CN106779306A - The structure of user's disaggregated model, electric energy efficiency analysis user classification method and device - Google Patents
The structure of user's disaggregated model, electric energy efficiency analysis user classification method and device Download PDFInfo
- Publication number
- CN106779306A CN106779306A CN201611046235.XA CN201611046235A CN106779306A CN 106779306 A CN106779306 A CN 106779306A CN 201611046235 A CN201611046235 A CN 201611046235A CN 106779306 A CN106779306 A CN 106779306A
- Authority
- CN
- China
- Prior art keywords
- user
- training
- model
- neural network
- default
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 20
- 238000012549 training Methods 0.000 claims abstract description 113
- 238000013528 artificial neural network Methods 0.000 claims abstract description 80
- 238000013145 classification model Methods 0.000 claims abstract description 61
- 238000010276 construction Methods 0.000 claims abstract description 36
- 238000005265 energy consumption Methods 0.000 claims description 98
- 230000007935 neutral effect Effects 0.000 claims description 13
- 238000003062 neural network model Methods 0.000 claims description 5
- 210000004218 nerve net Anatomy 0.000 claims description 3
- 230000005611 electricity Effects 0.000 description 10
- 238000012360 testing method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000013459 approach Methods 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000013100 final test Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000009738 saturating Methods 0.000 description 1
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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/80—Management or planning
- Y02P90/82—Energy audits or management systems therefor
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- General Engineering & Computer Science (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Computational Linguistics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Evolutionary Biology (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of construction method and device of user's disaggregated model, and building input includes that three parameters, output include 3 neural network classification models of node;The neural network classification model for building is trained according to default maximum frequency of training, default training objective precision, default learning rate, each layer connection weight neural network classification model, training for being built;According to each layer connection weight for training and the neural network classification model of the structure, user's disaggregated model is determined.Therefore, the user's disaggregated model for being built by the model building method and device is carried out user and classifies and can provide data foundation to formulate energy-saving and emission-reduction scheme.The present invention also provides electric energy efficiency analysis user classification method and the device of a kind of construction method for applying above-mentioned user's disaggregated model and device, and carrying out user by the method and device classifies and can provide data foundation to formulate energy-saving and emission-reduction scheme.
Description
Technical field
The present invention relates to technical field of electric power, more particularly to a kind of user's disaggregated model construction method and device,
A kind of electric energy efficiency analyzes user classification method and device.
Background technology
Artificial neural network (ArtificialNeuralNetworks, ANN) system is appearance after the forties in 20th century.
It is formed by connecting by numerous adjustable connection weights of neuron, with MPP, distributed information storage, it is good
The features such as self-organizing self-learning capability got well.BP (BackPropagation) algorithm is also called error backpropagation algorithm, is people
A kind of learning algorithm of the supervised in artificial neural networks.BP neural network algorithm can approach arbitrary function, base in theory
This structure is made up of nonlinear change unit, with very strong non-linear mapping capability.
At present, the Energy Saving Industry of China designs effective Energy Efficiency Analysis technology and will be helpful to reality also in the stage at initial stage
The emission reduction of existing national energy-saving, the target for making full use of the energy.And the user's classification in electric energy efficiency analysis is for formulating energy-saving and emission-reduction
Scheme tool is of great significance, and can provide data foundation to formulate energy-saving and emission-reduction scheme.
The content of the invention
Based on this, it is necessary to provide a kind of electric energy efficiency that data foundation is provided to formulate energy-saving and emission-reduction scheme and analyze user
Sorting technique and device and build the electric energy efficiency analysis user classification method and device model family disaggregated model structure
Construction method and device.
