CN114757393A - Method and device for predicting project expansion duration of power industry - Google Patents

Method and device for predicting project expansion duration of power industry Download PDF

Info

Publication number
CN114757393A
CN114757393A CN202210269439.9A CN202210269439A CN114757393A CN 114757393 A CN114757393 A CN 114757393A CN 202210269439 A CN202210269439 A CN 202210269439A CN 114757393 A CN114757393 A CN 114757393A
Authority
CN
China
Prior art keywords
duration
prediction
model
predictor
original
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210269439.9A
Other languages
Chinese (zh)
Inventor
林镜星
周鑫
林其雄
谢志炜
许斌斌
黄智豪
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangzhou Power Supply Bureau of Guangdong Power Grid Co Ltd
Original Assignee
Guangzhou Power Supply Bureau of Guangdong Power Grid Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangzhou Power Supply Bureau of Guangdong Power Grid Co Ltd filed Critical Guangzhou Power Supply Bureau of Guangdong Power Grid Co Ltd
Priority to CN202210269439.9A priority Critical patent/CN114757393A/en
Publication of CN114757393A publication Critical patent/CN114757393A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Software Systems (AREA)
  • Tourism & Hospitality (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • General Business, Economics & Management (AREA)
  • Molecular Biology (AREA)
  • Artificial Intelligence (AREA)
  • Marketing (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Primary Health Care (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a method for predicting the project duration of power industry expansion, which comprises the following steps: acquiring initial data of the project duration of the power industry expansion to be predicted; inputting initial data into a trained hybrid prediction model to obtain a duration prediction value output by the hybrid prediction model, wherein the hybrid prediction model comprises a BP (Back propagation) prediction sub-model, a DBN (direct Back propagation) prediction sub-model and an ELM (element-by-element) prediction sub-model; the duration prediction value comprises a first duration prediction value output by the BP prediction submodel, a second duration prediction value output by the DBN prediction submodel and a third duration prediction value output by the ELM prediction submodel; and obtaining the project duration of the power industry expansion project to be predicted according to the first duration predicted value, the second duration predicted value, the third duration predicted value and a preset weight value. The method and the device determine the preset weight value through the information entropy and the average mutual information, obtain a trained hybrid prediction model by using the weight value, and obtain the electric power industry expansion project duration according to the trained hybrid prediction model.

