CN106778841A - The method for building up of abnormal electricity consumption detection model - Google Patents
The method for building up of abnormal electricity consumption detection model Download PDFInfo
- Publication number
- CN106778841A CN106778841A CN201611081533.2A CN201611081533A CN106778841A CN 106778841 A CN106778841 A CN 106778841A CN 201611081533 A CN201611081533 A CN 201611081533A CN 106778841 A CN106778841 A CN 106778841A
- Authority
- CN
- China
- Prior art keywords
- electricity consumption
- detection model
- user
- consumption detection
- abnormal electricity
- 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
- 230000005611 electricity Effects 0.000 title claims abstract description 95
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 59
- 238000001514 detection method Methods 0.000 title claims abstract description 47
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000012549 training Methods 0.000 claims abstract description 28
- 230000008569 process Effects 0.000 claims abstract description 23
- 230000009467 reduction Effects 0.000 claims abstract description 12
- 238000013135 deep learning Methods 0.000 claims abstract description 11
- 238000000605 extraction Methods 0.000 claims abstract description 6
- 230000005856 abnormality Effects 0.000 claims description 15
- 238000000354 decomposition reaction Methods 0.000 claims description 9
- 230000035772 mutation Effects 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 5
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 230000006399 behavior Effects 0.000 abstract description 6
- 238000005516 engineering process Methods 0.000 abstract description 6
- 238000012544 monitoring process Methods 0.000 abstract description 6
- 238000013473 artificial intelligence Methods 0.000 abstract description 4
- 230000005540 biological transmission Effects 0.000 abstract description 2
- 238000005520 cutting process Methods 0.000 abstract description 2
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 230000007935 neutral effect Effects 0.000 description 3
- 230000003542 behavioural effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000000205 computational method Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000017105 transposition Effects 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
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Water Supply & Treatment (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Public Health (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to a kind of method for building up of abnormal electricity consumption detection model, the power information to training set user carries out the extraction of validity feature, sets up the initial training data set for deep learning;In initial training data set, by the data without the whether abnormal electricity consumption of demarcation, as the input of the abnormal electricity consumption detection model, the successively unsupervised learning from bottom to top layer is carried out in depth noise reduction autoencoder network, obtain the parameter of each layer coder of network and decoder;Network top is provided with Gaussian process grader, whether the data of abnormal electricity consumption exercise supervision study by having demarcated, and the transmission error from top layer to bottom, the parameter to each layer coder of network and decoder is adjusted.The present invention is analyzed with reference to artificial intelligence field cutting edge technology to multi-platform electricity consumption data, and depth excavates the user power utilization behavior pattern hidden in mass data, and the abnormal electricity consumption suspicion user of positioning makes abnormal electricity consumption monitoring more intelligent, more efficient.
Description
Technical field
The present invention relates to a kind of method for building up of abnormal electricity consumption detection model.
Background technology
Abnormal electricity consumption(Stealing, metering device exception)It is the key factor for causing line loss abnormal, can pacifies to electricity consumption and power network
Bring hidden danger entirely.But at present, there is following problem in exception electro-detection:It was found that difficult, the abnormal power consumption of difficult, evidence obtaining measures difficult.It is special
It is not low-voltage platform area, user is more and disperses, it is difficult to effective detection, this brings difficulty to Low-voltage Line Loss management, compromises power supply enterprise
Industry interests.
Specifically, under the overall background of battalion's auxiliary tone, electric power information degree is improved constantly, with electricity consumption data volume
It is swift and violent therewith to increase, power information acquisition system, sales service system, line loss become more meticulous the systems such as platform have accumulated largely,
Abundant, complete user data, but lack effective and reasonable means of numerical analysis.Electricity consumption data is carried out by technical staff currently
Manual analysis, is compared using electricity mutation threshold value or shallow-layer learns, and carries out exception electro-detection research, less effective.And threshold value
Need to be manually set, with larger uncertainty.
The line loss unusual fluctuation that abnormal electricity consumption causes is a key factor for influenceing line losses indices, and with low-voltage power supply, user is
Example, its power consumption occupies the vast scale of power consumption always.However, at present line loss level of the low-voltage customer caused by abnormal electricity consumption compared with
Greatly, local line loss level is even up to more than 50%, and these low pressure are distributed in old city residential block and business management more in the line loss platform area high
Concentration zones, all exist power utilization environment complexity be difficult grasp, lack effective Prevention Stealing Electricity Technology means situations such as, it has also become influence
The severely afflicated area of line losses indices, lacks effective line loss unusual fluctuation monitoring and abnormal electricity consumption location technology at present.