A kind of construction method of user's disaggregated model, including:
Building input includes that three parameters, output include 3 neural network classification models of node;Described three of input
Parameter is respectively peak period of the user in preset time period with electric energy consumption, paddy period electric energy consumption and usually section electric energy consumption;
Whether whether 3 nodes of output identify whether as low energy consumption user, be middle energy consumption user and be high energy consumption user respectively;
According to default maximum frequency of training, default training objective precision, default learning rate to build nerve net
Network disaggregated model is trained, each layer connection weight neural network classification model, training for being built;
According to each layer connection weight for training and the neural network classification model of the structure, determine that user classifies
Model.
A kind of construction device of user's disaggregated model, including:
Model construction module, includes that three parameters, output include 3 neural network classification moulds of node for building input
Type;Three parameters of input are respectively peak period of the user in preset time period with electric energy consumption, paddy period electric energy consumption
And usually section electric energy consumption;Output 3 nodes identify whether respectively for low energy consumption user, whether be middle energy consumption user and
Whether it is high energy consumption user;
Model training module, for according to default maximum frequency of training, default training objective precision, default study
Rate is trained to the neural network classification model for building, each layer neural network classification model, training for being built
Connection weight;
Model determining module, for each layer connection weight and the neural network classification of the structure that are trained according to
Model, determines user's disaggregated model.
The model building method and device, building input includes that three parameters, output include 3 neutral nets of node point
Class model;According to default maximum frequency of training, default training objective precision, default learning rate to build neutral net
Disaggregated model is trained, each layer connection weight neural network classification model, training for being built;According to the instruction
Each layer connection weight and the neural network classification model of the structure perfected, determine user's disaggregated model.Therefore, by the mould
User's disaggregated model that type construction method and device build carries out user's classification and can provide data to formulate energy-saving and emission-reduction scheme
Foundation.
A kind of electric energy efficiency analyzes user classification method, including:
Obtain user data to be sorted;The user data to be sorted includes that user uses the peak period in preset time period
Electric energy consumption, paddy period electric energy consumption and usually section electric energy consumption;
User's disaggregated model is built using the construction method of above-mentioned user's disaggregated model;
The input of the user data to be sorted as user's disaggregated model is determined into the user data to be sorted
The type of corresponding user.
A kind of electric energy efficiency analyzes user's sorter, including:
Data acquisition module, for obtaining user data to be sorted;The user data to be sorted includes user default
Peak period in time period is with electric energy consumption, paddy period electric energy consumption and usually section electric energy consumption;
Disaggregated model builds module, for the structure using the user's disaggregated model described in claim 6-9 any one
Device builds user's disaggregated model;
Classification determining module, for the input of the user data to be sorted as user's disaggregated model to be determined into institute
State the type of the corresponding user of user data to be sorted.
The electric energy efficiency analyzes user classification method and device, obtains user data to be sorted;The number of users to be sorted
According to the peak period including user in preset time period with electric energy consumption, paddy period electric energy consumption and usually section electric energy consumption;Utilize
The construction method and device of above-mentioned user's disaggregated model build user's disaggregated model;Using the user data to be sorted as institute
The input for stating user's disaggregated model determines the type of the corresponding user of the user data to be sorted.Therefore, by the method and
Device carries out user's classification and can provide data foundation to formulate energy-saving and emission-reduction scheme.
Brief description of the drawings
Fig. 1 is the flow chart of the construction method of user's disaggregated model of an embodiment;
One particular flow sheet of step of the construction method of user's disaggregated model of Fig. 2 Fig. 1;
The curvilinear motion figure of the BP neural network study in mono- specific embodiment of Fig. 3;
Fig. 4 is the flow chart of the electric energy efficiency analysis user classification method of an embodiment;
Fig. 5 is the structure chart of the construction device of user's disaggregated model of an embodiment;
Fig. 6 is the cellular construction figure of the model training module of the construction device of user's disaggregated model of Fig. 5;
Fig. 7 is the structure chart of electric energy efficiency analysis user's sorter of an embodiment.
Specific embodiment
For the ease of understanding the present invention, the present invention is described more fully below with reference to relevant drawings.In accompanying drawing
Give preferred embodiment of the invention.But, the present invention can be realized in many different forms, however it is not limited to herein
Described embodiment.On the contrary, the purpose for providing these embodiments is to make the understanding to the disclosure more saturating
It is thorough comprehensive.