Description

Method and device for predicting project duration of power industry expansion
Technical Field
The invention relates to the technical field of electric power, in particular to a method and a device for predicting electric power industry expansion project duration based on information entropy.
Background
The business expansion installation is an important work of marketing and distribution network management of power supply enterprises, and with the rapid increase of economy, business expansion installation businesses are more and more.
The existing business expansion installation process is relatively single, the mode is relatively fixed, the involvement link is wide, the process is relatively mechanized, the operation is relatively simplified, the operation efficiency of the work process is low, and the value-added activity time is long. In the prediction of the duration of the electric power industry expansion project, the related data has the characteristics of large scale and complex uncertainty, and meanwhile, certain errors necessarily exist in the data due to the limitation of human or monitoring level.
Disclosure of Invention
Based on the information entropy, the invention provides a method and a device for predicting the project duration of the power industry expansion project based on the information entropy. Information entropy, because it can process the entire data, has advantages in systematicness, integrity and objectivity, and therefore can be used to process such data for electric power industry expansion project duration prediction. The accuracy of the predicted duration can be improved.
According to a first aspect of some embodiments of the present application, there is provided a method for predicting a project duration of an electric power business expansion project, the method comprising the steps of:
acquiring initial data of the project duration of the power industry expansion to be predicted, wherein the initial data comprises power industry expansion project nodes and power industry expansion project cost;
inputting the initial data into a trained hybrid prediction model to obtain a duration prediction value output by the hybrid prediction model, wherein the hybrid prediction model comprises a BP (back propagation) predictor sub-model, a DBN (direct propagation network) predictor sub-model and an ELM (element-by-element) predictor sub-model; the time length prediction value comprises a first time length prediction value output by the BP prediction submodel, a second time length prediction value output by the DBN prediction submodel and a third time length prediction value output by the ELM prediction submodel;
and obtaining the project duration of the power industry expansion project to be predicted according to the first duration predicted value, the second duration predicted value, the third duration predicted value and a preset weight value.
Further, the determining the project duration of the power industry expansion project to be predicted according to the duration predicted value and the weight comprises the following calculation:
VMPB＝w1VA+w2VB+w3VC
wherein VMPB represents the project duration of the power business expansion to be predicted, V A,VB,VCRespectively obtaining a first time length predicted value, a second time length predicted value and a third time length predicted value, wherein A represents the BP predictor model, B represents the DBN predictor model, and C represents the ELM predictor model; w1 denotes a first weight, w2 denotes a second weight, and w3 denotes a third weight.
Further, the step of training the hybrid predictive model includes:
acquiring historical data, wherein the historical data comprises original nodes, original cost and original duration, and the historical data is randomly divided into a training set and a verification set according to a proportion;
training each predictor model by taking the original nodes and the original cost in the training set as target input and the original duration of the training set as target output to obtain each trained predictor model;
respectively inputting the verification sets into each trained predictor model to obtain information entropy and average mutual information of each trained predictor model, wherein the information entropy comprises first information entropy of the BP predictor model, second information entropy of the DBN predictor model and third information entropy of the ELM predictor model, and the average mutual information comprises first average mutual information of the BP predictor model, second average mutual information of the DBN predictor model and third average mutual information of the ELM predictor model;
Obtaining the preset weight value according to the ratio of the information entropy to the average mutual information, wherein the preset weight value comprises a first weight value of a BP (back propagation) predictor model, a second weight value of the DBN (direct propagation network) predictor model and a third weight value of the ELM predictor model;
and obtaining a mixed prediction model corresponding to the first weight value, the second weight value and the third weight value, namely the trained mixed prediction model.
Further, the inputting the verification set into each trained predictor model to obtain the information entropy of each predictor model includes:
respectively inputting the verification set into each trained predictor model for prediction to obtain the prediction accuracy of each trained predictor model, wherein the calculation formula of the accuracy is as follows:
Figure BDA0003554011770000021
wherein, aijUsing any one predictor model j to predict the accuracy of the ith verification data of the verification set; r isiRepresenting the original duration of the ith verification data; fijRepresenting the duration predicted value of the j prediction submodel to the ith verification data;
for m items of verification data, the j predictor model generates m corresponding precision values, and an integer part is screened to obtain a matrix R mn：
Figure BDA0003554011770000031
Wherein r isijRepresents a in the jth columnijThe number of occurrences of (c);
according to a matrix RmnAnd calculating the information entropy of the j prediction submodel, wherein the calculation formula is as follows:
Figure BDA0003554011770000032
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0003554011770000033
Ejis the information entropy of the j predictor model, wi′jAnd pijlogpijCorresponds to, NjA in the j-th column of the matrixijGreater than the amount of accuracy X%.
Further, the inputting the verification set into each trained predictor model to obtain average mutual information of the predictor models includes:
preset rk＝r1Then define the step function UN (r)k,r1) Otherwise, UN (r)k,r1)＝0，
Wherein r isk,r1(k 1) is the matrix RmnTwo rows in (1);
definition of
Figure BDA0003554011770000034
Obtain vector Cm T＝(c1,c2,…,cm)T，
Obtaining the average mutual information of the predictor models, wherein the calculation formula is as follows:
Figure BDA0003554011770000035
wherein the content of the first and second substances,
Figure BDA0003554011770000036
any two of the predictor models.
Further, obtaining the preset weight according to the ratio of the information entropy to the average mutual information, including the following steps:
obtaining the weight of the predictor model according to the information entropy and the average mutual information of each predictor model, wherein the calculation formula is as follows:
Figure BDA0003554011770000041
where Z is a parameter used to normalize the weights of all basic algorithms to ensure that the sum of all weights is 1, EjThe information entropy of the predictor model is I, and the average mutual information of the predictor model is I.
Further, the BP predictor model comprises an input layer consisting of n neurons, a hidden layer consisting of q neurons and an output layer consisting of m neurons;
