1. A power distribution network construction target risk assessment method is characterized by comprising the following steps: the method comprises the following steps:

step S1: building target definition and influence factor identification analysis are carried out by adopting system dynamics;

step S2: carrying out influence factor measurement analysis;

step S3: carrying out influence element index value prediction;

step S4: and (6) carrying out risk assessment.

2. The power distribution network construction target risk assessment method according to claim 1, characterized in that:

the specific content of step S2 is:

the principal component analysis method is usually used for dimensionality reduction of the indexes, and the weight value of each index can be calculated through the obtained variance contribution rate and the load matrix; the theory of principal component analysis is as follows:

the principal component analysis method firstly establishes p evaluation indexes, and then collects n groups of data for each index to obtain a matrix:

then processing the matrix by a Z-score standardization method to obtain a matrix Z ═ (Z)_ij)_n*pThen according to the formula:

obtaining a matrix of correlation coefficients, R ═ R_ij)_p*p(ii) a Note that the correlation coefficient matrix is equal to the covariance matrix, so there is:

A^TRA＝Λ＝diag(λ₁,λ₂,…,λ_p)

in the formula of lambda₁,λ₂,…,λ_pP eigenvalues of the matrix R; a ═ a_ij)_p*pThe normal orthogonal eigenvectors corresponding to the p eigenvalues are used;

let Y be A^TZ, written in matrix form as follows:

in the formula, y_iThe component I is the main component I, and the main components are sequentially arranged from big to small according to numerical values; z is a radical of₁,z₂,…,z_pIs an n-dimensional row vector in the matrix Z; and carrying out covariance operation on the principal component matrix Y to obtain:

obviously, the variance of the ith principal component is equal to its corresponding eigenvalue, and the correlation between any two different principal components is zero; at this point in time,the evaluation indexes are replaced by the principal component indexes, and the correlation among the evaluation indexes in the original index system is completely eliminated; to further simplify the set of indices, y for the principal component is defined_iContribution rate of variance to total varianceThe calculated fluctuation weight value is obtained; defining the cumulative contribution rate of the first m principal component variancesVisible, ω_iThe percentage of the ith main component index bearing the information quantity of the original index system is reflected, and rho reflects the percentage of the first m main component indexes bearing the information quantity of the original index system in an accumulated manner; selecting the first m main components to comprehensively evaluate the engineering target;

carrying out principal component analysis on each influence factor by using a principal component analysis method to obtain a weight sorting matrix of each influence factor;

obtaining a variance table and a principal component load matrix table through calculation; selecting the number of corresponding principal components according to the fact that the total contribution rate of the accumulated variance is greater than 85%; the weight of each influencing factor is calculated according to the following formula.

In the formula, a_ijRepresenting the load value of the ith principal component corresponding to the jth influencing factor,the ratio weight of each main component is regarded as; sigma_iRepresents the variance contribution rate of the ith principal component,which is considered as the fluctuating weight of each principal component. k is a radical of_jThe final weight index representing the jth influencing factor is a weighted average of the ratio weights of the principal components.

However, principal component analysis is particularly relatively adaptive depending on the analysis object, and requires that the KMO statistic be used to verify whether the object is suitable for analysis with principal components; the KMO test is a factor analysis mainly applied to multivariate statistics, and is used for comparing indexes of simple correlation coefficients and partial correlation coefficients among variables;

in the formula, r_jkIs a simple correlation coefficient, p, of the j element to the k element_jkIs the partial correlation coefficient of j and k; the KMO statistic takes a value between 0 and 1; when the simple correlation coefficient square sum among all variables is far greater than the partial correlation coefficient square sum, the KMO value is close to 1, the KMO value is closer to 1, which means that the correlation among the variables is stronger, and the original variables are more suitable for factor analysis; when the sum of squares of simple correlation coefficients among all variables is close to 0, the KMO value is close to 0, the more the KMO value is close to 0, the weaker the correlation among the variables is, the more the original variables are not suitable for the analysis of the cooperation factor;

kaiser gives the usual KMO metric, above 0.9 indicating that it is well suited; 0.8 indicates suitability; 0.7 represents normal; 0.6 means less suitable; 0.5 or less means extremely unsuitable.

3. The power distribution network construction target risk assessment method according to claim 1, characterized in that:

the specific content of step S3 is:

firstly, classifying indexes of influencing elements, and preliminarily classifying different indexes according to whether the identified influencing factors and related elements causing the influencing factors to change have linear structural relations or not, wherein the indexes with the linear relations adopt a multiple linear regression model, the indexes without the linear relations adopt a neural network model;

the specific content of the multiple linear regression is as follows: let the variables x1, x2, …, xp be p (p >1) linearly independent controlled variables, y be random variables, and the relationship between them is:

in the formula: b₀,b₁,…,b_p，σ²All are unknown parameters to be solved, epsilon is a random error, and the random error is a p element linear regression model;

for variable x₁,x₂,…,x_pAnd y are observed n times independently, and a sample with the capacity of n can be obtained

(x_i1,x_i2…,x_ip,y_i)_{(i＝1,2,…,n)}

In load prediction, the p-element linear regression relationship is:

the above formula is expressed in a matrix form and recorded

The linear regression model can be rewritten as:

