CN108171379B - Power load prediction method - Google Patents

Power load prediction method Download PDF

Info

Publication number
CN108171379B
CN108171379B CN201711464222.9A CN201711464222A CN108171379B CN 108171379 B CN108171379 B CN 108171379B CN 201711464222 A CN201711464222 A CN 201711464222A CN 108171379 B CN108171379 B CN 108171379B
Authority
CN
China
Prior art keywords
model
regression
data
output
variable
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.)
Active
Application number
CN201711464222.9A
Other languages
Chinese (zh)
Other versions
CN108171379A (en
Inventor
王椭
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.)
Wuxi Yingzhen Technology Co ltd
Original Assignee
Wuxi Yingzhen Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Wuxi Yingzhen Technology Co ltd filed Critical Wuxi Yingzhen Technology Co ltd
Priority to CN201711464222.9A priority Critical patent/CN108171379B/en
Publication of CN108171379A publication Critical patent/CN108171379A/en
Application granted granted Critical
Publication of CN108171379B publication Critical patent/CN108171379B/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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply

Landscapes

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

Abstract

The invention discloses an electric load forecasting method, which comprises the following steps: collecting historical data, collecting power utilization data of the power utilization unit for as long as possible in the past, and storing the power utilization data in a time sequence; secondly, time series conversion is carried out, and historical data are divided into an input data set and an output data set; training the model, namely training a training set by respectively using a linear regression model and a Boost regression model, wherein independent variables of the regression model comprise dummy variables of month, day, week, time and the like; checking the models and screening variables, and selecting 1 model with higher precision from the linear regression model and the Boost regression model by using the accuracy of the new data checking model; and step five, predicting subsequent new data by using the selected model and the selected variable.