Unless otherwise defined, all of technologies and scientific terms used here by the article with belong to technical field of the invention
The implication that technical staff is generally understood that is identical.The term for being used in the description of the invention herein is intended merely to description tool
The purpose of the embodiment of body, it is not intended that in the limitation present invention.Term as used herein " and/or " include one or more phases
The arbitrary and all of combination of the Listed Items of pass.
Refer to Fig. 1, the construction method of user's disaggregated model of an embodiment, including:
S110:Building input includes that three parameters, output include 3 neural network classification models of node.The institute of input
Peak period of three parameters respectively user in preset time period is stated with electric energy consumption, paddy period electric energy consumption and usually section electricity consumption
Energy consumption.Whether whether 3 nodes of output identify whether as low energy consumption user, be middle energy consumption user and be high energy consumption respectively
User.
When preset time period can be of 1 year, a season, one month, a ten days, one week, one day, or fixation
Between point to another time point time period.In the present embodiment, preset time period can be divided into multiple shorter short time again
Section.The duration of short time period can be 1 hour, 30 minutes, 10 minutes, 3 minutes, 1 minute, 30 seconds, 15 seconds and other settings
Duration.In these short time periods, according to electricity consumption energy consumption the peak period can be determined with electric energy consumption, paddy period electric energy consumption
And usually section electric energy consumption.Wherein, peak period electric energy consumption can be in preset time period in the most short time period of electricity consumption
Use electric energy consumption, or user's uses electric energy consumption in average power consumption highest short time period;Paddy period electric energy consumption
Can be the use electric energy consumption in preset time period in the minimum short time period of electricity consumption, or when average power consumption is minimum
Short time period in user use electric energy consumption;Usually section electric energy consumption is stable one of the electricity consumption in preset time period in short-term
Between use electric energy consumption in section, or user's uses electric energy consumption in the short time period when average power consumption is average value.Can
With understand ground, in other embodiments, preset time period, peak time section electric energy consumption, paddy period electric energy consumption and usually section use
Electric energy consumption can be defined as needed.
Wherein in one embodiment, it, with the less user of electric energy consumption, such as can be less than average use that low energy consumption user is
The user of the 50% of electric energy consumption;Middle energy consumption user is the user with electric energy consumption medium level, such as can be from averagely using electric energy consumption
50% to twice the average user with electric energy consumption;It, with electric energy consumption user higher, such as can be average that high energy consumption user is
Average user with electric energy consumption of the electricity consumption energy consumption higher than twice.It is to be appreciated that in other embodiments, specific low energy consumption is used
Family, middle energy consumption user, the definition of high energy consumption user can determine as needed.
Wherein in one embodiment, in three parameters of input, each parameter includes 32 nodes.The structure
Neutral net input layer include 96 nodes, hidden layer include 32 nodes.
That is, three parameters that will be input into, are converted into the vector of 96, such as:[11001010 ... 0101110] (totally 96
Position);Output is the vector of 3, respectively [1 0 0], [0 1 0], [0 0 1], and corresponding user's classification is respectively A classes
(low energy consumption electricity consumption user), B classes (middle energy consumption electricity consumption user), C classes (high energy consumption electricity consumption user).
Wherein in one embodiment, neural network classification model is BP neural network.Specifically, corresponding BP god is built
Through network, wherein input layer has 32 nodes, and hidden layer has 15 nodes, and output layer has 3 nodes, i.e. output is individual 3
Vector, correspond to for user to be divided into 3 class users, be respectively low energy consumption user, middle energy consumption user and high energy consumption user.
S130:According to default maximum frequency of training, default training objective precision, default learning rate to build god
It is trained through network class model, each layer connection weight neural network classification model, training for being built.