and training each predictor model by taking the original nodes and the original cost in the training set as target input and the original time length of the training set as target output to obtain each trained predictor model, wherein the method comprises the following steps of:
inputting the original node and the original cost into the input layer to obtain Sk,
Figure BDA0003554011770000042
wherein x isi(i-1, 2, …, n) represents input data, vki(i ═ 1,2, …, n ═ 1,2, …, q) represents the connection weights of the input layer to the hidden layer;
inputting the Sk to the hidden layer to obtain zk＝f(Sk)；
Inputting the Sj to the output layer to obtain yj, wherein
Figure BDA0003554011770000043
Wherein, wjk(k 1,2, …, q; j 1,2, …, m) represents the hidden-to-output layer connection weight, yj represents the original duration;
and obtaining the trained BP prediction model.
Further, the DBN prediction model comprises n layers of RBMs and a layer of BP neural network, wherein the RBMs comprise a layer of input layer and a layer of 2n-1 hidden layer which are sequentially connected with each other;
taking the original nodes and the original cost in the training set as target input, taking the original duration of the training set as target output, training each predictor model to obtain each trained predictor model, and the method comprises the following steps:
Inputting the original node and the original cost into the input layer, and abstracting and extracting input data of the input layer by a hidden layer connected with the input layer;
abstracting and extracting the input data of the previous hidden layer in sequence according to the connection sequence until the 2n hidden layer abstracts and extracts the input data of the 2n-1 hidden layer to obtain the abstracted and extracted input data;
and inputting the abstracted and extracted input data into a BP network, and outputting the original time length by the BP network to obtain a trained DBN model.
Further, the ELM predictive model comprises M hidden layers, an input layer and an output layer;
and training each predictor model by taking the original nodes and the original cost in the training set as target input and the original time length of the training set as target output to obtain each trained predictor model, wherein the method comprises the following steps of:
inputting the original node and the original cost into the hidden layer, and calculating an output matrix H of the original node in the hidden layer:
Figure BDA0003554011770000051
wherein the function G (a) is activatedj,bjX) is a Sigmoid function, aj,bjRespectively the connection weight and the offset value between the input layer and the hidden layer, N is the number of sampling samples, x NFor sampling data, tjTo classify the tag numbers, the relationship between each other satisfies:
Figure BDA0003554011770000052
wherein R isn×RmRepresentative is the total real number;
calculating an optimal output weight matrix of the original nodes of the hidden layer by a least square method:
Figure BDA0003554011770000053
wherein, betaL×MIs a weight matrix between the hidden layer and the output layer;
when the number of samples N is greater than or equal to the number of original nodes L of the hidden layer, calculating an optimal output weight matrix:
Figure BDA0003554011770000054
otherwise, calculating an optimal output weight matrix:
Figure BDA0003554011770000055
wherein the content of the first and second substances,
Figure BDA0003554011770000056
and the ELM prediction model corresponding to the optimal weight matrix is the trained ELM prediction model.
According to a second aspect of some embodiments of the present application, there is provided an electric power business expansion project duration prediction apparatus, including:
the system comprises an initial data acquisition module, a data processing module and a data processing module, wherein the initial data acquisition module is used for acquiring initial data of the project duration of the power business expansion to be predicted, and the initial data comprises power business expansion project nodes and power business expansion project cost;
the duration prediction value obtaining module is used for inputting the initial data into a trained hybrid prediction model to obtain a duration prediction value output by the hybrid prediction model, wherein the hybrid prediction model comprises a BP (Back propagation) prediction sub-model, a DBN (Back propagation network) prediction sub-model and an ELM (element-to-metal) prediction sub-model; the duration prediction value comprises a first duration prediction value output by the BP prediction submodel, a second duration prediction value output by the DBN prediction submodel and a third duration prediction value output by the ELM prediction submodel;
And the electric power industry expansion project duration obtaining module is used for obtaining the electric power industry expansion project duration to be predicted according to the first duration predicted value, the second duration predicted value, the third duration predicted value and a preset weight value.
According to the method and the device for predicting the project duration of the power industry expansion project, the nodes of the power industry expansion project and the corresponding cost are input into the trained hybrid prediction model, and the project duration of the power industry expansion project to be predicted can be obtained according to the preset weight value. And the training of the hybrid prediction model is to acquire original power industry expansion project nodes and cost as input, train the BP prediction sub-model, the DBN prediction sub-model and the ELM prediction sub-model respectively by using the original power industry expansion project nodes, the cost and the time length, determine the information entropy and the average mutual information of the BP prediction model, the DBN prediction model and the ELM prediction model, determine a preset weight value by using the information entropy and the average mutual information, and form the hybrid prediction model by using the obtained weight value and the trained prediction sub-model. The most concise expression of information entropy is a measure of uncertainty, and the main idea is to describe the information quantity by means of probability distribution. In the research of the duration prediction of the power industry expansion project, related data have the characteristics of large scale and complex uncertainty, and certain errors of the data are inevitable due to the limitation of human or monitoring level. The information entropy has the advantages of systematicness, integrity and objectivity because the information entropy can process the whole data, and the weight value of each predictor model in the hybrid prediction model is determined by adopting the ratio based on the mutual information and the information entropy, so that the corresponding time length of the hybrid prediction model is predicted according to different project nodes and cost. The method not only avoids the problem of local optimization of the weight in the hybrid prediction model, but also has better interpretability, enables the model with higher precision and better stability to be more important, and can accurately predict the duration, thereby effectively improving the efficiency of the business process and improving the core competitiveness of the power enterprise.
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Drawings
Fig. 1 is a flowchart illustrating steps of a method for predicting a project duration of an electric power business expansion project according to an embodiment of the present application;
fig. 2 is a schematic structural diagram of a device for predicting a duration of an electric power business expansion project in an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.
It should be understood that the embodiments described are only some embodiments of the present application, and not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application without any creative effort belong to the protection scope of the embodiments in the present application.
