Power distribution network construction material demand analysis and prediction method

Abstract

The invention discloses a power distribution network construction material demand analysis and prediction method, which comprises the following analysis and prediction steps: S1, collecting historical material data of a power distribution network construction project, the historical material data comprising project attribute data and material usage data; s2, preprocessing the material consumption data, including data cleaning, standardization and normalization processing; s3, carrying out dimension reduction processing on the material data based on a principal component analysis method; s4, performing clustering analysis on the material project attributes based on a clustering analysis algorithm; s5, parameters of the extreme learning machine are optimized based on a particle swarm algorithm, the optimized extreme learning machine is adopted to construct a power distribution network construction material demand prediction model, and the number and amount of power distribution network construction material demands are predicted; the system can effectively improve the material management level of a power enterprise, and improves the efficiency and quality of a material demand plan.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power distribution network construction, and particularly relates to a power distribution network construction material demand analysis and prediction method. BACKGROUND

[0002] With the rapid development of power enterprises, material management is becoming more and more important in the process of enterprise fine operation. Material demand prediction is an important part of power enterprise material management, which can guide the enterprise's material investment and use. High-quality material management is conducive to the rational use of enterprise materials, helps to reduce enterprise production costs, speeds up the flow and turnover of funds, improves enterprise economic and management efficiency, and optimizes enterprise resource allocation.

[0003] Taking power distribution network construction as an example, the power distribution network project has large material demand and various types. The annual material demand planning work needs to go through personnel organization, demand collection, review and summary, plan formation and other processes, which consumes a lot of manpower and material resources, and has the problems of long cycle and low accuracy, which can easily affect the construction progress and cause the phenomenon of excess inventory or insufficient materials. The traditional demand planning method restricts the efficiency and benefit of power distribution network project construction in many aspects such as material purchase, storage and use.

[0004] Therefore, in order to solve the above problems, it is necessary to develop a power distribution network construction material demand analysis and prediction method. SUMMARY

[0005] The purpose of the application is to overcome the shortcomings of the prior art and provide a power distribution network construction material demand analysis and prediction method, which can effectively improve the material management level of power enterprises and improve the efficiency and quality of material demand planning.

[0006] The purpose of the application is achieved by a power distribution network construction material demand analysis and prediction method, which comprises the following analysis and prediction steps:

[0007] S1, collecting historical material data of the power distribution network construction project, wherein the historical material data includes project attribute data and material consumption data;

[0008] S2, preprocessing the material consumption data, including data cleaning, standardization and normalization processing;

[0009] S3, dimension reduction processing of the material data based on principal component analysis method;

[0010] S4, clustering analysis of the material project attributes based on clustering analysis algorithm;

[0011] S5. The parameters of the Extreme Learning Machine are optimized based on the Particle Swarm Optimization algorithm. The optimized Extreme Learning Machine is used to construct a prediction model for the demand of materials for power distribution network construction, and to predict the quantity and amount of materials required for power distribution network construction.

[0012] Furthermore, in step S2, data cleaning is used to handle duplicate, missing, and outlier data to ensure data integrity and accuracy; data standardization and normalization are used to scale the data to a uniform scale, wherein data standardization is used to eliminate the dimensions between feature attributes and convert the data into a normal distribution with a mean of 0 and a standard deviation of 1. In the formula: For standardized data, This is the original data. The mean of the original data. Standard deviation; data normalization is used to scale the data to... Interval.

[0013] Furthermore, the dimensionality reduction processing of the material data using principal component analysis in step S3 specifically includes the following steps: ① Assume that the original material data has There are 10 samples, each sample has 100 samples. If there are 100 observations, the original variable matrix is ​​represented as: In the formula: For the first The first sample ① 1 observation value; ② Standardize the raw data. , , ③ Construct the correlation coefficient matrix of the observed objects: , ④ The characteristic equation is calculated using the Jacobi method, and the corresponding correlation coefficient matrix is ​​obtained. There are 1 non-negative eigenvalues, and each eigenvalue corresponds to a different eigenvalue: Then the corresponding eigenvectors And satisfy: ⑤ Then the first Cumulative variance of each principal component The cumulative variance closest to 1 is the first The factor is the first Principal components, selected accordingly The factors are used as the first principal component, the second principal component, and so on up to the third principal component. Principal component.

