CN110212520A - A kind of power predicating method based on convolutional neural networks - Google Patents

A kind of power predicating method based on convolutional neural networks Download PDF

Info

Publication number
CN110212520A
CN110212520A CN201910441311.4A CN201910441311A CN110212520A CN 110212520 A CN110212520 A CN 110212520A CN 201910441311 A CN201910441311 A CN 201910441311A CN 110212520 A CN110212520 A CN 110212520A
Authority
CN
China
Prior art keywords
neural networks
convolutional neural
data
electricity
power
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201910441311.4A
Other languages
Chinese (zh)
Inventor
贾晓亮
董佳霖
李志�
魏昕喆
张哲宇
刘尧
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.)
State Grid Corp of China SGCC
State Grid Tianjin Electric Power Co Ltd
Original Assignee
State Grid Corp of China SGCC
State Grid Tianjin Electric Power Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by State Grid Corp of China SGCC, State Grid Tianjin Electric Power Co Ltd filed Critical State Grid Corp of China SGCC
Priority to CN201910441311.4A priority Critical patent/CN110212520A/en
Publication of CN110212520A publication Critical patent/CN110212520A/en
Pending legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
    • H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand

Landscapes

  • Engineering & Computer Science (AREA)
  • Power Engineering (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The present invention relates to a kind of power predicating method based on convolutional neural networks, technical characterstic is: the following steps are included: step 1, building power quantity predicting convolutional neural networks；Step 2 pre-processes input layer electricity, temperature, festivals or holidays data, after converting dimensionless relative quantity for above-mentioned three classes data by normalized, by the power quantity predicting convolutional neural networks of mass data input step 1；Step 3, initialization multichannel convolutive neural network weight and biasing；Step 4 successively calculates input data by convolutional neural networks；Step 5, the back-propagation algorithm based on error gradient adjust weight and the biasing of every layer network；Step 6 reaches deconditioning after setting the number of iterations, and input test sample set obtains prediction result.The present invention improves mass data processing efficiency during power quantity predicting, comprehensively considers the related informations such as temperature, and overcomes the problems such as prediction process excessively relies on personal experience, and then can reduce personnel requirement.

