CN111340273A - Short-term load prediction method for power system based on GEP parameter optimization XGboost - Google Patents

Short-term load prediction method for power system based on GEP parameter optimization XGboost Download PDF

Info

Publication number
CN111340273A
CN111340273A CN202010096168.2A CN202010096168A CN111340273A CN 111340273 A CN111340273 A CN 111340273A CN 202010096168 A CN202010096168 A CN 202010096168A CN 111340273 A CN111340273 A CN 111340273A
Authority
CN
China
Prior art keywords
sample
xgboost
model
samples
day
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.)
Granted
Application number
CN202010096168.2A
Other languages
Chinese (zh)
Other versions
CN111340273B (en
Inventor
邓松
董霞
蔡清媛
岳东
袁新雅
陈福林
张建堂
祝展望
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.)
Nanjing University of Posts and Telecommunications
Original Assignee
Nanjing University of Posts and Telecommunications
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 Nanjing University of Posts and Telecommunications filed Critical Nanjing University of Posts and Telecommunications
Priority to CN202010096168.2A priority Critical patent/CN111340273B/en
Publication of CN111340273A publication Critical patent/CN111340273A/en
Application granted granted Critical
Publication of CN111340273B publication Critical patent/CN111340273B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • 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
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/067—Enterprise or organisation modelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • 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

Landscapes

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

Abstract

The invention discloses a short-term load forecasting method of an electric power system based on GEP parameter optimization XGboost, which belongs to the technical field of loads of the electric power system, researches a short-term load forecasting method in electric power load data with large information amount to obtain the electric power load characteristics of a day to be forecasted, and mainly comprises three parts: selecting similar days, optimizing parameters by a GEP algorithm, and training and predicting an XGboost model; and performing weight calculation on the key influence factors by selecting the similar days to form a weighted panel data gray correlation judgment matrix. Calculating and selecting data with high similarity to the day to be predicted as a similar day set; the GEP algorithm optimization parameters are subjected to iterative genetic operation on important parameters in the prediction model through a GEP algorithm to obtain fitness so as to obtain the optimal solution of the parameters, and the accuracy of the model is improved; and the XGboost model training and prediction inputs the similar day set as a training set into the XGboost model for training so as to obtain a model with complete functions and accurate prediction. And finally, inputting the day to be predicted into the model to obtain a predicted value and outputting the predicted value.

