CN115860263A - Power grid infrastructure planning project construction period prediction method and prediction system - Google Patents
Power grid infrastructure planning project construction period prediction method and prediction system Download PDFInfo
- Publication number
- CN115860263A CN115860263A CN202310044387.XA CN202310044387A CN115860263A CN 115860263 A CN115860263 A CN 115860263A CN 202310044387 A CN202310044387 A CN 202310044387A CN 115860263 A CN115860263 A CN 115860263A
- Authority
- CN
- China
- Prior art keywords
- construction period
- power grid
- project
- grid infrastructure
- distribution
- 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
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a method for predicting the construction period of a power grid infrastructure planning project, which comprises the steps of obtaining historical data information of the power grid infrastructure project; analyzing and modeling factors influencing the power grid infrastructure project; performing key risk feature identification; constructing a preliminary construction period prediction model and training to obtain a construction period prediction model; and inputting the data into a construction period prediction model to obtain a final construction period prediction result of the power grid infrastructure planning project. The invention also discloses a prediction system for realizing the method for predicting the construction period of the power grid infrastructure planning project. According to the method and the system for predicting the construction period of the power grid infrastructure planning project, the method for predicting the construction period of the power grid infrastructure planning project is innovated, so that the prediction of the construction period of the power grid infrastructure planning project can be realized, and the method and the system are high in reliability, good in accuracy and objective and scientific.
Description
Technical Field
The invention belongs to the field of electrical automation, and particularly relates to a method and a system for predicting the construction period of a power grid infrastructure planning project.
Background
With the development of economic technology and the improvement of living standard of people, electric energy becomes essential secondary energy in production and life of people, and brings endless convenience to production and life of people. Therefore, ensuring a stable and reliable supply of electric energy is one of the most important tasks of an electric power system.
The power grid infrastructure project comprises a power transmission and transformation project, a power transformation project, a line project and the like; in the implementation process of power transmission and transformation, power transformation and line engineering projects, the influence of various reasons such as diversified development of power loads, climate and environmental conditions can be caused, so that the project construction period is influenced. Therefore, it is very important to predict project construction period of power grid infrastructure project.
At present, a prediction scheme aiming at the construction period of a power grid infrastructure project is in a relatively primary manual prediction stage; however, the manual prediction mode has poor reliability and accuracy, and the prediction process is subjective, so that the deviation between the prediction construction period and the actual construction period is large.
Disclosure of Invention
One of the purposes of the invention is to provide a method for objectively and scientifically predicting the construction period of a power grid infrastructure planning project with high reliability and good accuracy.
The invention also aims to provide a prediction system for realizing the method for predicting the construction period of the power grid infrastructure planning project.
The method for predicting the construction period of the power grid infrastructure planning project comprises the following steps:
s1, acquiring historical data information of a power grid infrastructure project;
s2, analyzing and modeling factors influencing the power grid infrastructure project;
s3, performing key risk characteristic identification on the data information acquired in the step S1;
s4, constructing a preliminary construction period prediction model based on the Bayesian neural network model, and training to obtain a construction period prediction model;
and S5, inputting the data obtained in the steps S1-S3 into the construction period prediction model obtained in the step S4 to obtain a final construction period prediction result of the power grid infrastructure planning project.
The historical data information of the power grid infrastructure project in the step S1 specifically comprises a construction progress curve of a historical power grid infrastructure planning project.
The step S2 of analyzing and modeling the factors influencing the power grid infrastructure project specifically comprises the following steps:
analyzing to obtain factors influencing the power grid infrastructure project, wherein the factors comprise climate change factors, power supply and demand change factors and policy change factors;
describing the climate change factors by adopting triangular distribution, wherein the relevant parameters of the triangular distribution are (0, 0.2, 1); describing the power supply and demand variation factors by adopting triangular distribution, wherein the relevant parameters of the triangular distribution are (-0.5, 0, 0.5); for the policy variation factor, a triangular distribution is used for description, and relevant parameters of the triangular distribution are (-0.5, 0, 0.5).
