item exists in each candidate area.

As shown in fig. 1, the method begins at step S210. In step S210, a three-level network structure of the power grid project is established, wherein the three-level network structure comprises a three-level demand layer, a two-level demand layer and a one-level demand layer. Wherein, tertiary demand layer is also called demand index layer, and tertiary index layer includes at least one demand index, and second grade demand layer includes at least one second grade demand, and first grade demand layer includes economic nature demand, security demand and policy demand.

Fig. 2 shows a schematic diagram of a three-level network structure according to a specific embodiment of the present invention. As shown in fig. 2, the demand index layer includes one or more demand indexes of N-1 passage rate, outage rate, capacity-load ratio, old equipment ratio, equipment availability factor, low-voltage household number, available power generation power, frontier defense area, electric vehicle occupancy, smart meter household number, per-capita office equipment number, and information security frequency. The second-level demand layer comprises one or more second-level demands of meeting newly increased load demands, sending out conventional power supplies, accessing power supplies, marketing propaganda, optimizing grid structure, improving reliability of a power grid, solving heavy overload of equipment, eliminating potential safety hazards, changing coal into electricity, transforming rural power grids, electrifying projects in remote areas and sending out new energy.

There may be multiple categories of general grid projects, and there may be multiple projects under each category. The project category may include one or more of a production improvement/major repair project, a non-generation improvement/major repair project, a marketing input project, a management consultation project, a right-of-stock investment project, a sporadic purchase project, an industry infrastructure/industry improvement/industry major repair project, a research and development project, and an educational and training project.

Subsequently, in step S220, a multi-objective project screening model is constructed, the screening model having an objective function and constraint conditions, wherein the objective function includes an economic requirement satisfaction maximum, a security requirement satisfaction maximum and a policy requirement satisfaction maximum.

In some embodiments, the objective function includes the following three objectives:

In some embodiments, the constraints include one or more of a capital constraint, an item constraint, a regional constraint, and a minimum demand constraint.

Capital constraints are investment resources for all projectsThe sum of the funds does not exceed the upper limit of the total funds I_maxThe calculation formula is as follows:

wherein, I_iTo invest the funds required for the ith project.

As shown in fig. 1, in step S230, the satisfaction degree of each candidate item in each candidate area to each primary requirement is respectively calculated, and the screening model is solved by using a predetermined algorithm, so as to obtain an optimal planning scheme, where the optimal planning scheme includes a plurality of target items in the k candidate areas.

The invention is not limited to a specific implementation mode, and all methods capable of solving the multi-objective optimization screening model are within the protection scope of the invention. According to one embodiment, the predetermined algorithm is a modified particle swarm algorithm according to which the multi-objective item screening model is decided. The algorithm improves the PSO by applying methods such as cross operators, dynamic inertia factors, learning factor parameter improvement, chaos mutation operation addition and the like, and solves the problems of PSO local extreme value and premature convergence.

Specifically, a crossover operator is introduced into a PSO algorithm to generate a measurement position X corresponding to a particle i_i,t+1Then measuring the position X_i,t+1Discretely intersecting with the individual historical optimal positions to generate test positions

D is the dimension of the decision variable. The cross formula is:

in the formula, rand_j(0,1) is [0,1 ]]Random numbers satisfying uniform distribution; j is a function of_randIs [1, D ]]Randomly and uniformly generated integers; p is a radical of_iAnd is the cross probability. This interleaving is the same as the binomial interleaving in differential evolution. And finally, updating the individual historical optimal positions of the particles as follows:

wherein f () is a fitness function,

is the optimal particle position. The invention adopts a self-adaptive method to directly code the parameter p into each particle so as to realize self-adaptive control. The particles i in the extended encoded population can be described as:

for each particle in the population, the cross probability is updated using the following rule:

in the formula, λ is an update probability of the parameter.