The terminology used in the embodiments of the present application is for the purpose of describing particular embodiments only and is not intended to be limiting of the embodiments of the present application. As used in the examples of this application and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.
When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the application, as detailed in the appended claims. In the description of the present application, it is to be understood that the terms "first," "second," "third," and the like are used solely to distinguish one from another similar human body, and are not necessarily used to describe a particular order or sequence, nor are they to be construed as indicating or implying relative importance. The specific meaning of the above terms in the present application can be understood by those of ordinary skill in the art as appropriate.
In addition, in the description of the present application, "a plurality" means two or more unless otherwise specified. "and/or" describes an association relationship associated with a human body, and indicates that three relationships may exist, for example, a and/or B, and may indicate: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the context of the associated body is an "or" relationship.
The power business expansion project duration is used for indicating the consumption duration of the power business expansion project. The electric power industry expansion project mainly comprises all links of electric power engineering, such as: designing, auditing, constructing, checking, completing and checking and the like. In order to ensure reasonable planning of the power business expansion project, a planning time length of the power business expansion project is generally required, and in order to ensure the accuracy of the planning time length, the time length of the power business expansion project is required to be predicted. In the prediction of the duration of the electric power industry expansion project, the related data has the characteristics of large scale and complex uncertainty, and certain errors necessarily exist in the data due to the limitation of human or monitoring level.
To solve the above problem, please refer to fig. 1, in which fig. 1 provides a method for predicting a duration of an electric power expansion project, the method includes the following steps:
In step S101, initial data of a power business expansion project duration to be predicted is obtained, where the initial data includes power business expansion project nodes and power business expansion project costs.
The influence factors of the electric power industry expansion project duration are many, and the major influence on the duration is the node number of the electric power industry expansion project and the cost of the electric power industry expansion project. The larger the quantity, the more the cost, and the longer the corresponding power business expansion project may be. Thus, the duration can be predicted based on project node and project cost.
In step S102, inputting the initial data into a trained hybrid prediction model to obtain a duration prediction value output by the hybrid prediction model, wherein the hybrid prediction model includes a BP prediction sub-model, a DBN prediction sub-model and an ELM prediction sub-model; the duration prediction value comprises a first duration prediction value output by the BP prediction submodel, a second duration prediction value output by the DBN prediction submodel and a third duration prediction value output by the ELM prediction submodel.
The trained hybrid prediction model can predict and obtain a duration prediction value of the power industry expansion project duration according to the initial data. Since the hybrid model includes the three predictor models described above, three duration predictors are obtained accordingly. The characteristics of the three predictor models are different, so the weights of the three duration prediction values are different, and the measured duration prediction value is not the most accurate.
In step S103, the electric power industry expansion project duration to be predicted is obtained according to the first duration predicted value, the second duration predicted value, the third duration predicted value and a preset weight value.
The preset weight value is used for indicating the weight value of each trained predictor model in the hybrid prediction model, the weight value represents the specific gravity of each predictor model in the hybrid prediction model, and the importance degree and the accuracy degree of each predictor model are determined. Therefore, the weighted value of each predictor model is combined with the corresponding duration prediction value, and a more reasonable and accurate duration prediction value can be obtained.
According to the method and the device, the duration corresponding to different project nodes and expenses is predicted through the preset weight value and the trained hybrid prediction model. The method not only avoids the problem that the weight value of the sub-prediction model in the hybrid prediction model is locally optimal, but also has better interpretability, enables the model with higher precision and better stability to be more important, and can accurately predict the duration of the electric power industry expansion project, thereby effectively improving the efficiency of the business process.
In a specific embodiment, in step S103, determining the to-be-predicted electric power business expansion project duration according to the duration prediction value and the weight includes the following calculation:
VMPB＝w1VA+w2VB+w3VC
Wherein, VMPBIndicating the duration of the power business expansion project to be predicted, VA,VB,VCRespectively obtaining the first time length predicted value, the second time length predicted value and the third time length predicted value, wherein A represents the BP predictor model, B represents the DBN predictor model, and C represents the ELM predictor model; w is a1Represents a first weight, w2Represents a second weight, w3Representing a third weight.
The first weight is used for indicating the importance degree of the BP predictor model in the hybrid prediction model and the proportion of the first time length prediction value; the second weight is used for indicating the importance degree of the DBN predictor model in the hybrid prediction model and the proportion of the second duration prediction value; the third weight is used to indicate the importance of the ELM predictor model in the hybrid prediction model, and the specific gravity of the third duration prediction value.
In the hybrid prediction model, the proportion and weight of each predictor model in the hybrid prediction model are not determined, so that the importance degree of each predictor model in the hybrid prediction model needs to be obtained in order to ensure the accuracy of the prediction result of the hybrid prediction model and avoid the situation that a certain predictor model is locally optimal, that is, a trained hybrid prediction model needs to be obtained, in a specific embodiment, the step of training the hybrid prediction model comprises the following steps:
In step S201, historical data is obtained, where the historical data includes original nodes, original costs, and original durations, and the historical data is randomly divided into a training set and a verification set according to a proportion.
Specifically, the historical data may be randomly divided into a training set and a validation set in an 8:2 ratio.
In step S202, the original nodes and the original costs in the training set are used as target inputs, the original time length of the training set is used as target output, and each predictor model is trained to obtain each trained predictor model.