There is following difficult point in current exception electro-detection:
Abnormal electricity consumption finds difficult:The low pressure electricity consumption information of opposing electricity-stealing mostlys come from reports and is generally investigated with business, and reports are very
Hardly possible all area's electricity consumption situations of covering are uncertain big;Business generaI investigation cost manpower and materials are huge, and efficiency is low, it is impossible to normalization
Carry out, and the problem that power supply enterprise's generally existing power customer is more, power utility check personnel amount is few, stealing discovery is difficult to have turned into mesh
The greatest problem that preceding work of electricity anti-stealing is present.
Abnormal electricity consumption evidence obtaining is difficult:Because electric energy is different from other commodity, its production, transport and sale are completed simultaneously, thus
Electric energy is stolen stolen different from other commodity, there is the performance of its uniqueness:1. stealing scene is difficult to keep, and electricity filching person completely can be with fast
Speed destroys stealing instrument, does not leave trace;2. the stealing time do not fix, stealing evidence is difficult to be caught.
Abnormal power consumption metering is difficult:Because of the particularity of electric energy, even if exception electricity consumption at present is found, specific electricity is also difficult
Accurately to calculate, can only by inquiry inquire and determine energy data again after collecting relevant information.《Power supply and business rules》Regulation,
When that cannot find out the stealing time, stealing number of days was at least calculated by 180 days.But this computational methods is simple and crude, easily meets with and query.
Result is:The loss of power supply enterprise can not by faster, in full recover;Electricity filching person does not obtain due sanction.It is more tight
Weight, electricity filching behavior is increasingly serious, the electricity consumption order of the whole society is brought and is had a strong impact on.
The content of the invention
It is an object of the invention to provide a kind of method for building up of abnormal electricity consumption detection model, before artificial intelligence field
Multi-platform electricity consumption data is analyzed along technology, depth excavates the user power utilization behavior pattern hidden in mass data, positioning
Abnormal electricity consumption suspicion user, makes abnormal electricity consumption monitoring more intelligent, more efficient.
In order to achieve the above object, the technical scheme is that providing a kind of foundation side of abnormal electricity consumption detection model
Method, the power information to training set user carries out the extraction of validity feature, sets up the initial training data set for deep learning;
In initial training data set, by the data without the whether abnormal electricity consumption of demarcation, as the input of the abnormal electricity consumption detection model,
The successively unsupervised learning from bottom to top layer is carried out in depth noise reduction autoencoder network, each layer coder of network and decoding is obtained
The parameter of device;Network top is provided with Gaussian process grader, whether the data of abnormal electricity consumption exercise supervision by having demarcated
Study, error is transmitted from top layer to bottom, and the parameter to each layer coder of network and decoder is adjusted.
Preferably, when the input of abnormal electricity consumption detection model is power information-frequency feature.
Preferably, extract power information when-frequency feature, comprising in user profile single index time series difference
When carrying out overall experience mode decomposition and small echo-frequency division solution.
Preferably, the single index includes one or more following:Daily power consumption, platform area line loss, user's property, work(
Rate factor, contract hold proportion, zero power, day electricity mutation, moon electricity mutation, typical industry user's index, history power supply service
Index, electricity, electric energy meter dead electricity record, electric energy meter cover opening record before and after day of checking meter.
Preferably, the abnormal electricity consumption detection model output abnormality degree suspicion coefficient, and according to abnormality degree suspicion coefficient pair
The doubtful probability of abnormality degree of user is ranked up;The doubtful probability of abnormality degree of the bigger user of abnormality degree suspicion coefficient is higher.
Preferably, according to low pressure resident and the power information of the non-resident user of low pressure, the electricity consumption of corresponding exception is set up
Detection model.
Preferably, it is input into the power information of the abnormal electricity consumption detection model, comprising user's information about power, user property
And meter event information.
Preferably, by or many in power information acquisition system, sales service system, line loss fine-grained management platform
It is individual, provide the power information to the abnormal electricity consumption detection model.