, it is necessary to according to maximum frequency of training, training objective precision, learning rate pair after neural network classification model is built
The neutral net of structure is trained, so as to obtain each layer connection weight neural network classification model, training of the structure
Value.It should be noted that in the present embodiment, maximum frequency of training, training objective precision, learning rate pre-set.
S150:According to each layer connection weight for training and the neural network classification model of the structure, it is determined that with
Family disaggregated model.
Each layer connection weight that will be trained substitutes into the former lifting network assignment model for building and just can determine that final user
Disaggregated model.User is carried out by user's disaggregated model to classify and can provide data foundation to formulate energy-saving and emission-reduction scheme.
The model building method, building input includes that three parameters, output include 3 neural network classification moulds of node
Type;According to default maximum frequency of training, default training objective precision, default learning rate to build neural network classification
Model is trained, each layer connection weight neural network classification model, training for being built;Trained according to described
Each layer connection weight and the structure neural network classification model, determine user's disaggregated model.Therefore, by the model structure
User's disaggregated model that construction method builds carries out user's classification and can provide data foundation to formulate energy-saving and emission-reduction scheme.
Fig. 2 is referred to, it is described according to default maximum frequency of training, default training objective precision, default learning rate
Neural network classification model to building is trained, and each layer neural network classification model, training for being built connects
The step of connecing weights, including:
S231:The connection weight of each layer of neural network classification model of structure is initialized as non-zero random number, is instructed
Experienced neural network model, and initialize default maximum frequency of training, default training objective precision and default learning rate.
Connection weight (or being weight coefficient) Wij of each layer of neural network classification model to building is initialized, if
A less non-zero random number is set to, so as to the neural network model trained, with the neural network model to the training
It is trained.
Default maximum frequency of training can be 1400, and default training objective precision can be 0.001, default study
Rate can be 0.01.
S233:Receive learning sample successively, calculate the error of each layer of neural network classification model of the training, and according to
The error and the default learning rate of each layer of the neural network classification model of the training are calculated to the training
The connection weight of each layer of neural network classification model is modified.
Each learning sample is received successively.If current input is for P sample, the output of each layer is calculated successively:Oj,Ok。
Wherein, OjIt is j-th output of neuron, O on hidden layerkIt is k-th output of neuron on output layer.And then, according to each layer
Output, calculate each layer error.
After the error of each layer of the neural network classification model of training is calculated, can be according to being calculated
The neural network classification of the error of each layer of the neural network classification model of training and the default learning rate to the training
The connection weight of each layer of model is modified.
S235:When the training each layer of neural network classification model error be less than default training objective precision when or
When the quantity of the learning sample of reception reaches the default maximum frequency of training, the neural network classification mould of the training is determined
Type, each layer connection weight for training.
In a specific embodiment, the default maximum frequency of training is 1400.The default training objective essence
Spend is 0.001.The default learning rate is 0.01.
When training each layer of neural network classification model error be less than 0.001 when, or receive learning sample number
When amount reaches 1400, then terminate training, it is determined that each layer connection weight neural network classification model, training of training;It is no
Then, S233 is continued executing with, the training of a new round is carried out.
In a specific embodiment, the sample data of the Guangzhou power supply administration that will be handled well is to the BP neural network that builds
It is trained, has 2000 groups of sample data, sample data is as follows:
Customs Assigned Number | Power consumption during peak | Power consumption during paddy | Usually power consumption | Class of subscriber |
100001 | 32.65 | 56.46 | 38.76 | A classes |
100002 | 23.89 | 47.51 | 65.23 | B classes |
100003 | 70.34 | 34.68 | 45.37 | C classes |
....... | ....... | ...... | ....... | ....... |
According to training sample, it is 1400 to set default maximum frequency of training, and default training objective precision is 0.001,
Default learning rate is 0.01.The curvilinear motion of BP neural network study is as shown in Figure 3.
