CN115577856A - Method and system for predicting construction cost and controlling balance of power transformation project - Google Patents

Method and system for predicting construction cost and controlling balance of power transformation project Download PDF

Info

Publication number
CN115577856A
CN115577856A CN202211413315.XA CN202211413315A CN115577856A CN 115577856 A CN115577856 A CN 115577856A CN 202211413315 A CN202211413315 A CN 202211413315A CN 115577856 A CN115577856 A CN 115577856A
Authority
CN
China
Prior art keywords
power transformation
cost
transformation project
project
lstm
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
CN202211413315.XA
Other languages
Chinese (zh)
Inventor
马莉
周明
周蠡
卢生炜
孙利平
王枫
许汉平
蔡杰
熊川羽
熊一
廖晓红
高晓晶
李智威
陈然
周英博
张赵阳
舒思睿
李吕满
李俊卿
彭云
熊建武
陈东
汤雷
张俊健
于乐
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.)
China Power Engineering Consultant Group Central Southern China Electric Power Design Institute Corp
Economic and Technological Research Institute of State Grid Hubei Electric Power Co Ltd
Original Assignee
China Power Engineering Consultant Group Central Southern China Electric Power Design Institute Corp
Economic and Technological Research Institute of State Grid Hubei Electric Power 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 China Power Engineering Consultant Group Central Southern China Electric Power Design Institute Corp, Economic and Technological Research Institute of State Grid Hubei Electric Power Co Ltd filed Critical China Power Engineering Consultant Group Central Southern China Electric Power Design Institute Corp
Priority to CN202211413315.XA priority Critical patent/CN115577856A/en
Publication of CN115577856A publication Critical patent/CN115577856A/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
    • 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
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/06—Asset management; Financial planning or analysis
    • 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Abstract

The invention provides a method and a system for predicting the construction cost of a power transformation project and controlling balance, which comprises the following steps: measuring and calculating a correlation coefficient of the influence factors and the construction cost of the power transformation project, and correcting the correlation coefficient based on the entropy weight; constructing cost influence factor indexes; taking the influence factor indexes of the construction cost of the transformation project as input nodes of the LSTM, taking the predicted construction cost as output nodes, and establishing an LSTM transformation project construction cost prediction model; acquiring a construction cost influence factor index value of the to-be-tested power transformation project and inputting the construction cost influence factor index value into an LSTM power transformation project construction cost prediction model to obtain the predicted construction cost of the to-be-tested power transformation project; establishing an LSTM power transformation project investment balance rate prediction model by taking the power transformation project cost influence factor index and the prediction cost output by the LSTM power transformation project cost prediction model as input nodes of the LSTM and the investment balance rate as output nodes; and calculating the fund balance according to the predicted manufacturing cost and the investment balance rate. The invention improves the accuracy of the prediction of the cost level of the power transformation project.