Step S3, performing key risk feature identification on the data information acquired in step S1, specifically including the following steps:
aiming at the construction progress curve of the power grid infrastructure planning project acquired in the step S1, selecting the characteristics extracted from the construction progress curve by using a characteristic filtering method by taking the construction period as a filtering label:
scoring each extracted feature according to divergence or correlation, and if the extracted feature is a Boolean type, adopting KS test; if the extracted features are continuous value features, checking by using Kendall correlation coefficients;
the KS test value D for any two samples was calculated using the following equation:
in the formulaIs extracted when the characteristic value is 0The accumulated distribution function of the characteristics of the arrived construction period;the accumulated distribution function of the construction period is extracted when the characteristic value is 1; n is 0 The number of the samples of the construction period characteristics extracted when the characteristic value is 0; n is 1 The number of samples of the construction period features extracted when the feature value is 1; y is the construction period;
setting the extracted characteristic of the construction progress curve as X = (X) 1 ,...,x n ) Construction period of Y = (Y) 1 ,...,y n ) (ii) a X and Y coexist inGroup element pair (x) i ,x j ) And (y) i ,y j ) I is not equal to j and is not less than 1, and j is not less than n; and calculating a Kendall correlation coefficient tau between the characteristic X and the construction period Y by adopting the following formula:
in the formula N 1 The number of the positively correlated element pairs in X and Y; n is a radical of 2 The number of the element pairs which are negatively related in X and Y; n is a radical of hydrogen 3 The number of identical element pairs in X and Y; n is a radical of 4 Is the number of identical pairs of elements in Y; n is a radical of 0 Is the total number of element pairs; the positively correlated element pair is defined as x i >x j Time y i >y j And x i <x j Time y i <y j The pair of elements of (1); the negative correlation element pair is defined as x i >x j Time y i <y j And x i <x j Time y i >y j The pair of elements of (1);
selecting the characteristic that the KS inspection value D and the Kendall correlation coefficient tau are larger than a set value;
and finally, performing key risk feature identification on the selected features by adopting a principal component analysis method to obtain final construction time sequence features.
S4, constructing a preliminary construction period prediction model based on the Bayesian neural network model, and training to obtain a construction period prediction model, wherein the method specifically comprises the following steps:
the preliminary construction period prediction model comprises an input layer, a hidden layer and an output layer; the hidden layer is a probability layer, so that the network has the capability of describing uncertainty; setting the weight and bias in the probability layer to obey normal distribution;
setting the network parameter of the Bayesian neural network as W, and giving observation data D = (X, Y); x is input data which comprises the affiliated section of the capital construction project, the voltage level, the project attribute, the construction scale, the time sequence progress curve characteristic, the climate change weather factor, the power supply and demand change factor and the policy change factor; y is label data, including construction period of the capital construction project; the following distribution was obtained:
in the formula, P (Y) * |X * D) probability distribution of the construction period under given input data after network parameter training; p (Y) * |X * Omega) is the probability distribution of the construction period under the condition of determining the weight of the probability layer and the input data; p (ω | D) is the posterior distribution of probability layer weights, andp (ω) is the prior distribution of weights; p (D | omega) is probability distribution of the construction period under the condition of determining the weight and input of a probability layer; p (D) is the probability distribution of the construction period under the determined input;
the posterior distribution of the network weight is approximated by a variational inference method: introducing a distribution q (ω | θ) to approximate a distribution P (ω | D), where θ = (μ, σ), each weight ω i Obey normal distribution (mu) i ,θ i ) (ii) a The difference between the two distributions is measured by KL divergence; when q (ω | θ) is known, the weight distribution is optimized by minimizing the KL divergence:
in the formula [ theta ] * Is the optimized value of the distribution (μ, σ);is distributed under q (ω | θ)The expected value of (d);
then, the lower evidence bound needs to be introduced in the variation reasoning:
L=logP(Y|X)-D KL (q(ω|θ)||P(ω|D))
in the formula D KL (q (ω | θ) | P (ω | D)) is the KL divergence between the two distributions of q (ω | θ) and P (ω | D);
and (3) adopting a method of carrying out weight parameter on the weight to carry out derivation on the expectation in the calculation formula of the L to obtain:
finally, by a number of different epsilon i To obtainTo approximate the derivative of the KL divergence to θ;
during training, optimizing and training the weight of the probability layer by a variational reasoning method; and after the training is finished, predicting the construction period of the power grid infrastructure project by adjusting the probability of the occurrence of the construction period risks in the input quantity.