Y＝XB+ε

the estimated vector of BETA is

Thus, the following steps are obtained:

will obtainSubstituting into p element linear regression relation to obtain:

this equation is called p-element linear regression equation,coefficients called regression equations;

the neural network structure of the neural network model is as follows:

aiming at a three-layer BP network structure, assuming that n input nodes are provided, m output nodes of the network are provided, and the number of the nodes is q; the input vector is X ═ X₁,x₂,x₃,....x_m) The output vector is Y ═ Y₁,y₂,y₃,....y_m) The hidden layer unit input vector is S ═ S₁,s₂,s₃,....s_q) The output vector of the hidden layer unit is B ═ B₁,b₂,b₃,....b_q) (ii) a Output layer unit input direction L ═ L₁,l₂,l₃,....l_m) The output vector of the output layer unit is C ═ C₁,c₂,c₃,....c_m) Further carrying out output calculation of an output layer and a hidden layer of the three-layer BP neural network; the method comprises the following specific steps:

determining related influence factors (independent variables) influencing input indexes; the number of nodes of the input layer is related to work influencing factors;

step b: establishing a BP network model: selecting an input layer node of a BP network as 9 nodes, selecting an output layer node of the BP network as 1 node, and selecting an output variable as an independent variable under corresponding input conditions; in the construction of the BP neural network, in order to search for more reasonable hidden layer unit number and achieve the precision requirement, 10, 15, 20 and 25 hidden layers are respectively established and trained, compared and analyzed; considering from training errors, selecting the number of hidden nodes which enable the network errors to be minimum and enable the prediction accuracy to be high; creating a network in MATLAB by calling function newff ();

step c: training the BP network: the standard BP algorithm divides the learning process into 2 phases: in the forward propagation process, the information of the input variable is processed by a hidden layer through an input layer, and the actual output value of each unit is calculated; in the back propagation process, if the output layer fails to obtain the expected output value, the difference value between the actual output and the expected output, namely the error is calculated, and the weight value and the threshold value are recursively adjusted layer by layer according to the difference value, so that the error value is gradually reduced until the requirement of network precision is met; trained by calling the function train ();

step d: utilizing the trained BP network to carry out independent variable pair on the project to be built; and calling a sim () function to predict the manufacturing cost.

4. The power distribution network construction target risk assessment method according to claim 1, characterized in that:

the specific content of step S4 is:

and judging the difference between the predicted value and the average predicted value of the index according to the predicted value of the index, finally comprehensively evaluating the risk of realizing the project construction target, and if the predicted value is greater than the average value of the index, representing that the construction target is implemented and having the risk.

Background

In recent years, along with the gradual increase of the investment scale of the power distribution network, the construction management work of the power distribution network is gradually emphasized, and because the current internal and external operation situation is severe, the serious challenge is provided for the input and output level and the fine management level of the power distribution network construction, the standardization and normalization management mode of the power distribution network construction must be continuously strengthened in the future, and the quality, the efficiency and the benefit of the power distribution network project construction are improved. Still have a great deal of problem among the current distribution network construction management, along with the continuous increase of distribution network construction scale, the construction task is constantly aggravated, engineering construction project quantity is many, the kind is miscellaneous, relate to the face extensively, distribution network engineering quantity is many, it is wide to relate to the region, potential hidden danger exists in the security quality, consequently, for further realizing the distribution network construction target degree of realizing, further promote distribution network engineering standardization, normalization, lean management level, promote the distribution network engineering by "low-level engineering subcontract type" to "senior project management type" disintegration, the adaptation distribution network construction and the structural reform transform requirement of transformation project market supply side. The construction target risk assessment method is provided, the realization degree of the current project management target is monitored in a targeted mode, the construction risk is avoided, and the refinement level of the project construction management of the power distribution network is improved.

At present, the risk assessment of the construction target of a power distribution network project is mostly qualitative assessment, the support of quantitative data is lacked, and the target realization degree is not well decomposed and evaluated. The probability of occurrence of the risk factors of the power grid construction project and the influence degree of the risk factors on the engineering project are researched by a multi-application multi-level fuzzy comprehensive evaluation method, then the risk in the power grid construction project is analyzed and evaluated, the current research focuses on the mutual comparison relationship between the risk and the risk, but neglects the system influence of the risk on the project target, and therefore the evaluation model and the thinking need to be further improved.

Disclosure of Invention

In view of the above, the invention aims to provide a power distribution network construction target risk assessment method, which emphasizes direct influence of risks on construction targets of various projects and weakens mutual comparison relationship between the risks, so that an accurate and easy-to-operate risk assessment method is provided for power grid construction projects, and a powerful hand is provided for fine management of power distribution network projects.

The invention is realized by adopting the following scheme: the embodiment provides a power distribution network construction target risk assessment method, which comprises the following steps:

step S1: building target definition and influence factor identification analysis are carried out by adopting system dynamics;

step S2: carrying out influence factor measurement analysis;

step S3: carrying out influence element index value prediction;

step S4: and (6) carrying out risk assessment.