Description

Method and system for predicting construction cost and controlling balance of power transformation project
Technical Field
The invention belongs to the technical field of power grid engineering, and particularly relates to a method and a system for predicting construction cost and controlling balance of a power transformation project.
Background
The transformer substation is used as an important place for power conversion and electric energy distribution, and plays an important role in stability and reliability of a power system. Meanwhile, the transformer substation is an important component part with centralized investment of power grid engineering. However, the construction cost of similar substation projects may have great differences due to different voltage levels and different construction conditions of the substations.
The current research on the construction cost of the transformer substation mainly adopts the traditional regression statistical method to realize the prediction of the construction cost level. In recent years, along with the progress of artificial intelligence technology, power transformation project cost prediction methods based on AI technologies such as neural networks and support vector machines have been proposed. Compared with the traditional cost prediction method, the AI technology improves the accuracy of cost prediction and the adaptability of the prediction method.
The collection of mass project cost data establishes a perfect basic project database for power enterprises, historical project data serve for cost control of new projects in order to further develop the value of the transformer project data, the value of the historical project data needs to be further dug deeply, more scientific methods for predicting the cost of the transformer project and controlling balance level are established, and scientific methods for utilizing and processing the project data are provided for the new transformer projects.
Disclosure of Invention
The invention aims to solve the defects in the background technology, and provides a method and a system for predicting the construction cost of a power transformation project and controlling balance, so that the accuracy of predicting the construction cost level of the power transformation project is improved, and the future fund balance level of the project is effectively predicted.
The technical scheme adopted by the invention is as follows: a method for predicting the cost of power transformation project and controlling balance comprises the following steps:
determining influence factors of the construction cost of the power transformation project based on project information of historical power transformation projects of different types and grades;
respectively measuring and calculating the associated coefficients of the cost influence factors of each historical power transformation project and the cost of the power transformation project, and correcting the associated coefficients based on the entropy weight; constructing a cost influence factor index library based on the corrected correlation coefficient; the cost influence factor index library comprises each historical power transformation project and corresponding cost influence factor indexes thereof;
using historical power transformation project cost influence factor indexes as input nodes of the LSTM, using predicted cost as output nodes, optimizing LSTM parameters through historical project samples, and establishing an LSTM power transformation project cost prediction model;
using historical power transformation project cost influence factor indexes and prediction cost output by an LSTM power transformation project cost prediction model as input nodes of an LSTM, using investment balance rate as output nodes, optimizing LSTM parameters through historical project samples, and establishing an LSTM power transformation project investment balance rate prediction model;
acquiring cost influence factor indexes of the power transformation project to be tested and inputting the cost influence factor indexes into an LSTM power transformation project cost prediction model to obtain the predicted cost of the power transformation project to be tested;
inputting the cost influence factor index of the power transformation project to be tested and the predicted cost of the power transformation project to be tested into an LSTM power transformation project investment balance rate prediction model to obtain the investment balance rate of the power transformation project to be tested;
and calculating the possible fund balance generated after the project of the power transformation project to be tested is put into production according to the predicted construction cost and the investment balance rate of the power transformation project to be tested.
In the above technical solution, determining the influence factor of the manufacturing cost based on the project information of the power transformation project includes: general engineering, transformation scale, relevant factors of plant sites and cable engineering quantity. The project overview is divided into 4 factors of voltage level, region, construction property and landform; the transformation scale is divided into 5 factors of single main transformer capacity, the number of main transformers, the type of a transformer station, the number of outgoing lines and the scale of a reactor; the relevant factors of the plant site comprise 3 factors of land acquisition area, foundation treatment mode and site leveling cost; cable engineering quantities include power cables and control cables.
In the above technical solution, the process of measuring and calculating the correlation coefficient between the influence factor and the construction cost of the power transformation project, correcting the correlation coefficient based on the entropy weight, and constructing the construction cost influence factor index library based on the corrected correlation coefficient includes:
the determined influence factors of the construction cost are uniformly quantized, the correlation between the influence factors and the construction cost of the power transformation project is measured and calculated by adopting the correction correlation coefficient based on the entropy weight,
wherein, 5 factors of the region, the construction property, the landform, the transformer type and the foundation treatment mode are further quantified and then calculated;
calculating an initial correlation coefficient of each index based on the Pearson correlation coefficient, and correcting through the entropy weight to obtain a corrected correlation coefficient; eliminating the influence factors with low correlation degree with the construction cost of the power transformation project, and taking the rest influence factors as the indexes of the construction cost influence factors.
In the technical scheme, the influence factor indexes of the construction cost of the power transformation project are used as input nodes of the LSTM, the predicted construction cost is used as output nodes, the LSTM parameters are optimized through historical project samples, and the process of establishing the LSTM power transformation project construction cost prediction model comprises the following steps:
based on a historical project sample, obtaining numerical values of a plurality of cost influence factor indexes and corresponding cost data thereof; inputting the numerical values of a plurality of cost influence factor indexes into the LSTM model through an input gate, performing iterative optimization on the parameters of the LSTM model based on corresponding cost data, stopping iteration when the expected precision requirement is met, and generating an LSTM power transformation project cost prediction model.
In the technical scheme, the historical power transformation project cost influence factor index and the prediction cost output by the LSTM power transformation project cost prediction model are used as input nodes of the LSTM, the investment balance rate is used as output nodes, the LSTM parameters are optimized through historical project samples, and the process of establishing the LSTM power transformation project investment balance rate prediction model comprises the following steps:
respectively inputting numerical values of the cost influence factor indexes of the historical power transformation projects into the LSTM power transformation project cost prediction model to obtain the predicted cost of each historical power transformation project; calculating to obtain the investment balance rate of each historical power transformation project based on the project information of each historical power transformation project; inputting the numerical value of the cost influence factor index of each historical power transformation project and the predicted cost into the LSTM model through an input gate, performing iterative optimization on the parameters of the LSTM model based on the corresponding investment balance rate, stopping iteration when the expected precision requirement is met, and generating the LSTM power transformation project investment balance rate model.
In the above technical solution, the process of obtaining the cost impact factor index of the power transformation project to be tested and inputting the cost impact factor index into the LSTM power transformation project cost prediction model to obtain the predicted cost of the power transformation project to be tested includes:
acquiring cost influence factors of the power transformation project to be tested through project information of the power transformation project to be tested, measuring and calculating correlation coefficients of the cost influence factors of the power transformation project to be tested and the cost of the power transformation project, and correcting the correlation coefficients based on entropy weight; and acquiring the cost influence factor index of the power transformation project to be tested based on the corrected correlation coefficient, inputting the cost influence factor index into the LSTM power transformation project cost prediction model after iterative optimization, and outputting the predicted power transformation project cost from an output gate through an input gate, a forgetting gate and a cell state.
In the technical scheme, the historical project sample is continuously updated, and an iterative LSTM power transformation project cost prediction model and an LSTM power transformation project investment balance rate prediction model are optimized.
The invention provides a copper material price forecasting system based on price index adjustment, which is used for realizing the power transformation project cost forecasting and balance control method in the technical scheme.
The invention provides computer equipment, which comprises a memory and a processor, wherein the memory stores a computer program, and the computer program is executed by the processor, so that the processor executes the steps of the power transformation project cost prediction and balance control method in the technical scheme.
The invention has the beneficial effects that: the invention provides a power transformation project cost prediction and balance control method based on correction correlation analysis and deep learning, which optimizes and selects power transformation project cost influence factors, improves the accuracy of power transformation project cost level prediction, realizes effective prediction of future fund balance level of a project, assists in controlling the balance level of the power transformation project, improves project fund utilization efficiency, provides support for the establishment of a power transformation project investment plan, and ensures the stability of power transmission and distribution prices. In order to further improve the investment efficiency of the power transformation project, the capital construction investment of the power grid is scientifically and efficiently arranged, and the optimization and adjustment of the structure of the power grid are supported. The invention constructs a joint prediction method of the construction cost and the balance rate of the power transformation project, thereby realizing accurate measurement and calculation of investment and effective control of the balance level of the power transformation project. By correcting the index correlation coefficient, influence factors influencing the power transformation project are further mined, a power transformation project investment prediction method under a long-time memory neural network (LSTM) frame is established, accurate prediction of the power transformation project investment level and effective control of the investment balance level are achieved, and optimal configuration of power grid resources is supported.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a schematic diagram of the LSTM model of the present invention;
FIG. 3 is a diagram of the cost prediction of a power transformation project according to the present invention;
fig. 4 is a diagram for predicting the investment balance of the power transformation project.
Detailed Description
The invention will be further described in detail with reference to the following drawings and specific examples, which are not intended to limit the invention, but are for clear understanding.
As shown in fig. 1, the invention provides a method for predicting the construction cost of a power transformation project and controlling balance, which comprises the following steps:
determining influence factors of the construction cost of the power transformation project based on project information of historical power transformation projects of different types and grades;
respectively measuring and calculating the associated coefficients of the cost influence factors of each historical power transformation project and the cost of the power transformation project, and correcting the associated coefficients based on the entropy weight; constructing a cost influence factor index library based on the corrected correlation coefficient; the cost influence factor index library comprises each historical power transformation project and corresponding cost influence factor indexes thereof;
using historical power transformation project cost influence factor indexes as input nodes of the LSTM, using predicted cost as output nodes, optimizing LSTM parameters through historical project samples, and establishing an LSTM power transformation project cost prediction model;
taking the influence factor indexes of historical power transformation project cost and the prediction cost output by the LSTM power transformation project cost prediction model as input nodes of the LSTM, taking the investment balance rate as output nodes, optimizing LSTM parameters through historical project samples, and establishing an LSTM power transformation project investment balance rate prediction model;
acquiring cost influence factor indexes of the power transformation project to be tested and inputting the cost influence factor indexes into an LSTM power transformation project cost prediction model to obtain the predicted cost of the power transformation project to be tested;
inputting the cost influence factor index of the power transformation project to be tested and the predicted cost of the power transformation project to be tested into an LSTM power transformation project investment balance rate prediction model to obtain the investment balance rate of the power transformation project to be tested;
and calculating the fund balance possibly generated after the project of the power transformation project to be tested is put into operation according to the predicted construction cost and the investment balance rate of the power transformation project to be tested.
The specific embodiment 1 of the present invention provides a method for predicting the cost of a power transformation project and controlling the balance based on correction correlation analysis and deep learning, which comprises the following steps:
A. the influence factor collecting module: establishing the indexes of the influence factors of the construction cost of the power transformation project; and determining the influence factors of the construction cost of the power transformation project by analyzing the historical project information of the power transformation projects of different types and grades.
B. The influence factor analysis module: the method comprises the steps of respectively measuring and calculating the correlation coefficients of the influence factors of a plurality of power transformation projects and the construction costs of the power transformation projects, correcting the measured original correlation coefficients by taking the entropy values of index sample data as weights, and constructing a construction cost influence factor index database based on the corrected correlation coefficients. Respectively deleting the influence factors with smaller correlation coefficient aiming at each power transformation project, and taking the remaining influence factors as influence factor indexes; the construction cost influence factor index library comprises each power transformation project and corresponding influence factor indexes thereof; namely, the influence factor indexes of each power transformation project are different.
C. An LSTM model construction and optimization module: establishing an optimized LSTM power transformation project cost prediction model and an LSTM power transformation project investment balance rate prediction model; the method comprises the steps of taking the influence factor indexes of the construction cost of the transformation project as input nodes of the LSTM, taking the predicted construction cost as output nodes, optimizing LSTM parameters through historical project samples, and establishing an LSTM transformation project construction cost prediction model.
And (3) taking the historical transformation project cost influence factor index and the prediction cost output by the LSTM transformation project cost prediction model as input nodes of the LSTM, taking the investment balance rate as output nodes, optimizing LSTM parameters through historical project samples, and establishing the LSTM transformation project investment balance rate prediction model.
Specifically, numerical values of cost influence factor indexes of each historical power transformation project are respectively input into an LSTM power transformation project cost prediction model, and the predicted cost of each historical power transformation project is obtained; calculating to obtain the investment balance rate of each historical power transformation project based on the project information of each historical power transformation project; inputting the numerical value of the cost influence factor index of each historical power transformation project and the predicted cost into the LSTM model through an input gate, performing iterative optimization on the parameters of the LSTM model based on the corresponding investment balance rate, stopping iteration when the expected precision requirement is met, and generating the LSTM power transformation project investment balance rate model.