The invention also provides a prediction system for realizing the method for predicting the construction period of the power grid infrastructure planning project, which comprises a data acquisition module, an analysis and modeling module, a characteristic identification module, a model construction and training module and a prediction module; the data acquisition module, the analysis and modeling module, the feature recognition module, the model construction and training module and the prediction module are sequentially connected in series; the data acquisition module is used for acquiring historical data information of a power grid infrastructure project and uploading the data to the analysis and modeling module; the analysis and modeling module is used for analyzing and modeling factors influencing the power grid infrastructure project according to the received data and uploading the data to the feature recognition module; the characteristic identification module is used for carrying out key risk characteristic identification on the acquired data information according to the received data and uploading the data to the model construction and training module; the model building and training module is used for building a preliminary construction period prediction model based on a Bayesian neural network model according to the received data, training the preliminary construction period prediction model to obtain a construction period prediction model, and uploading the data to the prediction module; and the prediction module is used for inputting the data into the construction period prediction model according to the received data and obtaining a final construction period prediction result of the power grid infrastructure planning project.
According to the method and the system for predicting the construction period of the power grid infrastructure planning project, the method for predicting the construction period of the power grid infrastructure planning project is innovated, so that the prediction of the construction period of the power grid infrastructure planning project can be realized, and the method and the system are high in reliability, good in accuracy and objective and scientific.
Drawings
FIG. 1 is a flow chart of a prediction method according to the present invention.
FIG. 2 is a functional block diagram of a prediction system according to the present invention.
Detailed Description
FIG. 1 is a schematic flow chart of the prediction method of the present invention: the invention provides a method for predicting the construction period of a power grid infrastructure planning project, which comprises the following steps:
s1, acquiring historical data information of a power grid infrastructure project; similar to pattern recognition, image processing and the like in machine learning, the main purpose of extracting time series features is dimension reduction; by acquiring the characteristics with high identifiability and non-redundancy from a large amount of original data, the performance of the algorithm can be improved, and some interpretable effective information which is convenient to understand can be acquired from the characteristics; the method comprises the steps that original data extracted by the method are power grid infrastructure planning project construction progress curves, and the sequence is a construction progress sequence which is continuously carried out in time; compared with the traditional feature extraction method, the method has the advantages that the algorithm is high in efficiency and wide in range, and a large number of time series features can be automatically calculated;
s2, analyzing and modeling factors influencing the power grid infrastructure project; the method specifically comprises the following steps:
analyzing to obtain factors influencing the power grid infrastructure project, wherein the factors comprise climate change factors, power supply and demand change factors and policy change factors;
aiming at climate change factors, the construction of power grid capital construction projects is mainly operated in open air, high altitude and underground, and factors such as climate environment have great influence on the construction period; the power supply and demand change:
aiming at the power supply and demand variation factors: new energy becomes a main supply source of the increment of the electric power demand in China, the existing wind power and solar power generation installation and the generated energy are increased particularly rapidly, and the uncertainty of the electric power demand is increased by the diversified development of a load structure under the background of a novel electric power system
For policy change factors: relevant laws, regulations and policies revisions such as environmental protection and land requisition affect the construction of power grid infrastructure projects;
factors influencing the construction period of a capital construction project generally have the characteristics of uncertainty and randomness, the application considers that the influencing factors can be regarded as random variables, and the probability of the factors is subject to the respective probability distribution at any time; quantification of the risk of the uncertainty factors, selecting for each factor an appropriate probability distribution that shows the possible outcomes of the variable of the factor and the probability of each possible outcome, which is in effect a summary of the historical data of the uncertainty factors;