Further, the specific content of step S2 is:

the principal component analysis method is usually used for dimensionality reduction of the indexes, and the weight value of each index can be calculated through the obtained variance contribution rate and the load matrix; the theory of principal component analysis is as follows:

the principal component analysis method firstly establishes p evaluation indexes, and then collects n groups of data for each index to obtain a matrix:

then processing the matrix by a Z-score standardization method to obtain a matrix Z ═ (Z)_ij)_n*pThen according to the formula:

obtaining a matrix of correlation coefficients, R ═ R_ij)_p*p(ii) a Note that the correlation coefficient matrix is equal to the covariance matrix, so there is:

A^TRA＝Λ＝diag(λ₁,λ₂,…,λ_p)

in the formula of lambda₁,λ₂,…,λ_pP eigenvalues of the matrix R; a ═ a_ij)_p*pThe normal orthogonal eigenvectors corresponding to the p eigenvalues are used;

let Y be A^TZ, written in matrix form as follows:

in the formula, y_iThe component I is the main component I, and the main components are sequentially arranged from big to small according to numerical values; z is a radical of₁,z₂,…,z_pIs an n-dimensional row vector in the matrix Z; and carrying out covariance operation on the principal component matrix Y to obtain:

obviously, the ith principal componentIs equal to its corresponding eigenvalue, while the correlation between any two different principal components is zero; so far, the evaluation indexes are replaced by the principal component indexes, and the correlation among the evaluation indexes in the original index system is completely eliminated; to further simplify the set of indices, y for the principal component is defined_iContribution rate of variance to total varianceThe calculated fluctuation weight value is obtained; defining the cumulative contribution rate of the first m principal component variancesVisible, ω_iThe percentage of the ith main component index bearing the information quantity of the original index system is reflected, and rho reflects the percentage of the first m main component indexes bearing the information quantity of the original index system in an accumulated manner; selecting the first m main components to comprehensively evaluate the engineering target;

carrying out principal component analysis on each influence factor by using a principal component analysis method to obtain a weight sorting matrix of each influence factor;

obtaining a variance table and a principal component load matrix table through calculation; selecting the number of corresponding principal components according to the fact that the total contribution rate of the accumulated variance is greater than 85%; the weight of each influencing factor is calculated according to the following formula.

In the formula, a_ijRepresenting the load value of the ith principal component corresponding to the jth influencing factor,the ratio weight of each main component is regarded as; sigma_iRepresents the variance contribution rate of the ith principal component,which is considered as the fluctuating weight of each principal component. k is a radical of_jThe final weight index representing the jth influence factor is a weighted average of the ratio weights of the principal componentsAll the above steps are carried out.

However, principal component analysis is particularly relatively adaptive according to different analysis objects, and whether the objects are suitable to be analyzed by the principal components needs to be verified through KMO (Kaiser-Meyer-Olkin) statistics; the KMO test is a factor analysis mainly applied to multivariate statistics, and is used for comparing indexes of simple correlation coefficients and partial correlation coefficients among variables;

in the formula, r_jkIs a simple correlation coefficient, p, of the j element to the k element_jkIs the partial correlation coefficient of j and k; the KMO statistic takes a value between 0 and 1; when the simple correlation coefficient square sum among all variables is far greater than the partial correlation coefficient square sum, the KMO value is close to 1, the KMO value is closer to 1, which means that the correlation among the variables is stronger, and the original variables are more suitable for factor analysis; when the sum of squares of simple correlation coefficients among all variables is close to 0, the KMO value is close to 0, the more the KMO value is close to 0, the weaker the correlation among the variables is, the more the original variables are not suitable for the analysis of the cooperation factor;

kaiser gives the usual KMO metric, above 0.9 indicating that it is well suited; 0.8 indicates suitability; 0.7 represents normal; 0.6 means less suitable; 0.5 or less means extremely unsuitable.

Further, the specific content of step S3 is:

firstly, classifying indexes of influencing elements, and preliminarily classifying different indexes according to whether the identified influencing factors and related elements causing the influencing factors to change have linear structural relations or not, wherein the indexes with the linear relations adopt a multiple linear regression model, the indexes without the linear relations adopt a neural network model;

the specific content of the multiple linear regression is as follows: let the variables x1, x2, …, xp be p (p >1) linearly independent controlled variables, y be random variables, and the relationship between them is:

in the formula: b₀,b₁,…,b_p，σ²All are unknown parameters to be solved, epsilon is a random error, and the random error is a p element linear regression model;

for variable x₁,x₂,…,x_pAnd y are observed n times independently, and a sample with the capacity of n can be obtained

(x_i1,x_i2…,x_ip,y_i)_{(i＝1,2,…,n)}

In load prediction, the p-element linear regression relationship is:

the above formula is expressed in a matrix form and recorded

The linear regression model can be rewritten as:

Y＝XB+ε

the estimated vector of BETA is

Thus, the following steps are obtained:

will obtainSubstituting into the p-element linear regression relation,obtaining:

this equation is called p-element linear regression equation,coefficients called regression equations;

the neural network structure of the neural network model is as follows:

aiming at a three-layer BP network structure, assuming that n input nodes are provided, m output nodes of the network are provided, and the number of the nodes is q; the input vector is X ═ X₁,x₂,x₃,....x_m) The output vector is Y ═ Y₁,y₂,y₃,....y_m) The hidden layer unit input vector is S ═ S₁,s₂,s₃,....s_q) The output vector of the hidden layer unit is B ═ B₁,b₂,b₃,....b_q) (ii) a Output layer unit input direction L ═ L₁,l₂,l₃,....l_m) The output vector of the output layer unit is C ═ C₁,c₂,c₃,....c_m) Further carrying out output calculation of an output layer and a hidden layer of the three-layer BP neural network; the method comprises the following specific steps:

determining related influence factors (independent variables) influencing input indexes; the number of nodes of the input layer is related to work influencing factors;

step b: establishing a BP network model: selecting an input layer node of a BP network as 9 nodes, selecting an output layer node of the BP network as 1 node, and selecting an output variable as an independent variable under corresponding input conditions; in the construction of the BP neural network, in order to search for more reasonable hidden layer unit number and achieve the precision requirement, 10, 15, 20 and 25 hidden layers are respectively established and trained, compared and analyzed; considering from training errors, selecting the number of hidden nodes which enable the network errors to be minimum and enable the prediction accuracy to be high; creating a network in MATLAB by calling function newff ();

step c: training the BP network: the standard BP algorithm divides the learning process into 2 phases: in the forward propagation process, the information of the input variable is processed by a hidden layer through an input layer, and the actual output value of each unit is calculated; in the back propagation process, if the output layer fails to obtain the expected output value, the difference value between the actual output and the expected output, namely the error is calculated, and the weight value and the threshold value are recursively adjusted layer by layer according to the difference value, so that the error value is gradually reduced until the requirement of network precision is met; trained by calling the function train ();

step d: utilizing the trained BP network to carry out independent variable pair on the project to be built; and calling a sim () function to predict the manufacturing cost.

Further, the specific content of step S4 is:

and judging the difference between the predicted value and the average predicted value of the index according to the predicted value of the index, finally comprehensively evaluating the risk of realizing the project construction target, and if the predicted value is greater than the average value of the index, representing that the construction target is implemented and having the risk.

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

the risk assessment based on the predicted values is to combine the results of the prediction model, compare the predicted values of all indexes with the average target values of corresponding factors, and comprehensively evaluate the main risk of the current construction target. The management system is used for strengthening project construction management and control and professional management work subsequently, standardizing the whole process management of power distribution network construction, and providing a hand grip and guarantee for further popularizing the floor implementation of an advanced mode. The refinement level of project construction management of the power distribution network can be further improved.

Drawings

FIG. 1 is a flow chart of a method according to an embodiment of the present invention.

Fig. 2 is a structural diagram of the impact element identification and the impact mechanism according to the embodiment of the present invention.

Detailed Description

The invention is further explained below with reference to the drawings and the embodiments.

It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.

As shown in fig. 1, the present embodiment provides a method for evaluating risk of a power distribution network construction target, including the following steps:

step S1: building target definition and influence factor identification analysis are carried out by adopting system dynamics;

step S2: carrying out influence factor measurement analysis;

step S3: carrying out influence element index value prediction;

step S4: and (6) carrying out risk assessment.

In this embodiment, the specific content of step S2 is:

the principal component analysis method is usually used for dimensionality reduction of the indexes, and the weight value of each index can be calculated through the obtained variance contribution rate and the load matrix; the theory of principal component analysis is as follows:

the principal component analysis method firstly establishes p evaluation indexes, and then collects n groups of data for each index to obtain a matrix:

then processing the matrix by a Z-score standardization method to obtain a matrix Z ═ (Z)_ij)_n*pThen according to the formula:

obtaining a matrix of correlation coefficients, R ═ R_ij)_p*p(ii) a Note that the correlation coefficient matrix is equal to the covariance matrix, so there is:

A^TRA＝Λ＝diag(λ₁,λ₂,…,λ_p)

in the formula of lambda₁,λ₂,…,λ_pP eigenvalues of the matrix R; a ═ a_ij)_p*pThe normal orthogonal eigenvectors corresponding to the p eigenvalues are used;

let Y be A^TZ, written in matrix form as follows:

in the formula, y_iThe component I is the main component I, and the main components are sequentially arranged from big to small according to numerical values; z is a radical of₁,z₂,…,z_pIs an n-dimensional row vector in the matrix Z; and carrying out covariance operation on the principal component matrix Y to obtain:

obviously, the variance of the ith principal component is equal to its corresponding eigenvalue, and the correlation between any two different principal components is zero; so far, the evaluation indexes are replaced by the principal component indexes, and the correlation among the evaluation indexes in the original index system is completely eliminated; to further simplify the set of indices, y for the principal component is defined_iContribution rate of variance to total varianceThe calculated fluctuation weight value is obtained; defining the cumulative contribution rate of the first m principal component variancesVisible, ω_iThe percentage of the ith principal component index bearing the information of the original index system is reflected, and the rho reflects the accumulated bearing of the previous m principal component indexesPercentage of amount of information; selecting the first m main components to comprehensively evaluate the engineering target;

carrying out principal component analysis on each influence factor by using a principal component analysis method to obtain a weight sorting matrix of each influence factor;

obtaining a variance table and a principal component load matrix table through calculation; selecting the number of corresponding principal components according to the fact that the total contribution rate of the accumulated variance is greater than 85%; the weight of each influencing factor is calculated according to the following formula.

In the formula, a_ijRepresenting the load value of the ith principal component corresponding to the jth influencing factor,the ratio weight of each main component is regarded as; sigma_iRepresents the variance contribution rate of the ith principal component,which is considered as the fluctuating weight of each principal component. k is a radical of_jThe final weight index representing the jth influencing factor is a weighted average of the ratio weights of the principal components.