D. A predicted cost output module: outputting the predicted construction cost of the power transformation project based on the LSTM model; acquiring the cost influence factor of the to-be-measured power transformation project through the project information of the to-be-measured power transformation project, measuring and calculating the correlation coefficient between the cost influence factor of the to-be-measured power transformation project and the cost of the power transformation project, and correcting the correlation coefficient based on the entropy weight; and acquiring the cost influence factor index of the to-be-detected power transformation project based on the corrected correlation coefficient, inputting the cost influence factor index into the LSTM power transformation project cost prediction model subjected to iterative optimization, and outputting the predicted power transformation project cost from an output gate through an input gate, a forgetting gate and a cell state.
E. The cost prediction evaluation module: evaluating the effectiveness of the prediction model;
F. the investment balance level prediction output module: inputting the cost influence factor index of the power transformation project to be tested and the predicted cost of the power transformation project to be tested into an LSTM power transformation project investment balance rate prediction model to obtain the investment balance rate of the power transformation project to be tested; and calculating the possible fund balance generated after the project of the power transformation project to be tested is put into production according to the predicted construction cost and the investment balance rate of the power transformation project to be tested, and realizing effective prediction of the future balance level of the project.
The step A specifically comprises the following steps:
in order to analyze and predict the cost level of the historical power transformation project, main factors influencing the cost of the power transformation project need to be collected, wherein the main factors mainly comprise project overview, power transformation scale, plant site related factors and cable project quantity. The engineering overview is divided into 4 factors of voltage level, belonged area, construction property and landform; the transformation scale is divided into 5 factors of single main transformation capacity, the number of main transformers, the type of the transformer station, the number of outgoing lines and the size of the reactor; the relevant factors of the plant site comprise 3 factors of land acquisition area, foundation treatment mode and site leveling cost; the cable work includes power cables and control cables. The influence factors of the construction cost of the power transformation project acquired by the influence factor collecting module are shown in table 1.
TABLE 1 influence factor of cost of power transformation project
Figure BDA0003938852510000071
Figure BDA0003938852510000081
The step B specifically comprises the following steps:
in order to accurately measure and calculate the influence degree of the influence factors on the construction cost, the factors obtained in the step A are uniformly quantized, the correlation between the influence factors and the construction cost of the power transformation project is measured and calculated by adopting the correction correlation coefficient based on the entropy weight, and a construction cost influence factor index database is constructed.
Wherein, 5 factors of the region, the construction property, the landform, the transformer type and the foundation treatment mode need to be further quantized and then calculated.
According to n indexes (indexes refer to influence factors) of m power transformation project items after standardization, constructing a correlation matrix R:
R＝(r ij ) m×n
if the index is a forward index, the correlation coefficient r ij Comprises the following steps:
Figure BDA0003938852510000082
if the index is a negative index, the correlation coefficient r ij Comprises the following steps:
Figure BDA0003938852510000083
calculating the entropy value K of each index according to the constructed matrix j ：
Figure BDA0003938852510000084
Figure BDA0003938852510000085
Wherein f is ij The weight of the jth influencing factor for the ith item.
Calculating the information entropy of the j index:
Figure BDA0003938852510000091
correcting the information entropy to obtain the weight of the jth influence factor:
Figure BDA0003938852510000092
calculating the initial correlation coefficient of each index based on the pearson correlation coefficient, and correcting through the entropy weight to obtain the corrected correlation coefficient, as shown in table 2:
TABLE 2 correction of correlation coefficient for cost influence factor of power transformation project
Figure BDA0003938852510000093
And according to the corrected correlation coefficient, the reactor scale and the correlation degree between the affiliated area and the construction cost of the power transformation project are low, and the reactor scale and the affiliated area are eliminated. The voltage class, the construction property, the landform, the capacity of a single main transformer, the number of main transformers, the type of a transformer station, the number of outgoing lines, the land acquisition area, the foundation treatment mode, the site leveling cost and 12 factors of a power cable and a control cable form a manufacturing cost influence factor index library. The project information of a plurality of historical power transformation projects is calculated by adopting the scheme, and the cost influence factor indexes of each historical power transformation project are respectively obtained.
The step C specifically comprises the following steps:
based on a deep learning theory, the influence factor indexes of the construction cost of the transformation project are used as input nodes of the LSTM, the prediction construction cost is used as an output node, the LSTM parameters are optimized through historical project samples, and a stable LSTM transformation project construction cost prediction model is established.
Long-short term memory networks (LSTM) are a special Recurrent Neural Network (RNN) that enables selective memory of historical information. Conventional RNNs fail to capture long-term dependencies of information and may experience gradient explosions or disappearance.
LSTM overcomes the shortcomings of traditional RNNs by forgetting the design of gates, input gates, output gates, and cell states. LSTM has a self-assessment mechanism to retain useful information by judging the importance of the information. It is not only easier to train than traditional RNNs, but also can learn long-term correlations of information. The LSTM includes four interaction layers, which are a forgetting gate, an input gate, a cellular state, and an output gate.
Forgetting to record the door:
output h from previous time t-1 And the input x of the current time t Generating f t To filter the information. The symbol ω denotes a weight, b denotes an offset, and σ denotes an activation function sigmoid.
f t ＝σ(ω f [h t-1 ,x t ]+b f )
An input gate:
the value to be updated is determined by a sigmoid function, candidate value C' t Generated by the tanh function.
i t ＝σ(ω i [h t-1 ,x t ]+b i )
C' t ＝tanh(ω c [h t-1 ,x t ]+b c )
The cell state:
the cell state is determined by the calculation results of the forgetting gate and the input gate. The cell state may control the transmission of information to the next time, discarding unnecessary information and updating information.
C t ＝f t C t-1 +i t C' t
An output gate:
initial output o t Obtained from sigmoid layer, then C t The values are scaled by the tanh function. The final output can be obtained by multiplying these two values.
o t ＝σ(ω 0 [h t-1 ,x t ]+b 0 )
h t ＝o t tanh(C t )
There are three main phases inside the LSTM model:
forget the stage. In the stage, input information transmitted from the previous node is mainly selectively forgotten, unimportant information is forgotten, and important information is remembered; forget outdated information and remember updated information. In particular by calculating f t To control which information received by the last node needs to be forgotten.
A selective memory phase. It is mainly at this stage that the information received by the last node is selectively memorized. Mainly to x t The input information is selectively memorized, important information is kept as much as possible, unimportant information is selectively forgotten, and i is obtained by calculation through an input door t And finally, performing standardized transformation on the obtained information through the tanh function, and further recording the information in the model.
And (5) an output stage. At this stage the model will perform the final processing of the information and the results of the model are output through the output gates. Converting the calculated result through the sigmoid activation function of the activation function to obtain o t . Then o is put t The output result of (a) is scaled by the tanh function, and finallyTo obtain an output result h t 。