According to Fig. 3 as can be seen that reaching default training objective precision when training is to 709 step.Followed by one group
Test experiments are tested the BP neural network that this is trained, and select 1000 groups of test sample data, wherein, A classes user 345
Individual, B classes user 320, C classes user 335 is tested using the BP neural network, and final test result table is as follows:
Can be obtained by test result table, the accuracy rate of user classification results of the neutral net in data test is still
Compare high, the discrimination of wherein B classes user is minimum, error rate has reached the discrimination highest of 11.87%, C class users, reaches
92.2%, user classification results that contrast neutral net is obtained or satisfied, error rate is relatively small, it was demonstrated that should
BP neural network is have feasibility relatively higher in user classifies.In addition, the practicality of the BP neural network is higher.
Fig. 4 is referred to, the present invention also provides a kind of electric energy efficiency point of the construction method for applying above-mentioned user's disaggregated model
Analysis user classification method, including:
S420:Obtain user data to be sorted.
The user data to be sorted includes peak period of the user in preset time period with electric energy consumption, paddy period electric energy
Consumption and usually section electric energy consumption.
S440:The construction method of the above-mentioned user's disaggregated model of exploitation right builds user's disaggregated model.
S460:The input of the user data to be sorted as user's disaggregated model is determined into the user to be sorted
The type of the corresponding user of data.
The electric energy efficiency analyzes user classification method, obtains user data to be sorted;The user data to be sorted includes
Peak period of the user in preset time period is with electric energy consumption, paddy period electric energy consumption and usually section electric energy consumption;Using above-mentioned
The construction method of user's disaggregated model builds user's disaggregated model;Using the user data to be sorted as user classification mould
The input of type determines the type of the corresponding user of the user data to be sorted.Therefore, carrying out user's classification by the method can
Data foundation is provided with to formulate energy-saving and emission-reduction scheme.
Fig. 5 is referred to, the application provides a kind of user's disaggregated model corresponding with the structure of above-mentioned user's disaggregated model
Construction device, including:
Model construction module 510, includes that three parameters, output include 3 neutral nets of node point for building input
Class model.Three parameters of input are respectively peak period of the user in preset time period with electric energy consumption, the electricity consumption of paddy period
Energy consumption and usually section electric energy consumption.Whether 3 nodes of output identify whether as low energy consumption user, are that middle energy consumption is used respectively
Family and whether be high energy consumption user.
Model training module 530, for according to default maximum frequency of training, default training objective precision, default
Learning rate is trained to the neural network classification model for building, built it is neural network classification model, train
Each layer connection weight.
Model determining module 550, for each layer connection weight and the neutral net of the structure that are trained according to
Disaggregated model, determines user's disaggregated model.
The model construction device, building input includes that three parameters, output include 3 neural network classification moulds of node
Type;According to default maximum frequency of training, default training objective precision, default learning rate to build neural network classification
Model is trained, each layer connection weight neural network classification model, training for being built;Trained according to described
Each layer connection weight and the structure neural network classification model, determine user's disaggregated model.Therefore, by the model structure
Build device structure user's disaggregated model carry out user classification can for formulation energy-saving and emission-reduction scheme data foundation be provided.
Refer to Fig. 6, the model training module, including:
Initialization unit 631, for the connection weight of each layer of neural network classification model of structure to be initialized as into non-zero
Random number, the neural network model trained, and initialize default maximum frequency of training, default training objective precision and
Default learning rate.
Modified weight unit 633, for receiving learning sample successively, the neural network classification model for calculating the training is each
The error of layer, and the error according to each layer of the neural network classification model for being calculated the training and the default study
Rate is modified to the connection weight of each layer of neural network classification model of the training.
Weights determining unit 635, for each layer of neural network classification model when the training error less than default
During training objective precision or when the quantity of learning sample of reception reaches the default maximum frequency of training, the training is determined
Each layer connection weight neural network classification model, training.
Wherein in one embodiment, the default maximum frequency of training is 1400.The default training objective essence
Spend is 0.001.The default learning rate is 0.01.