In addition, the invention also adds dynamic inertia factors and learning factors in the particle swarm algorithm. Along with the iterative process of the algorithm, the inertia weight linearity is reduced, and the traditional algorithm cannot adapt to the problems of complexity and nonlinearity. Therefore, the particle swarm optimization of the dynamic inertia factor is adopted in the invention, and the particle evolution degree V is defined firstly₁And degree of polymerization V of particles₂Two variables, the calculation formulas of which are respectively:

in the formula, p_g,i-1、p_g,iRespectively obtaining a global optimal value in the previous generation iteration process and a current global optimal value; p is a radical of_sIs the population size.

Degree of particle evolution V₁The initial value of the particle is small, and the evolution speed is high. Adding the evolution degree V containing particles₁And degree of polymerization V of particles₂Can improve the algorithmic performance of ion packets. Magnitude of inertial weight w and degree of evolution V of the example₁And degree of polymerization V of the particles₂Closely related, the expression w ═ f (V)₁,V₂)＝w₀-0.6×V₁+0.15×V₂In the formula, w₀Is an initialThe weight value is typically 0.9.

Learning factor (c)₁,c₂) The memory and learning capabilities of the particles are shown. c. C₁、c₂Is p_id、p_gdWherein p is_idIs the optimal position, p, that the particle i has experienced_gdIs the optimal position that all particles have experienced. At the beginning of the iteration, the particles are searched in a larger space, the diversity of the particles is maintained, c₁Larger c₂And the method is small so as to accelerate the particle swarm search speed. In the later iteration stage, the gravity center of the particle swarm is in the global optimal solution, c₁Smaller c₂Large to keep the population of particles capable of fine searching. According to one embodiment, c₁、c₂The expression of (a) is:

in the formula, c₁₀、c₁₁Are respectively c₁Initial and final values of (a); c. C₂₀、c₂₁Are respectively c₂Initial and final values of (a); g. g_mRespectively taking the iteration times and the maximum value of the particle swarm.

Furthermore, in order to expand the search space of the particle swarm and prevent the particles from falling into a local optimal value in the later iteration stage, the variation operation is performed on the particles based on the mode that the chaotic system generates the chaotic sequence. Firstly, the PSO algorithm is subjected to convergence analysis, necessary conditions for ensuring the overall convergence of the algorithm can be obtained, and the particle x is assumed_iJ (t) converges on k_i＝(k_i1,k_i2,…,k_in)^TThe expression is as follows:

in the formula,

p_i,j(t) searching for an optimal location for the history of particle i; p is a radical of_g,j(t) is the optimal particle position. First, assuming that the dimension of a particle is 1, by calculating the position x of the ith particle_i(t) and k_iAnd (t) establishing a chaotic mapping relation by the distance, and dynamically adjusting the chaotic search range of each generation of particles in the iterative process. X of particles_i(t) the search ranges are as follows:

in the formula, x_iminAnd x_imaxRespectively search range minimum and maximum, z, u ∈ (0, 1). Then, the particle x is subjected to a passing formula_i(t) normalizing to obtain an initial value G of the chaotic sequence_i,1The following were used:

generating a new chaotic sequence G by a chaotic mapping structure_i＝(G_i1,G_i2,…,G_im) The length of the chaotic sequence is v. These sequences are inverse transformed into the original search space (r ═ 1,2 … …, v) by:

L(x_i,r(t))＝x_imin+G_i,r(x_imax-x_imin)

and finally, selecting the particles with the optimal evaluation result in the sequence as the next generation of particles, wherein the iteration formula of the APSO algorithm is as follows:

according to the technical scheme, a certain type of project is screened based on a multi-objective decision algorithm, so that the optimal project decision is determined to meet the company requirements. In order to realize the optimization of the degree of meeting the project requirements, the invention respectively considers the requirements of an economic target, a safety target and a technical policy type target, establishes a multi-target optimization model, and determines the project screening by considering the constraint conditions such as capital constraint, project constraint and the like.

Example 2:

on the basis of embodiment 1, as shown in fig. 3, the calculation of the satisfaction degree of each candidate item in each candidate area to each primary requirement in step S230 can be realized through steps S231-S233.