Description

Power load prediction method
Technical Field
The invention relates to the technical field of power consumption prediction, in particular to a power consumption load prediction method based on a regression model and a Boost regression model.
Background
The power load prediction is to analyze the supply and demand of power market by power generation enterprise marketers and master the load condition of the power market. Because the power generation enterprises and the power consumers are related through the power grid, and one power grid simultaneously comprises a plurality of power generation enterprises, when a certain power generation consumer predicts the load of the power market consumer, the grasped information is limited, and analysis can be performed only according to the historical information of the power market of the whole power grid, which is released by related organizations.
At present, there are two main types of methods for predicting the electrical load: linear models such as time series models, and machine learning models such as artificial neural network models. Although the artificial neural network model can comprehensively consider all factors influencing load change, the established model parameters are more, the logic structure among variables is more complex, the defects of easy falling into local minimum, large operand and the like exist, and how to combine the advantages of two models by using less operand is the problem to be solved at present.
Disclosure of Invention
The invention aims to provide an electricity load prediction method aiming at the defects in the prior art, which can improve the prediction precision and the operation speed of the existing model through historical data and effectively predict the future electricity consumption of an electricity unit.
In order to achieve the purpose, the invention adopts the following technical scheme: an electrical load prediction method comprises the following steps:
step one, collecting historical data, collecting power utilization data of the power utilization unit for as long as possible in the past, and storing the power utilization data in a time sequence;
step two, time series conversion is carried out, historical data are divided into an input data set and an output data set, the output of each sample is the power consumption of a power consumption unit at a certain moment, and the input is the power consumption power value of all points 21 days before the point; if a longer interval time (such as more than 1 day) needs to be predicted, the time distance between the input and the output also adopts a corresponding time interval;
training the model, namely training a training set by respectively using a linear regression model and a Boost regression model, wherein independent variables of the regression model comprise dummy variables of month, day, week, time and the like;
checking the model and screening variables, checking the accuracy of the model by using new data, selecting 1 with higher precision from the linear regression model and the Boost regression model, calculating an importance list of each variable, and selecting the variable with higher importance;
step five, predicting subsequent new data by using the selected model and the selected variable; and when the prediction deviation is found to be large, repeating the third step and the fourth step.
Preferably, the expression of the linear regression model in step three is as shown in formula (1):
yi＝b1×yi-1+b2×yi-2+…+bn×yi-n+c1×x1i+c2×x2i+…+cm×xmi (1)
wherein, yiIs the load value, x, at time i1i，…，xmiDummy variables generated for time variables such as month, week, day, hour, and minute, b1，…，bn，c1，…，cmAs a parameter, eiFor residual errors, the model is solved using a least squares method.
Preferably, the algorithm of the Boost regression model in step three is as follows:
s31: from the ensemble of samples D, non-replaced random samples n1If the number of samples is less than n, obtaining a set D1 used for training a weak regression classifier C1;
s32: extracting n from the sample ensemble set D2Combining half of samples which are classified wrongly by C1 to obtain a sample set D2 for training a weak regression classifier C2;
s33: repeating the steps of S1 and S2 to obtain a set of n classifiers (C1, C2, … and Cn), and taking the weighted average of the output results of all regression classifiers as output when the data to be identified enters;
assuming that the set of n classifiers is (C1, C2, …, Cn), the prediction result of each classifier is P1，P2，…，Pn-1，PnAnd then, the output result of the electric appliance A to be identified is as follows:
Figure BDA0001530808280000021
preferably, a regression tree model is derived through the weak regression classifier, and the algorithm is as follows:
s41: selecting a maximum entropy-increasing variable according to i ═ argmax IG (xi), wherein IG (xi) is an information gain function;
s42: generating a multi-classification decision by taking the value of the variable i as a classification standard and corresponding to the interval of the output value;
s43: and repeating the steps S41 and S42 until the significant entropy-increasing variables cannot be selected, wherein the sequence of all the classification rules is a regression tree.
Preferably, the regression tree model has the following algorithm:
s51: in an input space where a training data set D is located, recursively dividing each region into two sub-regions, determining an output value on each sub-region, constructing a binary decision tree, selecting an optimal segmentation variable j and a segmentation point s, and solving a formula (2):
Figure BDA0001530808280000022
segmenting a variable j, scanning a segmentation point s for the fixed segmentation variable j, and selecting to enable the formula to reach a minimum value pair (j, s);
s52: dividing the region by the selected pair (j, s) and determining the corresponding output value, as in equation (3):
R1(j，s)＝{x|xj≤s}，R2(j，s)＝{x|xj＞s}
Figure BDA0001530808280000031
s53: continuing to call the steps S51 and S52 for the two sub-regions until the stop condition is met;
s54: dividing an input space into M regions R1，R2，…，RmGenerating a decision regression tree formula (4):
Figure BDA0001530808280000032
and obtaining an output f (x) through a regression tree model, wherein the output f (x) is the predicted value of the electric load model.
The invention achieves the following beneficial effects: the invention provides a power load prediction method based on a Boost regression model, which can improve the prediction precision and the calculation speed of the existing model through historical data, effectively predict the future power consumption of a power consumption unit, and the model obtained by the invention has higher prediction precision and lower calculation complexity, is easy to perform programmed automation processing, and can be quickly connected with the actual demand.
Drawings
FIG. 1 is a flow chart of a method for predicting an electrical load according to the present invention.
Detailed Description
The following further describes the embodiments of the present invention with reference to the drawings.
As shown in fig. 1, the technical solution adopted by the present invention is: an electrical load prediction method comprises the following steps:
step one, collecting historical data, collecting power utilization data of the power utilization unit for as long as possible in the past, and storing the data in a time sequence, wherein for the power utilization unit MT1, the data time is divided from 2011 1 month to 2011 3 months at a fixed time interval delta t;
step two, time series conversion is carried out, historical data are divided into an input data set and an output data set, the output of each sample is the power consumption of a power consumption unit at a certain moment, and the input is the power consumption power value of all points 21 days before the point; if a longer interval time (such as more than 1 day) needs to be predicted, the time distance between the input and the output also adopts a corresponding time interval;
the unit of electricity in the present invention is represented by MT1, and the historical data is represented by D1，D2，D3，…Dn-1，Dn,DiA set of power data representing the power usage on day i,
Figure BDA0001530808280000041
the power consumption at the j time of the ith day is the time interval delta t, and the time interval delta t is 15 minutes;