However, principal component analysis is particularly relatively adaptive according to different analysis objects, and whether the objects are suitable to be analyzed by the principal components needs to be verified through KMO (Kaiser-Meyer-Olkin) statistics; the KMO test is a factor analysis mainly applied to multivariate statistics, and is used for comparing indexes of simple correlation coefficients and partial correlation coefficients among variables;

in the formula, r_jkIs a simple correlation coefficient, p, of the j element to the k element_jkIs the partial correlation coefficient of j and k; the KMO statistic takes a value between 0 and 1; when the simple sum of the squares of the correlation coefficients between all variables is much larger than the biasWhen the correlation coefficients are summed up in squares, the KMO value is close to 1, the KMO value is closer to 1, which means that the correlation between variables is stronger, and the original variables are more suitable for factor analysis; when the sum of squares of simple correlation coefficients among all variables is close to 0, the KMO value is close to 0, the more the KMO value is close to 0, the weaker the correlation among the variables is, the more the original variables are not suitable for the analysis of the cooperation factor;

kaiser gives the usual KMO metric, above 0.9 indicating that it is well suited; 0.8 indicates suitability; 0.7 represents normal; 0.6 means less suitable; 0.5 or less means extremely unsuitable.

In this embodiment, the specific content of step S3 is:

firstly, classifying indexes of influencing elements, and preliminarily classifying different indexes according to whether the identified influencing factors and related elements causing the influencing factors to change have linear structural relations or not, wherein the indexes with the linear relations adopt a multiple linear regression model, the indexes without the linear relations adopt a neural network model; (Classification Standard)

For example, the price factor of the equipment material, the key influencing factors of which are the price of the raw material, the labor cost and the currency policy, respectively, therefore, the price of the raw material (copper price), the labor cost and the currency policy are selected as independent variables in the price prediction of the equipment material based on the multiple linear regression, the data of the dependent variables and the dependent variables are subjected to linear regression model,

and obtaining the prediction result of the material price of the dependent variable equipment. As the main basis for risk assessment.

The specific content of the multiple linear regression is as follows: let the variables x1, x2, …, xp be p (p >1) linearly independent controlled variables, y be random variables, and the relationship between them is:

in the formula: b₀,b₁,…,b_p，σ²All are unknown parameters to be solved, epsilon is a random error, and the random error is a p element linear regression model;

for variable x₁,x₂,…,x_pAnd y are observed n times independently, and a sample with the capacity of n can be obtained

(x_i1,x_i2…,x_ip,y_i)_{(i＝1,2,…,n)}

In load prediction, the p-element linear regression relationship is:

the above formula is expressed in a matrix form and recorded

The linear regression model can be rewritten as:

Y＝XB+ε

the estimated vector of BETA is

Thus, the following steps are obtained:

will obtainSubstituting into p element linear regression relation to obtain:

this equation is called p-element linear regression equation,coefficients called regression equations;

the neural network structure of the neural network model is as follows:

aiming at a three-layer BP network structure, assuming that n input nodes are provided, m output nodes of the network are provided, and the number of the nodes is q; the input vector is X ═ X₁,x₂,x₃,....x_m) The output vector is Y ═ Y₁,y₂,y₃,....y_m) The hidden layer unit input vector is S ═ S₁,s₂,s₃,....s_q) The output vector of the hidden layer unit is B ═ B₁,b₂,b₃,....b_q) (ii) a Output layer unit input direction L ═ L₁,l₂,l₃,....l_m) The output vector of the output layer unit is C ═ C₁,c₂,c₃,....c_m) Further carrying out output calculation of an output layer and a hidden layer of the three-layer BP neural network; the method comprises the following specific steps:

determining related influence factors (independent variables) influencing input indexes; the number of nodes of the input layer is related to work influencing factors; taking mechanical price elements as an example, the number of nodes of the input layer is related to the factors influencing the labor, and the main factors influencing the mechanical price are finally determined by analyzing the factors influencing the mechanical price, including the number of loops, the type of a lead, the topographic condition, the geological condition, the unit number of a kilometer tower, the weight of a kilometer tower, the strain proportion, the unit price of the tower, the unit price of the lead and the like.

Step b: establishing a BP network model: selecting 9 nodes as input layer of BP network, wherein the input variables are respectively loop number, terrain condition, geological condition, single kilometer tower base number, single kilometer tower material weight, wire type, strain proportion, tower material unit price and wire unit price, and respectively making them be x₁～x₉(ii) a The need for textual descriptions in the influencing factors translates them into quantifiable criteria (e.g., ratings). The output layer of the network is selected as 1 node, and the output variable is an independent variable (namely, mechanical price) under the corresponding input condition; the number of hidden nodes is generally determined by multiple experiments or empirical formulasDetermining that in the construction of the BP neural network, in order to search for more reasonable number of hidden layer units and meet the precision requirement, 10, 15, 20 and 25 hidden layers are respectively established and trained, compared and analyzed; considering from training errors, selecting the number of hidden nodes which enable the network errors to be minimum and enable the prediction accuracy to be high; creating a network in MATLAB by calling function newff ();

step c: training the BP network: the standard BP algorithm divides the learning process into 2 phases: in the forward propagation process, the information of the input variable is processed by a hidden layer through an input layer, and the actual output value of each unit is calculated; in the back propagation process, if the output layer fails to obtain the expected output value, the difference value between the actual output and the expected output, namely the error is calculated, and the weight value and the threshold value are recursively adjusted layer by layer according to the difference value, so that the error value is gradually reduced until the requirement of network precision is met; trained by calling the function train ();

step d: using the trained BP network to carry out independent variable (namely mechanical price) on the project to be built; and calling a sim () function to predict the manufacturing cost.