distributions commonly used for probability description of random variables, such as: empirical distribution, trapezoidal distribution, triangular distribution, normal distribution, trapezoidal distribution, binomial distribution, poisson distribution, logarithmic distribution, negative exponential distribution and the like can all describe uncertainty factors; the method comprises the steps of carrying out statistical modeling by combining a historical power grid infrastructure project library, proposing a project construction period risk evaluation method under the influence of multiple factors based on PERT (plan review technology), quantifying uncertain risk factors, starting from analyzing the risk factors influencing the project construction period, giving different assumed distributions to the delay time of the power grid infrastructure project construction period caused by each risk factor according to the characteristics of the delay time, determining the influence of each risk factor on operation, and determining the distribution of the construction period of each path of the power grid infrastructure project through Monte Carlo simulation;
therefore, the climate change factor is described by adopting triangular distribution, and the relevant parameters of the triangular distribution are (0, 0.2, 1); describing the power supply and demand variation factors by adopting triangular distribution, wherein the relevant parameters of the triangular distribution are (-0.5, 0, 0.5); aiming at policy variation factors, triangular distribution is adopted for description, and relevant parameters of the triangular distribution are (-0.5, 0 and 0.5);
s3, performing key risk characteristic identification on the data information acquired in the step S1; the method specifically comprises the following steps:
the extracted time sequence characteristics of the construction progress curve usually comprise noise, redundant or irrelevant information; in order to avoid extracting irrelevant features, tsfresh has a built-in filtering process; in machine learning and statistics, feature filtering is a process of selecting valid features from all features for model training; when the model is trained, the larger the characteristic quantity is, the higher the accuracy rate is, and too many characteristics often bring about the defects of long training time consumption, high model complexity, poor model overfitting and model generalization capability and the like; feature selection is one of effective means for solving the above problems; the high-dimensional construction progress data contains a large amount of information, the extracted time series features are usually high in dimension, but the correlation between part of the extracted features and the construction period is low, or the extracted features are insufficient in distinguishing degree; in order to improve the efficiency and effect of the model, feature selection needs to be carried out on the extracted features; aiming at the construction progress curve of the power grid infrastructure planning project acquired in the step S1, selecting the characteristics extracted from the construction progress curve by using a characteristic filtering method by taking the construction period as a filtering label:
scoring each extracted feature according to divergence or correlation, and if the extracted feature is a Boolean type, adopting KS test; if the extracted features are continuous value features, checking by using Kendall correlation coefficients;
the KS inspection obtains two accumulated distribution curves related to the construction period according to the characteristic value, then the maximum difference of the two accumulated distribution curves is compared, namely the KS inspection value, and the KS inspection value is equal to 0, so that the characteristic X is completely independent of the construction period Y; the KS inspection value is equal to 1, so that the characteristic X has strong correlation with the construction period Y; the KS test value D for any two samples was calculated using the following equation:
in the formula F 0,n0 (y) a cumulative distribution function of the construction period features extracted when the feature value is 0;the accumulated distribution function of the construction period is extracted when the characteristic value is 1; n is 0 The number of the samples of the construction period characteristics extracted when the characteristic value is 0; n is 1 The number of samples of the construction period features extracted when the feature value is 1; y is the construction period;
setting the characteristic obtained by extracting the construction progress curve as X = (X) 1 ,...,x n ) Construction period of Y = (Y) 1 ,...,y n ) (ii) a X and Y coexist inGroup element pair (x) i ,x j ) And (y) i ,y j ) I is not equal to j and is not less than 1, and j is not less than n; calculating the characteristic by the following formulaThe Kendall correlation coefficient tau between X and the construction period Y is characterized as follows:
in the formula N 1 The number of the positively correlated element pairs in X and Y; n is a radical of 2 The number of the element pairs which are negatively related in X and Y; n is a radical of 3 The number of the same element pairs in X and Y; n is a radical of 4 Is the number of identical pairs of elements in Y; n is a radical of 0 Is the total number of element pairs; the positively correlated element pair is defined as x i >x j Time y i >y j And x i <x j Time y i <y j The pair of elements of (1); the negative correlation element pair is defined as x i >x j Time y i <y j And x i <x j Time y i >y j The pair of elements of (1);