the power utilization unit MT1 outputs data set as:
Figure BDA0001530808280000042
the input data is: di-21，Di-20，…，Di-2，Di-1。
Values for 21 days: if the electricity consumption power is predicted in 2011 in the early morning of No. 1/22, the input data is No. 1-21, and the electricity consumption power value is calculated every 15 minutes.
Training the model, namely training a training set by respectively using a linear regression model and a Boost regression model, wherein independent variables of the regression model comprise dummy variables of month, day, week, time and the like; the number of months, days, weeks, and times, where the number of hours refers to hours and minutes of the day, here 16 hours and 15 minutes, are changed to dummy variables according to the date format of the data, such as 2015/11/14: 15.
And step four, checking the model and screening variables, checking the accuracy of the model by using new data, selecting 1 with higher precision from the linear regression model and the Boost regression model, calculating an importance list of each variable, and selecting the variable with higher importance.
The model is tested by performing 1-fold cross validation on the selected model, namely a linear regression model, a random forest model and a boost regression model, and finally rejecting the poor model according to a scoring criterion MAPE. And the variable screening is to select a model with high prediction precision from the rest models according to the prediction result of the new data, and calculate the importance of the variable according to the prediction result of the model. The model we finally choose here is the boost model, so the computation of feature importance based on the tree model is as follows:
the global importance of feature j is measured by the average of the importance of feature j in a single tree:
Figure BDA0001530808280000043
where M is the number of trees.
The importance of feature j in a single tree is as follows:
Figure BDA0001530808280000044
wherein L is the leaf node number of the tree, L-1 is the non-leaf node number of the tree (the constructed trees are all binary trees with left and right children), vtIs a feature associated with the node t,
Figure BDA0001530808280000045
is the reduction in the square penalty after splitting of the node t.
Scoring criteria: the minimum is selected according to the median of MPAE calculated for each electricity unit by each model.
Figure BDA0001530808280000051
y' is a predicted value, y is a true value, and n is a sample size.
Step five, predicting subsequent new data by using the selected model and the selected variable; and when the prediction deviation is found to be large, repeating the third step and the fourth step.
The explicitly selected model is a boost regression model, the variables of the model are different because of the selected time interval, and thus the determined variables are different, for example, the time interval is 15 minutes, the input variables are the power consumption power every 15 minutes in the first 21 days of the predicted point, and the variables are selected from the input variables according to the variable importance (each variable has a value, the variable is determined according to the selected threshold value, here we select the variable > 20).
When the deviation is large: also determined according to a defined threshold, such as defining MAPE mean < 10%.
Preferably, the expression of the linear regression model in step three is as shown in formula (1):
yi＝b1×yi-1+b2×yi-2+…+bn×yi-n+c1×x1i+c2×x2i+…+cm×xmi (1)
wherein, yiIs the load value, x, at time i1i，…，xmiDummy variables generated for time variables such as month, week, day, hour, and minute, b1，…，bn，c1，…，cmAs a parameter, eiFor residual errors, the model is solved using a least squares method.
The square loss function:
Figure BDA0001530808280000052
determination of variable parameter b by minimizing the solution of the quadratic loss function1，…，bn，c1，…，cmMinimize we can byAnd obtaining the extreme value of the function. By solving the system of equations (5), we can obtain the parameter bi，cj，i＝1，2，…，n，j＝1，2，…，m。
Figure BDA0001530808280000053
Preferably, the algorithm of the Boost regression model in step three is as follows:
s31: from the ensemble of samples D, non-replaced random samples n1If the number of samples is less than n, obtaining a set D1 used for training a weak regression classifier C1;
s32: extracting n from the sample ensemble set D2Combining half of samples which are classified wrongly by C1 to obtain a sample set D2 for training a weak regression classifier C2;
s33: repeating the steps of S1 and S2 to obtain a set of n classifiers (C1, C2, … and Cn), and taking the weighted average of the output results of all regression classifiers as output when the data to be identified enters;
assuming that the set of n classifiers is (C1, C2, …, Cn), the prediction result of each classifier is P1，P2，…，Pn-1，PnAnd then, the output result of the electric appliance A to be identified is as follows:
Figure BDA0001530808280000064
preferably, a regression tree model is derived through the weak regression classifier, and the algorithm is as follows:
s41: selecting a maximum entropy-increasing variable according to i ═ argmax IG (xi), wherein IG (xi) is an information gain function;
s42: generating a multi-classification decision by taking the value of the variable i as a classification standard and corresponding to the interval of the output value;
s43: and repeating the steps S41 and S42 until the significant entropy-increasing variables cannot be selected, wherein the sequence of all the classification rules is a regression tree.
Preferably, the regression tree model has the following algorithm:
s51: in an input space where a training data set D is located, recursively dividing each region into two sub-regions, determining an output value on each sub-region, constructing a binary decision tree, selecting an optimal segmentation variable j and a segmentation point s, and solving a formula (2):
Figure BDA0001530808280000061
segmenting a variable j, scanning a segmentation point s for the fixed segmentation variable j, and selecting to enable the formula to reach a minimum value pair (j, s);
s52: dividing the region by the selected pair (j, s) and determining the corresponding output value, as in equation (3):
R1(j，s)＝{x|xj≤s}，R2(j，s)＝{x|xj＞s}
Figure BDA0001530808280000062
s53: continuing to call the steps S51 and S52 for the two sub-regions until the stop condition is met;
s54: dividing an input space into M regions R1，R2，…，RmGenerating a decision regression tree formula (4):
Figure BDA0001530808280000063
and obtaining an output f (x) through a regression tree model, wherein the output f (x) is the predicted value of the electric load model.
And (3) verifying the experimental results, verifying the results by adopting linear regression, RNN, LSTM, MLP, RBF and ELMAN models, and predicting the results at 15 minutes, 1 hour and 4 hours.
The method comprises the steps of total 140256 samples of source data, 15min of resolution, 135000 of samples as a training set, 5256 of samples as a testing set, selecting the last sample of the training set as the training set for a model with overlarge operation amount, adjusting parameters and recording an optimal result.
(1) Prediction (15 min) for the previous 1 step, the results are shown in table 1:
(R is speed, H is hidden number, E is day number, M is maximum iteration)
Figure BDA0001530808280000071
Table 1: training and prediction values for each model at 15 minutes
(2) The prediction was carried out 4 steps forward (1 hour), and the results are shown in Table 2:
(R is speed, H is hidden number, E is day number, M is maximum iteration)
Figure BDA0001530808280000081
Table 2: training and prediction values for 1 hour models
(3) Prediction (4 hours) for the 16 forward steps, the results are shown in table 3:
(R is speed, H is hidden number, E is day number, M is maximum iteration)
Figure BDA0001530808280000082
Table 3: and (3) evaluating the training values and the predicted values of the models in 1 hour:
(1) for units with larger electricity consumption, the prediction accuracy is obviously higher than that of the previous household electricity consumption data.
(2) The error of multi-step prediction is higher than that of single-step prediction.
(3) The MLP effect was best in the experiment. Average error percentage: 15 minutes 6.16%, 1 hour 9.61%, 4 hours 16.82%.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (3)