The method for building up of abnormal electricity consumption detection model of the present invention, the advantage is that:
The present invention is performed based on overall experience mode decomposition-wavelet transformation-entropy by the analysis to abnormal electricity consumption behavioral mechanism
The extraction of abnormal user behavioural characteristic, cancelling noise component of signal obtains the useful feature of source signal.The present invention considers
Noise, deep learning and Gaussian process, establish the abnormal electricity consumption behavior based on depth noise reduction autoencoder network-Gaussian process
Identification model.Model training is carried out by mass data, recycles model to automatically analyze user according to user power utilization information data
Electricity consumption behavior, automatic discrimination multiplexing electric abnormality suspicion user reduces power utility check scope.The input of model is the electricity of mass users
Information, user property and meter event information, are output as abnormal electricity consumption suspicion user list.The present invention is based on exception electro-detection
Model, is that abnormal electricity consumption monitoring sets up step testing mechanism with positioning, and first examination suspicion user again targetedly monitor by collection,
Power utility check scope can be effectively reduced, the manpower and materials of field monitoring are saved, company operation cost is reduced.
Brief description of the drawings
Fig. 1 is a kind of principle schematic of autocoder;
Fig. 2 a are traditional neural network structure schematic diagrames;
Fig. 2 b are the schematic diagrames of deep learning structure;
Fig. 3 is exception electro-detection modeling procedure;
Fig. 4 a, Fig. 4 b refer respectively to mark EEMD mode decompositions and Hilbert that power factor (PF) is obtained by EEMD-HT Time-frequency Decompositions
Huang energy spectrum schematic diagram;
Fig. 5 is the Wavelet time-frequency decomposing schematic representation of index power factor (PF);
Fig. 6 a, Fig. 6 b refer respectively to indicate EEMD mode decompositions and Martin Hilb that the total electric energy of work(is obtained by EEMD-HT Time-frequency Decompositions
Special Huang energy spectrum schematic diagram;
Fig. 7 refers to the Wavelet time-frequency decomposing schematic representation for indicating the total electric energy of work(.
Specific embodiment
Abnormal electricity consumption detection model of the present invention, is applied to a kind of abnormal electricity consumption monitoring and localization process system.From
Power information acquisition system, sales service system, line loss become more meticulous platform etc., mass users power information are obtained, as different
The input of conventional electro-detection model, is automatically extracted validity feature by model and is classified, and exports the list of suspicion abnormal user;So as to
Power utility check personnel carry out site inspection to the suspicion user for filtering out, and abnormal user is investigated and prosecuted.Wherein, using artificial intelligence
Can field cutting edge technology --- deep learning(Deep Learning, hereinafter referred to as DL)Algorithm, builds of the present invention being based on
The artificial intelligence exception electricity consumption detection model of depth noise reduction autoencoder network-Gaussian process.
It is as shown in Figure 1 a kind of autocoder(Auto-Encoder, AE)Realization principle, input signal(input)
Send an encoder to(encoder)Obtain representing the corresponding encoded of the input signal(code), one is connected afterwards
Decoder(decoder)A reconfiguration information is obtained from code conversion(reconstruction)If, reconfiguration information with input
Signal like(Ideally both are the same), then have reason to believe that this coding is reliable.Therefore, by adjusting volume
The parameter of code device and decoder so that reconstructed error is minimum, has at this time just obtained first expression of input signal, i.e., foregoing
Coding information.Input be without label data when, the source of error is exactly to be obtained compared with original input after directly reconstruct.
On the basis of autocoder shown in Fig. 1, a certain proportion of ambient noise is added in input signal, then existed
During noise is removed in study, study to input vector is most essential, most stable of feature representation, so as to obtain a kind of noise reduction
Autocoder(Denoise Auto-Encoder, DAE).After training is completed, noise reduction autoencoder network containing having powerful connections to making an uproar
The input data of sound has stronger adaptability, the expression for having more robust to input signal.
Depth noise reduction autoencoder network of the present invention(Deep Denoise Auto-Encoder Network,
DDAEN)It is made up of several noise reduction autocoders DAE, initial data is successively trained as the input of ground floor DAE, will
The output of L-1 layers for training is used as L layers of input.When successively training, training criterion is to minimize reconstructed error, and is instructed
Neutral net can be expanded into encoder and decoder when practicing, that is, expand into three layers of neutral net, propagated by direction of error
Algorithm changes weights.The encoder for obtaining by this method and decoder weights, in the absence of the restriction relation of transposition.Successively
Complex distributions in training process since initial data in feature space, low-dimensional, smooth distribution are successively transformed into, that is, ignored
Fall some small noises, remove the redundancy of initial data so that data distribution is simpler.Instruction is gone by gradient descent algorithm
Practice DDAEN, multilayer study can be carried out to the appraisement system index of abnormal electricity consumption detection model, and then obtain exception electro-detection
The evaluation target abstract characteristics of model.