Specifically, in step S231, for the three-level network structure, an ANP network analysis method is adopted, the upper-level network layer is used as a criterion, and each element set in the current-level network layer is used as a secondary criterion, so as to respectively calculate a first weight matrix a of all the requirement indexes and a second weight matrix B of all the secondary requirements.

The method utilizes the project cause diagnosis method to construct the mapping relation from the influence indexes to the project secondary requirements, and then utilizes the same method to construct the mapping relation from the project secondary requirements to the project primary requirements, thereby determining the requirements of the current project on the aspects of economy, safety, policy and the like. Therefore, the quantitative mapping relation from different projects to the influence indexes can be established according to the current requirements, and finally the meeting condition of each project to the primary requirement is calculated through a project requirement cause diagnosis method.

When the mapping relation from the project to the requirement index and the mapping relation from the requirement index to the secondary requirement and then to the primary requirement are established, in order to quantitatively determine the influence degree of the project construction index, the method carries out quantitative evaluation by using an ANP (artificial neural network) method, and further determines the benefit satisfaction condition of each project to various secondary requirements and various primary requirements.

The ANP method, namely the network analysis method, is provided on the basis of the AHP method, a network structure is established according to the relationship among all risk factors, and the influence relationship among the factors in each hierarchy and the feedback influence of the upper-layer factors on the lower-layer factors are considered. The ANP model structure divides the index factors into a control layer and a network layer, wherein the control factor layer comprises a problem target and a decision criterion, and the network layer is formed by index elements according to mutual dependence and feedback relation networks of the index elements.

In order to establish the mapping relationship from the project to the requirement index and the mapping relationship from the requirement index to the secondary requirement and then to the primary requirement, a weight matrix of the requirement index and the secondary requirement is calculated first. Taking the calculation of the weight matrix of the demand index as an example, the second-level demand layer is taken as the criterion of the control layer, and each element set U in the demand index layer is set_i(i-1, 2, …, n) is a sub-criterion, and other elements are set into a U_j(j ═ 1,2, …, n) for U_iAnd (4) scoring the dominance to obtain a comparison matrix. For example, the element set U can be obtained by a five-scale method or a nine-scale method_iThe comparison matrix of (2). Then, calculating the feature vector of each comparison matrix by using a feature root algorithm to obtain a weight matrix P of the layer:

similarly, element sets U are respectively_iEach element in (i ═ 1,2, …, n) is a secondary criterion, and a super matrix W between elements is calculated:

multiplying each submatrix in the super matrix by each element in the matrix P to obtain a weighted super matrix

i, j ═ 1,2, …, n. Then, corresponding feature vectors are calculated based on the weighted hypermatrixAnd the normalization is carried out,

then

The obtained feature vector is the weight matrix of each demand index. The weight matrix for the second level requirement can be obtained by the same method asWherein

Is the weight value of the i-th index,

is the weight value of the r-th demand.

Subsequently, in step S232, a satisfaction degree matrix X of the project to each requirement index is established, and a first mapping relationship between the requirement index and the secondary requirement and a second mapping relationship from the secondary requirement to the primary requirement are established. The satisfaction degree matrix of the project to each demand index can be generated by manually defining or directly taking the normalized index value or automatically scoring according to the index value.

According to one embodiment, the first mapping relationship is Y ═ X omica, and the second mapping relationship is Z ═ Y omicb [ -Z — ]₁,Z₂,Z₃]. Wherein, Y and Z represent the satisfaction degree of the project to the secondary requirement and the primary requirement respectively, and Z₁、Z₂、Z₃Represents the satisfaction degree of the project on economic requirement, safety requirement and policy requirement, wherein omicron is a generalized model synthesis operator, and the commonly used operators are (+, ×) operators, namely weighted average operators.

On the basis of obtaining the weight value, a satisfaction degree matrix X of each demand index of the project and a mapping relation Y from the demand index to the secondary demand are respectively as follows:

subsequently, in step S233, based on the established two mapping relationships, the satisfaction of the primary requirement and the secondary requirement of each item in each candidate area is calculated. For example, for a certain production technical improvement and major repair project in a certain candidate area, a satisfaction degree matrix of the project to each requirement index is obtained, and then a satisfaction degree Y of the project to the secondary requirement can be obtained according to the first weight matrix a. On the basis, according to the second weight matrix B, the satisfaction degree Z of the item to the first-level requirement can be further obtained.