Description

Short-term load prediction method for power system based on GEP parameter optimization XGboost
Technical Field
The invention belongs to the technical field of loads of power systems, and particularly relates to a short-term load prediction method for a power system based on GEP parameter optimization XGboost.
Background
The load prediction problem relates to the aspects of power system planning and design, the economy, reliability and safety of power system operation, power market trading and the like, and has become an important field in the operation and management of modern power systems. The main task of the power system is to provide economic, reliable and high-quality electric energy for various users so as to meet the requirements of the load demand and the load characteristics of the users at any time. For this reason, in power system planning design, operation management, and power market trading, an accurate prediction of load demand variation and load characteristics is necessary. This is an important reason for people to continuously research and develop the load prediction theory of the power system.
In the past, short, medium and long-term power load prediction is important research work of power supply companies and researchers, and a large number of theoretical methods are also proposed. The traditional method is a prediction method established on the basis of a time series prediction principle, and is represented by a series of methods such as an autoregressive method and an accumulative autoregressive moving average method. In recent years, with the rise of artificial intelligence, another type of power load prediction algorithm based on machine learning has appeared, and the artificial neural network and the support vector machine method are highlighted as representatives. The series of machine learning algorithms have many outstanding advantages, including self-adaptive function for a large number of non-structural and non-precise rules, better generalization performance of the predicted object, but also have a series of defects of complex optimization process, slow convergence speed or large error and the like, so that a reasonable and accurate power load prediction method is particularly important in practical application.
The existing method for predicting the short-term load mostly has the defects of low prediction precision, complex process and the like, and cannot realize accurate prediction of the power load, so that the new and effective method for predicting the load of the power system has great significance to the operation and management of the power system.
And establishing a short-term load prediction model according to the XGboost model, and training the model by taking historical power load data as training sample data. And selecting optimal parameters from some important parameters in the XGboost model through a gene expression programming algorithm.
Disclosure of Invention
The purpose of the invention is as follows: the invention aims to provide a short-term load forecasting method of an electric power system based on GEP parameter optimization XGboost, which comprises the steps of selecting similar samples on a day to be forecasted and optimizing model parameters, using historical sample data to train an XGboost model, inputting the samples to be forecasted after the training is finished, and obtaining the characteristics of the electric power load to be forecasted; the method is a strategic method, can quickly and accurately obtain the load demand and the load characteristic of the day to be predicted, and ensures the high efficiency and stability of the operation and management of the power system.
The technical scheme is as follows: in order to achieve the purpose, the invention provides the following technical scheme:
a short-term load prediction method for an electric power system based on GEP parameter optimization XGboost comprises the following steps:
1) performing weighted panel data gray similarity correlation analysis on all historical samples and the days to be predicted, selecting similar days, and selecting samples with high feature similarity with the days to be predicted as a similar day set;
2) performing iterative genetic operation on important parameters in the XGboost model through optimizing the parameters by using a gene expression programming algorithm to obtain fitness so as to obtain the optimal solution of the parameters;
3) putting the similar day set as a training set into an XGboost model for training to obtain an XGboost load prediction model;
4) after the XGboost load prediction model is built, the day to be predicted is input, and the load characteristics of the day to be predicted are calculated and output through the XGboost load prediction model.
Further, in step 1), the similar day selection is performed, and the method comprises the following steps:
1.1) taking the history sample as panel data, X ═ X1,X2,X3…Xn]Wherein X isiRepresenting historical data of the ith day, selecting some key factors influencing power load change, including temperature, humidity, rainfall and the like of meteorological factors, constructing a feature vector, dividing the time sequence into a plurality of sections, selecting the feature vector of the influencing factors in the sample observation value in each section, and entering the second step;
1.2) constructing a panel data gray correlation judgment matrix XiD，Xi(m, q) represents a load observation corresponding to the mth factor in the ith sample with respect to the qth time period; and processing the data in the matrix through a grey correlation operator, and entering the step 1.3);
1.3) calculating the weight of each influence factor by using a principal component analysis method to obtain a weight vector; weighting the weight vector to the panel data gray correlation judgment matrix to obtain a weighted panel data gray correlation decision matrix Fi；
And 1.4) taking all samples with similarity higher than the threshold as a similar day set, and selecting similar day sample set data as samples for model training.
Further, in step 1.1), the historical samples are used as panel data, where N historical samples and m key influencing factors are set, the time sequence is divided into q segments, and if a value of a sample i in the panel data with respect to the influencing factor s at time p is xi(s, p), i ═ 1,2, …, N, s ═ 1,2, …, m, p ═ 1,2, …, q: defining:
Xi＝(xi(1),xi(2),…,xi(m))T；
wherein: x is the number ofi(s)＝(xi(s,1),xi(s,2),…,xi(s, q)), T represents a transpose of the matrix; scale xi(s) a time series of samples i for the set of panel data with respect to the influencing factor; let the matrix form of the samples be expressed as:
Figure BDA0002385422370000031
wherein x isi(m, q) represents the X-thiThe influence factor value of the qth time section of each sample under the mth influence factor;
in step 1.2), the gray correlation judgment matrix X of the panel data is constructediD is expressed as:
Figure BDA0002385422370000032
wherein D is a panel data mean value fluctuation initialization operator,
Figure BDA0002385422370000033
Figure BDA0002385422370000034
let two different samples, sample i and sample j (i, j ≠ 1,2, …, N, i ≠ j), whose time sequence after initialization by D under s influence factor is:
xi'(s)＝xi(s)d＝(xi(s,1)d,xi(s,2)d,…,xi(s,q)d)；
xj'(s)＝xj(s)d＝(xj(s,1)d,xj(s,2)d,…,xj(s,q)d)；
in step 1.3), the method uses principal component analysis to calculate each influencing factorThe weight vector is obtained as W ═ W1,w2,…wm]In the formula wmThe weight value of the mth influence factor; weighted panel data gray correlation decision matrix FiUsing the obtained weight vector to judge the matrix X for the grey correlation of the panel dataiD, weighting to obtain weighted panel data gray correlation decision matrix FiExpressed as:
Figure BDA0002385422370000041
the weighted time series of sample i and sample j under s influence factor is: x is the number ofi”(s)、xjAnd the included angle of two vectors is shown as follows(s):
Figure BDA0002385422370000042
wherein<xi”(s),xj”(s)>Is the included angle of two vectors, and the unit is degree (degree) | | xi”(s)||、||xj"(s) | | is the modulo length of the two vectors, respectively;
in step 1.4), the said taking all samples with similarity higher than the threshold as the similar day set is based on
Figure BDA0002385422370000043
Calculating the sample similarity correlation degree of the historical sample matrix and the day matrix to be predicted, wherein rho (X)0,Xj) Representing X in a sequence of panel data samples0And XjThe relevance degrees are sorted from large to small, a threshold value is set, and all samples with the similarity degrees higher than the threshold value are used as a similarity day set.
Further, in step 2), the optimizing parameters by the gene expression programming algorithm includes the following steps:
2.1) taking parameters required by the XGboost model as input, initializing a population through a gene expression programming algorithm, calculating the fitness, storing an optimal individual, then performing selection operation to output population data R, and entering the step 2.2);
2.2) generating a next generation population by carrying out a series of basic genetic operations including crossing, mutation and recombination on the population R after the selection operation, and entering the step 2.3);
2.3) calculating the fitness of the population generated in the step 2.1), if the fitness meets a termination condition, outputting the obtained optimal parameters, entering sample set data of similar days as samples for model training, and otherwise returning to the step 2.1).
Further, in step 3), the step of putting the similar day set as a training set into the XGBoost model for training includes the following steps:
3.1) performing model training by taking similar day sample set data as samples, wherein an XGboost model objective function is as follows:
Figure BDA0002385422370000051
wherein: n 'is the total number of samples in the similar days, N' is less than or equal to N, i 'is the ith' sample, wherein i 'is 1 … N', K is the total number of decision trees, and omega (f)k) Complexity of the kth tree, yi′'is the resulting predicted value for the i' th sample; the method comprises two parts, namely training loss l and a regularization term, wherein the regularization term is the complexity sum of all K trees, and the step 3.2) is carried out;
3.2) performing a second order Taylor expansion of the loss function of the kth ongoing tree, wherein the first order partial derivatives
Figure BDA0002385422370000052
Second order partial derivative
Figure BDA0002385422370000053
Entering step 3.3);
3.3) substituting the loss function into the objective function to obtain an approximate value, and removing the constant term to obtain the objective function
Figure BDA0002385422370000054
Entering step 3.4);
3.3) defining a tree fk(x)＝wk(x)Indicating that the sample falls into the leaf of tree kCombining the nodes with the weights, and entering step 3.4);
3.4) complexity of defining a tree:
Figure BDA0002385422370000055
gamma and lambda are constants, U is the number of leaf nodes, and the step 3.5) is carried out;
3.5) simplifying the function according to the first two steps to obtain
Figure BDA0002385422370000056
To relate to wj'The one-dimensional quadratic equation of (3) is calculated to obtain the minimum value of the objective function, and the step is carried out in step 3.6);
3.6) establishing a model based on the loss function, and segmenting the nodes; gain according to loss function ObjL+R-(ObjL+ObjR) Searching an optimal splitting point; and finishing the establishment of the XGboost load prediction model.
The invention principle is as follows: the invention mainly comprises the following steps: selecting similar days, optimizing parameters by a GEP algorithm, and training and predicting an XGboost model; the similar day selection is mainly to use all historical data as panel data, and to perform weight calculation on key influence factors to form a weighted panel data gray correlation judgment matrix. And selecting the date set with high similarity as a similar date set by calculating the similarity between the date to be predicted and the historical sample. The GEP optimization parameters are mainly obtained by carrying out iterative genetic operation on important parameters in the XGboost model through the algorithm to obtain fitness so as to obtain the optimal solution of the parameters. The XGboost model training and prediction mainly comprises the steps of training a similar day set as a training set, inputting a day to be predicted into a model after the training is finished, and obtaining and outputting a predicted value.
Has the advantages that: compared with the prior art, the method for predicting the short-term load of the power system based on the GEP parameter optimization XGboost is mainly used for predicting the short-term load of the power system. By the method and the device, the load demand and the load characteristic of the day to be predicted can be predicted more accurately, so that the purposes of providing economic, reliable and high-quality electric energy for various users and meeting the users are achieved.
Drawings
FIG. 1 is a system structure diagram of a short-term load prediction method for an electrical power system based on GEP parameter optimization XGboost;
fig. 2 is a schematic flow diagram of a short-term load prediction method of an electric power system based on GEP parameter optimization XGBoost.
Detailed Description
For a better understanding of the contents of the present patent application, the technical solutions of the present invention will be further described below with reference to the accompanying drawings and specific examples.
As shown in fig. 1, the method mainly includes: selecting similar days, optimizing parameters by a GEP algorithm, and training and predicting an XGboost model; the similar day selection in fig. 1 is mainly used for performing panel data gray similarity correlation analysis on all historical samples and days to be predicted, and selecting a sample with high similarity to the characteristic of the day to be predicted as a training set of a model to improve prediction accuracy; the GEP algorithm optimization parameters are obtained by carrying out iterative genetic operation on important parameters in the XGboost model through the algorithm to obtain fitness so as to obtain the optimal solution of the parameters; the XGboost model training and prediction is to train the model by taking the similar day set as a training set to predict the load characteristics of the day to be predicted.
A short-term load prediction method for an electric power system based on GEP parameter optimization XGboost comprises the following steps:
1) performing weighted panel data gray similarity correlation analysis on all historical samples and the days to be predicted, selecting similar days, and selecting samples with high feature similarity with the days to be predicted as a similar day set;
2) performing iterative genetic operation on important parameters in the XGboost model through optimizing the parameters by using a gene expression programming algorithm to obtain fitness so as to obtain the optimal solution of the parameters;
3) putting the similar day set as a training set into an XGboost model for training to obtain an XGboost load prediction model;