Specifically, the process of deep learning training is carried out in the present invention, comprising:
1)From the unsupervised learning of lower rising(Since bottom, past top layer training in layer), using whether different without demarcating
The data hierarchy for commonly using electricity trains each layer parameter, and this step can be regarded as a unsupervised training process, be and traditional neural
Network distinguishes the best part:Specifically, first training ground floor with without nominal data, the parameter of ground floor is first learnt during training
(This layer can be regarded as obtaining a hidden layer for causing to export and be input into the minimum three-layer neural network of difference), due to model
Capacity(capacity)Limitation and sparsity constraints so that the model for obtaining can learn the structure to data in itself, from
And obtain the feature than input with more expression ability;After study obtains (n-1)th layer, using n-1 layers of output as n-th layer
Input, trains n-th layer, thus respectively obtains the parameter of each layer.
2)Top-down supervised learning(It is trained by the data of tape label, the top-down transmission of error, to network
It is finely adjusted);Based on the 1st)Each layer parameter that step is obtained further is adjusted(fine-tune)The parameter of whole multilayered model, this
One step is a Training process;The random initializtion initial value process of the similar neutral net of the first step, due to the first of DL
Step is not random initializtion, but by learning what the structure of input data was obtained, thus this initial value is closer to global optimum,
So as to obtain more preferable effect;So modelling effect well largely gives the credit to the feature learning of the first step(feature
learning)Process.The comparing of traditional neural network and deep learning DL structures, referring to shown in Fig. 2 a, Fig. 2 b.
As shown in figure 3, being exception electro-detection workflow.The course of work bag of abnormal electricity consumption detection model in the present invention
Include training and test two parts.Training process is abnormal electricity consumption detection model and sets up process, and test process is to be according to foundation
Model carry out user power utilization abnormality degree prediction.
A training process:
A1)The information such as electricity, ammeter event index to training set mass users, when carrying out-frequency feature extraction, set up depth
The initial training data set of habit;
A2)For the abnormal electricity consumption detection model based on depth noise reduction autoencoder network-Gaussian process, by the initial instruction without label
Practice data set input model, the network initial weight for more optimizing is obtained by unsupervised Level by level learning;
A3)In the Gaussian process grader of the additional supervised learning of network top(It is trained by the data of tape label, error
Transmitted to bottom from top layer;Whether tag representation is abnormal electricity consumption user), the overall network parameter of model is carried out further thin
Fine adjustment, it is final to obtain abnormal electricity consumption detection model.
It is lift scheme forecasting accuracy, different models can be set up for different types of user, is directed in invention
Low pressure resident and the non-resident user of low pressure model respectively.
B, test process:
1)During the information such as the tested user's electricity of extraction, ammeter event index-frequency feature;
2)The feature that will be extracted, input to abnormal electricity consumption detection model, the tested user's exception suspicion coefficient of model output.Suspicion system
Number is bigger, and abnormal electricity consumption possibility is bigger.
Deep learning algorithm is based in the present invention, is held with daily power consumption, platform area line loss, user's property, power factor (PF), contract
Proportion, zero power, day electricity mutation, moon electricity mutation, typical industry user's index, history power supply service index, check meter a few days ago
Afterwards electricity, electric energy meter dead electricity record, electric energy meter cover opening record etc. as exception electricity consumption detection model single index, to these lists
When the time series of index carries out overall experience mode decomposition and small echo respectively-frequency division solution, time and frequency domain characteristics are extracted respectively to be made
It is the original input data collection of deep learning model(Fig. 4 a and Fig. 4 b, Fig. 5 provide an overall experience mode for index point respectively
Solution and wavelet decomposition spectrogram;Fig. 6 a and Fig. 6 b, Fig. 7 provide the overall experience mode decomposition and Wavelet time-frequency of another index respectively
Exploded view).
On this basis, commenting for the abnormal electricity consumption detection model based on depth noise reduction autoencoder network-Gaussian process is set up
Valency system, using the time-frequency characteristics of the bottom single index of appraisement system as input vector project training sample, sets up depth drop
Autoencoder network structure, selection Gaussian process model make an uproar as top-level categories precaution device.Application training sample is to deep learning net
Network is trained, and finally using abnormality degree suspicion coefficient as output, the user power utilization situation big by analyzing suspicion coefficient is found out
User property feature and judgment rule with abnormal electricity consumption behavior.