In this embodiment, the specific content of step S4 is:

and judging the difference between the predicted value and the average predicted value of the index according to the predicted value of the index, finally comprehensively evaluating the risk of realizing the project construction target, and if the predicted value is greater than the average value of the index, representing that the construction target is implemented and having the risk. For example, the area rainfall prediction value is higher than the average value, which indicates that the construction risk of the project is increased. Project management and risk prevention are to be enhanced.

Preferably, in the present embodiment,

the method comprises the following steps: construction target definition and influence factor identification analysis

Firstly, construction target definition is carried out from multiple dimensions such as engineering progress, quality, cost, safety and the like, then, a system dynamics method is applied to scientifically and comprehensively identify construction target influence elements from the aspects of technical elements, natural elements, policy elements and the like, and influence mechanisms are determined. The specific system is shown in figure 2.

Step two: impact factor metric analysis

The principal component analysis method is generally used for the dimensionality reduction of the indicators, and the weight values of the indicators can be calculated by the obtained variance contribution rates and load matrices. The theory of principal component analysis is as follows:

the principal component analysis method firstly establishes p evaluation indexes, and then collects n groups of data for each index to obtain a matrix:

then processing the matrix by a Z-score standardization method to obtain a matrix Z ═ (Z)_ij)_n*pThen according to the formula:

obtaining a matrix of correlation coefficients, R ═ R_ij)_p*p。

Note that the correlation coefficient matrix is equal to the covariance matrix, so there is:

A^TRA＝Λ＝diag(λ₁,λ₂,…,λ_p)

in the formula of lambda₁,λ₂,…,λ_pP eigenvalues of the matrix R; a ═ a_ij)_p*pAnd the normal orthogonal eigenvectors corresponding to the p eigenvalues are used.

Let Y be A^TZ, written in matrix form as follows:

in the formula, y_iThe component I is the main component I, and the main components are sequentially arranged from big to small according to numerical values; z is a radical of₁,z₂,…,z_pIs an n-dimensional row vector in matrix Z.

The covariance operation of the principal component matrix Y can be obtained:

it is clear that the variance of the ith principal component is equal to its corresponding eigenvalue, while the correlation between any two different principal components is zero. So far, the evaluation indexes are replaced by the principal component indexes, and the correlation among the evaluation indexes in the original index system is completely eliminated.

To further simplify the set of indices, y for the principal component is defined_iContribution rate of variance to total varianceI.e. the calculated fluctuation weight. Defining the cumulative contribution rate of the first m principal component variancesVisible, ω_iAnd the percentage of the ith main component index bearing the information quantity of the original index system is reflected, and the rho reflects the percentage of the first m main component indexes bearing the information quantity of the original index system in an accumulated manner. The first m main components can be selected to comprehensively evaluate the engineering target.

Principal component analysis is an objective method. In the report, principal component analysis is performed on each influence factor to obtain a weight sorting matrix of each influence factor.

And calculating to obtain a variance table and a principal component load matrix table. The number of corresponding principal components is generally selected according to the cumulative total variance contribution rate of more than 85%. The weight of each influencing factor is calculated according to the following formula.

In the formula, a_ijRepresenting the load value of the ith principal component corresponding to the jth influencing factor,can be regarded as the proportion weight of each main component; sigma_iRepresents the variance contribution rate of the ith principal component,can be regarded as the fluctuating weight of each principal component. k is a radical of_jThe final weight index representing the jth influencing factor is a weighted average of the ratio weights of the principal components.

However, principal component analysis is particularly relatively adaptive depending on the object to be analyzed, and it is necessary to verify whether the object is suitable for principal component analysis by using KMO (Kaiser-Meyer-Olkin) statistic. The KMO test mentioned here is a factor analysis in which an index for comparing simple correlation coefficients and partial correlation coefficients between variables is mainly applied to multivariate statistics.

In the formula, r_jkIs a simple correlation coefficient, p, of the j element to the k element_jkIs the partial correlation coefficient of j and k. The KMO statistic takes on a value between 0 and 1. When the simple correlation coefficient square sum among all variables is far greater than the partial correlation coefficient square sum, the KMO value is close to 1, the KMO value is closer to 1, which means that the correlation among the variables is stronger, and the original variables are more suitable for factor analysis; when the sum of the squares of the simple correlation coefficients between all variables is close to 0, the KMO value is close to 0, the closer the KMO value is to 0, meaning that the weaker the correlation between the variables, the more inappropriate the original variables are for the analysis of the co-factor.

Kaiser gives the usual KMO metric, above 0.9 indicating that it is well suited; 0.8 indicates suitability; 0.7 represents normal; 0.6 means less suitable; 0.5 or less means extremely unsuitable.