Wherein in one embodiment, in three parameters of input, each parameter includes 32 nodes.The structure
Neutral net input layer include 96 nodes, hidden layer include 32 nodes.
Fig. 7 is referred to, the application also provides a kind of electric power energy corresponding with above-mentioned electric energy efficiency analysis user classification method
Effect analysis user's sorter, including:
Data acquisition module 720, for obtaining user data to be sorted.The user data to be sorted includes user pre-
If peak period in the time period is with electric energy consumption, paddy period electric energy consumption and usually section electric energy consumption.
Disaggregated model builds module 740, and the construction device for above-mentioned user's disaggregated model builds user's disaggregated model.
Classification determining module 760, for the input of the user data to be sorted as user's disaggregated model is true
Determine the type of the corresponding user of the user data to be sorted.
The electric energy efficiency analyzes user's sorter, obtains user data to be sorted;The user data to be sorted includes
Peak period of the user in preset time period is with electric energy consumption, paddy period electric energy consumption and usually section electric energy consumption;Using above-mentioned
The construction device of user's disaggregated model builds user's disaggregated model;Using the user data to be sorted as user classification mould
The input of type determines the type of the corresponding user of the user data to be sorted.Therefore, carrying out user's classification by the device can
Data foundation is provided with to formulate energy-saving and emission-reduction scheme.
Because said apparatus are corresponding with the above method, therefore, the details technical characteristic for device is no longer repeated one by one.
Above example only expresses several embodiments of the invention, and its description is more specific and detailed, but can not
Therefore it is interpreted as the limitation to the scope of the claims of the present invention.It should be pointed out that for the person of ordinary skill of the art,
Without departing from the inventive concept of the premise, multiple deformations can also be made and is improved, these belong to protection model of the invention
Enclose.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (10)
1. a kind of construction method of user's disaggregated model, it is characterised in that including:
Building input includes that three parameters, output include 3 neural network classification models of node;Three parameters of input
Respectively peak period of the user in preset time period is with electric energy consumption, paddy period electric energy consumption and usually section electric energy consumption;Output
3 nodes identify whether respectively as low energy consumption user, whether be middle energy consumption user and whether be high energy consumption user;
The neutral net for building is divided according to default maximum frequency of training, default training objective precision, default learning rate
Class model is trained, each layer connection weight neural network classification model, training for being built;
According to each layer connection weight for training and the neural network classification model of the structure, user's classification mould is determined
Type.
2. the construction method of user's disaggregated model according to claim 1, it is characterised in that described according to default maximum
Frequency of training, default training objective precision, default learning rate are trained to the neural network classification model for building, and obtain
The step of each layer connection weight neural network classification model, training of structure, includes:
The connection weight of each layer of neural network classification model of structure is initialized as non-zero random number, the nerve net trained
Network model, and initialize default maximum frequency of training, default training objective precision and default learning rate;
Receive learning sample successively, calculate the error of each layer of neural network classification model of the training, and according to being calculated
The neutral net of the error of each layer of the neural network classification model of the training and the default learning rate to the training
The connection weight of each layer of disaggregated model is modified;
When the training each layer of neural network classification model error be less than default training objective precision when or reception
When the quantity for practising sample reaches the default maximum frequency of training, neural network classification model, the instruction of the training are determined
Each layer connection weight perfected.
3. the construction method of user's disaggregated model according to claim 1, it is characterised in that the default maximum training
Number of times is 1400;The default training objective precision is 0.001;The default learning rate is 0.01.
4. the construction method of user's disaggregated model according to claim 1, it is characterised in that three parameters of input
In, each parameter includes 32 nodes;The input layer of the neutral net of the structure includes 96 nodes, and hidden layer includes 32
Node.