And training the hybrid prediction model through a known training set, namely respectively inputting the training set into a BP (Back propagation) predictor model, a DBN (Back propagation network) predictor model and an ELM (element-element model) predictor model in the hybrid prediction model. And respectively obtaining a trained BP predictor model, a DBN predictor model and an ELM predictor model, wherein the combination of the three predictor models is a trained hybrid prediction model.
In step S203, the verification sets are respectively input to each trained predictor model to obtain an information entropy and average mutual information of each trained predictor model, where the information entropy includes a first information entropy of the BP predictor model, a second information entropy of the DBN predictor model, and a third information entropy of the ELM predictor model, and the average mutual information includes a first average mutual information of the BP predictor model, a second average mutual information of the DBN predictor model, and a third average mutual information of the ELM predictor model.
The information entropy is used for describing the uncertainty of the information source, is a mathematical abstract concept and can represent the occurrence probability of certain specific information. The most concise expression of information entropy is a measure of uncertainty, and the main idea is to describe the amount of information by means of probability distribution.
Average mutual information is a useful information measure in information theory, which can be seen as the amount of information contained in a random variable about another random variable, or the unsuitability of a random variable to be reduced by knowing another random variable.
And acquiring the information entropy and the average mutual information of each predictor model through the verification set, wherein the purpose of acquiring is to obtain the weight value corresponding to each predictor model through the information entropy and the average mutual information.
In a specific example, obtaining the information entropy of each predictor model comprises the following steps:
respectively inputting the verification set into each trained predictor model for prediction to obtain the prediction accuracy of each trained predictor model, wherein the calculation formula of the accuracy is as follows:
Figure BDA0003554011770000091
wherein, aijUsing any one predictor model j to predict the accuracy of the ith verification data of the verification set; r iRepresenting the original time length of the ith verification data; fijAnd representing the predicted value of the duration of the j prediction submodel on the ith verification data. The predictor model j represents any one of the predictor models in the hybrid prediction model.
For m items of verification data, the j predictor model generates m corresponding precision values, and an integer part is screened to obtain a matrix Rmn：
Figure BDA0003554011770000092
Wherein r isijRepresents a in the j-th columnijThe number of occurrences of (c).
According to a matrix RmnAnd calculating the information entropy of the j prediction submodel, wherein the calculation formula is as follows:
Figure BDA0003554011770000093
wherein the content of the first and second substances,
Figure BDA0003554011770000101
Ejis the information entropy of the j predictor model, wi′jAnd pijlogpijCorresponds to, NjA in the j-th column of the matrixijThe quantity is larger than the accuracy rate X%, and X takes a proper value according to actual needs.
In a specific example, obtaining the average mutual information of the predictor models comprises the following steps:
preset rk＝r1Then define the step function UN (r)k,r1) Otherwise, UN (r)k,r1)＝0，
Wherein r isk,r1(k 1) is the matrix RmnTwo rows in (a).
Definition of
Figure BDA0003554011770000102
Obtain vector Cm T＝(c1,c2,…,cm)TAnd obtaining the average mutual information of the predictor models, wherein the calculation formula is as follows:
Figure BDA0003554011770000103
wherein the content of the first and second substances,
Figure BDA0003554011770000104
any two of the predictor models. Specifically, J' is selected as the group that works best in any two predictor model combinations.
In step S204, the preset weight is obtained according to a ratio of the information entropy to the average mutual information, where the preset weight includes a first weight of a BP predictor model, a second weight of the DBN predictor model, and a third weight of the ELM predictor model.
In a specific example, obtaining the preset weight includes the following steps:
obtaining the weight of the predictor model according to the information entropy and the average mutual information of each predictor model, wherein the calculation formula is as follows:
Figure BDA0003554011770000105
where Z is a parameter used to normalize the weights of all basic algorithms to ensure that the sum of all weights is 1, EjInformation entropy for predictor submodel, I predictor submodelAverage mutual information of (3).
In step S205, a hybrid prediction model corresponding to the first weight, the second weight, and the third weight is obtained, that is, the trained hybrid prediction model.
In a specific embodiment, the BP predictor model comprises an input layer consisting of n neurons, a hidden layer consisting of q neurons, and an output layer consisting of m neurons;
in step S202, training each predictor model by using the original nodes and the original costs in the training set as target inputs and the original duration of the training set as target outputs, to obtain each trained predictor model, including:
inputting the original node and the original cost into the input layer to obtain Sk,
Figure BDA0003554011770000111
wherein x isi(i-1, 2, …, n) represents input data, v ki(i ═ 1,2, …, n; (k ═ 1,2, …, q) represents the connection weights of the input layer to the hidden layer.
Inputting the Sk to the hidden layer to obtain zk＝f(Sk)。
Inputting the Sj to the output layer to obtain yj, wherein
Figure BDA0003554011770000112
Wherein, wjkAnd (k is 1,2, …, q, j is 1,2, …, m) represents the connection weight value from the hidden layer to the output layer, and yj represents the original time length, so that the trained BP prediction model is obtained.
In a specific embodiment, the DBN prediction model comprises n layers of RBMs and a layer of BP neural network, wherein the RBMs comprise a layer of input layer and a layer of 2n-1 hidden layer which are sequentially connected with each other;
in step S202, training each predictor model by using the original nodes and the original costs in the training set as target inputs and the original duration of the training set as target outputs, to obtain each trained predictor model, including:
and inputting the original node and the original cost into the input layer, and abstracting and extracting the input data of the input layer by a hidden layer connected with the input layer.
And abstracting and extracting the input data of the previous hidden layer in sequence according to the connection sequence until the 2n hidden layer abstracts and extracts the input data of the 2n-1 hidden layer to obtain the abstracted and extracted input data.
And inputting the abstracted and extracted input data into a BP network, and outputting the original time length by the BP network to obtain a trained DBN model.
In a particular embodiment, the ELM predictive model includes M hidden layers, an input layer, and an output layer;
in step S202, training each predictor model by using the original nodes and the original costs in the training set as target inputs and the original duration of the training set as target outputs, to obtain each trained predictor model, including:
inputting the original node and the original cost into the hidden layer, and calculating an output matrix H of the original node in the hidden layer:
Figure BDA0003554011770000113
wherein the function G (a) is activatedj,bjX) is a Sigmoid function, aj,bjRespectively, the connection weight and the bias value between the input layer and the hidden layer;
n is the number of sampling samples, xNFor sampling data, tjTo classify the tag numbers, the relationship between each other satisfies:
Figure BDA0003554011770000121
wherein R isn×RmRepresentative is the total real number.
Calculating an optimal output weight matrix of the original nodes of the hidden layer by a least square method:
Figure BDA0003554011770000122
wherein, betaL×MIs a weight matrix between the hidden layer and the output layer;
when the number of samples N is greater than or equal to the number of original nodes L of the hidden layer, calculating an optimal output weight matrix:
Figure BDA0003554011770000123
Otherwise, calculating an optimal output weight matrix:
Figure BDA0003554011770000124
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0003554011770000125
and the ELM prediction model corresponding to the optimal weight matrix is the trained ELM prediction model.
Corresponding to the above method for predicting the duration of the power business expansion project, please refer to fig. 2, the present application further provides a device 200 for predicting the duration of the power business expansion project, including:
the initial data acquisition module 210 is configured to acquire initial data of a project duration of power business expansion to be predicted, where the initial data includes power business expansion project nodes and power business expansion project costs;
a duration prediction value obtaining module 220, configured to input the initial data into a trained hybrid prediction model to obtain a duration prediction value output by the hybrid prediction model, where the hybrid prediction model includes a BP prediction sub-model, a DBN prediction sub-model, and an ELM prediction sub-model; the time length prediction value comprises a first time length prediction value output by the BP prediction submodel, a second time length prediction value output by the DBN prediction submodel and a third time length prediction value output by the ELM prediction submodel;
the power industry expansion project duration obtaining module 230 is configured to obtain the power industry expansion project duration to be predicted according to the first duration predicted value, the second duration predicted value, the third duration predicted value, and a preset weight value.
In an optional embodiment, the electric power business expansion project duration prediction module 230 further includes:
the electric power industry expansion project duration prediction unit is used for obtaining the electric power industry expansion project duration to be predicted according to a calculation formula, and the calculation formula is as follows:
VMPB＝w1VA+w2VB+w3VC
wherein VMPB represents the project duration of the power business expansion to be predicted, VA,VB,VCRespectively obtaining a first time length predicted value, a second time length predicted value and a third time length predicted value, wherein A represents the BP predictor model, B represents the DBN predictor model, and C represents the ELM predictor model; w1 denotes the first weight, w2 denotes the second weight, and w3 denotes the third weight.
In an alternative embodiment, the apparatus 200 further comprises a training module for training the hybrid predictive model, the training module comprising:
and the historical data acquisition unit is used for acquiring historical data, wherein the historical data comprises original nodes, original cost and original duration, and is randomly divided into a training set and a verification set according to a proportion.
And the predictor model training unit is used for taking the original nodes and the original cost in the training set as target input, taking the original time length of the training set as target output, and training each predictor model to obtain each trained predictor model.
An information entropy and average mutual information obtaining unit, configured to input the verification set to each trained predictor model, respectively, to obtain an information entropy and average mutual information of each trained predictor model, where the information entropy includes a first information entropy of the BP predictor model, a second information entropy of the DBN predictor model, and a third information entropy of the ELM predictor model, and the average mutual information includes a first average mutual information of the BP predictor model, a second average mutual information of the DBN predictor model, and a third average mutual information of the ELM predictor model.
A preset weight value obtaining unit, configured to obtain the preset weight value according to a ratio between the information entropy and the average mutual information, where the preset weight value includes a first weight value of a BP predictor model, a second weight value of the DBN predictor model, and a third weight value of the ELM predictor model.
And the mixed prediction model obtaining unit is used for obtaining a mixed prediction model corresponding to the first weight value, the second weight value and the third weight value, namely the trained mixed prediction model.
According to the method and the device for predicting the project duration of the power industry expansion project, the nodes of the power industry expansion project and the corresponding cost are input into the trained hybrid prediction model, and the project duration of the power industry expansion project to be predicted can be obtained according to the preset weight value. And the training of the hybrid prediction model is to acquire original power industry expansion project nodes and cost as input, train the BP prediction sub-model, the DBN prediction sub-model and the ELM prediction sub-model by using the original power industry expansion project nodes, the cost and the time length respectively, determine the information entropy and the average mutual information of the BP prediction model, the DBN prediction model and the ELM prediction model, determine a preset weight value by using the information entropy and the average mutual information, and form the hybrid prediction model by using the obtained weight value and the trained prediction sub-model. The most concise expression of information entropy is a measure of uncertainty, and the main idea is to describe the amount of information by means of probability distribution. In the research of the time length prediction of the power industry expansion project, the related data has the characteristics of large scale and complex uncertainty, and certain errors necessarily exist in the data due to the limitation of human or monitoring level. The information entropy has the advantages of systematicness, integrity and objectivity because the information entropy can process the whole data, and the weight value of each predictor model in the hybrid prediction model is determined by adopting the ratio based on the mutual information and the information entropy, so that the corresponding duration of the prediction model is predicted according to different project nodes and cost. The method not only avoids the problem that the weight in the hybrid prediction model is locally optimal, but also has better interpretability, enables the model with higher precision and better stability to be more important, and can accurately predict the duration, thereby effectively improving the efficiency of the business process and improving the core competitiveness of the power enterprise.
It is to be understood that the embodiments of the present application are not limited to the precise arrangements which have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the embodiments of the present application is limited only by the following claims. The above embodiments only express several implementation manners of the embodiments of the present application, and the descriptions are specific and detailed, but should not be construed as limiting the scope of the invention. It should be noted that, for those skilled in the art, without departing from the concept of the embodiments of the present application, several variations and modifications can be made, which all fall within the scope of the embodiments of the present application.