selecting the characteristic that the KS inspection value D and the Kendall correlation coefficient tau are larger than a set value;
finally, performing key risk feature identification on the selected features by adopting a principal component analysis method to obtain final construction time sequence features;
the resulting build timing characteristics include: absolute energy values, first order difference absolute sums, aggregate statistical features of autocorrelation coefficients of respective orders, linear regression based on block-wise time series aggregate values, approximate entropy, autoregressive coefficients, ADF tests, lag order autocorrelation, grouping entropy, construction time series data nonlinear measures, time series data description statistics of a given interval, construction time series data complexity, block local entropy ratios, spectral statistics of absolute fourier transforms, longest running self-column length under the mean, largest fixed point of langevin models, average of absolute difference values between time series values, average of difference values between time series values, and value of partial autocorrelation function at a given lag;
s4, constructing a preliminary construction period prediction model based on the Bayesian neural network model, and training to obtain a construction period prediction model; the method specifically comprises the following steps:
the preliminary construction period prediction model comprises an input layer, a hidden layer and an output layer; the hidden layer is a probability layer, so that the network has the capability of describing uncertainty; setting the weight and bias in the probability layer to obey normal distribution;
setting the network parameter of the Bayesian neural network as W, and giving observation data D = (X, Y); x is input data which comprises the affiliated section of the capital construction project, the voltage level, the project attribute, the construction scale, the time sequence progress curve characteristic, the climate change weather factor, the power supply and demand change factor and the policy change factor; y is label data comprising construction period of the capital construction project; the following distribution was obtained:
in the formula P (Y) * |X * D) probability distribution of the construction period under given input data after network parameter training; p (Y) * |X * Omega) is the probability distribution of the construction period under the condition of determining the weight of the probability layer and the input data; p (ω | D) is the posterior distribution of probability layer weights, anP (ω) is the prior distribution of weights; p (D | omega) is probability distribution of the construction period under the condition of determining the weight and input of a probability layer; p (D) is the probability distribution of the construction period under the determined input;
the core of the probability modeling and prediction of project construction period by using the Bayesian neural network lies in making high-efficiency approximate posterior inference, and the variational inference is a very suitable method; the posterior distribution of the network weight is approximated by a variational inference method: introducing a distribution q (ω | θ) to approximate a distribution P (ω | D), where θ = (μ, σ), each weight ω i Obey normal distribution (mu) i ,θ i ) (ii) a The difference between the two distributions is measured by KL divergence; when q (ω | θ) is known, the weight distribution is optimized by minimizing the KL divergence:
in the formula theta * Is the optimized value of the distribution (μ, σ);is distributed under q (ω | θ)The expected value of (d);
then, in the variation reasoning, the lower evidence bound is introduced:
L=logP(Y|X)-D KL (q(ω|θ)||P(ω|D))
in the formula D KL (q (ω | θ) | P (ω | D)) is the KL divergence between the two distributions of q (ω | θ) and P (ω | D);
and (3) adopting a method of carrying out weight parameter on the weight to carry out derivation on the expectation in the calculation formula of the L to obtain:
finally, by a number of different epsilon i To obtainTo approximate the derivative of the KL divergence to θ;
during training, optimizing and training the weight of the probability layer by a variational reasoning method; after training is finished, predicting the construction period of the power grid infrastructure project by adjusting the probability of occurrence of the construction period risks in the input quantity;
s5, inputting the data obtained in the steps S1-S3 into the construction period prediction model obtained in the step S4 to obtain a final construction period prediction result of the power grid infrastructure planning project; in specific implementation, a power grid infrastructure planning project construction period prediction model based on a Bayesian neural network is input to a probability layer of the Bayesian neural network by taking construction period risk factors, power grid infrastructure project construction time sequence characteristics and project basic characteristics as variables, optimization training is performed on the weight of the probability layer through a variational inference method, probability prediction is performed on the power grid infrastructure project construction period by adjusting the occurrence probability of the work period risk in input quantity after training is completed, and power grid infrastructure planning project construction period probability distribution under severe climate, power demand and policy change risk is output.