The present invention studies mass data in depth by depth noise reduction autoencoder network-Gaussian process model, quickly from difference
Dimension extracts the validity feature of data, and provides the doubtful probability sorting of abnormality degree.Abnormality degree sequence need to be only monitored using this model
Forward a few users, you can find most of abnormal user, solve abnormal electricity consumption and find difficult problem.
Although present disclosure is discussed in detail by above preferred embodiment, but it should be appreciated that above-mentioned
Description is not considered as limitation of the present invention.After those skilled in the art have read the above, for of the invention
Various modifications and substitutions all will be apparent.Therefore, protection scope of the present invention should be limited to the appended claims.
Claims (8)
1. a kind of method for building up of abnormal electricity consumption detection model, it is characterised in that
Power information to training set user carries out the extraction of validity feature, sets up the initial training data for deep learning
Collection;
In initial training data set, by the data without the whether abnormal electricity consumption of demarcation, as the abnormal electricity consumption detection model
Input, the successively unsupervised learning from bottom to top layer is carried out in depth noise reduction autoencoder network, obtains each layer coder of network
With the parameter of decoder;
Network top is provided with Gaussian process grader, whether the data of abnormal electricity consumption exercise supervision study by having demarcated,
Error is transmitted from top layer to bottom, the parameter to each layer coder of network and decoder is adjusted.
2. the method for building up of exception electricity consumption detection model as claimed in claim 1, it is characterised in that
When the input of the abnormal electricity consumption detection model is power information-frequency feature.
3. the method for building up of exception electricity consumption detection model as claimed in claim 2, it is characterised in that
Extract power information when-frequency feature, carry out overall warp respectively comprising the time series to single index in user profile
When testing mode decomposition and small echo-frequency division solution.
4. the method for building up of exception electricity consumption detection model as claimed in claim 3, it is characterised in that
The single index includes one or more following:Daily power consumption, platform area line loss, user's property, power factor (PF), contract
Hold proportion, zero power, day electricity mutation, moon electricity mutation, typical industry user's index, history power supply service index, check meter day
Front and rear electricity, electric energy meter dead electricity record, electric energy meter cover opening record.
5. the method for building up of exception electricity consumption detection model as claimed in claim 1, it is characterised in that
The abnormal electricity consumption detection model output abnormality degree suspicion coefficient, and according to abnormality degree suspicion coefficient to the abnormality degree of user
Doubtful probability is ranked up;The doubtful probability of abnormality degree of the bigger user of abnormality degree suspicion coefficient is higher.
6. the method for building up of exception electricity consumption detection model as claimed in claim 3, it is characterised in that
According to low pressure resident and the power information of the non-resident user of low pressure, corresponding exception electricity consumption detection model is set up.
7. as described in claim 1 or 3 abnormal electricity consumption detection model method for building up, it is characterised in that
It is input into the power information of the abnormal electricity consumption detection model, comprising user's information about power, user property and meter event
Information.
8. the method for building up of exception electricity consumption detection model as claimed in claim 7, it is characterised in that
By one or more in power information acquisition system, sales service system, line loss fine-grained management platform, to described different