Example 3:

based on the above embodiments, fig. 4 shows a flowchart of a project requirement index determining method 500 according to an embodiment of the invention, and as shown in fig. 4, the method 500 starts with step S510.

In step S510, a set of candidate metrics for each item of the grid company is determined, the set of candidate metrics including one or more candidate metrics.

According to another embodiment, the candidate indexes of each item can be, for example, N-1 passage rate, outage rate, capacity-to-load ratio, old equipment occupancy, equipment availability factor, low-voltage number of households, available generated power, frontier power shortage area, electric vehicle occupancy, number of households of smart meters, number of per-person office equipment, frequency of information security check, voltage qualification rate, frequency qualification rate, loss of electricity sales revenue, cost of exemptable power generation, average power outage frequency, average power outage time, and the like. For different items, other candidate indexes may also be set as needed, which is not limited by the present invention.

Subsequently, in step S520, the historical index value and the current index value of each candidate index are collected, and the standard value interval of each candidate index is calculated from the historical index values using a machine learning algorithm.

It should be understood that the historical index value and the current index value of each candidate index may be calculated according to the basic data of the power grid system to obtain a corresponding result value. The collection range is determined according to the index type and the requirement of the algorithm, and comprises annual data, monthly data or daily average data and the like. The skilled person can obtain the index values according to the prior art, and detailed description is omitted here. In addition, for the situations of data missing and data abnormality during the collection process, the collected data may be preprocessed in step S520, such as clearing dirty data, abnormal data, supplementing missing data by difference method, and the like.

According to one embodiment, considering that a machine learning algorithm is sensitive to data scale, after a historical index value and a current index value are obtained, normalization processing can be performed on the index values to obtain a value of a dimensionless value. Specifically, the MinMax method can be adopted for normalization processing to transform the data value domain into [0,1 ]]The calculation formula is as follows: upsilon is_no_rm＝(υ-υ_min)/(υ_max-υ_min). Wherein upsilon is_normIs a value of some index value of some item after upsilon normalization_maxAnd upsilon_minRespectively the maximum value and the minimum value of all index values of the index in the item.

Based on the collected index historical data, the data are processed through a machine learning algorithm, an index interval value can be determined and used as a diagnostic standard for diagnosing whether the index of the power grid is qualified, and therefore the specific requirements of the power grid project are further judged.

According to one embodiment, the machine learning algorithm may employ a long-short term memory neural network LSTM model algorithm. In the information transmission process of the LSTM, the information in the memory cell state is added or deleted through the current time input, the previous time hidden layer state, the previous time memory cell state and the gate structure. The gate structure is used to control the extent of adding or deleting information, the input gate i being common_tForgetting door f_tAnd an output gate o_tThree kinds of structure doors. The input gate is used for controlling whether the current input information is stored in the memory unit, the forgetting gate is used for determining how much information in the memory unit at the last moment can be transmitted to the current moment, and the output gate is used for controlling how much memory information is output. At time t, according to its input value x_tAnd the output value h at the previous moment_t-1And memory cell state c_t-1The input gate i at that moment can be calculated_tForgetting door f_tOutput gate o_t。

The updated cell state is mainly composed of the state information of the old cell from the previous time and the newly generated information of the current input. The hidden state is then output based on the updated memory cell state. In the calculation process of the invention, the updating calculation process is repeated until the difference value between the output value and the actual training sample is minimum, then the Boostrap self-sampling method is utilized to obtain the estimation interval based on the data sample, and the uncertainty brought by the LSTM algorithm is effectively quantified. The detailed training process of the model and the calculation process of the standard interval value will be described later.

Then, in step S530, a demand indicator for each item is screened from the candidate indicator set, and for any candidate indicator, if its current indicator value is outside the standard value interval of the indicator, the candidate indicator is determined as the demand indicator.