FIG. 2 is a schematic diagram of system functional modules of the prediction system of the present invention: the prediction system for realizing the power grid infrastructure planning project period prediction method comprises a data acquisition module, an analysis and modeling module, a feature recognition module, a model construction and training module and a prediction module; the data acquisition module, the analysis and modeling module, the feature recognition module, the model construction and training module and the prediction module are sequentially connected in series; the data acquisition module is used for acquiring historical data information of a power grid infrastructure project and uploading the data to the analysis and modeling module; the analysis and modeling module is used for analyzing and modeling factors influencing the power grid infrastructure project according to the received data and uploading the data to the feature recognition module; the characteristic identification module is used for carrying out key risk characteristic identification on the acquired data information according to the received data and uploading the data to the model construction and training module; the model building and training module is used for building a preliminary construction period prediction model based on a Bayesian neural network model according to the received data, training the preliminary construction period prediction model to obtain a construction period prediction model, and uploading the data to the prediction module; and the prediction module is used for inputting the data into the construction period prediction model according to the received data and obtaining a final construction period prediction result of the power grid infrastructure planning project.
Claims (6)
1. A power grid infrastructure planning project construction period prediction method comprises the following steps:
s1, acquiring historical data information of a power grid infrastructure project;
s2, analyzing and modeling factors influencing the power grid infrastructure project;
s3, performing key risk characteristic identification on the data information acquired in the step S1;
s4, constructing a preliminary construction period prediction model based on the Bayesian neural network model, and training to obtain a construction period prediction model;
and S5, inputting the data obtained in the steps S1-S3 into the construction period prediction model obtained in the step S4 to obtain a final construction period prediction result of the power grid infrastructure planning project.
2. The method for predicting the construction period of the power grid infrastructure planning project according to claim 1, wherein the historical data information of the power grid infrastructure project in the step S1 specifically includes a construction progress curve of the historical power grid infrastructure planning project.
3. The method for predicting the construction period of the power grid infrastructure planning project according to claim 2, wherein the step S2 of analyzing and modeling factors influencing the power grid infrastructure project specifically comprises the following steps:
analyzing to obtain factors influencing the power grid infrastructure project, wherein the factors comprise climate change factors, power supply and demand change factors and policy change factors;
describing the climate change factors by adopting triangular distribution, wherein the relevant parameters of the triangular distribution are (0, 0.2, 1); describing the power supply and demand variation factors by adopting triangular distribution, wherein the relevant parameters of the triangular distribution are (-0.5, 0, 0.5); for the policy variation factor, a triangular distribution is used for description, and relevant parameters of the triangular distribution are (-0.5, 0, 0.5).
4. The method for predicting the construction period of the power grid infrastructure planning project according to claim 3, wherein the step S3 of identifying key risk features of the data information obtained in the step S1 specifically comprises the following steps:
aiming at the construction progress curve of the power grid infrastructure planning project acquired in the step S1, selecting the characteristics extracted from the construction progress curve by using a characteristic filtering method by taking the construction period as a filtering label:
scoring each extracted feature according to divergence or correlation, and if the extracted feature is a Boolean type, adopting KS test; if the extracted features are continuous value features, checking by using Kendall correlation coefficients;
the KS test value D for any two samples was calculated using the following equation:
in the formulaThe accumulated distribution function of the construction period features extracted when the feature value is 0 is obtained; />The accumulated distribution function of the construction period is extracted when the characteristic value is 1; n is 0 The number of the samples of the construction period characteristics extracted when the characteristic value is 0; n is 1 The number of samples of the construction period features extracted when the feature value is 1; y is the construction period;
setting the characteristic obtained by extracting the construction progress curve as X = (X) 1 ,...,x n ) Construction period of Y = (Y) 1 ,...,y n ) (ii) a X and Y coexist inGroup element pair (x) i ,x j ) And (y) i ,y j ) I is not equal to j and is not less than 1, and j is not less than n; and calculating a Kendall correlation coefficient tau between the characteristic X and the construction period Y by adopting the following formula:
in the formula N 1 The number of positively correlated element pairs in X and Y; n is a radical of 2 The number of the element pairs which are negatively related in X and Y; n is a radical of 3 The number of the same element pairs in X and Y; n is a radical of 4 Is the number of identical pairs of elements in Y; n is a radical of 0 Is the total number of element pairs; the positively correlated element pair is defined as x i >x j Time y i >y j And x i <x j Time y i <y j The pair of elements of (1); the negative correlation element pair is defined as x i >x j Time y i <y j And x i <x j Time y i >y j The pair of elements of (1);
selecting the characteristic that the KS inspection value D and the Kendall correlation coefficient tau are larger than a set value;
and finally, performing key risk feature identification on the selected features by adopting a principal component analysis method to obtain final construction time sequence features.