And establishing a characteristic LSTM in the sample data through the processing of the LSTM power transformation project prediction model, and optimizing the model in continuous iterative updating. And (3) along with the lapse of time, forgetting to update the original partial invalid information, and establishing the latest sample data characteristics in the model, thereby realizing that the model is continuously updated and iterated along with the time, reconstructing the latest change rule and incidence relation of the construction cost of the power transformation project in the model, and improving the stability, effectiveness and adaptability of the LSTM model.
The LSTM controls the transmission state by gating the state, and remembering important information requires a long time and forgetting unimportant information also requires a certain time. Compared with the traditional recurrent neural network which only can apply a memory stack mode, the LSTM model overcomes the defect of a single memory mode, and input information is finally fed back to important parameters of the model through selective memory. And a certain time is required for the loss and update of information.
Because the relation between the characteristics of the power transformation project and the cost level is in a relatively stable state in a short period, the influence of different factors on the cost level is likely to change in a long period, the influence of part of factors on the cost is weakened continuously, and the influence of some factors on the cost is obviously enhanced in a specific time due to certain reasons. The characteristics of LSTM on the long-term memory of information are adapted to the requirements of the cost prediction of the power transformation project. The LSTM updates the parameter base of the model through the continuous model of the sample data of the power transformation project, so that the LSTM can continuously adapt to the continuous change of the construction cost of the power transformation project and adaptively adjust the model.
In the LSTM model, sigmoid and tanh two activation functions exist, wherein the sigmoid is used on various gates (including an input gate, a forgetting gate and an output gate) to generate a value between 0 and 1, and is generally applied; tanh is used for status and output, and data is processed. The advantages of the two activation functions are complementary, and the applicability of the LSTM model is improved.
The LSTM model well solves the problems of fleet disappearance and gradient explosion, and greatly influences the performance of the model by taking huge historical data as support for power transformation project cost prediction and possibly causing the gradient explosion or gradient disappearance of the model in the presence of a large amount of data by a traditional neural network structure. And the LSTM model analyzes the characteristics of the cost data of the power transformation project through long-term memory selection, long-term memory is carried out on important information, and model parameters of the LSTM model are continuously updated along with the updating of the samples, so that the LSTM model can always adapt to the latest sample data of the power transformation project, and a model structure suitable for the current cost prediction of the power transformation project is built. The establishment of the cell state effectively controls the rate of information transmitted to the next node, and the cell state can be used for reselecting the characteristic information of the sample, so that the truly useful information can be transmitted to the output gate unit of the LSTM. In long-term recursive learning, a plurality of invalid information is discarded through forgetting to remember a gate, a long-term dependence relationship between the construction cost influence factor of the power transformation project and the construction cost level is gradually established, and the long-term dependence relationship recorded by the LSTM is iteratively updated along with the time lapse and the change of sample data, so that the robustness of the LSTM model is improved, and the predictability of the construction cost level of the power transformation project is enhanced.
The step D specifically comprises the following steps:
and C, inputting 12 factors of the voltage grade, the construction property, the landform, the capacity of a single main transformer, the number of the main transformers, the type of the transformer station, the number of outgoing lines, the land acquisition area, the foundation treatment mode, the field flattening cost, the power cable and the control cable of the power transformation project into the LSTM model through an input door based on the power transformation project exemplified in the step C, performing iterative optimization on parameters of the LSTM model based on the sample project cost data, and stopping iteration when the expected precision requirement is met.
The LSTM model gradually strengthens the influence of main factors on the model through continuous learning of samples, and establishes long-term dependence between the influencing factors and the power transformation engineering indexes.
And re-inputting the cost influence factors of the power transformation project to be tested into the optimized LSTM model, and outputting the predicted power transformation project cost from the output gate through the optimized structure and parameters, the input gate, the forgetting gate and the cell state by the model.
By adding the latest sample data of the power transformation project and updating the information base of the LSTM model, the LSTM model can continuously update the parameters of the LSTM model along with the time, long-term relation between the construction cost influence factors and the construction cost is established, and the LSTM model is ensured to have long-term effectiveness.
The step E specifically comprises the following steps:
in order to further evaluate the prediction result of the LSTM model, the actual cost of the power transformation project sample is compared with the prediction cost, and the absolute error and the relative error of the LSTM prediction result are measured and calculated.
And comparing error levels of different types of projects of different voltage grades and the like, verifying that the prediction precision of the LSTM model meets the expected precision requirement, and effectively predicting the cost levels of different types of power transformation projects.
The step F specifically comprises the following steps:
and establishing long-term dependence between the construction cost level and the balance level of the power transformation project based on the prediction result of the LSTM on the construction cost of the power transformation project, and taking the investment balance rate as the index of the investment balance level of the power transformation project.
Introducing 12 index data of a power transformation project cost influence factor index library, combining with a power transformation project cost predicted value, using the 13 data as input data of an LSTM model, and training the prediction capability and performance of the LSTM model on the investment balance index of the power transformation project through sample data.
And continuously improving the prediction accuracy of the LSTM model on the investment balance level of the power transformation project through continuous iterative optimization. Meanwhile, by continuously updating sample data, long-term relation between LSTM model influence factors and the investment balance rate is established, and the change rule of the investment balance level of the power transformation project is reflected.
The construction cost and the investment balance rate predicted by the LSTM model are combined, the possible fund balance generated after the power transformation project is put into operation can be calculated, and reference is provided for the investment planning and the control of the project cost of the power transformation project.
In the embodiment 1, from the perspective of controlling the cost and balance level of the power transformation project, a method for predicting the cost and controlling the investment balance of the LSTM power transformation project, which is constructed by performing factor correlation measurement based on correction correlation analysis and through a deep learning neural network, is established. Based on the specific embodiment, the following can be realized:
(1) The correlation between the influence factors and the construction cost of the power transformation project is effectively measured and calculated through the corrected correlation coefficients, an information entropy theory is applied, the influence of the magnitude of the sample information entropy on the correlation coefficients is considered in the measurement and calculation of the correlation coefficients, the information entropy is adopted to correct the Pearson correlation coefficients, and therefore the correlation between the influence factors and the construction cost is compared.
(2) The LSTM model is provided with a forgetting gate unit, information obtained by an input gate is selectively received, and part of unimportant information is selectively forgotten, so that the adaptability, stability and prediction accuracy of the LSTM model are improved.