That is, the current index value of each power grid project is compared with the standard value interval of the index, if the index value is in the standard value interval, the index value is qualified, and the power grid project has no such requirement. Otherwise, the index value is unqualified, the power grid project has the requirement, and the index is the requirement index of the power grid project. If the standard value interval of a candidate index is [0.4, 0.7] and the current index value is 0.9, the candidate index is a demand index, and if the current index value is 0.5, the candidate index is not a demand index.

Subsequently, in step S540, the deviation of the current index value of the demand index from the standard value interval is calculated as the degree of demand of the item on the demand index. Therefore, according to the requirements of the power grid project on various indexes, the comprehensive requirements of the project can be theoretically analyzed by combining the actual situation. And for the index with the requirement, further calculating to obtain a deviation value of the index and the diagnosis standard interval, thereby judging the requirement degree of the power grid project investment on the index.

According to one embodiment, when calculating the deviation between the current index value and the standard value interval, a boundary value closest to the determination of the current index value may be selected from the standard value interval, and the absolute value of the difference between the boundary value and the current index value is the deviation. Assuming that the standard value interval is [0.4, 0.7] and the current index value is 0.9, which is closer to 0.7, the absolute value of the difference between the two values is 0.2, which is the deviation between the current index value and the standard value interval. In another implementation, a median value of the standard value interval may be calculated, and an absolute value of a difference between the current index value and the median value may be taken as the deviation.

Based on the size of the deviation value, a weight value can be set for each index item, the weight value can be in positive correlation with the deviation value, and if the deviation value is large, the corresponding weight value is also large. In addition, the multiple demand indicators of each project can be sorted in descending order according to the size of the deviation value, so as to obviously distinguish the influence weight of each demand indicator on the project.

In addition, in the above step S520, the machine learning algorithm is used to calculate the standard value interval of each candidate index according to the historical index value, which can be specifically realized by the following steps S521-S527:

first, in step S521, for a candidate index, the historical index value is defined as a first training set { (x)₁,y₁),(x₂,y₂),…,(x_N,y_N) The LSTM model is trained for the first time. Specifically, the corresponding weights and biases in the LSTM memory cells are obtained based on training sample data. The detailed training process of the model is a relatively mature technique in the field, and a person skilled in the art can select a training mode to train the model by himself, which is not limited by the present invention.

Thereafter, in step S522, a model estimated value is calculated by using the model after the first training

And calculates the actual value y_gAnd an estimateEstimated error of

Wherein,

that is, x is₁、x₂、……、x_NAnd the estimated value obtained by inputting the estimated value into the model after the first training.

According to one embodiment, the prediction error is calculated by the formula

g is 1,2, … …, N. Then, the error needs to be estimated from

Eliminating systematic errors

To obtain a corrected error

Namely:

then, in step S523, the output value y at time t is predicted from the model after the first training and the input value including the known history information_t. Namely, the input value at the time t is input into the model after the first training to obtain the corresponding output value y_t。

Thereafter, in step S524, a second training set is created based on the estimated values and estimated errors in the first training set

The training samples are boottrap data samples,

the method can be obtained by calculation through an external Boostrap method, and the calculation method comprises the following steps:

wherein,

is from

Is obtained by randomly selecting a numerical value with the probability of 1/N, η_gTo follow an equal distribution with a mean of 0 and a variance of 1.

Then, in step S525, a second training set is used to obtain a model after the second training, and new estimated values, estimated errors, and output values y at time t are calculated_t。

Then, in step S526, other training sets are continuously generated according to the above steps, and the model trained by the other training sets, the estimated value, the estimated error and the output value y at time t under the model are obtained_tAnd until the model prediction error reaches the minimum value.

That is, a third training set is generated based on the second training set, the model is trained by the third training set, and the model after the third training is used for calculating the estimated value, the estimated error and the output value y at the time t_t. Generating a fourth training set based on the third training set, and calculating the estimated value, the estimated error and the output value y at the time t of the model trained by the fourth training set_t. And analogizing in turn until the model prediction error reaches the expected minimum value, and finishing the training process of the model.