Step three: influence element index value prediction

Firstly, the indexes of the influencing elements are classified, and different index rows can be preliminarily classified according to whether the variable and the dependent variable have a linear structure relationship or not. The indexes with linear relation adopt a multiple linear regression model, the indexes without linear relation adopt a neural network model.

(1) Multiple linear regression

The multiple linear regression prediction technology mainly researches the correlation between a dependent variable and a plurality of independent variables. A phenomenon is often associated with multiple factors, and predicting or estimating a dependent variable from an optimal combination of multiple independent variables together is more efficient and more practical than predicting or estimating with only one independent variable.

Let the variables x1, x2, …, xp be p (p >1) linearly independent controlled variables, y be random variables, and the relationship between them is:

in the formula: b₀,b₁,…,b_p，σ²Are unknown parameters to be solved, and epsilon is a random error, which is a p-element linear regression model.

For variable x₁,x₂,…,x_pAnd y are observed n times independently, and a sample with the capacity of n can be obtained

(x_i1,x_i2…,x_ip,y_i)_{(i＝1,2,…,n)}

In load prediction, these constants are historical data from the past, and are obtained from p-element linear regression relations

For the convenience of mathematical processing, the above formula is expressed in a matrix form. Note the book

The linear regression model can be rewritten as:

Y＝XB+ε

the estimated vector of BETA is

Thus, it is possible to obtain:

will obtainSubstituting into p-element linear regression equation to obtain

This equation is called a p-element linear regression equation.Referred to as the coefficients of the regression equation.

(2) Neural network model

The neural network has many models, and the BP neural network is one of the most widely used models. The BP neural network continuously adjusts the weight and the threshold value of the network by learning and storing a large number of input-output mode mapping relations and using a gradient descent method and back propagation, so that the error of the network is minimized.

1) Neural network architecture

Taking a three-layer BP network structure as an example, assume that n input nodes are provided, m output nodes of the network are provided, and the number of the nodes is q. The input vector is X ═ X₁,x₂,x₃,....x_m) The output vector is Y ═ Y₁,y₂,y₃,....y_m)

The hidden layer unit input vector is S ═ S₁,s₂,s₃,....s_q) The output vector of the hidden layer unit is B ═ B₁,b₂,b₃,....b_q) (ii) a Output layer unit input direction L ═ L₁,l₂,l₃,....l_m) The output vector of the output layer unit is C ═ C₁,c₂,c₃,....c_m) And further carrying out output calculation of an output layer and a hidden layer of the three-layer BP neural network.

2) Network training

Initialization of the network. The connection weight between input layer neuron i and hidden layer neuron j is w_ijThe connection weight between hidden layer neuron j and output layer neuron t is v_ij(ii) a The threshold for hidden layer neuron j is θ_jThe threshold of the output layer neuron t is r_t. A learning rate and a neuron excitation function are given.

The hidden layer outputs the computation. And calculating the hidden layer output according to the input vector, the connection weight between the input layer and the hidden layer and the threshold of the hidden layer.

The output layer outputs the calculation. And calculating the predicted output of the neural network according to the hidden layer output, the connection weight and the threshold.

And (4) error calculation. A net prediction error is calculated based on the net prediction output and the expected output.

And updating the weight and the threshold. And updating the network connection weight and the threshold according to the network prediction error.

And judging whether the algorithm iteration is finished or not, and returning to output calculation if the algorithm iteration is not finished.

And when the algorithm is finished, saving the weight value and the threshold value as the initial weight value and the threshold value in the prediction.

Step four: risk assessment

And judging the difference between the predicted value and the average value of the index according to the predicted value of the index, and finally comprehensively evaluating the risk and the potential problem of realizing the project construction target. For example, the rainfall is obviously higher than other values, and the project is subjected to important management and control and avoidance due to the adverse effects on the project construction safety and progress caused by overlarge rainfall.

Preferably, in this embodiment, the construction target definition is first performed from multiple dimensions such as engineering progress, quality, cost, and safety, and then the system dynamics method is applied to scientifically and comprehensively identify the construction target influence elements from the aspects of technical elements, natural elements, policy elements, and the like, so as to clarify the influence mechanism.

And then, carrying out influence factor measurement analysis, combining with relevant index data collection and sorting, obtaining the load of the factors on the principal component by adopting a principal component analysis method, and analyzing and judging the influence degree of different factors on the construction target. And selecting the main influence factors as main indexes for risk judgment.

And secondly, forecasting and calculating the influence factors, selecting a proper forecasting method to forecast and calculate the factors by combining different factor types and characteristics, and taking the forecasting result as a main basis of risk assessment.

And finally, carrying out construction target risk assessment based on the predicted values, namely, combining the prediction model results, comparing the predicted values of all indexes with the average target values of corresponding factors, and comprehensively evaluating the main risk of the current construction target.

Preferably, the method can analyze the construction target indexes based on the pressure, the management problems and the risk points of the power distribution network management, accurately predict and evaluate the change condition of each important influence factor, form a scientific evaluation result, and pointedly reflect the internal and external situations faced by the current construction management, so that the risk of the project construction target is scientifically judged and evaluated.