5. The method for predicting the construction period of the power grid infrastructure planning project according to claim 4, wherein the step S4 of constructing a preliminary construction period prediction model based on the Bayesian neural network model and training the preliminary construction period prediction model to obtain the construction period prediction model specifically comprises the following steps:
the preliminary construction period prediction model comprises an input layer, a hidden layer and an output layer; the hidden layer is a probability layer, so that the network has the capability of describing uncertainty; setting the weight and bias in the probability layer to obey normal distribution;
setting the network parameter of the Bayesian neural network as W, and giving observation data D = (X, Y); x is input data which comprises the affiliated section of the capital construction project, the voltage level, the project attribute, the construction scale, the time sequence progress curve characteristic, the climate change weather factor, the power supply and demand change factor and the policy change factor; y is label data, including construction period of the capital construction project; the following distribution was obtained:
P(Y * |X * ,D)=∫P(Y * |X * ,ω)P(ω|D)dω
in the formula, P (Y) * |X * D) probability distribution of the construction period under given input data after network parameter training; p (Y) * |X * Omega) is the probability distribution of the construction period under the condition of determining the weight of the probability layer and the input data; p (ω | D) is the posterior distribution of probability layer weights, andp (ω) is the prior distribution of weights; p (D | omega) is the probability distribution of the construction period under the condition of determining the weight and the input of the probability layer; p (D) is the probability distribution of the construction period under the determined input;
the posterior distribution of the network weight is approximated by a variational inference method: introducing a distribution q (ω | θ) to approximate a distribution P (ω | D), where θ = (μ, σ), each weight ω i Obey normal distribution (mu) i ,θ i ) (ii) a The difference between the two distributions is measured by KL divergence; when q (ω | θ) is known, the weight distribution is optimized by minimizing the KL divergence:
in the formula [ theta ] * Is the optimized value of the distribution (μ, σ);is distributed under q (ω | θ)The expected value of (d);
then, in the variation reasoning, the lower evidence bound is introduced:
L=logP(Y|X)-D KL (q(ω|θ)||P(ω|D))
in the formula D KL (q (ω | θ) | P (ω | D)) is the KL divergence between the two distributions of q (ω | θ) and P (ω | D);
and (3) adopting a method of carrying out weight parameter on the weight to carry out derivation on the expectation in the calculation formula of the L to obtain:
finally, by a number of different epsilon i To obtainTo approximate the derivative of the KL divergence to θ;
during training, optimizing and training the weight of the probability layer by a variational reasoning method; and after the training is finished, predicting the construction period of the power grid infrastructure project by adjusting the probability of the occurrence of the construction period risks in the input quantity.