Finally, in step S527, the obtained plurality y_tAnd sorting is carried out, and the numerical value in the preset interval is obtained as the standard value interval of the candidate index. If the sorted y_tIs recorded as Q₍₁₎,Q₍₂₎,…Q_(c)If the predetermined interval is Q_(cα/2)<y_t<Q_(c-cα/2))α is a coefficient according to one embodiment, ten thousand model trainings, ranked y, can be performed_tThe total number of the units is 1 ten thousand, c is 10000, and α may be 0.5, but is not limited thereto and may be other values.

Example 4:

FIG. 5 illustrates a block diagram of a multi-project screening system 600, which apparatus 600 may reside in a computing device, in accordance with an embodiment of the present invention. As shown in fig. 5, the apparatus 600 includes: a relationship building module 610, a model building module 620, and a model calculation module 630.

The relationship building module 610 builds a three-level network structure of the power grid project, which includes a demand indicator layer, a secondary demand layer, and a primary demand layer. The demand index layer comprises n demand indexes, the second-level demand layer comprises m second-level demands, and the first-level demand layer comprises economic demands, safety demands and policy demands. The relationship construction module 610 may perform processing corresponding to the processing described above in step S210, and the detailed description thereof will not be repeated.

The model construction module 620 constructs a multi-objective project screening model having an objective function and constraint conditions, the objective function including the maximum economic demand satisfaction, the maximum security demand satisfaction and the maximum policy demand satisfaction. The relationship construction module 620 may perform processing corresponding to the processing described above in step S220, and the detailed description thereof will not be repeated.

The model calculation module 630 calculates the satisfaction degree of each candidate item in each candidate region to each primary requirement, and solves the screening model by using a predetermined algorithm to obtain an optimal planning scheme, where the optimal planning scheme includes multiple target items in the k candidate regions. The model calculation module 630 may perform processing corresponding to the processing described above in step S230, and the detailed description thereof will not be repeated here.

Example 5:

on the basis of embodiment 4, according to an embodiment of the present invention, the apparatus 600 may further include a demand indicator determining module (not shown in the figure) adapted to determine a demand indicator of each item according to the following steps: determining a candidate index set of each item of a power grid company, wherein the candidate index set comprises one or more candidate indexes; collecting historical index values and current index values of the candidate indexes, and calculating standard value intervals of the candidate indexes according to the historical index values by adopting a machine learning algorithm; for any candidate index, if the current index value is outside the standard value interval of the index, determining the candidate index as a demand index; and calculating the deviation between the current index value of the demand index and the standard value interval as the demand degree of the project on the demand index. The detailed calculation process of the module is disclosed in detail in the description based on fig. 4, and is not described in detail herein.

According to the technical scheme of the invention, in order to realize the optimization of the project requirement satisfaction degree, a three-level network structure is constructed, the requirements of an economic target, a safety target and a technical policy target are respectively considered, a multi-objective optimization model is established, and the project screening is determined by considering capital constraints, project constraints and the like. Moreover, the invention establishes the mapping relation from the requirement index to the secondary requirement and from the secondary requirement to the primary requirement in the three-level network structure. Based on the mapping relation, the accurate value of the project requirement satisfaction degree can be obtained, so that the staff can compare, analyze and judge a plurality of projects, and judgment conditions are provided for subsequent project investment planning.

In addition, the method diagnoses the demand cause of each index based on a machine learning algorithm to obtain the standard value interval of each candidate demand index of each item. And then, determining whether various demand indexes of the project are qualified on the basis of data analysis, and selecting unqualified indexes as demand indexes to perform next analysis, thereby determining demand conditions to be met by the investment decision of the power grid project. The index items determined in the mode are necessary index items for power grid project demand evaluation, and on the basis of ensuring the demand evaluation accuracy, the calculation amount and complexity of the algorithm are reduced as much as possible.