4) after the XGboost load prediction model is built, the day to be predicted is input, and the load characteristics of the day to be predicted are calculated and output through the XGboost load prediction model.
1. Similar day selection
In the power system load prediction, in order to improve the prediction accuracy, a sample with high similarity to a daily sample to be predicted is selected from historical samples as a precondition of a training set of a prediction model. In this patent, the historical data of selecting among the electric power system is taken as the panel data sequence, selects some key factors that influence the change of power load, including temperature, humidity, rainfall, etc. of meteorological factor to become the time series with the time segmentation of observing, show panel data indifference. And initializing the data and calculating the weight of the influence factors through a gray correlation operator, and weighting the panel data gray correlation judgment matrix by using the weight vector to obtain a weighted panel data gray correlation decision matrix F. By rho (X)0,Xj) Representing X in a sequence of data samples representing a panel0And XjThe similarity between two samples is determined by the similarity correlation of the two samples, p (X)0,Xj) The larger the value of the similarity value is, the higher the similarity between the sample to be predicted and the sample is, after all the similarities are calculated, the samples are sorted from large to small, a threshold value is selected, and all the samples with the similarities higher than the threshold value are used as a similarity day set.
If the number of the historical samples is N, the number of the key influence factors is m, the time sequence is divided into q sections, and if the value of the sample i in the panel data on the influence factor s in time p is xi(s, p), i ═ 1,2, …, N, s ═ 1,2, …, m, p ═ 1,2, …, q. Defining:
Xi＝(xi(1),xi(2),…,xi(m))T；
wherein: x is the number ofi(s)＝(xi(s,1),xi(s,2),…,xi(s, q)), T represents a transpose of the matrix; scale xi(s) is a time series of samples i of the set of panel data with respect to the influencing factor.
(1) Let the matrix form of the samples be expressed as:
Figure BDA0002385422370000071
wherein x isi(m, q) represents the X-thiThe influential value of the qth time period for each sample under the mth influential factor.
(2) Let D be the panel data mean value fluctuation initialization operator, then XiD is expressed as:
Figure BDA0002385422370000072
wherein
Figure BDA0002385422370000081
Let two different samples, sample i and sample j (i, j ≠ 1,2, …, N, i ≠ j), whose time sequence after initialization by D under s influence factor is:
xi'(s)＝xi(s)d＝(xi(s,1)d,xi(s,2)d,…,xi(s,q)d)；
xj'(s)＝xj(s)d＝(xj(s,1)d,xj(s,2)d,…,xj(s,q)d)；
(3) determining the weight of each influence factor according to a principal component analysis method to obtain a weight vector W ═ W1,w2,…wm]In the formula wmIs the weighted value of the mth influencing factor. Weighting the initialized matrix by using the weight vector to obtain a weighted panel data gray correlation decision matrix Fi＝XiDW, expressed as:
Figure BDA0002385422370000082
the weighted time series of sample i and sample j under s influence factor is: x is the number ofi”(s)、xjAnd the included angle of two vectors is shown as follows(s):
Figure BDA0002385422370000083
wherein<xi”(s),xj”(s)>Is the included angle of two vectors, and the unit is degree (degree) | | xi”(s)||、||xj"(s) | | is the modulo length of the two vectors, respectively;
(4) according to
Figure BDA0002385422370000084
ξ∈ (0, infinity), calculating the similarity association degree of the historical samples and the samples to be predicted, after the calculation is finished, sorting all the association degrees from large to small, selecting a threshold value, and taking all the samples with the similarity higher than the threshold value as a similar day set.
2. Gene expression programming algorithm optimization parameters
The genetic algorithm simulates the phenomena of reproduction, crossing and variation in the natural selection and genetic process, and according to the natural rules of survival, excellence and disqualification of fittest provided by the Darwinian evolution theory, under the action of genetic operators (selection, crossing and variation), the population is optimized generation by generation, excellent individuals are generated, and finally, the optimal solution is searched. The genetic expression programming algorithm is an improvement of the genetic algorithm, integrates the advantages of the genetic programming and the genetic algorithm, and provides a novel evolutionary algorithm: and (4) GEP. In the expression form, the method inherits the characteristics of simple and quick fixed-length linear coding of GA; in the gene expression, the characteristic that the tree structure of GP is flexible and changeable is inherited, the load problem is solved by using simple codes, and the method is compared with the traditional evolutionary computation. In the patent, the algorithm is adopted to carry out iteration selection on more important parameters in the XGboost model in the algorithm to obtain the effect of improving the prediction precision of the model.
The more important parameters of the XGBoost model are: the method comprises the steps of initializing a population, calculating the fitness, storing an optimal individual, performing selection operation to output population data, generating a next generation population after the population subjected to the selection operation is subjected to a series of basic genetic operations including crossing, mutation and recombination, iteratively calculating the fitness, finishing the algorithm when a termination condition is met, and outputting the optimal solution of each parameter.
3. XGboost model training and prediction
The XGboost is one of Boosting algorithms and is a lifting tree model, the idea of the Boosting algorithm is to integrate a plurality of weak classifiers together, try to correct the residual errors of all the previous weak classifiers by adding a new weak learner, and finally add a plurality of learners together for final prediction, so that the accuracy is higher than that of a single one, and a strong classifier is formed. And the addition of regularization is used to reduce errors when adding new models. In the patent, firstly, the optimal parameters obtained by the GEP algorithm are set in the model, then the similar day sample set data is used as a sample for model training, and a tree is added from one tree under the residual error of the previous tree to ensure that the model effect is better. The objective function of the XGboost model consists of two parts, namely training loss l and a regularization term, wherein the regularization term is the complexity sum of all K trees. The objective function is expressed as:
Figure BDA0002385422370000091
wherein: n 'is the total number of similar day samples (N' ≦ N), i '(i' ═ 1 … N ') is the ith' sample, K is the total number of decision trees, Ω (f)k) Complexity of the kth tree, yi′'is the resulting predicted value for the ith' sample. Since XGboost is an additive model, the predicted value is the sum of the predicted values of each tree, fkAs a function of each tree, i.e.:
Figure BDA0002385422370000092
when a kth (K is more than or equal to 1 and less than or equal to K) tree is trained, performing second-order Taylor expansion on the loss function of the ongoing tree, wherein:
first order partial derivative
Figure BDA0002385422370000093
Second order partial derivative
Figure BDA0002385422370000101