Description

A kind of power predicating method based on convolutional neural networks
Technical field
The invention belongs to electric system demand side management technology fields, are related to the side of the anticipation of electric system demand electricity Method, especially a kind of power predicating method based on convolutional neural networks.
Background technique
Currently, power information technology reaches its maturity so that electric power big data application has wide prospect.Electric power data kind Class is more, and data area is wide, runs through entire power generation consumptive link, electric power demand forecasting is to power construction planning, dispatching of power netwoks Control, evaluation of electricity market are significant.Conventional electric power requirement forecasting is carried out by simple information about power, lacks external data pass The analysis of connection property causes to lack the prediction to variation tendency other than fitting rule, and precision of prediction is not high.
Summary of the invention
It is an object of the invention to overcome the deficiencies in the prior art, propose a kind of design rationally, precision of prediction it is high based on The power predicating method of convolutional neural networks.
The present invention solves its realistic problem and adopts the following technical solutions to achieve:
A kind of power predicating method based on convolutional neural networks, comprising the following steps:
Step 1, using convolutional neural networks, for hour electricity, daily electricity, all electricity and temperature, the outside of festivals or holidays Multivariable complex nonlinear mapping relations between data extract electricity proximity, periodicity, tendency feature, construct electricity Predict convolutional neural networks；
Step 2 pre-processes input layer electricity, temperature, festivals or holidays data, by normalized by above-mentioned three classes After data are converted into dimensionless relative quantity, by the power quantity predicting convolutional neural networks of mass data input step 1；
Step 3, initialization multichannel convolutive neural network weight and biasing；
Step 4 successively calculates input data by convolutional neural networks；
Step 5, the back-propagation algorithm based on error gradient adjust weight and the biasing of every layer network；
Step 6 reaches deconditioning after setting the number of iterations, and input test sample set obtains prediction result.
Moreover, the dimensionless number evidence of the step 2 become above-mentioned three classes data after normalization, its calculation formula is:
Wherein:
Y: data value after normalization
ymax=1
ymin=-1
xmax: data maximums before normalizing
xmin: data minimum value before normalizing
Moreover, the step 3 method particularly includes: weight is initialized using Xavier normal distribution:
E (w)=0
Realizing that weight obeys mean value is 0 to be uniformly distributed, wherein E indicates mean value, and Var indicates variance, njIndicate jth layer Node number, nj+1Indicate+1 node layer number of jth；
Moreover, the convolutional layer calculation formula of the step 4 are as follows:
N=1,2 ... C0
In above formula, ymIt is exported for m-th of convolutional layer, a is convolutional layer activation primitive, xjIt is inputted for j-th of channel, wmFor m A convolution kernel, bmFor bias, C0For convolution kernel total number.
The advantages of the present invention:
1, the present invention utilizes convolutional neural networks, outer for hour electricity, daily electricity, all electricity and temperature, festivals or holidays etc. Multivariable complex nonlinear mapping relations between portion's data extract electricity proximity, periodicity, tendency feature, building electricity Amount prediction convolutional neural networks.This method covers the three classes data such as electricity, temperature, festivals or holidays.By normalized by three classes Data are converted into dimensionless relative quantity, and mass data input convolutional neural networks are successively calculated, anti-using error gradient Weight and biasing in every layer network are adjusted to propagation algorithm, reaches deconditioning output power quantity predicting knot after trained the number of iterations Fruit.
2, the present invention constructs the calculation of convolutional neural networks power quantity predicting according to the three classes data such as electricity, temperature, festivals or holidays Method improves mass data processing efficiency during power quantity predicting, comprehensively considers the related informations such as temperature, and overcomes and predicted Journey excessively relies on the problems such as personal experience, and then can reduce personnel requirement, simplifies the course of work.
Detailed description of the invention
Fig. 1 is processing flow schematic diagram of the invention.
Specific embodiment
The embodiment of the present invention is described in further detail below in conjunction with attached drawing:
A kind of power predicating method based on convolutional neural networks, as shown in Figure 1, comprising the following steps:
Step 1, using convolutional neural networks, for hour electricity, daily electricity, all electricity and temperature, the outside of festivals or holidays Multivariable complex nonlinear mapping relations between data extract electricity proximity, periodicity, tendency feature, construct electricity Predict convolutional neural networks；
Step 2 pre-processes input layer electricity, temperature, festivals or holidays data, by normalized by above-mentioned three classes Data are converted into the power quantity predicting convolutional neural networks of mass data input step 1 after dimensionless relative quantity；
In the present embodiment, by initialize input layer electricity, temperature, festivals or holidays data become normalization after dimensionless Data, its calculation formula is:Wherein:
Y: data value after normalization
ymax=1
ymin=-1
xmax: data maximums before normalizing
xmin: data minimum value before normalizing
Step 3, initialization multichannel convolutive neural network weight and biasing, are initialized using Xavier normal distribution and are weighed Value:
E (w)=0
Realizing that weight obeys mean value is 0 to be uniformly distributed, wherein E indicates mean value, and Var indicates variance, njIndicate jth layer Node number, nj+1Indicate+1 node layer number of jth；
Step 4 successively calculates input data by convolutional neural networks, and wherein convolutional layer calculation formula is
N=1,2 ... C0
ymIt is exported for m-th of convolutional layer, a is convolutional layer activation primitive, xjIt is inputted for j-th of channel, wmFor m-th of convolution Core, bmFor bias, C0For convolution kernel total number；
Step 5, the back-propagation algorithm based on error gradient adjust weight and the biasing of every layer network；
Step 6 reaches deconditioning after setting the number of iterations, and input test sample set obtains prediction result.
It is emphasized that embodiment of the present invention be it is illustrative, without being restrictive, therefore the present invention includes It is not limited to embodiment described in specific embodiment, it is all to be obtained according to the technique and scheme of the present invention by those skilled in the art Other embodiments, also belong to the scope of protection of the invention.

Claims (4)

1. a kind of power predicating method based on convolutional neural networks, it is characterised in that: the following steps are included:
Step 1, using convolutional neural networks, for hour electricity, daily electricity, all electricity and temperature, the external data of festivals or holidays Between multivariable complex nonlinear mapping relations, extract electricity proximity, periodicity, tendency feature, construct power quantity predicting Convolutional neural networks；
Step 2 pre-processes input layer electricity, temperature, festivals or holidays data, by normalized by above-mentioned three classes data After being converted into dimensionless relative quantity, by the power quantity predicting convolutional neural networks of mass data input step 1；
Step 3, initialization multichannel convolutive neural network weight and biasing；
Step 4 successively calculates input data by convolutional neural networks；
Step 5, the back-propagation algorithm based on error gradient adjust weight and the biasing of every layer network；
Step 6 reaches deconditioning after setting the number of iterations, and input test sample set obtains prediction result.
2. a kind of power predicating method based on convolutional neural networks according to claim 1, it is characterised in that: the step The rapid 2 dimensionless number evidence become above-mentioned three classes data after normalization, its calculation formula is:
Wherein:
Y: data value after normalization
ymax=1
ymin=-1
xmax: data maximums before normalizing
xmin: data minimum value before normalizing.
3. a kind of power predicating method based on convolutional neural networks according to claim 1, it is characterised in that: the step Rapid 3 method particularly includes: weight is initialized using Xavier normal distribution:
E (w)=0
Realizing that weight obeys mean value is 0 to be uniformly distributed, wherein E indicates mean value, and Var indicates variance, njIndicate jth node layer Number, nj+1Indicate+1 node layer number of jth.
4. a kind of power predicating method based on convolutional neural networks according to claim 1, it is characterised in that: the step Rapid 4 convolutional layer calculation formula are as follows:
N=1,2 ... C0
In above formula, ymIt is exported for m-th of convolutional layer, a is convolutional layer activation primitive, xjIt is inputted for j-th of channel, wmIt is m-th volume Product core, bmFor bias, C0For convolution kernel total number.
CN201910441311.4A 2019-05-24 2019-05-24 A kind of power predicating method based on convolutional neural networks Pending CN110212520A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910441311.4A CN110212520A (en) 2019-05-24 2019-05-24 A kind of power predicating method based on convolutional neural networks