[0014] Furthermore, step S4 employs K-means clustering analysis to analyze the material types, specifically including the following steps: ① Analyzing the material usage data... Data points Randomly generated Clusters ② Calculate the Euclidean distance from each sample to the cluster center. Find the minimum value of each distance. And assign the samples to the cluster to which the nearest cluster center belongs; ③ Calculate the average distance from the sample to the cluster center in each cluster. In the formula: Indicates the first Number of samples in the class ④ Represent the updated cluster centers; ③ Reclassify the sample categories and repeat step ③, calculating the distance from the new cluster sample to the cluster center, until the distance to the cluster center is minimized, the cluster center no longer changes, and the clustering iteration is completed.

[0015] Furthermore, the optimal number of clusters for the K-means clustering analysis method using the silhouette coefficient in step S4 includes... The overall silhouette coefficient of the clustering of a dataset of samples is expressed as: Clustering profile coefficient In the formula: Let the set of clusters represent the cohesion of sample points within a cluster. For sample points If it belongs to a cluster, then , Let the set represent the separation degree of samples between different clusters. , When the cluster profile coefficient The closer the value is to 1, The closer the K value is to 0, the greater the separation between different clusters is than the cohesion between samples in the same cluster, resulting in the best clustering effect. Plotting the K value of the clustering silhouette coefficient on a graph using a broken line, the number of clusters corresponding to the first inflection point is the optimal number of clusters, which is the optimal number of clusters.

[0016] Furthermore, the step S5, which uses an optimized extreme learning machine to construct a power distribution network construction material demand forecasting model, includes the following steps: given a training dataset containing M types of material demand forecasts for N project samples. , Indicates project sample Project sample description attributes, This represents the target true value of the material demand in M ​​dimensions for this sample, and the activation function of the hidden layer neuron. If the number of hidden neurons is L, then: ① Randomly generate the parameter matrix of the hidden neurons. ② Output matrix of hidden layer neurons Here, the neuron activation function is an additive function. denote the hidden layer node parameters, denote the input weight, denote the hidden layer neuron threshold; ③ the output weight vector , denote the pseudo generalized inverse matrix of matrix , ; ④ the decision function .

[0017] Further, the particle swarm algorithm is used in the step S5 to optimize the input and output weights and the bias threshold of the extreme learning machine, so as to obtain the optimal input and output weights and the bias threshold; the training error of the extreme learning machine is used as the fitness function of the particle swarm algorithm in the step S5, and is expressed as: .

[0018] Further, the optimization of the extreme learning machine parameters by the particle swarm algorithm in the step S5 specifically includes the following steps: ① initializing the particle swarm and related parameters; ② calculating the fitness value of the particle, obtaining the initial individual optimal position, and obtaining the global optimal position of the initial population according to the individual optimal position; ③ updating the particle speed and position according to the speed and position updating formula; ④ calculating the fitness value of the updated particle; ⑤ updating the individual optimal position and the global optimal position of each particle; ⑥ judging whether the termination condition is reached, and if the termination condition is reached, ending to obtain the global optimal solution, and if the termination condition is not reached, returning to step ④ to continue iteration.

[0019] Further, the speed and position updating formula of the particle swarm algorithm in the step S5 is expressed as: , , wherein, and denote the speed of the particle, and denote the position of the particle, denote the individual optimal position in the particle, denote the population optimal position, and denote the learning coefficient, and are random numbers, denote the weight coefficient, denote the maximum search weight, and the value is 0.8, denote the minimum search weight, and the value is 0.2, denote the current iteration number, denote the maximum iteration number.

[0020] Due to the adoption of the above technical solutions, the application has the following beneficial effects:

[0021] (1) by principal component analysis method based on material data dimension reduction processing, and based on clustering analysis algorithm for clustering analysis of material project attributes, get a variety of material type use condition similar project cluster, reduce the prediction model output space dimension and complexity, improve data density and prediction model precision;

[0022] (2) by using extreme learning machine to build power distribution network construction material demand prediction model, while ensuring the operation precision, it has faster learning speed, and the parameters of the extreme learning machine are optimized by particle swarm algorithm, which effectively improves the prediction accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 is the flow chart of the present application. DETAILED DESCRIPTION

[0024] The technical solutions of the present application will be further specifically described below by examples, and combined with the drawings.