In order to solve the above problem, according to an aspect of the present invention, there is provided a power distribution network construction target risk assessment method, including:

(1) the method mainly comprises the steps of defining a construction target and analyzing influence factors, emphasizing control targets in the aspects of power distribution network engineering quality, progress, construction cost and safety, using the analysis and identification basis from the aspects, applying a system dynamics principle, identifying main factors influencing the control targets from a plurality of problems such as technical elements, management elements, market elements and policy elements, and determining influence mechanisms.

(2) And (3) influence factor measurement analysis, namely judging and measuring the influence degrees of different factors on the target by combining the recognition and analysis results of the influence factors and applying a principal component analysis method, and processing and simplifying a factor set to obtain corresponding key factors serving as important data sources for risk assessment.

(3) The influence element index prediction module is used for dividing the types of the influence elements according to the decomposition and integration concept, different structural characteristics are provided between different elements and corresponding influence parameters, the element indexes can be preliminarily classified according to whether linear structural relations exist between variables and dependent variables, a proper method is selected from a model according to a prediction model optimization mechanism for fitting prediction, and statistics and analysis of historical data are given to obtain an element index value prediction result.

(4) And (4) risk evaluation, namely judging the difference between the predicted value and the average value of the index according to the predicted value of the index, and finally comprehensively evaluating the risk and the potential problem of realizing the project construction target. For example, the rainfall is obviously higher than other values, and the project is subjected to important management and control and avoidance due to the adverse effects on the project construction safety and progress caused by overlarge rainfall.

Preferably, the embodiment is based on the current power distribution network management situation, emphasizes the actual conditions in the aspects of power distribution network engineering quality, progress, cost, safety and the like, and identifies the main elements influencing the project construction target from different dimensions such as technical elements, management elements, environmental elements and the like by using a system dynamics model, and the main elements are used as the basic basis for risk assessment. And secondly, by combining the historical data condition of the indexes and applying a principal component analysis method, the influence degree of different elements on the construction target is determined, and the core element indexes are determined. Secondly, defining applicable prediction methods of different types of indexes, performing prediction analysis on index values by combining data collection, and finally, evaluating the risk of the construction target by a system by combining the predicted values of the indexes and referring to the standard values of the related indexes. In order to further strengthen the management and control, promote the project management normative, provide prospective reference and reference.

The construction target definition and the influence factor analysis are as follows: the system combs the current management situation, excavates management problems and promotion directions, highlights the width and breadth of research, highlights the control targets in the aspects of power distribution network engineering quality, progress, construction cost and safety, takes analysis and identification as the basis from the aspects, follows construction principles such as guidance, quantifiability, comprehensiveness and the like, applies the system dynamics principle, identifies main factors influencing the control targets from a plurality of problems such as technical elements, management elements, market elements, policy elements and the like, and determines the influence mechanism.

And the influence factor measurement analysis comprises the following steps: and by combining the recognition and analysis results of the influence factors, the influence degrees of different factors on the target are judged and measured by using a principal component analysis method, and the original data is converted into a group of representations which are linearly independent of each dimension, so that the correlation among the factors is effectively eliminated. The principal component concentrates most of the information of the original variables, reducing the workload. And processing and simplifying the factor set by using a principal component analysis method to obtain corresponding key elements serving as important data sources for risk assessment.

The influence element index prediction module is used for: according to the concept of decomposition and integration, firstly, the types of the influence elements are divided, different elements and corresponding influence parameters have different structural characteristics, element indexes can be preliminarily classified according to whether linear structural relations exist between the variables and the dependent variables, a proper method is selected from a model according to a prediction model optimization mechanism for fitting prediction, and statistics and analysis of historical data are given to obtain an element index value prediction result.

The risk assessment module: and judging the difference between the predicted value and the average value of the index according to the predicted value of the index, and finally comprehensively evaluating the risk and the potential problem of realizing the project construction target. For example, the rainfall is obviously higher than other values, and the project is subjected to important management and control and avoidance due to the adverse effects on the project construction safety and progress caused by overlarge rainfall.

Preferably, the construction target definition and the influence factor analysis in this embodiment define the construction target from multiple dimensions such as engineering progress, quality, cost, safety, and the like, and then scientifically and comprehensively identify the influence factors of the construction target from the aspects such as technical factors, natural factors, policy factors, and the like by using a system dynamics method, thereby defining the influence mechanism. And then, carrying out influence factor measurement analysis, combining with relevant index data collection and sorting, obtaining the load of the factors on the principal component by adopting a principal component analysis method, and analyzing and judging the influence degree of different factors on the construction target. And selecting the main influence factors as main indexes for risk judgment. And the method comprises the following steps of forecasting and calculating the influence factors, selecting a proper forecasting method to forecast and calculate the factors by combining different factor types and characteristics, and taking a forecasting result as a main basis of risk assessment. And risk evaluation based on the predicted values is to combine the results of the prediction model, compare the predicted values of all indexes with the average target values of the corresponding factors, and comprehensively evaluate the main risk of the current construction target. The management system is used for strengthening project construction management and control and professional management work subsequently, standardizing the whole process management of power distribution network construction, and providing a hand grip and guarantee for further popularizing the floor implementation of an advanced mode. The refinement level of project construction management of the power distribution network can be further improved.

The above description is only a preferred embodiment of the present invention, and all equivalent changes and modifications made in accordance with the claims of the present invention should be covered by the present invention.