The loss function is brought into the objective function to obtain an approximate value, and a constant term is removed to obtain a new objective function form:
Figure BDA0002385422370000102
defining a tree fk(x)＝wk(x)Indicating that the sample falls into a leaf node of a tree k and then is combined with the weight; complexity of defining a tree:
Figure BDA0002385422370000103
gamma and lambda are constants, and U is the number of leaf nodes. Simplifying the function to obtain a final objective function:
Figure BDA0002385422370000104
to relate to wj'The quadratic equation of the first order is calculated to obtain the minimum value of the target function, a model is built based on the loss function, and the nodes are segmented. Gain according to loss function ObjL+R-(ObjL+ObjR) And finding the optimal splitting point. And finishing the establishment of the whole model.
(4) After the XGboost load prediction model is built, the day to be predicted is input, and the load characteristics of the day to be predicted are calculated and output through the XGboost load prediction model. And inputting the day to be predicted into the trained model, and outputting a predicted value.
As shown in fig. 2, a short-term load prediction method for an electric power system based on GEP parameter optimization XGBoost includes the following steps:
the method comprises the following steps: taking the historical sample as panel data, X ═ X1,X2,X3…Xn]Wherein X isiRepresenting the historical data of the ith day, selecting some key factors influencing the change of the power load, including the temperature, the humidity, the rainfall and the like of meteorological factors, constructing a characteristic vector, dividing the time sequence into a plurality of sections, and selecting an observed value of the influence factor in a sample in each sectionEntering the step two;
step two: building a panel data gray correlation judgment matrix, Xi(m, q) represents a load observation corresponding to the mth factor in the ith sample with respect to the qth time period. Processing the data in the matrix through a grey correlation operator, and entering a third step;
step three: calculating the weight of each influence factor by using a principal component analysis method to obtain a weight vector, weighting the panel data gray correlation judgment matrix by using the weight vector to obtain a weighted panel data gray correlation decision matrix F, and entering the step eight;
step four: according to
Figure BDA0002385422370000111
Calculating the sample similarity correlation degree of the historical sample matrix and the day matrix to be predicted, wherein rho (X)0,Xj) Representing X in a sequence of panel data samples0And XjThe similarity relevance degrees are ranked from large to small, a threshold value is set, a large historical sample is selected as a similar day sample set, and the step five is carried out;
step five: taking parameters required by the XGboost model as input, initializing a population through a gene expression programming algorithm, calculating the fitness, storing the optimal individual, performing selection operation to output population data R, and entering the sixth step;
step six: generating a next generation population by carrying out a series of basic genetic operations including crossing, mutation and recombination on the population R after the selection operation, and entering the seventh step;
step seven: calculating the fitness of the generated population in the step five, if the fitness meets a termination condition, outputting the obtained optimal parameters, entering the step eight, and otherwise, returning to the step five;
step eight: performing model training by taking the sample set data of the similar day as a sample, and increasing a tree from one tree under the residual error of the previous tree to ensure that the model effect is better; XGboost model objective function
Figure BDA0002385422370000112
The method comprises two parts, namely training loss l and a regularization term, wherein the regularization term is the sum of the complexity of all K trees, and the step nine is carried out;
step nine: performing a second order Taylor expansion on the loss function of the k-th ongoing tree, wherein the first order partial derivatives
Figure BDA0002385422370000113
Second order partial derivative
Figure BDA0002385422370000114
Entering the step ten;
step ten: the loss function is brought into the objective function to obtain an approximate value, and the constant term is removed to obtain the objective function
Figure BDA0002385422370000115
Entering the step eleven;
step eleven: defining a tree fk(x)＝wk(x)After the sample falls into the leaf node of the tree k, combining the sample with the weight, and entering the step twelve;
step twelve: complexity of defining a tree:
Figure BDA0002385422370000116
gamma and lambda are constants, U is the number of leaf nodes, and step thirteen is carried out;
step thirteen: simplifying the function according to the first two steps to obtain
Figure BDA0002385422370000121
To relate to wj'Calculating the minimum value of the objective function by using the unitary quadratic equation, and entering a step fourteen;
fourteen steps: and establishing a model based on the loss function, and segmenting the nodes. Gain according to loss function ObjL+R-(ObjL+ObjR) And finding the optimal splitting point. And finishing the establishment of the whole model. Entering a step fifteen;
step fifteen; and inputting the day to be predicted into the trained model, outputting a predicted value, and ending.
Examples
With the reformation and development of electric power marketization, the load information in the electric power system has multiple dimensions, large data volume and complex data types. The information of the load has great value to users, enterprises and social economy, and relates to multiple aspects of power system planning and design, power system operation economy, reliability and safety, power market transaction and the like. How to build a model on the basis of the existing large amount of historical data and perform predictive analysis on the power load becomes an important research direction. Suppose there is accurate historical load data for a period of time in a certain area and a determination is to be made of the load characteristics for a future day. Firstly, selecting a sample with characteristics similar to those of the day to be predicted through similarity correlation analysis, obtaining the optimal parameter solution of the prediction model by using a gene expression programming algorithm, and finally, performing power load prediction of the day to be predicted after the XGboost model is trained through the sample data.
The specific implementation scheme is as follows:
(1) and taking all historical data as panel data, and carrying out weight calculation on key influence factors to form a weighted panel data gray correlation judgment matrix. And selecting the date set with high similarity as a similar date set by calculating the similarity between the date to be predicted and the historical sample.
(2) The method mainly comprises the steps of carrying out iterative genetic operation on important selected parameters in a prediction model through a gene expression programming algorithm to obtain fitness so as to obtain the optimal solution of the parameters, and improving the accuracy of model prediction.
(3) And inputting the similar day set serving as a training set into the XGboost model for training, and obtaining a model with complete functions and accurate prediction after training. And finally, inputting the day to be predicted into the model to obtain a predicted value and outputting the predicted value.