The various techniques described herein may be implemented in connection with hardware or software or, alternatively, with a combination of both. Thus, the methods and apparatus of the present invention, or certain aspects or portions thereof, may take the form of program code (i.e., instructions) embodied in tangible media, such as removable hard drives, U.S. disks, floppy disks, CD-ROMs, or any other machine-readable storage medium, wherein, when the program is loaded into and executed by a machine, such as a computer, the machine becomes an apparatus for practicing the invention. In the case of program code execution on programmable computers, the computing device will generally include a processor, a storage medium readable by the processor (including volatile and non-volatile memory and/or storage elements), at least one input device, and at least one output device. Wherein the memory is configured to store program code; the processor is configured to execute the multi-project screening model construction method of the present invention in accordance with instructions in the program code stored in the memory.

By way of example, and not limitation, readable media may comprise readable storage media and communication media. Readable storage media store information such as computer readable instructions, data structures, program modules or other data. Communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media. Combinations of any of the above are also included within the scope of readable media. In the description provided herein, algorithms and displays are not inherently related to any particular computer, virtual system, or other apparatus. Various general purpose systems may also be used with examples of this invention. The required structure for constructing such a system will be apparent from the description above. Moreover, the present invention is not directed to any particular programming language. It is appreciated that a variety of programming languages may be used to implement the teachings of the present invention as described herein, and any descriptions of specific languages are provided above to disclose the best mode of the invention. In the description provided herein, numerous specific details are set forth. It is understood, however, that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.

Similarly, it should be appreciated that in the foregoing description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. However, the disclosed method should not be interpreted as reflecting an intention that: that the invention as claimed requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention. Those skilled in the art will appreciate that the modules or units or components of the devices in the examples disclosed herein may be arranged in a device as described in this embodiment or alternatively may be located in one or more devices different from the devices in this example. The modules in the foregoing examples may be combined into one module or may be further divided into multiple sub-modules.

Those skilled in the art will appreciate that the modules in the device in an embodiment may be adaptively changed and disposed in one or more devices different from the embodiment. The modules or units or components of the embodiments may be combined into one module or unit or component, and furthermore they may be divided into a plurality of sub-modules or sub-units or sub-components. All of the features disclosed in this specification, claims, abstract and drawings, and all of the processes or elements of any method or apparatus so disclosed, may be combined in any combination, except combinations where at least some of such features and/or processes or elements are mutually exclusive. Each feature disclosed in the specification, claims, abstract, and drawings may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise. Furthermore, those skilled in the art will appreciate that while some embodiments described herein include some features included in other embodiments, rather than other features, combinations of features of different embodiments are meant to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments may be used in any combination.

Furthermore, some of the described embodiments are described herein as a method or combination of method elements that can be performed by a processor of a computer system or by other means of performing the described functions. A processor having the necessary instructions for carrying out the method or method elements thus forms a means for carrying out the method or method elements. Further, the elements of the apparatus embodiments described herein are examples of the following apparatus: the apparatus is used to implement the functions performed by the elements for the purpose of carrying out the invention.

Example 6:

FIG. 6 is a block diagram of a computing device 100, according to one embodiment of the invention. As shown in FIG. 6, in a basic configuration 102, a computing device 100 typically includes a system memory 106 and one or more processors 104. A memory bus 108 may be used for communication between the processor 104 and the system memory 106.

Depending on the desired configuration, the processor 104 may be any type of processing, including but not limited to: a microprocessor (μ P), a microcontroller (μ C), a Digital Signal Processor (DSP), or any combination thereof. The processor 104 may include one or more levels of cache, such as a level one cache 110 and a level two cache 112, a processor core 114, and registers 116. The example processor core 114 may include an Arithmetic Logic Unit (ALU), a Floating Point Unit (FPU), a digital signal processing core (DSP core), or any combination thereof. The example memory controller 118 may be used with the processor 104, or in some implementations the memory controller 118 may be an internal part of the processor 104.