(3) Further expanding the application of the LSTM in the investment balance of the power transformation project. The influence factors and the predicted construction cost are used as input data of the LSTM, long-term relations among the influence factors, the construction cost of the power transformation project and the investment balance are established, and the long-term dependence relation of the LSTM is updated by updating the learning samples and the like, so that the LSTM model is ensured to have good timeliness.
(4) And evaluating the performance of the LSTM model by measuring and calculating errors of the predicted construction cost and the actual construction cost of the power transformation project, including absolute errors and relative errors. Therefore, two different dimensions of absolute error and relative error are used for measuring and calculating whether the LSTM model can meet the actual requirement of cost prediction of the power transformation project. The investment balance value possibly generated after the power transformation project is put into operation is further predicted through the cost level and the investment balance rate index obtained through the LSTM model prediction, and reference is provided for cost control of the power transformation project.
The invention realizes the cost prediction and balance control of the power transformation project through computer equipment. The computer equipment comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the power transformation project cost prediction and balance control method in the technical scheme, so that the utilization rate of industrial data is effectively improved, the accurate prediction and balance control of the power transformation project cost are realized, and technical support is provided for the project cost work.
The specific embodiment 2 of the present invention selects 15 actual power transformation project projects, predicts the manufacturing costs of the selected 15 projects by using the LSTM model, and the comparison result between the predicted manufacturing costs and the actual manufacturing costs is shown in table 3.
TABLE 3 predicted value of cost of power transformation project
Figure BDA0003938852510000151
The absolute error and relative error index of the predicted result and the actual manufacturing cost are further compared, and the result is shown in table 4.
The relative prediction error of the project 2 is the largest and reaches 3.05%, the relative prediction errors of other 14 projects are all smaller than 3%, and the surface LSTM prediction model can effectively predict the manufacturing cost level of the power transformation project.
From the investment scale, the investment scale of the project 1, the project 2 and the project 4 is more than 1 hundred million, and generally the investment scale is over 220 kV. From the prediction effect, the prediction error of the item 2 is the largest, the absolute error is 440.56 ten thousand yuan, and the relative error is 3.05%. Except for the three projects with larger scale, the cost prediction errors of other power transformation project projects are within 200 ten thousand yuan, which shows that the prediction precision of the LSTM model can effectively support the cost control of the power transformation project.
TABLE 4 prediction error evaluation of power transformation project cost
Figure BDA0003938852510000161
In addition, in order to further improve the deep learning capability of the LSTM model and deepen the value of cost prediction, cost influence factor index database data and the cost obtained through prediction are used as the input of the LSTM model, the investment balance level of the power transformation project is obtained through deep learning, and the investment balance rate of the power transformation project is obtained from the output node of the LSTM model, as shown in the graph 4 and the table 5.
TABLE 5 prediction error evaluation of construction cost of power transformation project
Figure BDA0003938852510000162
Figure BDA0003938852510000171
By predicting the investment balance level, the possible fund balance level generated after the project is built can be integrally analyzed, so that measures can be timely taken for the project which is expected to possibly generate larger balance, the investment balance is reduced, and the fund use efficiency of the power transformation project is improved.
According to the prediction result of the LSTM model, the investment balance level of the projects 3, 6, 7, 9 and the like is high, and the predicted investment balance rates are 4.32%, 3.98%, 4.29% and 3.97% respectively. Aiming at the items, the LSTM deep learning model can effectively predict the investment balance risk which is possibly generated in the future, so that measures are taken in time to control the investment balance.
For project 5, the lstm prediction results indicate that the project may be at risk of investment over-estimation, with an investment over-budget level of around 4%. Therefore, important attention should be paid to the projects, the investment approximate calculation compilation work of the projects is refined, meanwhile, the project settlement is strictly controlled in the construction process, the risk caused by the high project investment which is possibly generated is avoided, the over-approximate calculation risk which is possibly generated after the project investment can be timely identified by the prediction result of the LSTM model, and therefore the project investment risk is reduced and eliminated from the source.
Specific embodiment 2 constructs a depth model prediction model based on LSTM for the construction cost and investment balance of the power transformation project. Based on key indexes and investment influence factors of the power transformation project, the manufacturing cost level and the balance rate of the power transformation project are effectively predicted through LSTM model optimization and model prediction, and the investigation design and construction of the power transformation project are supported.
The main achievements and model optimization work of the specific embodiment 2 are as follows:
(1) According to the characteristics of the power transformation project and the engineering requirements, 14 influence factors including four types, namely project overview, power transformation scale, plant site correlation and cable engineering quantity are taken as main lines. Calculating initial correlation coefficients of 14 factors and the construction cost of the power transformation project based on the Pearson correlation coefficients, acquiring information entropies of all the influence factors through entropy weights, standardizing the entropy values to obtain correction weights of 14 factors, and combining the initial correlation coefficients to obtain the correlation coefficients after the correction of 14 influence factors.
(2) Based on the corrected correlation coefficient, 12 factors of voltage grade, construction property, landform, capacity of a single main transformer, the number of the main transformers, the type of a transformer station, the number of outgoing lines, land acquisition area, foundation treatment mode, site leveling cost and power cable and control cable are determined to form a manufacturing cost influence factor index library.
(3) Based on deep learning and artificial intelligence theory, an LSTM model framework is adopted to establish a deep learning model between the investment influence factors of the power transformation project and the project cost and the balance rate. The influence factors are input from the input nodes of the LSTM model, the LSTM model outputs the predicted cost of the project, and the project cost level is predicted. Secondly, inputting the investment influence factors and the forecast cost of the power transformation project into an LSTM model, realizing the forecast of the project investment balance level and realizing the forecast of the project cost and the investment balance rate of the power transformation project in the same intelligent forecast model.
(4) And based on the actual engineering project, measuring and calculating the error between the predicted cost and the actual cost level of the LSTM model, thereby evaluating the prediction effect and the prediction stability of the LSTM model. And simultaneously, the prediction result is scientifically evaluated from the two angles of absolute error and relative error, and the LSTM model is verified to meet the prediction requirement of the actual construction cost level of the power transformation project.
(5) The method has the advantages that the balance level of the power transformation project can be predicted based on the manufacturing cost prediction result of the LSTM model, balance risks possibly brought by the project can be obtained in time, the investment balance level can be obtained in advance through the LSTM model prediction result for the project possibly bringing larger investment balance, project approximate calculation is further strictly controlled, project construction management is enhanced, process control of the investment balance of the power transformation project is supported, and the investment prediction value of the LSTM model is exerted. Meanwhile, for the risk that the investment of part of projects is beyond the approximate calculation in the process of construction, the LSTM model can timely early warn the possible future high investment risk of the projects through deep learning and optimized prediction, so that the construction and investment control of the projects are supported.
Those not described in detail in this specification are within the skill of the art.