Claims (5)

1. A short-term load prediction method for an electric power system based on GEP parameter optimization XGboost is characterized by comprising the following steps: the method comprises the following steps:
1) performing weighted panel data gray similarity correlation analysis on all historical samples and the days to be predicted, selecting similar days, and selecting samples with high feature similarity with the days to be predicted as a similar day set;
2) performing iterative genetic operation on important parameters in the XGboost model through optimizing the parameters by using a gene expression programming algorithm to obtain fitness so as to obtain the optimal solution of the parameters;
3) putting the similar day set as a training set into an XGboost model for training to obtain an XGboost load prediction model;
4) after the XGboost load prediction model is built, the day to be predicted is input, and the load characteristics of the day to be predicted are calculated and output through the XGboost load prediction model.
2. The method for predicting the short-term load of the power system based on the GEP parameter optimization XGboost of claim 1, wherein the method comprises the following steps: in the step 1), the similar day selection is carried out, and the method comprises the following steps:
1.1) taking the history sample as panel data, X ═ X1,X2,X3…Xn]Wherein X isiRepresenting historical data of the ith day, selecting some key factors influencing power load change, including temperature, humidity, rainfall and the like of meteorological factors, constructing a feature vector, dividing the time sequence into a plurality of sections, selecting the feature vector of the influencing factors in the sample observation value in each section, and entering the second step;
1.2) constructing a panel data gray correlation judgment matrix XiD，Xi(m, q) represents a load observation corresponding to the mth factor in the ith sample with respect to the qth time period; and processing the data in the matrix through a grey correlation operator, and entering the step 1.3);
1.3) calculating the weight of each influence factor by using a principal component analysis method to obtain a weight vector; weighting the weight vector to the panel data gray correlation judgment matrix to obtain a weighted panel data gray correlation decision matrix Fi；
And 1.4) taking all samples with similarity higher than the threshold as a similar day set, and selecting similar day sample set data as samples for model training.
3. The method for predicting the short-term load of the power system based on the GEP parameter optimization XGboost of claim 2, wherein the method comprises the following steps: in step 1.1), the historical samples are used as panel data, wherein N historical samples and m key influence factors are set, the time sequence is divided into q segments, and if the value of a sample i in the panel data on the influence factor s at time p is xi(s, p), i ═ 1,2, …, N, s ═ 1,2, …, m, p ═ 1,2, …, q: defining:
Xi＝(xi(1),xi(2),…,xi(m))T；
wherein: x is the number ofi(s)＝(xi(s,1),xi(s,2),…,xi(s, q)), T represents a transpose of the matrix; scale xi(s) a time series of samples i for the set of panel data with respect to the influencing factor; let the matrix form of the samples be expressed as:
Figure FDA0002385422360000021
wherein x isi(m, q) represents the X-thiThe influence factor value of the qth time section of each sample under the mth influence factor;
in step 1.2), the gray correlation judgment matrix X of the panel data is constructediD is expressed as:
Figure FDA0002385422360000022
wherein D is a panel data mean value fluctuation initialization operator,
Figure FDA0002385422360000023
Figure FDA0002385422360000024
let two different samples, sample i and sample j (i, j ≠ 1,2, …, N, i ≠ j), whose time sequence after initialization by D under s influence factor is:
xi'(s)＝xi(s)d＝(xi(s,1)d,xi(s,2)d,…,xi(s,q)d)；
xj'(s)＝xj(s)d＝(xj(s,1)d,xj(s,2)d,…,xj(s,q)d)；
in step 1.3), the weight of each influencing factor is calculated by using a principal component analysis method to obtain a weight vector W ═ W1,w2,…wm]In the formula wmThe weight value of the mth influence factor; weighted panel data gray correlation decision matrix FiUsing the obtained weight vector to judge the matrix X for the grey correlation of the panel dataiD, weighting to obtain weighted panel data gray correlation decision matrix FiExpressed as:
Figure FDA0002385422360000025
the weighted time series of sample i and sample j under s influence factor is: x is the number ofi”(s)、xjAnd the included angle of two vectors is shown as follows(s):
Figure FDA0002385422360000031
wherein<xi”(s),xj”(s)>Is the included angle of two vectors, and the unit is degree (degree) | | xi”(s)||、||xj"(s) | | is the modulo length of the two vectors, respectively;
in step 1.4), the said taking all samples with similarity higher than the threshold as the similar day set is based on
Figure FDA0002385422360000032
Calculating the sample similarity correlation degree of the historical sample matrix and the day matrix to be predicted, wherein rho (X)0,Xj) Representing X in a sequence of panel data samples0And XjThe similarity degree of the relation is sorted from big to small, a threshold value is set,all samples with similarity higher than the threshold are taken as a similar day set.
4. The method for predicting the short-term load of the power system based on the GEP parameter optimization XGboost of claim 1, wherein the method comprises the following steps: in step 2), the parameter optimization through the gene expression programming algorithm comprises the following steps:
2.1) taking parameters required by the XGboost model as input, initializing a population through a gene expression programming algorithm, calculating the fitness, storing an optimal individual, then performing selection operation to output population data R, and entering the step 2.2);
2.2) generating a next generation population by carrying out a series of basic genetic operations including crossing, mutation and recombination on the population R after the selection operation, and entering the step 2.3);
2.3) calculating the fitness of the population generated in the step 2.1), if the fitness meets a termination condition, outputting the obtained optimal parameters, entering sample set data of similar days as samples for model training, and otherwise returning to the step 2.1).
5. The method for predicting the short-term load of the power system based on the GEP parameter optimization XGboost of claim 1, wherein the method comprises the following steps: in step 3), the similar day set is used as a training set and put into an XGboost model for training, and the method comprises the following steps:
3.1) performing model training by taking similar day sample set data as samples, wherein an XGboost model objective function is as follows:
Figure FDA0002385422360000033
wherein: n 'is the total number of samples in the similar days, N' is less than or equal to N, i 'is the ith' sample, wherein i 'is 1 … N', K is the total number of decision trees, and omega (f)k) Complexity of the kth tree, yi’'is the resulting predicted value for the i' th sample; the method comprises two parts, namely training loss l and a regularization term, wherein the regularization term is the complexity sum of all K trees, and the step 3.2) is carried out;
3.2) on goingThe loss function of k trees is subjected to a second order Taylor expansion, in which the first order partial derivatives
Figure FDA0002385422360000041
Second order partial derivative
Figure FDA0002385422360000042
Entering step 3.3);
3.3) substituting the loss function into the objective function to obtain an approximate value, and removing the constant term to obtain the objective function
Figure FDA0002385422360000043
Entering step 3.4);
3.3) defining a tree fk(x)＝wk(x)After the sample falls into the leaf node of the tree k, combining the sample with the weight, and entering the step 3.4);
3.4) complexity of defining a tree:
Figure FDA0002385422360000044
gamma and lambda are constants, U is the number of leaf nodes, and the step 3.5) is carried out;
3.5) simplifying the function according to the first two steps
Figure FDA0002385422360000045
To relate to wj'The one-dimensional quadratic equation of (3) is calculated to obtain the minimum value of the objective function, and the step is carried out in step 3.6);
3.6) establishing a model based on the loss function, and segmenting the nodes; gain according to loss function ObjL+R-(ObjL+ObjR) Searching an optimal splitting point; and finishing the establishment of the XGboost load prediction model.
CN202010096168.2A 2020-02-17 2020-02-17 Short-term load prediction method for power system based on GEP parameter optimization XGboost Active CN111340273B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010096168.2A CN111340273B (en) 2020-02-17 2020-02-17 Short-term load prediction method for power system based on GEP parameter optimization XGboost