5. a kind of electric energy efficiency analyzes user classification method, it is characterised in that including:
Obtain user data to be sorted;The user data to be sorted includes peak period electric energy of the user in preset time period
Consumption, paddy period electric energy consumption and usually section electric energy consumption;
User's disaggregated model is built using the construction method of the user's disaggregated model described in claim 1-4 any one;
The input of the user data to be sorted as user's disaggregated model is determined the user data correspondence to be sorted
User type.
6. a kind of construction device of user's disaggregated model, it is characterised in that including:
Model construction module, includes that three parameters, output include 3 neural network classification models of node for building input;
Three parameters of input are respectively peak period of the user in preset time period with electric energy consumption, paddy period electric energy consumption and put down
Period electric energy consumption;Output 3 nodes identify whether respectively for low energy consumption user, whether be middle energy consumption user and whether
It is high energy consumption user;
Model training module, for according to default maximum frequency of training, default training objective precision, default learning rate pair
The neural network classification model of structure is trained, each layer neural network classification model, the training connection for being built
Weights;
Model determining module, for each layer connection weight and the neural network classification mould of the structure that are trained according to
Type, determines user's disaggregated model.
7. the construction device of user's disaggregated model according to claim 6, it is characterised in that the model training module,
Including:
Initialization unit, for the connection weight of each layer of neural network classification model of structure to be initialized as into non-zero random number,
The neural network model trained, and initialize default maximum frequency of training, default training objective precision and default
Learning rate;
Modified weight unit, for receiving learning sample successively, calculates the mistake of each layer of neural network classification model of the training
Difference, and according to the error and the default learning rate of each layer of the neural network classification model for being calculated the training to institute
The connection weight for stating each layer of neural network classification model of training is modified;
Weights determining unit, the error for each layer of neural network classification model when the training is less than default training objective
During precision or when the quantity of learning sample of reception reaches the default maximum frequency of training, the nerve net of the training is determined
Network disaggregated model, each layer connection weight for training.
8. the construction device of user's disaggregated model according to claim 6, it is characterised in that the default maximum training
Number of times is 1400;The default training objective precision is 0.001;The default learning rate is 0.01.
9. the construction device of user's disaggregated model according to claim 6, it is characterised in that three parameters of input
In, each parameter includes 32 nodes;The input layer of the neutral net of the structure includes 96 nodes, and hidden layer includes 32
Node.
10. a kind of electric energy efficiency analyzes user's sorter, it is characterised in that including:
Data acquisition module, for obtaining user data to be sorted;The user data to be sorted includes user in Preset Time
Peak period in section is with electric energy consumption, paddy period electric energy consumption and usually section electric energy consumption;
Disaggregated model builds module, for the construction device using the user's disaggregated model described in claim 6-9 any one
Build user's disaggregated model;
Classification determining module, for will be treated described in the user data to be sorted as the input determination of user's disaggregated model
The type of the corresponding user of sorted users data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611046235.XA CN106779306A (en) | 2016-11-22 | 2016-11-22 | The structure of user's disaggregated model, electric energy efficiency analysis user classification method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611046235.XA CN106779306A (en) | 2016-11-22 | 2016-11-22 | The structure of user's disaggregated model, electric energy efficiency analysis user classification method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106779306A true CN106779306A (en) | 2017-05-31 |
Family
ID=58975266