1. An electrical load prediction method is characterized by comprising the following steps:
collecting historical data, collecting power utilization data of a power utilization unit for a long time in the past, and storing the data in a time sequence;
step two, time series conversion is carried out, historical data are divided into an input data set and an output data set, the output of each sample is the power consumption of a power consumption unit at a certain moment, and the input is the power consumption power value of all points 21 days before the moment; if longer interval time needs to be predicted, the time distance between input and output also adopts corresponding time interval;
training the model, namely training a training set by respectively using a linear regression model and a Boost regression model, wherein independent variables of the regression model comprise dummy variables of month, day, week and time;
the expression of the linear regression model in the third step is shown as the formula (1):
yi＝b1×yi-1+b2×yi-2+…+bn×yi-b+c1×x1i+c2×x2i+…+cm×xmi+ei(1)
wherein, yiIs the load value, x, at time i1i，…，xmiDummy variables generated for time variables of month, week, day, hour, minute, b1，…，bn，c1，…，cmAs a parameter, eiFor residual errors, the model is solved by a least square method; the algorithm of the Boost regression model in the third step is as follows:
s31: from the ensemble of samples D, non-replaced random samples n1If the number of samples is less than n, obtaining a set D1 used for training a weak regression classifier C1;
s32: extracting n from the sample ensemble set D2< n samples, of which half the samples classified erroneously by C1 were merged,
obtaining a sample set D2 for training a weak regression classifier C2;
s33: repeating the steps of S1 and S2 to obtain a set of n classifiers (C1, C2, … and Cn), when the data to be identified enters,
using the weighted average of the output results of all regression classifiers as output;
assuming that the set of n classifiers is (C1, C2, …, Cn), the prediction result of each classifier is P1，P2，…，Pn-1，Pn，
The output result of the electric appliance A to be identified is as follows:
Figure FDA0003197395840000011
checking the model and screening variables, checking the accuracy of the model by using new data, and performing linear regression on the model and Boost
Selecting 1 with higher precision from the regression model, calculating an importance list of each variable, and selecting the variable with high importance;
step five, predicting subsequent new data by using the selected model and the selected variable; and when the prediction deviation is found to be large, repeating the third step and the fourth step.
2. The method of claim 1, wherein the weak regression classifier is used to derive a regression tree model, and the algorithm is as follows:
s41: selecting a maximum entropy-increasing variable according to i ═ argmax IG (xi), wherein IG (xi) is an information gain function;
s42: generating a multi-classification decision by taking the value of the variable i as a classification standard and corresponding to the interval of the output value;
s43: and repeating the steps S41 and S42 until the significant entropy-increasing variables cannot be selected, wherein the sequence of all the classification rules is a regression tree.
3. The method of claim 2, wherein the regression tree model is calculated as follows:
s51: in an input space where a training data set D is located, recursively dividing each region into two sub-regions, determining an output value on each sub-region, constructing a binary decision tree, selecting an optimal segmentation variable j and a segmentation point s, and solving a formula (2):
Figure FDA0003197395840000021
segmenting a variable j, scanning a segmentation point s for the fixed segmentation variable j, and selecting to enable the formula to reach a minimum value pair (j, s);
s52: dividing the region by the selected pair (j, s) and determining the corresponding output value, as in equation (3):
R1(j，s)＝{x|xj≤s}，R2(j，s)＝{x|xj>s}
Figure FDA0003197395840000022
s53: continuing to call the steps S51 and S52 for the two sub-regions until the stop condition is met;
s54: dividing an input space into M regions R1，R2，…，RmGenerating a decision regression tree formula (4):
Figure FDA0003197395840000023
and obtaining an output f (x) through a regression tree model, wherein the output f (x) is the predicted value of the electric load model.
CN201711464222.9A 2017-12-28 2017-12-28 Power load prediction method Active CN108171379B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711464222.9A CN108171379B (en) 2017-12-28 2017-12-28 Power load prediction method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711464222.9A CN108171379B (en) 2017-12-28 2017-12-28 Power load prediction method