[0025] As shown in the figure, a power distribution network construction material demand analysis and prediction method comprises the following analysis and prediction steps: Figure 1 S1, collecting historical material data of power distribution network construction projects, the historical material data includes project attribute data and material consumption data.

[0026] S2, preprocessing the material consumption data, including data cleaning, standardization and normalization processing.

[0027] Preferably, the data cleaning in step S2 is used to process repeated data, missing data and abnormal data to ensure data integrity and accuracy; data standardization and normalization are used to scale the data to a unified scale, wherein data standardization is used to eliminate the dimension between feature attributes, and the data is converted to a normal distribution with a mean of 0 and a standard deviation of 1,

[0028] , wherein: is the standardized data, is the original data, is the mean of the original data, is the standard deviation; data normalization is used to scale the data to interval.

[0029] S3, dimension reduction processing of material data based on principal component analysis method.

[0030] Preferably, the step S3 adopts principal component analysis method to process the dimension reduction of material data, which specifically comprises the following steps: ① assuming that the original material data has samples, each sample has observation values, then the original variable matrix is represented as:​ , wherein: is the first observation value of the first sample; 2, the original data is normalized, , , , , ; 3, the correlation coefficient matrix of the observation object is constructed: , ; 4, the Jacobi method is used to calculate the characteristic equation, and the non-negative eigenvalue corresponding to the correlation coefficient matrix is calculated, and the eigenvalue is one-to-one corresponding: , , then the characteristic vector of the corresponding is , and satisfies: ; 5, the cumulative variance of the first principal component is , the first factor whose cumulative variance is closest to 1 is the first principal component, and the first principal component, the second principal component and the first principal component are selected as the first principal component, the second principal component and the first principal component.

[0031] S4, the attribute of the material project is analyzed based on the clustering analysis algorithm.

[0032] Preferably, the K-means clustering analysis method is used to analyze the material type in step S4, which specifically includes the following steps: 1, the data points of the material consumption data are randomly generated , , , ; 2, the Euclidean distance of each sample to the cluster center is calculated , the minimum value of each distance is solved , and the sample is distributed to the cluster to which the nearest cluster center belongs; 3, the average value of the sample to the cluster center in each cluster is calculated, , wherein: represents the number of samples in the first class, represents the updated cluster center; 4, the sample class is redivided, and step 3 is repeated to calculate the distance of the new cluster sample to the cluster center until the distance to the cluster center is minimum and the cluster center no longer changes, and the clustering iteration is completed. Preferably, the optimal cluster number of the K-means clustering analysis method is determined by the silhouette coefficient in step S4, and the total silhouette coefficient of the data set containing

[0033] samples is represented as: , , ​​​​​In the formula: Let the set of clusters represent the cohesion of sample points within a cluster. For sample points If it belongs to a cluster, then , Let the set represent the separation degree of samples between different clusters. , When the cluster profile coefficient The closer the value is to 1, The closer the K value is to 0, the greater the separation between different clusters is than the cohesion between samples in the same cluster, resulting in the best clustering effect. Plotting the K value of the clustering silhouette coefficient on a graph using a broken line, the number of clusters corresponding to the first inflection point is the optimal number of clusters, which is the optimal number of clusters.

[0034] S5. The parameters of the Extreme Learning Machine are optimized based on the Particle Swarm Optimization algorithm. The optimized Extreme Learning Machine is used to construct a prediction model for the demand of materials for power distribution network construction, and to predict the quantity and amount of materials required for power distribution network construction.