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910441311.4A CN110212520A (en) 2019-05-24 2019-05-24 A kind of power predicating method based on convolutional neural networks

Publications (1)

Publication Number Publication Date
CN110212520A true CN110212520A (en) 2019-09-06

Family

ID=67788621

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910441311.4A Pending CN110212520A (en) 2019-05-24 2019-05-24 A kind of power predicating method based on convolutional neural networks

Country Status (1)

Country Link
CN (1) CN110212520A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112907062A (en) * 2021-02-08 2021-06-04 国网安徽省电力有限公司蚌埠供电公司 Power grid electric quantity prediction method, device, medium and terminal integrating temperature characteristics

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105809264A (en) * 2014-12-29 2016-07-27 西门子公司 Electrical load predicting method and device
CN106934497A (en) * 2017-03-08 2017-07-07 青岛卓迅电子科技有限公司 Wisdom cell power consumption real-time predicting method and device based on deep learning
CN107730059A (en) * 2017-11-29 2018-02-23 成都思晗科技股份有限公司 The method of transformer station's electricity trend prediction analysis based on machine learning

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105809264A (en) * 2014-12-29 2016-07-27 西门子公司 Electrical load predicting method and device
CN106934497A (en) * 2017-03-08 2017-07-07 青岛卓迅电子科技有限公司 Wisdom cell power consumption real-time predicting method and device based on deep learning
CN107730059A (en) * 2017-11-29 2018-02-23 成都思晗科技股份有限公司 The method of transformer station's electricity trend prediction analysis based on machine learning

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
XAVIER GLOROT等: "《Understanding the difficulty of training deep feedforward neural network》", 《APPEARING IN PROCEEDINGS OF THE 13TH INTERNATIONAL CONFERENCE》 *
徐志敏: "《人工神经网络法预测时用电量》", 《中国科技论文在线》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112907062A (en) * 2021-02-08 2021-06-04 国网安徽省电力有限公司蚌埠供电公司 Power grid electric quantity prediction method, device, medium and terminal integrating temperature characteristics
CN112907062B (en) * 2021-02-08 2024-04-02 国网安徽省电力有限公司蚌埠供电公司 Power grid electric quantity prediction method, device, medium and terminal integrating temperature characteristics

Similar Documents

Publication Publication Date Title
Ke et al. Short-term electrical load forecasting method based on stacked auto-encoding and GRU neural network
Xuan et al. A multi-energy load prediction model based on deep multi-task learning and ensemble approach for regional integrated energy systems
Succetti et al. Deep neural networks for multivariate prediction of photovoltaic power time series
Rai et al. Analysis of classical and machine learning based short-term and mid-term load forecasting for smart grid
CN113554466B (en) Short-term electricity consumption prediction model construction method, prediction method and device
CN110705743A (en) New energy consumption electric quantity prediction method based on long-term and short-term memory neural network
Wang et al. Cloud feature extraction and fluctuation pattern recognition based ultrashort-term regional PV power forecasting
CN112149890A (en) Comprehensive energy load prediction method and system based on user energy label
Liu et al. Heating load forecasting for combined heat and power plants via strand-based LSTM
Tian et al. An adaptive ensemble predictive strategy for multiple scale electrical energy usages forecasting
Song et al. An indoor temperature prediction framework based on hierarchical attention gated recurrent unit model for energy efficient buildings
CN115358437A (en) Power supply load prediction method based on convolutional neural network
Huang et al. Improved clustering and deep learning based short-term wind energy forecasting in large-scale wind farms
Wang et al. Multi-criteria comprehensive study on predictive algorithm of heating energy consumption of district heating station based on timeseries processing
Saber et al. IoT based online load forecasting
CN114819395A (en) Industry medium and long term load prediction method based on long and short term memory neural network and support vector regression combination model
CN115545333A (en) Method for predicting load curve of multi-load daily-type power distribution network
CN111091223A (en) Distribution transformer short-term load prediction method based on Internet of things intelligent sensing technology
CN110212520A (en) A kind of power predicating method based on convolutional neural networks
Abdelkader et al. Predicting energy demand peak using M5 model trees
Zhang Short‐Term Power Load Forecasting Based on SAPSO‐CNN‐LSTM Model considering Autocorrelated Errors
CN110852628B (en) Rural medium-long term load prediction method considering development mode influence
Chen et al. Accounting information disclosure and financial crisis beforehand warning based on the artificial neural network
CN116435987A (en) Ultra-short-term wind power prediction method for large-scale wind power clusters
CN109767037A (en) A kind of prediction technique of summer Daily treatment cost

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20190906