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611046235.XA Pending CN106779306A (en) | 2016-11-22 | 2016-11-22 | The structure of user's disaggregated model, electric energy efficiency analysis user classification method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106779306A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112183634A (en) * | 2020-09-29 | 2021-01-05 | 国网山东省电力公司冠县供电公司 | User demand response method, device, terminal and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102570609A (en) * | 2012-01-18 | 2012-07-11 | 山东工商学院 | Resident intelligent energy management system suitable for intelligent power grid |
CN102629106A (en) * | 2012-04-11 | 2012-08-08 | 广州东芝白云自动化***有限公司 | Water supply control method and water supply control system |
CN103942606A (en) * | 2014-03-13 | 2014-07-23 | 国家电网公司 | Residential electricity consumption customer segmentation method based on fruit fly intelligent optimization algorithm |
CN105023042A (en) * | 2015-07-10 | 2015-11-04 | 国家电网公司 | User electricity stealing suspicion analyzing device and method based on big data neural network algorithm |
-
2016
- 2016-11-22 CN CN201611046235.XA patent/CN106779306A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102570609A (en) * | 2012-01-18 | 2012-07-11 | 山东工商学院 | Resident intelligent energy management system suitable for intelligent power grid |
CN102629106A (en) * | 2012-04-11 | 2012-08-08 | 广州东芝白云自动化***有限公司 | Water supply control method and water supply control system |
CN103942606A (en) * | 2014-03-13 | 2014-07-23 | 国家电网公司 | Residential electricity consumption customer segmentation method based on fruit fly intelligent optimization algorithm |
CN105023042A (en) * | 2015-07-10 | 2015-11-04 | 国家电网公司 | User electricity stealing suspicion analyzing device and method based on big data neural network algorithm |
Non-Patent Citations (2)
Title |
---|
庆蕾: "济南供电公司CRM中的客户分类研究", 《中国优秀硕士学位论文全文数据库 经济与管理科学辑》 * |
杨劲锋 等: "基于过程回声状态网络的电力用户用电模式分类研究", 《电网技术》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112183634A (en) * | 2020-09-29 | 2021-01-05 | 国网山东省电力公司冠县供电公司 | User demand response method, device, terminal and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110991786B (en) | 10kV static load model parameter identification method based on similar daily load curve | |
CN107657243B (en) | Neural network Radar range profile's target identification method based on genetic algorithm optimization | |
CN109902874A (en) | A kind of micro-capacitance sensor photovoltaic power generation short term prediction method based on deep learning | |
CN105913150A (en) | BP neural network photovoltaic power station generating capacity prediction method based on genetic algorithm | |
CN104156464B (en) | Micro- video retrieval method and device based on micro- video frequency feature data storehouse | |
CN107194625A (en) | Wind power plant based on neutral net abandons wind-powered electricity generation amount appraisal procedure | |
CN106327357B (en) | A kind of load identification method based on improvement probabilistic neural network | |
CN110544177A (en) | Load identification method based on power fingerprint and computer readable storage medium | |
CN104636801A (en) | Transmission line audible noise prediction method based on BP neural network optimization | |
CN106803135A (en) | The Forecasting Methodology and device of a kind of photovoltaic power generation system output power | |
CN110472840A (en) | A kind of agricultural water conservancy dispatching method and system based on nerual network technique | |
CN109636009A (en) | It is a kind of to establish the method and system for determining the neural network model of grid line loss | |
CN110222908A (en) | A kind of short-term electricity demand forecasting method based on data-driven | |
CN106597154B (en) | Transformer fault diagnosis method for improving based on DAG-SVM | |
CN109934514A (en) | A kind of forest soil fertility artificial intelligence integrated evaluating method based on cloud model | |
CN108566253A (en) | It is a kind of based on the signal recognition method extracted to power spectrum signal fit characteristic | |
CN106971241A (en) | The method that sewage quality data are predicted based on fuzzy neural network | |
Cao et al. | Currency recognition modeling research based on BP neural network improved by gene algorithm | |
CN108520986A (en) | A kind of power battery method for group matching based on generation confrontation network | |
CN109408896A (en) | A kind of anerobic sowage processing gas production multi-element intelligent method for real-time monitoring | |
CN106779306A (en) | The structure of user's disaggregated model, electric energy efficiency analysis user classification method and device | |
CN104182854A (en) | Mixed energy consumption measuring method for enterprise energy management system | |
Mollaiy-Berneti | Developing energy forecasting model using hybrid artificial intelligence method | |
CN109726849A (en) | A kind of building microgrid load forecasting method of colored pollination algorithm optimization neural network | |
CN109086952A (en) | It is a kind of based on genetic algorithm-neural network heat 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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170531 |