[0035] Preferably, the step S5 of constructing a power distribution network construction material demand prediction model using an optimized extreme learning machine includes the following steps: given a training dataset containing M types of material demand prediction values ​​for N project samples. , Indicates project sample Project sample description attributes, This represents the target true value of the material demand in M ​​dimensions for this sample, and the activation function of the hidden layer neuron. If the number of hidden neurons is L, then: ① Randomly generate the parameter matrix of the hidden neurons. ② Output matrix of hidden layer neurons Here, the neuron activation function is an additive function. These are parameters for hidden layer nodes. Indicates the input weights. ③ Output weight vector , Representation matrix The inverse of the pseudo-generalized inverse matrix, ④ Decision function .

[0036] Preferably, in step S5, the particle swarm optimization algorithm is used to optimize the input-output weights and bias threshold of the extreme learning machine to obtain the optimal input-output weights and bias threshold; in step S5, the particle swarm optimization algorithm uses the training error of the extreme learning machine as the fitness function, expressed as: .

[0037] Preferably, the step S5 of optimizing the extreme learning machine parameters by using the particle swarm algorithm comprises the following steps: ① initializing the particle swarm and related parameters; ② calculating the particle fitness value, obtaining the initial individual optimal position, and obtaining the global optimal position of the initial population according to the individual optimal position; ③ updating the particle speed and position according to the speed and position updating formula; ④ calculating the fitness value of the updated particle; ⑤ updating the individual optimal position and the global optimal position of each particle; ⑥ judging whether the termination condition is reached, and if the termination condition is reached, ending to obtain the global optimal solution, and if the termination condition is not reached, returning to step ④ to continue iteration.

[0038] Preferably, the speed and position updating formula of the particle swarm algorithm in the step S5 is represented as: , wherein, and represent the speed of the particle, and represent the position of the particle, represents the individual optimal position in the particle, represents the optimal position of the population, and represent the learning coefficient, and are random numbers, represents the weight coefficient, represents the maximum search weight, and the value is 0.8, represents the minimum search weight, and the value is 0.2, represents the current iteration number, represents the maximum iteration number.

[0039] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the specific embodiments of the present application can still be modified or replaced by equivalents without departing from the spirit and scope of the present application, and any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered in the scope of the claims of the present application.​

Claims

1. A method for analyzing and predicting the demand for materials in power distribution network construction, characterized in that: The analysis and prediction steps include the following: S1. Collect historical material data for power distribution network construction projects, including project attribute data and material usage data; S2. Preprocess the material usage data, including data cleaning, standardization and normalization. S3. Dimensionality reduction of material data based on principal component analysis; S4. Perform cluster analysis on the attributes of material projects based on cluster analysis algorithms; S5. The parameters of the Extreme Learning Machine are optimized based on the particle swarm optimization algorithm. The optimized Extreme Learning Machine is used to construct a prediction model for the demand of materials for power distribution network construction, and to predict the quantity and amount of materials required for power distribution network construction.

2. The method for analyzing and predicting the demand for materials in power distribution network construction according to claim 1, characterized in that: In step S2, data cleaning is used to process duplicate, missing, and outlier data to ensure data integrity and accuracy. Data standardization and normalization are used to scale the data to a uniform scale. Specifically, data standardization eliminates the dimensions between feature attributes and transforms the data into a normal distribution with a mean of 0 and a standard deviation of 1. In the formula: For standardized data, The original data, The mean of the original data. Standard deviation; data normalization is used to scale the data to... Interval.

3. The method for analyzing and predicting the demand for materials in power distribution network construction according to claim 1, characterized in that: Step S3, which uses principal component analysis to reduce the dimensionality of the material data, specifically includes the following steps: ① Assume that the original material data has There are 10 samples, each sample has 100 samples. If there are 100 observations, the original variable matrix is ​​represented as: In the formula: For the first The first sample ① 1 observation value; ② Standardize the raw data. , , ③ Construct the correlation coefficient matrix of the observed objects: , ④ The characteristic equation is calculated using the Jacobi method, and the corresponding correlation coefficient matrix is ​​obtained. There are 1 non-negative eigenvalues, and each eigenvalue corresponds to a different eigenvalue: Then the corresponding eigenvectors And satisfy: ⑤ Then the first Cumulative variance of each principal component The cumulative variance closest to 1 is the first The factor is the first Principal components, selected accordingly The factors are used as the first principal component, the second principal component, and so on up to the third principal component. Principal component.