Claims (9)

1. A method for predicting the construction cost of a power transformation project and controlling balance is characterized in that: the method comprises the following steps:
determining influence factors of the construction cost of the power transformation project based on project information of historical power transformation projects of different types and grades;
respectively measuring and calculating the associated coefficients of the cost influence factors of each historical power transformation project and the cost of the power transformation project, and correcting the associated coefficients based on the entropy weight; constructing a cost influence factor index library based on the corrected correlation coefficient; the construction cost influence factor index library comprises various historical power transformation projects and corresponding construction cost influence factor indexes thereof;
using historical power transformation project cost influence factor indexes as input nodes of the LSTM, using predicted cost as output nodes, optimizing LSTM parameters through historical project samples, and establishing an LSTM power transformation project cost prediction model;
using historical power transformation project cost influence factor indexes and prediction cost output by an LSTM power transformation project cost prediction model as input nodes of an LSTM, using investment balance rate as output nodes, optimizing LSTM parameters through historical project samples, and establishing an LSTM power transformation project investment balance rate prediction model;
acquiring cost influence factor indexes of the power transformation project to be tested and inputting the cost influence factor indexes into an LSTM power transformation project cost prediction model to obtain the predicted cost of the power transformation project to be tested;
inputting the cost influence factor index of the power transformation project to be tested and the predicted cost of the power transformation project to be tested into an LSTM power transformation project investment balance rate prediction model to obtain the investment balance rate of the power transformation project to be tested;
and calculating the fund balance possibly generated after the project of the power transformation project to be tested is put into operation according to the predicted construction cost and the investment balance rate of the power transformation project to be tested.
2. The method for predicting construction cost and controlling balance of power transformation project according to claim 1, characterized in that: the influence factors for determining the manufacturing cost based on the project information of the power transformation project comprise: general engineering, transformation scale, relevant factors of plant sites and cable engineering quantity. The project overview is divided into 4 factors of voltage level, region, construction property and landform; the transformation scale is divided into 5 factors of single main transformer capacity, the number of main transformers, the type of a transformer station, the number of outgoing lines and the scale of a reactor; the relevant factors of the plant site comprise 3 factors of land acquisition area, foundation treatment mode and site leveling cost; cable engineering quantities include power cables and control cables.
3. The method for predicting construction cost and controlling balance of power transformation project according to claim 1, characterized in that: the method comprises the following steps of measuring and calculating a correlation coefficient of the influence factors and the construction cost of the power transformation project, correcting the correlation coefficient based on the entropy weight, and constructing a construction cost influence factor index library based on the corrected correlation coefficient, wherein the process comprises the following steps:
the determined influence factors of the construction cost are uniformly quantized, the correlation between the influence factors and the construction cost of the power transformation project is measured and calculated by adopting the correction correlation coefficient based on the entropy weight,
wherein, 5 factors of the region, the construction property, the landform, the transformer type and the foundation treatment mode are further quantified and then calculated;
calculating an initial correlation coefficient of each index based on the Pearson correlation coefficient, and correcting through the entropy weight to obtain a corrected correlation coefficient; eliminating influence factors with low correlation degree with the construction cost of the power transformation project, and taking the remaining influence factors as the indexes of the construction cost influence factors.
4. The method for predicting the construction cost of the power transformation project and controlling the balance as claimed in claim 1, wherein: the method comprises the following steps of taking historical power transformation project cost influence factor indexes as input nodes of the LSTM, taking predicted cost as output nodes, optimizing LSTM parameters through historical project samples, and establishing an LSTM power transformation project cost prediction model, wherein the process comprises the following steps:
acquiring numerical values of cost influence factor indexes of a plurality of historical power transformation projects and corresponding actual cost data based on historical project samples; inputting the numerical values of a plurality of cost influence factor indexes into the LSTM model through an input gate, performing iterative optimization on the parameters of the LSTM model based on corresponding cost data, stopping iteration when the expected precision requirement is met, and generating an LSTM power transformation project cost prediction model.
5. The method for predicting construction cost and controlling balance of power transformation project according to claim 1, characterized in that: the method comprises the following steps of taking historical power transformation project cost influence factor indexes and prediction cost output by an LSTM power transformation project cost prediction model as input nodes of an LSTM, taking investment balance rate as output nodes, optimizing LSTM parameters through historical project samples, and establishing the LSTM power transformation project investment balance rate prediction model, wherein the process comprises the following steps:
respectively inputting numerical values of the cost influence factor indexes of each historical power transformation project into an LSTM power transformation project cost prediction model to obtain the predicted cost of each historical power transformation project; calculating to obtain the investment balance rate of each historical power transformation project based on the project information of each historical power transformation project; inputting the numerical value of the cost influence factor index of each historical power transformation project and the predicted cost into the LSTM model through an input gate, performing iterative optimization on the parameters of the LSTM model based on the corresponding investment balance rate, stopping iteration when the expected precision requirement is met, and generating the LSTM power transformation project investment balance rate model.
6. The method for predicting the construction cost of the power transformation project and controlling the balance as claimed in claim 4, wherein the method comprises the following steps: the method comprises the following steps of obtaining cost influence factor indexes of the power transformation project to be tested and inputting the cost influence factor indexes into an LSTM power transformation project cost prediction model, wherein the process of obtaining the predicted cost of the power transformation project to be tested comprises the following steps:
acquiring the cost influence factor of the to-be-measured power transformation project through the project information of the to-be-measured power transformation project, measuring and calculating the correlation coefficient between the cost influence factor of the to-be-measured power transformation project and the cost of the power transformation project, and correcting the correlation coefficient based on the entropy weight; and acquiring the cost influence factor index of the to-be-detected power transformation project based on the corrected correlation coefficient, inputting the cost influence factor index into the LSTM power transformation project cost prediction model subjected to iterative optimization, and outputting the predicted power transformation project cost from an output gate through an input gate, a forgetting gate and a cell state.
7. The method for predicting the construction cost of the power transformation project and controlling the balance as claimed in claim 1, wherein: and continuously updating historical engineering samples, and optimizing an iterative LSTM power transformation engineering cost prediction model and an LSTM power transformation engineering investment balance rate prediction model.
8. The utility model provides a copper product price prediction system based on price index adjustment which characterized in that: the system is used for realizing the method for predicting the construction cost of the power transformation project and controlling the balance of the power transformation project as claimed in any one of claims 1 to 7.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program which, when executed by the processor, causes the processor to perform the method steps of the substation construction cost prediction and balance control method of any one of claims 1 to 7.
CN202211413315.XA 2022-11-11 2022-11-11 Method and system for predicting construction cost and controlling balance of power transformation project Pending CN115577856A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211413315.XA CN115577856A (en) 2022-11-11 2022-11-11 Method and system for predicting construction cost and controlling balance of power transformation project