Conventional electro-detection model provides the power information.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611081533.2A CN106778841A (en) | 2016-11-30 | 2016-11-30 | The method for building up of abnormal electricity consumption detection model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611081533.2A CN106778841A (en) | 2016-11-30 | 2016-11-30 | The method for building up of abnormal electricity consumption detection model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106778841A true CN106778841A (en) | 2017-05-31 |
Family
ID=58901439
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611081533.2A Pending CN106778841A (en) | 2016-11-30 | 2016-11-30 | The method for building up of abnormal electricity consumption detection model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106778841A (en) |
Cited By (24)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107703854A (en) * | 2017-10-26 | 2018-02-16 | 国网黑龙江省电力有限公司信息通信公司 | power load monitoring system and method |
CN108334907A (en) * | 2018-02-09 | 2018-07-27 | 哈尔滨工业大学(威海) | A kind of complex equipment point method for detecting abnormality and system based on deep learning |
CN108520357A (en) * | 2018-04-10 | 2018-09-11 | 北京中电普华信息技术有限公司 | A kind of method of discrimination, device and the server of line loss abnormal cause |
CN109242097A (en) * | 2018-07-12 | 2019-01-18 | 上海交通大学 | The visual representation learning system and method for unsupervised learning |
CN109284886A (en) * | 2018-02-02 | 2019-01-29 | 中领世能(天津)科技有限公司 | Electrical Safety management method and device based on artificial intelligence |
CN109543943A (en) * | 2018-10-17 | 2019-03-29 | 国网辽宁省电力有限公司电力科学研究院 | A kind of electricity price inspection execution method based on big data deep learning |
CN109636194A (en) * | 2018-12-14 | 2019-04-16 | 国网山东省电力公司电力科学研究院 | A kind of power transmission and transformation project significant movement multi-source collaborative detection method and system |
CN110084448A (en) * | 2018-01-24 | 2019-08-02 | 阿里巴巴集团控股有限公司 | A kind of analysis method of electric system, device and its equipment |
CN110082579A (en) * | 2019-05-21 | 2019-08-02 | 国网湖南省电力有限公司 | A kind of area's Intelligent power-stealing prevention monitoring method, system, equipment and medium |
CN110223196A (en) * | 2019-06-04 | 2019-09-10 | 国网浙江省电力有限公司电力科学研究院 | Analysis method of opposing electricity-stealing based on typical industry feature database and sample database of opposing electricity-stealing |
CN110363384A (en) * | 2019-06-03 | 2019-10-22 | 杭州电子科技大学 | Exception electric detection method based on depth weighted neural network |
WO2020019403A1 (en) * | 2018-07-26 | 2020-01-30 | 平安科技(深圳)有限公司 | Electricity consumption abnormality detection method, apparatus and device, and readable storage medium |
CN110749784A (en) * | 2019-08-05 | 2020-02-04 | 上海大学 | Line electricity stealing detection method based on electric power data wavelet analysis |
CN111710150A (en) * | 2020-05-14 | 2020-09-25 | 国网江苏省电力有限公司南京供电分公司 | Abnormal electricity consumption data detection method based on countermeasure self-coding network |
CN111898650A (en) * | 2020-07-08 | 2020-11-06 | 国网浙江省电力有限公司杭州供电公司 | Marketing and distribution data automatic clustering analysis equipment and method based on deep learning |
CN112396087A (en) * | 2020-09-28 | 2021-02-23 | 国网浙江省电力有限公司杭州供电公司 | Smart electric meter based method and device for analyzing electricity consumption data of elderly people living alone |
CN112637835A (en) * | 2020-12-13 | 2021-04-09 | 成都易书桥科技有限公司 | Technology for timely updating embedded behavior recognition model based on OTA |
CN113327008A (en) * | 2021-04-22 | 2021-08-31 | 同济大学 | Electricity stealing detection method, system and medium based on time sequence automatic encoder |
CN113487062A (en) * | 2021-05-31 | 2021-10-08 | 国网上海市电力公司 | Power load prediction method based on periodic automatic encoder |
CN113627821A (en) * | 2021-08-25 | 2021-11-09 | 深圳供电局有限公司 | Method and system for identifying abnormal electricity utilization based on electricity utilization behavior characteristics |
CN113792767A (en) * | 2021-08-27 | 2021-12-14 | 国网福建省电力有限公司 | Load electricity utilization characteristic monitoring and analyzing method based on graph signal processing |
CN113807527A (en) * | 2020-06-11 | 2021-12-17 | 华硕电脑股份有限公司 | Signal detection method and electronic device using same |
CN113902581A (en) * | 2021-08-04 | 2022-01-07 | 广西电网有限责任公司 | Power utilization abnormity detection method based on depth self-encoder Gaussian mixture model |
CN116089838A (en) * | 2023-03-01 | 2023-05-09 | 中南大学 | Training method and recognition method for intelligent recognition model of electricity stealing user |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2141898B1 (en) * | 2008-07-04 | 2011-03-02 | Postech Academy-Industry- Foundation | Method and apparatus for detecting abnormal power consumption of a battery in mobile devices |