Claims (10)

1. A method for predicting the project duration of power industry expansion is characterized by comprising the following steps of:
acquiring initial data of the project duration of the power industry expansion to be predicted, wherein the initial data comprises power industry expansion project nodes and power industry expansion project cost;
inputting the initial data into a trained hybrid prediction model to obtain a duration prediction value output by the hybrid prediction model, wherein the hybrid prediction model comprises a BP (back propagation) predictor sub-model, a DBN (direct propagation network) predictor sub-model and an ELM (element-by-element) predictor sub-model; the time length prediction value comprises a first time length prediction value output by the BP prediction submodel, a second time length prediction value output by the DBN prediction submodel and a third time length prediction value output by the ELM prediction submodel;
And obtaining the project duration of the power industry expansion project to be predicted according to the first duration predicted value, the second duration predicted value, the third duration predicted value and a preset weight value.
2. The method for predicting the electric power industry expansion project duration according to claim 1, wherein the step of determining the electric power industry expansion project duration to be predicted according to the duration prediction value and the weight comprises the following calculation:
VMPB＝w1VA+w2VB+w3VC
wherein, VMPBIndicating the duration of the power business expansion project to be predicted, VA,VB,VCRespectively obtaining a first time length predicted value, a second time length predicted value and a third time length predicted value, wherein A represents the BP predictor model, B represents the DBN predictor model, and C represents the ELM predictor model; w is a1Represents a first weight, w2Represents a second weight, w3Representing a third weight.
3. The method for predicting the project duration of the electric power industry according to claim 1, wherein the step of training the hybrid prediction model comprises the following steps:
acquiring historical data, wherein the historical data comprises original nodes, original cost and original duration, and the historical data is randomly divided into a training set and a verification set according to a proportion;
training each predictor model by taking the original nodes and the original cost in the training set as target input and the original duration of the training set as target output to obtain each trained predictor model;
Inputting the verification sets into each trained predictor model respectively to obtain information entropy and average mutual information of each trained predictor model, wherein the information entropy comprises first information entropy of a BP (back propagation) predictor model, second information entropy of the DBN (direct propagation) predictor model and third information entropy of the ELM predictor model, and the average mutual information comprises first average mutual information of the BP predictor model, second average mutual information of the DBN predictor model and third average mutual information of the ELM predictor model;
obtaining the preset weight value according to the ratio of the information entropy to the average mutual information, wherein the preset weight value comprises a first weight value of a BP (back propagation) predictor model, a second weight value of the DBN (direct propagation network) predictor model and a third weight value of the ELM predictor model;
and obtaining a mixed prediction model corresponding to the first weight value, the second weight value and the third weight value, namely the trained mixed prediction model.
4. The method for predicting the electric power industry extension project duration according to claim 3, wherein the step of inputting the verification set into each trained predictor model to obtain the information entropy of each predictor model comprises the following steps:
Respectively inputting the verification set into each trained predictor model for prediction to obtain the prediction accuracy of each trained predictor model, wherein the calculation formula of the accuracy is as follows:
Figure FDA0003554011760000021
wherein, aijUsing any one predictor model j to predict the accuracy of the ith verification data of the verification set; riRepresenting the original duration of the ith verification data; fijRepresenting the duration predicted value of the j prediction submodel to the ith verification data;
for m items of verification data, the j predictor model generates m corresponding precision values, and an integer part is screened to obtain a matrix Rmn：
Figure FDA0003554011760000022
Wherein r isijRepresents a in the j-th columnijThe number of occurrences of (c);
according to a matrix RmnAnd calculating the information entropy of the j prediction submodel, wherein the calculation formula is as follows:
Figure FDA0003554011760000023
wherein the content of the first and second substances,
Figure FDA0003554011760000024
Ejis the information entropy, w 'of the j prediction submodel'ijAnd pijlogpijCorresponds to, NjA in the j-th column of the matrixijGreater than the amount of accuracy X%.
5. The method for predicting the project duration of the power industry extension according to claim 3, wherein the step of inputting the validation set into each trained predictor model to obtain average mutual information of the predictor models comprises the following steps:
preset rk＝r1Then define the step function UN (r)k,r1) Otherwise, UN (r) k,r1)＝0，
Wherein r isk,r1(k 1) is a matrix RmnTwo rows in (a);
definition of
Figure FDA0003554011760000031
Obtain the vector Cm T＝(c1,c2,…,cm)T，
Obtaining the average mutual information of the predictor models, wherein the calculation formula is as follows:
Figure FDA0003554011760000032
wherein, the first and the second end of the pipe are connected with each other,
Figure FDA0003554011760000033
j, J' are any two of the predictor models.
6. The method for predicting the duration of the electric power business expansion project according to claim 3, wherein the preset weight is obtained according to the ratio of the information entropy to the average mutual information, and the method comprises the following steps:
obtaining the weight of the predictor model according to the information entropy and the average mutual information of each predictor model, wherein the calculation formula is as follows:
Figure FDA0003554011760000034
where Z is a parameter used to normalize the weights of all basic algorithms to ensure that the sum of all weights is 1, EjThe information entropy of the predictor model is I, and the average mutual information of the predictor model is I.
7. The method for predicting the duration of the electric power business expansion project according to claim 3, wherein:
the BP predictor model comprises an input layer consisting of n neurons, a hidden layer consisting of q neurons and an output layer consisting of m neurons;
taking the original nodes and the original cost in the training set as target input, taking the original duration of the training set as target output, training each predictor model to obtain each trained predictor model, and the method comprises the following steps:
Inputting the original node and the original cost into the input layer to obtain Sk，
Figure FDA0003554011760000035
Wherein x isi(i-1, 2, …, n) represents input data, vki(i ═ 1,2, …, n ═ 1,2, …, q) represents the connection weights of the input layer to the hidden layer;
will the SkInput to the hidden layer, resulting in zk＝f(Sk)；
Will the SjInput to the output layer to obtain yjThe above-mentioned
Figure FDA0003554011760000036
Wherein, wjk(k 1,2, …, q; j 1,2, …, m) represents the connection weight of the hidden layer to the output layer, yjRepresenting the original duration;
and obtaining the trained BP prediction model.
8. The method for predicting the duration of the electric power business expansion project according to claim 3, wherein:
the DBN prediction model comprises n layers of RBMs and a layer of BP neural network, wherein the RBMs comprise a layer of input layer and a layer of 2n-1 hidden layer which are sequentially connected with each other;
taking the original nodes and the original cost in the training set as target input, taking the original duration of the training set as target output, training each predictor model to obtain each trained predictor model, and the method comprises the following steps:
inputting the original node and the original cost into the input layer, and abstracting and extracting input data of the input layer by a hidden layer connected with the input layer;
Abstracting and extracting the input data of the previous hidden layer in sequence according to the connection sequence until the 2n hidden layer abstracts and extracts the input data of the 2n-1 hidden layer to obtain the abstracted and extracted input data;
and inputting the abstracted and extracted input data into a BP network, and outputting the original time length by the BP network to obtain a trained DBN model.
9. The method for predicting the project duration of the electric power industry according to claim 3, wherein the project duration of the electric power industry is the duration of the electric power industry;
the ELM prediction model comprises M hidden layers, an input layer and an output layer;
and training each predictor model by taking the original nodes and the original cost in the training set as target input and the original time length of the training set as target output to obtain each trained predictor model, wherein the method comprises the following steps of:
inputting the original node and the original cost into the hidden layer, and calculating the hidden layerOutput matrix H of the middle original node:
Figure FDA0003554011760000041
wherein the function G (a) is activatedj,bjX) is a Sigmoid function, aj,bjRespectively, the connection weight and offset between the input layer and the hidden layer, N is the number of sample samples, xNTo sample data, tjTo classify the tag numbers, the relationship between each other satisfies:
Figure FDA0003554011760000042
Wherein R isn×RmRepresentative is the total number of real numbers;
calculating an optimal output weight matrix of the original nodes of the hidden layer by a least square method:
Figure FDA0003554011760000043
wherein beta isL×MIs a weight matrix between the hidden layer and the output layer;
when the sample number N is larger than or equal to the original node number L of the hidden layer, calculating an optimal output weight matrix:
Figure FDA0003554011760000044
otherwise, calculating an optimal output weight matrix:
Figure FDA0003554011760000045
wherein, the first and the second end of the pipe are connected with each other,
Figure FDA0003554011760000046
and the ELM prediction model corresponding to the optimal weight matrix is the trained ELM prediction model.
10. An electric power industry expansion project duration prediction device, comprising:
the system comprises an initial data acquisition module, a data processing module and a data processing module, wherein the initial data acquisition module is used for acquiring initial data of the project duration of the power business expansion to be predicted, and the initial data comprises power business expansion project nodes and power business expansion project cost;
the duration prediction value obtaining module is used for inputting the initial data into a trained hybrid prediction model to obtain a duration prediction value output by the hybrid prediction model, wherein the hybrid prediction model comprises a BP (Back propagation) prediction sub-model, a DBN (Back propagation network) prediction sub-model and an ELM (element-to-metal) prediction sub-model; the duration prediction value comprises a first duration prediction value output by the BP prediction submodel, a second duration prediction value output by the DBN prediction submodel and a third duration prediction value output by the ELM prediction submodel;
And the electric power industry expansion project duration obtaining module is used for obtaining the electric power industry expansion project duration to be predicted according to the first duration predicted value, the second duration predicted value, the third duration predicted value and a preset weight value.
CN202210269439.9A 2022-03-18 2022-03-18 Method and device for predicting project expansion duration of power industry Pending CN114757393A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210269439.9A CN114757393A (en) 2022-03-18 2022-03-18 Method and device for predicting project expansion duration of power industry