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211413315.XA CN115577856A (en) 2022-11-11 2022-11-11 Method and system for predicting construction cost and controlling balance of power transformation project

Publications (1)

Publication Number Publication Date
CN115577856A true CN115577856A (en) 2023-01-06

Family

ID=84588954

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211413315.XA Pending CN115577856A (en) 2022-11-11 2022-11-11 Method and system for predicting construction cost and controlling balance of power transformation project

Country Status (1)

Country Link
CN (1) CN115577856A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116542401A (en) * 2023-07-05 2023-08-04 江南大学附属医院 Medical insurance hyperbranched prediction method and system for hospitalization diagnosis and treatment service unit

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116542401A (en) * 2023-07-05 2023-08-04 江南大学附属医院 Medical insurance hyperbranched prediction method and system for hospitalization diagnosis and treatment service unit
CN116542401B (en) * 2023-07-05 2023-09-19 江南大学附属医院 Medical insurance hyperbranched prediction method and system for hospitalization diagnosis and treatment service unit

Similar Documents

Publication Publication Date Title
CN112949945B (en) Wind power ultra-short-term prediction method for improving bidirectional long-term and short-term memory network
CN102269972A (en) Method and device for compensating pipeline pressure missing data based on genetic neural network
KR102181966B1 (en) Soft survey method and system for hydraulic cylinder comprehensive test station
CN116150897A (en) Machine tool spindle performance evaluation method and system based on digital twin
CN103885867B (en) Online evaluation method of performance of analog circuit
CN110807490A (en) Intelligent prediction method for construction cost of power transmission line based on single-base tower
CN110097205A (en) A kind of building load prediction weather forecast data preprocessing method
CN112990587A (en) Method, system, equipment and medium for accurately predicting power consumption of transformer area
CN115577856A (en) Method and system for predicting construction cost and controlling balance of power transformation project
CN115270239A (en) Bridge reliability prediction method based on dynamic characteristics and intelligent algorithm response surface method
CN115545333A (en) Method for predicting load curve of multi-load daily-type power distribution network
CN114330815A (en) Ultra-short-term wind power prediction method and system based on improved GOA (generic object oriented architecture) optimized LSTM (least Square TM)
CN116960962A (en) Mid-long term area load prediction method for cross-area data fusion
CN117114176A (en) Land utilization change prediction method and system based on data analysis and machine learning
CN116611785A (en) Power transmission and transformation project cost model construction method, system, equipment and medium based on big data
CN112581311B (en) Method and system for predicting long-term output fluctuation characteristics of aggregated multiple wind power plants
CN114861555A (en) Regional comprehensive energy system short-term load prediction method based on Copula theory
CN110414146B (en) Water environment treatment project design parameter optimization method based on deep learning
CN113673145A (en) Surface subsidence prediction method based on PCA-LSTM neural network
CN112765141A (en) Continuous large-scale water quality missing data filling method based on transfer learning
CN113361187A (en) Power grid line operation safety optimization method, system and storage medium
CN111401624A (en) Wind power prediction method and device and computer readable storage medium
CN117077546B (en) Power system load modeling method and system based on data driving
Wang et al. Evolutionary Optimization of Robust Multi-objective Expensive Problems Assisted by Incremental Kriging Models
CN112560898B (en) Load space-time prediction method based on deep learning

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