Publications (2)

Publication Number Publication Date
CN108171379A CN108171379A (en) 2018-06-15
CN108171379B true CN108171379B (en) 2021-12-17

Family

ID=62519689

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711464222.9A Active CN108171379B (en) 2017-12-28 2017-12-28 Power load prediction method

Country Status (1)

Country Link
CN (1) CN108171379B (en)

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109214601A (en) * 2018-10-31 2019-01-15 四川长虹电器股份有限公司 Household electric appliances big data Method for Sales Forecast method
CN109508820A (en) * 2018-11-01 2019-03-22 上海理工大学 Campus electricity demand forecasting modeling method based on differentiation modeling
CN110413657B (en) * 2019-07-11 2021-08-17 东北大学 Average response time evaluation method for seasonal non-stationary concurrency
CN110889544B (en) * 2019-11-20 2022-07-01 贵州电网有限责任公司电力科学研究院 Method and device for predicting operation indexes of power distribution network
CN112232985B (en) * 2020-10-15 2023-02-28 国网天津市电力公司 Power distribution and utilization data monitoring method and device for ubiquitous power Internet of things
CN112508301B (en) * 2020-12-21 2024-05-17 北京梧桐车联科技有限责任公司 Method, device and storage medium for predicting charge load of electric vehicle
WO2022246627A1 (en) * 2021-05-25 2022-12-01 罗伯特·博世有限公司 Method and apparatus for controlling refrigerating device
CN114842921A (en) * 2022-04-27 2022-08-02 江苏集萃工业过程模拟与优化研究所有限公司 Municipal sewage dephosphorization agent adding method based on linear regression and decision tree model
CN116109019A (en) * 2023-04-12 2023-05-12 成都英沃信科技有限公司 Method and equipment for predicting daily gas production of gas well based on random forest