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010096168.2A CN111340273B (en) 2020-02-17 2020-02-17 Short-term load prediction method for power system based on GEP parameter optimization XGboost

Publications (2)

Publication Number Publication Date
CN111340273A true CN111340273A (en) 2020-06-26
CN111340273B CN111340273B (en) 2022-08-26

Family

ID=71186979

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010096168.2A Active CN111340273B (en) 2020-02-17 2020-02-17 Short-term load prediction method for power system based on GEP parameter optimization XGboost

Country Status (1)

Country Link
CN (1) CN111340273B (en)

Cited By (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111931992A (en) * 2020-07-16 2020-11-13 广东电网有限责任公司 Power load prediction index selection method and device
CN111950622A (en) * 2020-08-10 2020-11-17 中国平安人寿保险股份有限公司 Behavior prediction method, behavior prediction device, behavior prediction terminal and storage medium based on artificial intelligence
CN111985678A (en) * 2020-07-06 2020-11-24 上海交通大学 Photovoltaic power short-term prediction method
CN112330050A (en) * 2020-11-20 2021-02-05 国网辽宁省电力有限公司营口供电公司 Power system load prediction method considering multiple features based on double-layer XGboost
CN112465250A (en) * 2020-12-08 2021-03-09 深圳供电局有限公司 Power load prediction method, power load prediction device, computer equipment and storage medium
CN112614010A (en) * 2020-12-07 2021-04-06 国网北京市电力公司 Load prediction method and device, storage medium and electronic device
CN112669173A (en) * 2021-01-07 2021-04-16 云南电网有限责任公司电力科学研究院 Short-term load prediction method based on multi-granularity features and XGboost model
CN112766569A (en) * 2021-01-19 2021-05-07 国网能源研究院有限公司 Integrated power consumption increase prediction method and device integrating high-frequency influence factors
CN113204857A (en) * 2021-03-15 2021-08-03 北京锐达芯集成电路设计有限责任公司 Method for predicting residual life of electronic device based on extreme gradient lifting tree algorithm
CN113239946A (en) * 2021-02-02 2021-08-10 广东工业大学 Method for checking current-carrying capacity of power transmission line
CN113807568A (en) * 2021-08-11 2021-12-17 南方电网科学研究院有限责任公司 Power load prediction method and device and terminal equipment
CN113947201A (en) * 2021-08-02 2022-01-18 国家电投集团电站运营技术（北京）有限公司 Training method and device for power decomposition curve prediction model and storage medium
CN114091782A (en) * 2021-11-30 2022-02-25 国网湖南省电力有限公司 Medium-and-long-term power load prediction method
CN114169595A (en) * 2021-11-25 2022-03-11 中国海洋大学 Learning situation prediction method based on big data
CN114595873A (en) * 2022-02-23 2022-06-07 安徽理工大学 Grey-correlation-based DA-LSTM short-term power load prediction method
CN114898818A (en) * 2022-04-06 2022-08-12 中国石油大学(北京) Mixed crude oil condensation point prediction model training method and device and application method
CN114912720A (en) * 2022-07-15 2022-08-16 石家庄科林电气股份有限公司 Memory network-based power load prediction method, device, terminal and storage medium
CN115132298A (en) * 2021-03-29 2022-09-30 宝山钢铁股份有限公司 Spring steel decarburization analysis and prediction method
CN115688588A (en) * 2022-11-04 2023-02-03 自然资源部第一海洋研究所 Sea surface temperature daily change amplitude prediction method based on improved XGB method
CN116760033A (en) * 2023-08-21 2023-09-15 南京博网软件科技有限公司 Real-time power demand prediction system based on artificial intelligence
CN116881854A (en) * 2023-09-08 2023-10-13 国际关系学院 XGBoost-fused time sequence prediction method for calculating feature weights
CN117236800A (en) * 2023-11-14 2023-12-15 齐信科技(北京)有限公司 Multi-scene self-adaptive electricity load prediction method and system
CN117370770A (en) * 2023-12-08 2024-01-09 江苏米特物联网科技有限公司 Hotel load comprehensive prediction method based on shape-XGboost

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108416366A (en) * 2018-02-06 2018-08-17 武汉大学 A kind of power-system short-term load forecasting method of the weighting LS-SVM based on Meteorological Index
CN108830418A (en) * 2018-06-14 2018-11-16 东北电力大学 A kind of Short-Term Load Forecasting Method
CN110516831A (en) * 2019-06-18 2019-11-29 国网（北京）节能设计研究院有限公司 A kind of short-term load forecasting method based on MWOA algorithm optimization SVM
CN110728401A (en) * 2019-10-10 2020-01-24 郑州轻工业学院 Short-term power load prediction method of neural network based on squirrel and weed hybrid algorithm

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108416366A (en) * 2018-02-06 2018-08-17 武汉大学 A kind of power-system short-term load forecasting method of the weighting LS-SVM based on Meteorological Index
CN108830418A (en) * 2018-06-14 2018-11-16 东北电力大学 A kind of Short-Term Load Forecasting Method
CN110516831A (en) * 2019-06-18 2019-11-29 国网（北京）节能设计研究院有限公司 A kind of short-term load forecasting method based on MWOA algorithm optimization SVM
CN110728401A (en) * 2019-10-10 2020-01-24 郑州轻工业学院 Short-term power load prediction method of neural network based on squirrel and weed hybrid algorithm