US20130173076A1 (en) * | 2012-01-02 | 2013-07-04 | Yang Pan | Power Management System and Method Using Power Limiter |
CN103778567A (en) * | 2014-01-21 | 2014-05-07 | 深圳供电局有限公司 | Method and system for discriminating abnormal electricity utilization of user |
CN103942453A (en) * | 2014-05-07 | 2014-07-23 | 华北电力大学 | Intelligent electricity utilization anomaly detection method for non-technical loss |
-
2016
- 2016-11-30 CN CN201611081533.2A patent/CN106778841A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2141898B1 (en) * | 2008-07-04 | 2011-03-02 | Postech Academy-Industry- Foundation | Method and apparatus for detecting abnormal power consumption of a battery in mobile devices |
US20130173076A1 (en) * | 2012-01-02 | 2013-07-04 | Yang Pan | Power Management System and Method Using Power Limiter |
CN103778567A (en) * | 2014-01-21 | 2014-05-07 | 深圳供电局有限公司 | Method and system for discriminating abnormal electricity utilization of user |
CN103942453A (en) * | 2014-05-07 | 2014-07-23 | 华北电力大学 | Intelligent electricity utilization anomaly detection method for non-technical loss |
Non-Patent Citations (1)
Title |
---|
[email protected]: "Deep Learning(深度学习)学习笔记整理系列之(四)", 《HTTP://BLOG.CSDN.NET/ZOUXY09》 * |
Cited By (32)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107703854A (en) * | 2017-10-26 | 2018-02-16 | 国网黑龙江省电力有限公司信息通信公司 | power load monitoring system and method |
CN110084448A (en) * | 2018-01-24 | 2019-08-02 | 阿里巴巴集团控股有限公司 | A kind of analysis method of electric system, device and its equipment |
CN109284886A (en) * | 2018-02-02 | 2019-01-29 | 中领世能(天津)科技有限公司 | Electrical Safety management method and device based on artificial intelligence |
CN108334907A (en) * | 2018-02-09 | 2018-07-27 | 哈尔滨工业大学(威海) | A kind of complex equipment point method for detecting abnormality and system based on deep learning |
CN108520357A (en) * | 2018-04-10 | 2018-09-11 | 北京中电普华信息技术有限公司 | A kind of method of discrimination, device and the server of line loss abnormal cause |
CN109242097B (en) * | 2018-07-12 | 2020-09-25 | 上海交通大学 | Visual representation learning system and method for unsupervised learning |
CN109242097A (en) * | 2018-07-12 | 2019-01-18 | 上海交通大学 | The visual representation learning system and method for unsupervised learning |
WO2020019403A1 (en) * | 2018-07-26 | 2020-01-30 | 平安科技(深圳)有限公司 | Electricity consumption abnormality detection method, apparatus and device, and readable storage medium |
CN109543943A (en) * | 2018-10-17 | 2019-03-29 | 国网辽宁省电力有限公司电力科学研究院 | A kind of electricity price inspection execution method based on big data deep learning |
CN109543943B (en) * | 2018-10-17 | 2023-07-25 | 国网辽宁省电力有限公司电力科学研究院 | Electric price checking execution method based on big data deep learning |
CN109636194A (en) * | 2018-12-14 | 2019-04-16 | 国网山东省电力公司电力科学研究院 | A kind of power transmission and transformation project significant movement multi-source collaborative detection method and system |
CN109636194B (en) * | 2018-12-14 | 2020-10-16 | 国网山东省电力公司电力科学研究院 | Multi-source cooperative detection method and system for major change of power transmission and transformation project |
CN110082579A (en) * | 2019-05-21 | 2019-08-02 | 国网湖南省电力有限公司 | A kind of area's Intelligent power-stealing prevention monitoring method, system, equipment and medium |
CN110363384A (en) * | 2019-06-03 | 2019-10-22 | 杭州电子科技大学 | Exception electric detection method based on depth weighted neural network |
CN110223196A (en) * | 2019-06-04 | 2019-09-10 | 国网浙江省电力有限公司电力科学研究院 | Analysis method of opposing electricity-stealing based on typical industry feature database and sample database of opposing electricity-stealing |
CN110223196B (en) * | 2019-06-04 | 2021-08-31 | 国网浙江省电力有限公司营销服务中心 | Anti-electricity-stealing analysis method based on typical industry feature library and anti-electricity-stealing sample library |
CN110749784A (en) * | 2019-08-05 | 2020-02-04 | 上海大学 | Line electricity stealing detection method based on electric power data wavelet analysis |
CN110749784B (en) * | 2019-08-05 | 2022-07-08 | 上海大学 | Line electricity stealing detection method based on electric power data wavelet analysis |
CN111710150A (en) * | 2020-05-14 | 2020-09-25 | 国网江苏省电力有限公司南京供电分公司 | Abnormal electricity consumption data detection method based on countermeasure self-coding network |
CN113807527A (en) * | 2020-06-11 | 2021-12-17 | 华硕电脑股份有限公司 | Signal detection method and electronic device using same |
CN111898650A (en) * | 2020-07-08 | 2020-11-06 | 国网浙江省电力有限公司杭州供电公司 | Marketing and distribution data automatic clustering analysis equipment and method based on deep learning |