Depending on the desired configuration, system memory 106 may be any type of memory, including but not limited to: volatile memory (such as RAM), non-volatile memory (such as ROM, flash memory, etc.), or any combination thereof. System memory 106 may include an operating system 120, one or more applications 122, and program data 124. In some embodiments, application 122 may be arranged to operate with program data 124 on an operating system. Program data 124 includes instructions, and in a computing device 100 according to the present invention, program data 124 includes instructions for performing multi-project screening model construction method 200 and/or project requirement index determination method 500 of the present invention.

Computing device 100 may also include an interface bus 140 that facilitates communication from various interface devices (e.g., output devices 142, peripheral interfaces 144, and communication devices 146) to the basic configuration 102 via the bus/interface controller 130. The example output device 142 includes a graphics processing unit 148 and an audio processing unit 150. They may be configured to facilitate communication with various external devices, such as a display or speakers, via one or more a/V ports 152. Example peripheral interfaces 144 may include a serial interface controller 154 and a parallel interface controller 156, which may be configured to facilitate communication with external devices such as input devices (e.g., keyboard, mouse, pen, voice input device, touch input device) or other peripherals (e.g., printer, scanner, etc.) via one or more I/O ports 158. An example communication device 146 may include a network controller 160, which may be arranged to facilitate communications with one or more other computing devices 162 over a network communication link via one or more communication ports 164.

A network communication link may be one example of a communication medium. Communication media may typically be embodied by computer readable instructions, data structures, program modules, and may include any information delivery media, such as carrier waves or other transport mechanisms, in a modulated data signal. A "modulated data signal" may be a signal that has one or more of its data set or its changes made in such a manner as to encode information in the signal. By way of non-limiting example, communication media may include wired media such as a wired network or private-wired network, and various wireless media such as acoustic, Radio Frequency (RF), microwave, Infrared (IR), or other wireless media. The term computer readable media as used herein may include both storage media and communication media.

Computing device 100 may be implemented as a server, such as a file server, a database server, an application server, a WEB server, etc., or as part of a small-form factor portable (or mobile) electronic device, such as a cellular telephone, a Personal Digital Assistant (PDA), a personal media player device, a wireless WEB-watch device, a personal headset device, an application specific device, or a hybrid device that include any of the above functions. Computing device 100 may also be implemented as a personal computer including both desktop and notebook computer configurations. In some embodiments, the computing device 100 is configured to perform the multi-objective project screening model construction method 200 and/or the project requirement index determination method 500 of the present invention.

While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.

It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims

1. A power grid multi-target project screening method is characterized by comprising the following steps:

2. The power grid multi-target project screening method of claim 1, wherein the constraint condition comprises one or more of a capital constraint, a project constraint, a regional constraint and a minimum requirement constraint.

3. The power grid multi-target project screening method according to claim 1, wherein the objective function comprises:

4. The power grid multi-target project screening method according to claim 3, wherein the step of calculating the satisfaction degree of each candidate project on each demand index in the primary demand layer by adopting the objective function of the screening model specifically comprises the following steps:

5. The power grid multi-target project screening method according to claim 1, wherein the multi-target project screening model is decided by adopting an improved particle swarm optimization.

6. The power grid multi-target project screening method according to any one of claims 1 to 5, wherein the secondary requirements in the secondary requirement layer comprise one or more of newly increased load requirements, conventional power supply output, power supply access, marketing promotion, grid structure optimization, power grid reliability improvement, equipment overload solving, potential safety hazard elimination, coal-to-electricity conversion, rural power grid transformation, remote power-on engineering and new energy output engineering.

7. The power grid multi-target project screening method according to any one of claims 1 to 5, wherein the three-level requirements in the three-level requirement layer comprise one or more of N-1 passing rate, outage rate, capacity-to-load ratio, old equipment occupation ratio, equipment availability factor, low-voltage household number, available generated power, frontier defense area, electric vehicle occupation ratio, intelligent electric meter household number, number of per capita office equipment and information security frequency.

8. The method for screening the multi-target projects of the power grid as claimed in claim 7, further comprising the step of determining three-level requirements of each project:

9. A power grid multi-target project screening system is characterized by comprising:

10. A readable storage medium storing program instructions that, when read and executed by a computing device, cause the computing device to perform the method of any of claims 1-8.