6. A prediction system for realizing the power grid infrastructure planning project period prediction method of one of claims 1 to 5 is characterized by comprising a data acquisition module, an analysis and modeling module, a feature recognition module, a model construction and training module and a prediction module; the data acquisition module, the analysis and modeling module, the feature recognition module, the model construction and training module and the prediction module are sequentially connected in series; the data acquisition module is used for acquiring historical data information of a power grid infrastructure project and uploading the data to the analysis and modeling module; the analysis and modeling module is used for analyzing and modeling factors influencing the power grid infrastructure project according to the received data and uploading the data to the feature recognition module; the characteristic identification module is used for carrying out key risk characteristic identification on the acquired data information according to the received data and uploading the data to the model construction and training module; the model building and training module is used for building a preliminary construction period prediction model based on a Bayesian neural network model according to the received data, training the preliminary construction period prediction model to obtain a construction period prediction model, and uploading the data to the prediction module; and the prediction module is used for inputting the data into the construction period prediction model according to the received data and obtaining a final construction period prediction result of the power grid infrastructure planning project.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310044387.XA CN115860263A (en) | 2023-01-30 | 2023-01-30 | Power grid infrastructure planning project construction period prediction method and prediction system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310044387.XA CN115860263A (en) | 2023-01-30 | 2023-01-30 | Power grid infrastructure planning project construction period prediction method and prediction system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115860263A true CN115860263A (en) | 2023-03-28 |
Family
ID=85657361
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310044387.XA Pending CN115860263A (en) | 2023-01-30 | 2023-01-30 | Power grid infrastructure planning project construction period prediction method and prediction system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115860263A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117745247A (en) * | 2024-02-21 | 2024-03-22 | 中国有色金属工业昆明勘察设计研究院有限公司 | Rock-soil construction wisdom building site system |
-
2023
- 2023-01-30 CN CN202310044387.XA patent/CN115860263A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117745247A (en) * | 2024-02-21 | 2024-03-22 | 中国有色金属工业昆明勘察设计研究院有限公司 | Rock-soil construction wisdom building site system |
CN117745247B (en) * | 2024-02-21 | 2024-06-11 | 中国有色金属工业昆明勘察设计研究院有限公司 | Rock-soil construction wisdom building site system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110232203B (en) | Knowledge distillation optimization RNN short-term power failure prediction method, storage medium and equipment | |
Alqahtani et al. | Artificial neural networks incorporating cost significant items towards enhancing estimation for (life-cycle) costing of construction projects | |
CN106408223A (en) | Short-term load prediction based on meteorological similar day and error correction | |
CN112990500B (en) | Transformer area line loss analysis method and system based on improved weighted gray correlation analysis | |
CN102522709B (en) | Decision-making method and decision-making system for state overhaul of transformers | |
CN104318482A (en) | Comprehensive assessment system and method of smart distribution network | |
CN111639823B (en) | Building cold and heat load prediction method constructed based on feature set | |
CN107067341B (en) | RBFNN power distribution automation system state operation evaluation method based on multi-level entropy weight | |
CN111178621A (en) | Parameter optimization method of electric heating load prediction support vector regression model | |
CN110852496A (en) | Natural gas load prediction method based on LSTM recurrent neural network | |
CN115860263A (en) | Power grid infrastructure planning project construction period prediction method and prediction system | |
CN117273440A (en) | Engineering construction Internet of things monitoring and managing system and method based on deep learning | |
CN113139701B (en) | Regional energy health evaluation method based on analytic hierarchy process | |
CN114548494A (en) | Visual cost data prediction intelligent analysis system | |
CN104598969A (en) | High-voltage electrical appliance operation quality evaluation method and system based on neural network | |
CN116842841A (en) | Immersed tunnel service state prediction method based on artificial intelligence | |
CN116823008A (en) | Park energy utilization efficiency evaluation method, system, equipment and storage medium | |
CN115495862A (en) | Power transmission network extension planning method and system considering extreme scenes of renewable energy sources | |
CN115511230A (en) | Electric energy substitution potential analysis and prediction method | |
CN114780609A (en) | Regional carbon emission calculation method based on electric power big data and time series | |
CN115310746A (en) | State evaluation method and system for main transmission system of wind generating set | |
Sanayha et al. | Remaining useful life prediction using enhanced convolutional neural network on multivariate time series sensor data | |
CN118246714B (en) | Water conservancy and hydropower engineering construction energy consumption analysis method and system | |
CN116720662B (en) | Distributed energy system applicability evaluation method based on set pair analysis | |
CN117522173B (en) | Natural gas hydrate depressurization exploitation productivity prediction method based on deep neural network |
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 |