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105205565A (en) * 2015-09-30 2015-12-30 成都民航空管科技发展有限公司 Controller workload prediction method and system based on multiple regression model
CN105225193A (en) * 2015-09-30 2016-01-06 中国民用航空总局第二研究所 A kind of method and system of the sector runnability aggregative index based on multiple regression model
CN106682764A (en) * 2016-12-02 2017-05-17 河海大学 Method for predicting other day air-conditioning load of public building based on parallel prediction strategy
CN107220851A (en) * 2017-05-25 2017-09-29 北京中电普华信息技术有限公司 Electricity sales amount Forecasting Methodology and device based on X13 seasonal adjustments and Cox regression

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20140043184A (en) * 2012-09-28 2014-04-08 한국전자통신연구원 Apparatus and method for forecasting an energy comsumption

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105205565A (en) * 2015-09-30 2015-12-30 成都民航空管科技发展有限公司 Controller workload prediction method and system based on multiple regression model
CN105225193A (en) * 2015-09-30 2016-01-06 中国民用航空总局第二研究所 A kind of method and system of the sector runnability aggregative index based on multiple regression model
CN106682764A (en) * 2016-12-02 2017-05-17 河海大学 Method for predicting other day air-conditioning load of public building based on parallel prediction strategy
CN107220851A (en) * 2017-05-25 2017-09-29 北京中电普华信息技术有限公司 Electricity sales amount Forecasting Methodology and device based on X13 seasonal adjustments and Cox regression

Also Published As

Publication number Publication date
CN108171379A (en) 2018-06-15

Similar Documents

Publication Publication Date Title
CN108171379B (en) Power load prediction method
US10606862B2 (en) Method and apparatus for data processing in data modeling
CN110163429B (en) Short-term load prediction method based on similarity day optimization screening
CN109726865A (en) Customer charge probability density prediction technique, device and storage medium based on EMD-QRF
CN111754024B (en) Prediction method and system of power industry time series based on regression analysis
CN110443417A (en) Multi-model integrated load prediction method based on wavelet transformation
Lin et al. Approximate mixed-integer programming solution with machine learning technique and linear programming relaxation
CN114169434A (en) Load prediction method
CN113344288A (en) Method and device for predicting water level of cascade hydropower station group and computer readable storage medium
CN115470862A (en) Dynamic self-adaptive load prediction model combination method
Fonseca et al. Unsupervised load shape clustering for urban building performance assessment
CN113159441A (en) Prediction method and device for implementation condition of banking business project
CN109840536A (en) A kind of power grid power supply reliability horizontal clustering method and system
CN113762591A (en) Short-term electric quantity prediction method and system based on GRU and multi-core SVM counterstudy
CN115759343A (en) E-LSTM-based user electric quantity prediction method and device
WO2022222230A1 (en) Indicator prediction method and apparatus based on machine learning, and device and storage medium
CN115358473A (en) Power load prediction method and prediction system based on deep learning
CN115130924A (en) Microgrid power equipment asset evaluation method and system under source grid storage background
CN112801415A (en) Ultra-short-term load prediction method and system based on Markov chain distribution model
Loseva et al. Ensembles of neural networks with application of multi-objective self-configurable genetic programming
Kuvayskova et al. Forecasting the Technical State of an Object Based on the Composition of Machine Learning Methods
CN113094636A (en) Interference user harmonic level estimation method based on massive monitoring data
CN111507777A (en) System model for predicting electricity price based on lightweight gradient lifting algorithm
FOLORUNSO Empirical Comparison of Time Series Data Mining Algorithms
US20220335271A1 (en) Information processing apparatus, information processing method, and non-transitory computer readable medium

Legal Events

Date Code Title Description
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
CP03 Change of name, title or address
CP03 Change of name, title or address

Address after: F4, 200 Linghu Avenue, Xinwu District, Wuxi City, Jiangsu Province, 214000

Patentee after: Wuxi Yingzhen Technology Co.,Ltd.

Address before: 214000 Tianan smart city 2-405, 406, New District, Wuxi City, Jiangsu Province

Patentee before: WUXI YINGZHEN TECHNOLOGY CO.,LTD.