Cited By (32)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111985678A (en) * 2020-07-06 2020-11-24 上海交通大学 Photovoltaic power short-term prediction method
CN111931992A (en) * 2020-07-16 2020-11-13 广东电网有限责任公司 Power load prediction index selection method and device
CN111950622B (en) * 2020-08-10 2023-08-15 中国平安人寿保险股份有限公司 Behavior prediction method, device, terminal and storage medium based on artificial intelligence
CN111950622A (en) * 2020-08-10 2020-11-17 中国平安人寿保险股份有限公司 Behavior prediction method, behavior prediction device, behavior prediction terminal and storage medium based on artificial intelligence
CN112330050A (en) * 2020-11-20 2021-02-05 国网辽宁省电力有限公司营口供电公司 Power system load prediction method considering multiple features based on double-layer XGboost
CN112614010A (en) * 2020-12-07 2021-04-06 国网北京市电力公司 Load prediction method and device, storage medium and electronic device
CN112465250A (en) * 2020-12-08 2021-03-09 深圳供电局有限公司 Power load prediction method, power load prediction device, computer equipment and storage medium
CN112669173A (en) * 2021-01-07 2021-04-16 云南电网有限责任公司电力科学研究院 Short-term load prediction method based on multi-granularity features and XGboost model
CN112766569A (en) * 2021-01-19 2021-05-07 国网能源研究院有限公司 Integrated power consumption increase prediction method and device integrating high-frequency influence factors
CN113239946A (en) * 2021-02-02 2021-08-10 广东工业大学 Method for checking current-carrying capacity of power transmission line
CN113239946B (en) * 2021-02-02 2023-10-27 广东工业大学 Checking method for current-carrying capacity of power transmission line
CN113204857A (en) * 2021-03-15 2021-08-03 北京锐达芯集成电路设计有限责任公司 Method for predicting residual life of electronic device based on extreme gradient lifting tree algorithm
CN115132298A (en) * 2021-03-29 2022-09-30 宝山钢铁股份有限公司 Spring steel decarburization analysis and prediction method
CN113947201A (en) * 2021-08-02 2022-01-18 国家电投集团电站运营技术（北京）有限公司 Training method and device for power decomposition curve prediction model and storage medium
CN113807568A (en) * 2021-08-11 2021-12-17 南方电网科学研究院有限责任公司 Power load prediction method and device and terminal equipment
CN114169595A (en) * 2021-11-25 2022-03-11 中国海洋大学 Learning situation prediction method based on big data
CN114169595B (en) * 2021-11-25 2024-08-02 中国海洋大学 Emotion prediction method based on big data
CN114091782A (en) * 2021-11-30 2022-02-25 国网湖南省电力有限公司 Medium-and-long-term power load prediction method
CN114091782B (en) * 2021-11-30 2024-06-07 国网湖南省电力有限公司 Medium-long term power load prediction method
CN114595873A (en) * 2022-02-23 2022-06-07 安徽理工大学 Grey-correlation-based DA-LSTM short-term power load prediction method
CN114595873B (en) * 2022-02-23 2024-07-19 安徽理工大学 Gray correlation-based DA-LSTM short-term power load prediction method
CN114898818A (en) * 2022-04-06 2022-08-12 中国石油大学(北京) Mixed crude oil condensation point prediction model training method and device and application method
CN114912720A (en) * 2022-07-15 2022-08-16 石家庄科林电气股份有限公司 Memory network-based power load prediction method, device, terminal and storage medium
CN115688588A (en) * 2022-11-04 2023-02-03 自然资源部第一海洋研究所 Sea surface temperature daily change amplitude prediction method based on improved XGB method
CN116760033B (en) * 2023-08-21 2024-04-12 南京博网软件科技有限公司 Real-time power demand prediction system based on artificial intelligence
CN116760033A (en) * 2023-08-21 2023-09-15 南京博网软件科技有限公司 Real-time power demand prediction system based on artificial intelligence
CN116881854B (en) * 2023-09-08 2023-12-22 国际关系学院 XGBoost-fused time sequence prediction method for calculating feature weights
CN116881854A (en) * 2023-09-08 2023-10-13 国际关系学院 XGBoost-fused time sequence prediction method for calculating feature weights
CN117236800A (en) * 2023-11-14 2023-12-15 齐信科技(北京)有限公司 Multi-scene self-adaptive electricity load prediction method and system
CN117236800B (en) * 2023-11-14 2024-02-20 齐信科技(北京)有限公司 Multi-scene self-adaptive electricity load prediction method and system
CN117370770A (en) * 2023-12-08 2024-01-09 江苏米特物联网科技有限公司 Hotel load comprehensive prediction method based on shape-XGboost
CN117370770B (en) * 2023-12-08 2024-02-13 江苏米特物联网科技有限公司 Hotel load comprehensive prediction method based on shape-XGboost

Also Published As

Publication number Publication date
CN111340273B (en) 2022-08-26

Similar Documents

Publication Publication Date Title
CN111340273B (en) Short-term load prediction method for power system based on GEP parameter optimization XGboost
CN110751318B (en) Ultra-short-term power load prediction method based on IPSO-LSTM
CN114065646B (en) Energy consumption prediction method based on hybrid optimization algorithm, cloud computing platform and system
CN113627070A (en) Short-term photovoltaic power prediction method
CN110738363B (en) Photovoltaic power generation power prediction method
CN114595884A (en) Genetic intelligent optimization neural network wind power generation equipment temperature prediction method
CN113052373A (en) Monthly runoff change trend prediction method based on improved ELM model
CN116187835A (en) Data-driven-based method and system for estimating theoretical line loss interval of transformer area
CN116646929A (en) PSO-CNN-BILSTM-based short-term wind power prediction method
CN115481788A (en) Load prediction method and system for phase change energy storage system
CN114093426B (en) Marker screening method based on gene regulation network construction
Abdelaziz et al. Convolutional neural network with genetic algorithm for predicting energy consumption in public buildings
Zhang et al. Error-feedback stochastic modeling strategy for time series forecasting with convolutional neural networks
Sun Real estate evaluation model based on genetic algorithm optimized neural network
Wu et al. Hybrid intelligent deep kernel incremental extreme learning machine based on differential evolution and multiple population grey wolf optimization methods
CN112183721B (en) Construction method of combined hydrological prediction model based on self-adaptive differential evolution
CN116245259B (en) Photovoltaic power generation prediction method and device based on depth feature selection and electronic equipment
CN117313795A (en) Intelligent building energy consumption prediction method based on improved DBO-LSTM
CN111563614A (en) Load prediction method based on adaptive neural network and TLBO algorithm
Rifa'i et al. Optimized fuzzy backpropagation neural network using genetic algorithm for predicting Indonesian stock exchange composite index
CN115759343A (en) E-LSTM-based user electric quantity prediction method and device
CN113420492A (en) Modeling method for frequency response model of wind-solar-fire coupling system based on GAN and GRU neural network
Luo et al. Prediction of the Stock Adjusted Closing Price Based On Improved PSO-LSTM Neural Network
He et al. Application of neural network model based on combination of fuzzy classification and input selection in short term load forecasting
CN113675844B (en) Multi-node power load prediction method

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
GR01 Patent grant
GR01 Patent grant