4. The method for analyzing and predicting the demand for materials in power distribution network construction according to claim 1, characterized in that: Step S4 uses K-means clustering analysis to analyze the material types, specifically including the following steps: ① Analyzing the material usage data Data points Randomly generated Clusters ② Calculate the Euclidean distance from each sample to the cluster center. Find the minimum value of each distance. And assign the samples to the cluster to which the nearest cluster center belongs; ③ Calculate the average distance from the sample to the cluster center in each cluster. In the formula: Indicates the first Number of samples in the class ④ Represent the updated cluster centers; ③ Reclassify the sample categories and repeat step ③, calculating the distance from the new cluster sample to the cluster center, until the distance to the cluster center is minimized, the cluster center no longer changes, and the clustering iteration is completed.

5. The method for analyzing and predicting the demand for materials in power distribution network construction according to claim 4, characterized in that: In step S4, the optimal number of clusters for the K-means clustering analysis method is determined by the silhouette coefficient, which includes... The overall silhouette coefficient of the clustering of a dataset of samples is expressed as: Clustering profile coefficient In the formula: Let the set of clusters represent the cohesion of sample points within a cluster. For sample points If it belongs to a cluster, then , Let the set represent the separation degree of samples between different clusters. , When the cluster profile coefficient The closer the value is to 1, The closer the K value is to 0, the greater the separation between different clusters is than the cohesion between samples in the same cluster, resulting in the best clustering effect. Plotting the K value of the clustering silhouette coefficient on a graph using a broken line, the number of clusters corresponding to the first inflection point is the optimal number of clusters, which is the optimal number of clusters.

6. The method for analyzing and predicting the demand for materials in power distribution network construction according to claim 1, characterized in that: The step S5, which uses an optimized extreme learning mechanism to build a power distribution network construction material demand prediction model, includes the following steps: given a training dataset containing M types of material demand prediction values ​​for N project samples. , Indicates project sample Project sample description attributes, This represents the target true value of the material demand in M ​​dimensions for this sample, and the activation function of the hidden layer neuron. If the number of hidden neurons is L, then: ① Randomly generate the parameter matrix of the hidden neurons. ② Output matrix of hidden layer neurons Here, the neuron activation function is an additive function. These are the parameters for the hidden layer nodes. Indicates the input weights. ③ Output weight vector , Representation matrix The inverse of the pseudo-generalized inverse matrix, ④ Decision function .

7. The method for analyzing and predicting the demand for materials in power distribution network construction according to claim 6, characterized in that: In step S5, the particle swarm optimization algorithm is used to optimize the input-output weights and bias threshold of the extreme learning machine to obtain the optimal input-output weights and bias threshold. In step S5, the particle swarm optimization algorithm uses the training error of the extreme learning machine as the fitness function, expressed as: .

8. The method for analyzing and predicting the demand for materials in power distribution network construction according to claim 1, characterized in that: Step S5, which uses the particle swarm optimization algorithm to optimize the parameters of the extreme learning machine, specifically includes the following steps: ① Initialize the particle swarm and related parameters; ② Calculate the particle fitness value to obtain the initial optimal position of each individual, and obtain the global optimal position of the initial population based on the optimal positions of the individuals; ③ Update the particle velocity and position according to the velocity and position update formulas; ④ Calculate the updated particle fitness value; ⑤ Update the optimal position of each particle and the global optimal position; ⑥ Determine whether the termination condition has been met. If the termination condition has been met, the process ends and the global optimal solution is obtained. If the termination condition has not been met, the process returns to step ④ to continue iterating.

9. The method for analyzing and predicting the demand for materials in power distribution network construction according to claim 8, characterized in that: The velocity and position update formulas for the particle swarm algorithm in step S5 are expressed as follows: , In the formula, and Indicates the velocity of the particle. and Indicates the particle position. This represents the optimal position of an individual particle. Indicates the optimal position of the population. and Represents the learning coefficient. and It is a random number. Indicates the weighting coefficient. This represents the maximum search weight, with a value of 0.

8. This represents the minimum search weight, with a value of 0.

2. Indicates the current iteration number. This indicates the maximum number of iterations.

Images