CN112396087A (en) * | 2020-09-28 | 2021-02-23 | 国网浙江省电力有限公司杭州供电公司 | Smart electric meter based method and device for analyzing electricity consumption data of elderly people living alone |
CN112396087B (en) * | 2020-09-28 | 2024-04-26 | 国网浙江省电力有限公司杭州供电公司 | Method and device for analyzing power consumption data of solitary old people based on intelligent ammeter |
CN112637835A (en) * | 2020-12-13 | 2021-04-09 | 成都易书桥科技有限公司 | Technology for timely updating embedded behavior recognition model based on OTA |
CN112637835B (en) * | 2020-12-13 | 2023-12-26 | 成都易书桥科技有限公司 | Method for timely updating embedded behavior recognition model based on OTA |
CN113327008A (en) * | 2021-04-22 | 2021-08-31 | 同济大学 | Electricity stealing detection method, system and medium based on time sequence automatic encoder |
CN113487062A (en) * | 2021-05-31 | 2021-10-08 | 国网上海市电力公司 | Power load prediction method based on periodic automatic encoder |
CN113902581A (en) * | 2021-08-04 | 2022-01-07 | 广西电网有限责任公司 | Power utilization abnormity detection method based on depth self-encoder Gaussian mixture model |
CN113627821A (en) * | 2021-08-25 | 2021-11-09 | 深圳供电局有限公司 | Method and system for identifying abnormal electricity utilization based on electricity utilization behavior characteristics |
CN113792767A (en) * | 2021-08-27 | 2021-12-14 | 国网福建省电力有限公司 | Load electricity utilization characteristic monitoring and analyzing method based on graph signal processing |
CN116089838A (en) * | 2023-03-01 | 2023-05-09 | 中南大学 | Training method and recognition method for intelligent recognition model of electricity stealing user |
CN116089838B (en) * | 2023-03-01 | 2023-09-26 | 中南大学 | Training method and recognition method for intelligent recognition model of electricity stealing user |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106707099B (en) | Monitoring and positioning method based on abnormal electricity consumption detection model | |
CN106778841A (en) | The method for building up of abnormal electricity consumption detection model | |
CN113837477B (en) | Method, device and equipment for predicting power grid faults under typhoon disasters driven by data | |
CN110097297B (en) | Multi-dimensional electricity stealing situation intelligent sensing method, system, equipment and medium | |
CN111562996B (en) | Method and system for detecting time sequence abnormality of key performance index data | |
CN106780115A (en) | Abnormal electricity consumption monitoring and alignment system and method | |
CN106650767B (en) | Flood forecasting method based on cluster analysis and real-time correction | |
CN112098714A (en) | ResNet-LSTM-based electricity stealing detection method and system | |
CN110108914B (en) | Intelligent decision-making method, system, equipment and medium for preventing electricity stealing | |
CN108008332A (en) | A kind of new energy Remote testing device method for diagnosing faults based on data mining | |
CN102185735B (en) | Network security situation prediction method | |
CN109635928A (en) | A kind of voltage sag reason recognition methods based on deep learning Model Fusion | |
CN110634080A (en) | Abnormal electricity utilization detection method, device, equipment and computer readable storage medium | |
CN102098180A (en) | Network security situational awareness method | |
CN103154845A (en) | Machine learning for power grids | |
CN108133225A (en) | A kind of icing flashover fault early warning method based on support vector machines | |
CN103953490A (en) | Implementation method for monitoring status of hydraulic turbine set based on HLSNE | |
CN110674858B (en) | Traffic public opinion detection method based on space-time correlation and big data mining | |
CN112036042A (en) | Power equipment abnormality detection method and system based on variational modal decomposition | |
CN107527223A (en) | A kind of method and device of Ticketing information analysis | |
CN112576312B (en) | Data collection and processing method for electric-hydraulic control support of intelligent fully-mechanized coal mining face | |
CN113188807A (en) | Abs result automatic judging algorithm | |
CN113554361A (en) | Comprehensive energy system data processing and calculating method and processing system | |
Hu et al. | A hybrid model for predicting the degradation trend of hydropower units based on deep learning | |
CN112507479A (en) | Oil drilling machine health state assessment method based on manifold learning and softmax |
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 |