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210269439.9A CN114757393A (en) 2022-03-18 2022-03-18 Method and device for predicting project expansion duration of power industry

Publications (1)

Publication Number Publication Date
CN114757393A true CN114757393A (en) 2022-07-15

Family

ID=82326774

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210269439.9A Pending CN114757393A (en) 2022-03-18 2022-03-18 Method and device for predicting project expansion duration of power industry

Country Status (1)

Country Link
CN (1) CN114757393A (en)

Similar Documents

Publication Publication Date Title
CN111124840B (en) Method and device for predicting alarm in business operation and maintenance and electronic equipment
CN109784806B (en) Supply chain control method, system and storage medium
He et al. Joint decision-making of parallel machine scheduling restricted in job-machine release time and preventive maintenance with remaining useful life constraints
CN108090788A (en) Ad conversion rates predictor method based on temporal information integrated model
CN104134159A (en) Method for predicting maximum information spreading range on basis of random model
US20200372342A1 (en) Systems and methods for predictive early stopping in neural network training
Hanga et al. A graph-based approach to interpreting recurrent neural networks in process mining
WO2017071369A1 (en) Method and device for predicting user unsubscription
CN114169412A (en) Federal learning model training method for large-scale industrial chain privacy calculation
Almeida et al. Life cycle cost optimisation in highway concrete bridges management
CN111415027A (en) Method and device for constructing component prediction model
Müller-Hannemann et al. Estimating the robustness of public transport schedules using machine learning
CN112632179A (en) Model construction method and device, storage medium and equipment
Anh et al. Effect of gradient descent optimizers and dropout technique on deep learning LSTM performance in rainfall-runoff modeling
CN112463532B (en) Method for constructing SNN workload automatic mapper and automatic mapper
CN113762470A (en) Prediction model construction method, prediction method, device, equipment and medium
Mohammed et al. Predicting performance measurement of residential buildings using machine intelligence techniques (MLR, ANN and SVM)
Elwakil et al. Construction knowledge discovery system using fuzzy approach
CN114757393A (en) Method and device for predicting project expansion duration of power industry
Mawlana et al. Joint probability for evaluating the schedule and cost of stochastic simulation models
Chou et al. Estimating software project effort for manufacturing firms
CN115204501A (en) Enterprise evaluation method and device, computer equipment and storage medium
Milewski Determination of the truss static state by means of the combined FE/GA approach, on the basis of strain and displacement measurements
CN114784795A (en) Wind power prediction method and device, electronic equipment and storage medium
CN115049458A (en) Commodity pushing method and device based on user crowd modeling, medium and equipment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination