CN107966600A - A kind of electricity anti-theft system and its electricity anti-theft method based on deep learning algorithm - Google Patents
A kind of electricity anti-theft system and its electricity anti-theft method based on deep learning algorithm Download PDFInfo
- Publication number
- CN107966600A CN107966600A CN201710919165.2A CN201710919165A CN107966600A CN 107966600 A CN107966600 A CN 107966600A CN 201710919165 A CN201710919165 A CN 201710919165A CN 107966600 A CN107966600 A CN 107966600A
- Authority
- CN
- China
- Prior art keywords
- power
- mrow
- deep learning
- msub
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R11/00—Electromechanical arrangements for measuring time integral of electric power or current, e.g. of consumption
- G01R11/02—Constructional details
- G01R11/24—Arrangements for avoiding or indicating fraudulent use
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A kind of electricity anti-theft system and its electricity anti-theft method based on deep learning algorithm, including the doubtful stealing line module of online recognition and off-line training deep learning mixed-media network modules mixed-media, the doubtful stealing line module of online recognition is used to identify doubtful stealing user, and off-line training deep learning mixed-media network modules mixed-media is used to determine each network parameter of deep learning network in the doubtful stealing line module of online recognition;The data of existing power information acquisition system can be efficiently used, improve the utilization rate of data;It is of low cost, it is only necessary to increase the anti-electricity-theft algoritic module based on deep learning in existing power information acquisition system, it is not necessary to increase other hardware;Identify that suspectable stealing user accuracy is higher.
Description
Technical field
The present invention relates to a kind of electricity anti-theft method, electricity anti-theft system specifically based on deep learning algorithm and its anti-
Stealing electricity method.
Background technology
Now, electric energy is highly important clean energy resource.But because people pursue economic interests, stealing electricity phenomenon starts
It is a large amount of to grow, all cause interests loss to each area.Also, the casualties produced because of the electricity stealing of electricity filching person
Also emerge in an endless stream with equipment damage, in fact, the indirect loss caused by stealing causes accident is then more huge.Thus may be used
See, the significance of anti-electricity-theft work.Constantly improved with anti-theft electricity technology and the improvement of measure, electricity filching means, it is thief-proof
Electric problem is, it is necessary to which the study of more personnel, and management profrssion personnel constantly go to study, crack.
Existing anti-electricity-theft means achieve certain effect really, such as by carrying out concentrated setting to electric energy meter and being tamping
The measure in jail is locked, avoids user from will contacting electric energy meter;The method for packing for uniformly taking sealing can not recover after depressing, and all
Seal up number and carry out record management together;The electric energy meter for forbidding reversing function using having, electric power supply plant are agreed to require producer's design during goods
Kilowatt-hour meter equipped with inverse-stopping functions, prevents negative-phase sequence curent stealing.But with the development of science and technology the style of high-tech stealing is also more next
More, existing anti-theft electricity technology can not take precautions against miscellaneous electricity stealing well.As intelligent grid key skill
One of art, power information acquisition system is by the electric energy meter of user terminal, host computer data management system and for its communication
Network forms, and has taken into account the basic function of information gathering and load monitoring.Its integration application is the inexorable trend of intelligent grid,
To meet the interactive platform for having built a higher of intelligent distribution network.Implement advanced power information acquisition system, pass through
Its electricity consumption data gathered analysis judges that stealing user is one of current research hotspot.However, it is since power information gathers
System will gather the electricity consumption data of user every a set time point, by taking the SCADA in power information acquisition system as an example, often
An electricity consumption data is gathered every 3~4s, 10000 telemetry stations will produce the data of 1.03TB every year.Therefore, power information is utilized
Acquisition system identification stealing user plane faces the predicament of mass data.
Deep learning is an emerging field in machine learning, namely a popular direction in artificial intelligence.
Its concept is based on neutral net, is increasing for hidden layer with the point that traditional artificial neural network distinguishes.So
One can train the data of magnanimity, learn useful feature, be finally reached the classification and prediction of pinpoint accuracy.Actually
It is the Feature Mapping by rearranging combination low layer, different Attribute class is represented after abstract high-rise Feature Mapping is formed
Not.Compared with the method for artificial rule construct feature, using big data come learning characteristic, the abundant inherence of data can be more portrayed
Information.
Compared to other methods, deep learning has very strong data learning ability and generalization ability, therefore can be used for real
Now intelligent is anti-electricity-theft, more efficiently solves the problems, such as stealing.
The content of the invention
The purpose of the present invention is to the stealing problem in life to provide a kind of anti-electricity-theft side based on deep learning algorithm
Method.
The technical solution adopted by the present invention is:A kind of electricity anti-theft system based on deep learning algorithm, it is characterised in that:Bag
Include the doubtful stealing line module of online recognition and off-line training deep learning mixed-media network modules mixed-media, the doubtful stealing line module of online recognition
For identifying doubtful stealing user, off-line training deep learning mixed-media network modules mixed-media is used to determine the doubtful stealing line module of online recognition
The middle each network parameter of deep learning network.
Further, the electricity anti-theft method of the electricity anti-theft system based on deep learning algorithm, it is characterised in that:Step
Suddenly include:
A, using power information acquisition system, user power utilization data, including active power W are obtainedIt is active, reactive power WIt is idle、
Valley power WValley, crest segment power WPeak value, level values power WLevel valuesAnd the data such as power;
Active power:WIt is active=(MThis month, is active-MLast month is active) m, m represent multiplying power, M represents meter reading value;
Reactive power:WIt is idle=(MThis month, is idle-MLast month is idle)·m;
Valley power:WValley=(MThis month valley-MLast month valley)·m;
Peak power:WPeak value=(MThis month peak value-MLast month peak value)·m;
Level values power:WLevel values=(MThis month level values-MLast month level values)·m;
Power factor (PF):
B, it is necessary to active power, reactive power, flat section power, paddy section power, peak before using deep learning algorithm
The normalization of the section electricity consumption data such as power and power, its value range is turned to [- 1,1], and normalization formula is:
Wherein, xmaxFor the maximum in this index value, xminFor the minimum value in this index value, xiIt is defeated for i-th
Enter data, xmidRepresent median,Represent xiNumerical value after normalization;
C, according to normalized active power, reactive power, flat section power, paddy section power, crest segment power and power etc.
Electricity consumption data, structure electricity consumption data matrix N × 6, wherein, N represents electricity consumption data sampling number, and 6 represent mentioned above active
The quantity of the parameters such as power, reactive power, flat section power, paddy section power, crest segment power and power, using deep learning algorithm
In convolutional neural networks algorithm, according to softmax classifier calculated user power utilization suspicion coefficients, in softmax graders,
Softmax values calculate as follows:
Wherein, SiRepresent the softmax values of i-th of data, viRepresent i-th of data of input softmax graders, j tables
Show data number, the computation model of stealing suspicion coefficient y is as follows:
Y=H (x1, x2, x3, x4, x5, x6) (4)
Wherein, H (x) represents deep learning computing
If stealing suspicion coefficient was more than or equal to 0.85 in d, this month, there is larger stealing suspicion the moon, and send warning information;
If stealing suspicion coefficient was less than 0.85 in this month, it is determined that the moon is without electricity stealing.
Beneficial effects of the present invention and feature are:1. the data of existing power information acquisition system can be efficiently used, carry
The utilization rate of high data;It is 2. of low cost, it is only necessary to which that increase is anti-based on deep learning in existing power information acquisition system
Stealing algoritic module, it is not necessary to increase other hardware;3. identify that suspectable stealing user accuracy is higher.
Brief description of the drawings
Fig. 1 is the overview flow chart of the method for the invention;
Figure label represents respectively:The doubtful stealing line module of 1- online recognitions, 2- off-line training deep learning network moulds
Block;
Embodiment
The present invention is further described below in conjunction with the accompanying drawings:
Patent of the present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.
The overall system structure of electricity anti-theft system of the invention based on deep learning algorithm, as shown in Figure 1,
The electricity anti-theft system based on deep learning algorithm, including the doubtful stealing line module 1 of online recognition and offline instruction
Practice deep learning mixed-media network modules mixed-media 2, the doubtful stealing line module of online recognition is used to identify doubtful stealing user, off-line training depth
Learning network module is used to determine each network parameter of deep learning network in the doubtful stealing line module of online recognition.
The electricity anti-theft method of the electricity anti-theft system based on deep learning algorithm, includes the following steps:
A, using power information acquisition system, user power utilization data, including active power W are obtainedIt is active, reactive power WIt is idle、
Valley power WValley, crest segment power WPeak value, level values power WLevel valuesAnd the data such as power;
Active power:WIt is active=(MThis month, is active-MLast month is active) m, m represent multiplying power, M represents meter reading value;
Reactive power:WIt is idle=(MThis month, is idle-MLast month is idle)·m;
Valley power:WValley=(MThis month valley-MLast month valley)·m;
Peak power:WPeak value=(MThis month peak value-MLast month peak value)·m;
Level values power:WLevel values=(MThis month level values-MLast month level values)·m;
Power factor (PF):
B, it is necessary to active power, reactive power, flat section power, paddy section power, peak before using deep learning algorithm
The normalization of the section electricity consumption data such as power and power, its value range is turned to [- 1,1], and normalization formula is:
Wherein, xmaxFor the maximum in this index value, xminFor the minimum value in this index value, xiIt is defeated for i-th
Enter data, xmidRepresent median,Represent xiNumerical value after normalization;
C, according to normalized active power, reactive power, flat section power, paddy section power, crest segment power and power etc.
Electricity consumption data, structure electricity consumption data matrix N × 6, wherein, N represents electricity consumption data sampling number, and 6 represent mentioned above active
The quantity of the parameters such as power, reactive power, flat section power, paddy section power, crest segment power and power, using deep learning algorithm
In convolutional neural networks algorithm, according to softmax classifier calculated user power utilization suspicion coefficients, in softmax graders,
Softmax values calculate as follows:
Wherein, SiRepresent the softmax values of i-th of data, viRepresent i-th of data of input softmax graders, j tables
Show data number, the computation model of stealing suspicion coefficient y is as follows:
Y=H (x1, x2, x3, x4, x5, x6) (0≤y≤1) (4)
Wherein, H (x) represents deep learning computing
If stealing suspicion coefficient was more than or equal to 0.85 in d, this month, there is larger stealing suspicion the moon, and send warning information;
If stealing suspicion coefficient was less than 0.85 in this month, it is determined that the moon is without electricity stealing.
The step of off-line training deep learning mixed-media network modules mixed-media specific implementation, includes:
A. user power utilization data are chosen, input convolutional neural networks;
B. the reality output of convolutional neural networks is calculated;
C. the reality output of convolutional neural networks and the difference of ideal value are calculated;
D. adjustment matrix is propagated according to minimization error approach direction;
E. training is completed, and calculates user power utilization suspicion coefficient;
During using convolutional neural networks, the specific implementation of the above process includes the following steps:
Step1. from selected training group, a certain number of samples are randomly selected in meter reading data is normalized as training sample
This collection;
Step2. initialize, by each weights vij, wjkAnd threshold θk,Be arranged to it is sufficiently small close to 0 it is random
Value, and the initial value of the parameter such as control parameter ε and learning rate α, v are setijThe weights of expression output unit i to implicit unit j,
wjkRepresent implicit unit j to the weights of output unit k, θkWithThe threshold value of output unit k and the threshold of implicit unit j are represented respectively
Value;
Step3. any one input pattern X is chosen from training group to be input in network, and specify its target export to
Measure D;
Step4. (5) are utilized to calculate intermediate layer output vector H=(h0, h1..., hL), recycle formula (6) to calculate
Reality output vector Y;
In formula, L represents the network number in network intermediate layer, and k represents output unit, ykRepresent the output of output unit k, hjTable
Show the output valve of intermediate layer j;.
Step5. the element y in output vector is calculatedkWith the element d in preferable object vectorkDifference, calculate M
Output error item:
δk=(dk-yk)yk(1-yk) (7)
Wherein, M represents the unit number of output layer;
Step6. the error term of the implicit unit in intermediate layer is calculated, common L is a:
Step7. the adjustment amount Δ w of each connection weight is calculated successivelyjk(n) and Δ vjk(n), n represents input layer unit
Number, Δ vijRepresent output unit i to the weighed value adjusting amount of implicit unit j, Δ wjkRepresent the power of implicit unit j to output unit k
It is worth adjustment amount;
Δwjk(n)=(α/(1+L)) * (Δ wjk(n-1)+1)*δk*hj
Δvjk(n)=(α/(1+L)) * (Δ vjk(n-1)+1)*δk*hj (9)
With the adjustment amount of threshold value, Δ θkWithThe threshold value of output unit k and the adjusting thresholds of implicit unit j are represented respectively
Amount:
Δθk(n)=(α/(1+L)) * (Δ θk(n-1)+1)*δk
Δφj(n)=(α/(1+L)) * Δs φj(n-1)+1)*δj (10)
Adjust weights:
wjk(n+1)=wjk(n)+Δwjk(n)
vjk(n+1)=vjk(n)+Δvjk(n) (11)
Adjust threshold value:
θk(n+1)=θk(n)+Δθk(n)
φj(n+1)=φj(n)+Δφj(n) (12)
After Step8.k repeats 1 to M, judge whether deliberated index meets required precision:E≤ε, wherein E are overall errors
Function, and
If not satisfied, being returned to the 3rd step, continue to calculate iteration, carried out in next step if meeting;
Step9. after training, weights and threshold value that training is obtained preserve in a network, it is believed that stablize available net
Network grader has been formed, and need to be such as trained again, then directly invokes existing weights and threshold value is trained, without again
Carry out initialization operation.
The method of two module cooperative work is as follows:First, off-line training deep learning mixed-media network modules mixed-media is according to the use of input
Electric data, determine the network parameter in convolutional neural networks;Then, the doubtful stealing line module of online recognition is according to definite net
Network parameter, doubtful stealing user is determined using softmax graders;Finally, regularly updated and rolled up according to the electricity consumption data of storage
Network parameter in product neutral net, adjusts relevant parameter, ensures convolution audit net in the doubtful stealing line module of online recognition
The real-time of network parameter.
The basic principle and main feature and advantages of the present invention of the present invention has been shown and described above.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and the above embodiments and description only describe this
The structural relation and principle of invention, without departing from the spirit and scope of the present invention, the present invention also have various change and
Improve, these changes and improvements all fall within the protetion scope of the claimed invention.The claimed scope of the invention is by appended power
Sharp claim and its equivalent thereof.
Claims (3)
- A kind of 1. electricity anti-theft system based on deep learning algorithm, it is characterised in that:Including the doubtful stealing user mould of online recognition Block (1) and off-line training deep learning mixed-media network modules mixed-media (2), the doubtful stealing line module of online recognition are used to identify that doubtful stealing is used Family, off-line training deep learning mixed-media network modules mixed-media are used to determine that deep learning network is each in the doubtful stealing line module of online recognition Network parameter.
- 2. the electricity anti-theft method of the electricity anti-theft system according to claim 1 based on deep learning algorithm, it is characterised in that: Step includes:A, using power information acquisition system, user power utilization data, including active power W are obtainedIt is active, reactive power WIt is idle, valley Power WValley, crest segment power WPeak value, level values power WLevel valuesAnd the data such as power;Active power:WIt is active=(MThis month, is active-MLast month is active) m, m represent multiplying power, M represents meter reading value;Reactive power:WIt is idle=(MThis month, is idle-MLast month is idle)·m;Valley power:WValley=(MThis month valley-MLast month valley)·m;Peak power:WPeak value=(MThis month peak value-MLast month peak value)·m;Level values power:WLevel values=(MThis month level values-MLast month level values)·m;Power factor (PF):B, it is necessary to active power, reactive power, flat section power, paddy section power, crest segment work(before using deep learning algorithm The normalization of the electricity consumption data such as rate and power, its value range is turned to [- 1,1], and normalization formula is:<mrow> <msub> <mi>x</mi> <mrow> <mi>m</mi> <mi>i</mi> <mi>d</mi> </mrow> </msub> <mo>=</mo> <mfrac> <mrow> <msub> <mi>x</mi> <mrow> <mi>m</mi> <mi>i</mi> <mi>n</mi> </mrow> </msub> <mo>+</mo> <msub> <mi>x</mi> <mrow> <mi>m</mi> <mi>a</mi> <mi>x</mi> </mrow> </msub> </mrow> <mn>2</mn> </mfrac> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow><mrow> <mover> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>&OverBar;</mo> </mover> <mo>=</mo> <mfrac> <mrow> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>-</mo> <msub> <mi>x</mi> <mrow> <mi>m</mi> <mi>i</mi> <mi>d</mi> </mrow> </msub> </mrow> <mrow> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mrow> <mi>m</mi> <mi>a</mi> <mi>x</mi> </mrow> </msub> <mo>-</mo> <msub> <mi>x</mi> <mrow> <mi>m</mi> <mi>i</mi> <mi>n</mi> </mrow> </msub> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>2</mn> <mo>)</mo> </mrow> </mrow>Wherein, xmaxFor the maximum in this index value, xminFor the minimum value in this index value, xiFor i-th of input number According to xmidRepresent median,Represent xiNumerical value after normalization;C, according to electricity consumptions such as normalized active power, reactive power, flat section power, paddy section power, crest segment power and power Data, structure electricity consumption data matrix N × 6, wherein, N expression electricity consumption data sampling numbers, 6 expressions active power mentioned above, The quantity of the parameters such as reactive power, flat section power, paddy section power, crest segment power and power, using in deep learning algorithm Convolutional neural networks algorithm, according to softmax classifier calculated user power utilization suspicion coefficients, in softmax graders, Softmax values calculate as follows:Wherein, SiRepresent the softmax values of i-th of data, viRepresent i-th of data of input softmax graders, j represents number According to numbering, the computation model of stealing suspicion coefficient y is as follows:Y=H (x1, x2, x3, x4, x5, x6) (4)Wherein, H (x) represents deep learning computingIf stealing suspicion coefficient was more than or equal to 0.85 in d, this month, there is larger stealing suspicion the moon, and send warning information;If should The moon, stealing suspicion coefficient was less than 0.85, it is determined that the moon is without electricity stealing.
- 3. the electricity anti-theft system according to claim 1 based on deep learning algorithm, it is characterised in that:The off-line training The step of deep learning mixed-media network modules mixed-media implements includes:A. user power utilization data are chosen, input convolutional neural networks;B. the reality output of convolutional neural networks is calculated;C. the reality output of convolutional neural networks and the difference of ideal value are calculated;D. adjustment matrix is propagated according to minimization error approach direction;E. training is completed, and calculates user power utilization suspicion coefficient;During using convolutional neural networks, the specific implementation of the above process includes the following steps:Step1. from selected training group, a certain number of samples are randomly selected in meter reading data is normalized as training sample Collection;Step2. initialize, by each weights vij, wjkAnd threshold θk,Be arranged to it is sufficiently small close to 0 random value, and set Put the initial value of the parameter such as control parameter ε and learning rate α, vijRepresent output unit i to the weights of implicit unit j, wjkRepresent hidden J containing unit is to the weights of output unit k, θkWithThe threshold value of output unit k and the threshold value of implicit unit j are represented respectively;Step3. any one input pattern X is chosen from training group to be input in network, and specifies its target output vector D;Step4. (5) are utilized to calculate intermediate layer output vector H=(h0,h1,…,hL), recycle formula (6) to calculate reality Output vector Y;In formula, L represents the network number in network intermediate layer, and k represents output unit, ykRepresent the output of output unit k, hjIn expression The output valve of interbed j;Step5. the element y in output vector is calculatedkWith the element d in preferable object vectorkDifference, calculate M output Error term:δk=(dk-yk)yk(1-yk) (7)Wherein, M represents the unit number of output layer;Step6. the error term of the implicit unit in intermediate layer is calculated, common L is a:Step7. the adjustment amount Δ w of each connection weight is calculated successivelyjk(n) and Δ vjk(n), n represents input layer unit number, Δ vijRepresent output unit i to the weighed value adjusting amount of implicit unit j, Δ wjkRepresent the weights tune of implicit unit j to output unit k Whole amount;Δwjk(n)=(α/(1+L)) * (Δ wjk(n-1)+1)*δk*hjΔvjk(n)=(α/(1+L)) * (Δ vjk(n-1)+1)*δk*hj (9)With the adjustment amount of threshold value, Δ θkWithThe threshold value of output unit k and the adjusting thresholds amount of implicit unit j are represented respectively:Δθk(n)=(α/(1+L)) * (Δ θk(n-1)+1)*δkΔφj(n)=(α/(1+L)) * (Δ φj(n-1)+1)*δj (10)Adjust weights:wjk(n+1)=wjk(n)+Δwjk(n)vjk(n+1)=vjk(n)+Δvjk(n) (11)Adjust threshold value:θk(n+1)=θk(n)+Δθk(n)φj(n+1)=φj(n)+Δφj(n) (12)After Step8.k repeats 1 to M, judge whether deliberated index meets required precision:E≤ε, wherein E are overall error letters Number, andIf not satisfied, being returned to the 3rd step, continue to calculate iteration, carried out in next step if meeting;Step9. after training, weights and threshold value that training is obtained preserve in a network, it is believed that stablize available network point Class device has been formed, and need to be such as trained again, then directly invokes existing weights and threshold value is trained, without carrying out again Initialization operation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710919165.2A CN107966600B (en) | 2017-09-30 | 2017-09-30 | Deep learning algorithm-based electricity stealing prevention system and electricity stealing prevention method thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710919165.2A CN107966600B (en) | 2017-09-30 | 2017-09-30 | Deep learning algorithm-based electricity stealing prevention system and electricity stealing prevention method thereof |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107966600A true CN107966600A (en) | 2018-04-27 |
CN107966600B CN107966600B (en) | 2021-03-19 |
Family
ID=61996690
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710919165.2A Active CN107966600B (en) | 2017-09-30 | 2017-09-30 | Deep learning algorithm-based electricity stealing prevention system and electricity stealing prevention method thereof |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107966600B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110082577A (en) * | 2019-04-18 | 2019-08-02 | 中国电力科学研究院有限公司 | It is a kind of for judging the diagnostic method and system of electricity-saving appliance electricity stealing |
CN110349050A (en) * | 2019-06-19 | 2019-10-18 | 国网江西省电力有限公司电力科学研究院 | A kind of intelligent stealing criterion method and device extracted based on electrical network parameter key feature |
CN111461923A (en) * | 2020-04-03 | 2020-07-28 | 国网山东省电力公司聊城供电公司 | Electricity stealing monitoring system and method based on deep convolutional neural network |
CN112733456A (en) * | 2021-03-17 | 2021-04-30 | 国网河南省电力公司营销服务中心 | Electricity stealing prevention behavior identification method and system |
CN113687176A (en) * | 2021-10-25 | 2021-11-23 | 广东电网有限责任公司湛江供电局 | Deep neural network-based power consumption abnormity detection method and system |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN105205531A (en) * | 2014-06-30 | 2015-12-30 | 国家电网公司 | Anti-electric-larceny prediction method based on machine learning and apparatus thereof |
CN106680557A (en) * | 2017-01-05 | 2017-05-17 | 云南电网有限责任公司电力科学研究院 | Electricity theft prevention method based on user behavior analysis |
-
2017
- 2017-09-30 CN CN201710919165.2A patent/CN107966600B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105205531A (en) * | 2014-06-30 | 2015-12-30 | 国家电网公司 | Anti-electric-larceny prediction method based on machine learning and apparatus thereof |
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 |
CN106680557A (en) * | 2017-01-05 | 2017-05-17 | 云南电网有限责任公司电力科学研究院 | Electricity theft prevention method based on user behavior analysis |
Non-Patent Citations (2)
Title |
---|
丁新立: "基于卷积神经网络的车辆前方障碍物识别", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
陈文瑛等: "应用大数据技术的反窃电分析", 《电子测量与仪器学报》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110082577A (en) * | 2019-04-18 | 2019-08-02 | 中国电力科学研究院有限公司 | It is a kind of for judging the diagnostic method and system of electricity-saving appliance electricity stealing |
CN110349050A (en) * | 2019-06-19 | 2019-10-18 | 国网江西省电力有限公司电力科学研究院 | A kind of intelligent stealing criterion method and device extracted based on electrical network parameter key feature |
CN110349050B (en) * | 2019-06-19 | 2022-06-14 | 国网江西省电力有限公司电力科学研究院 | Intelligent electricity stealing criterion method and device based on power grid parameter key feature extraction |
CN111461923A (en) * | 2020-04-03 | 2020-07-28 | 国网山东省电力公司聊城供电公司 | Electricity stealing monitoring system and method based on deep convolutional neural network |
CN112733456A (en) * | 2021-03-17 | 2021-04-30 | 国网河南省电力公司营销服务中心 | Electricity stealing prevention behavior identification method and system |
CN112733456B (en) * | 2021-03-17 | 2022-10-14 | 国网河南省电力公司营销服务中心 | Electricity stealing prevention behavior identification method and system |
CN113687176A (en) * | 2021-10-25 | 2021-11-23 | 广东电网有限责任公司湛江供电局 | Deep neural network-based power consumption abnormity detection method and system |
CN113687176B (en) * | 2021-10-25 | 2022-02-15 | 广东电网有限责任公司湛江供电局 | Deep neural network-based power consumption abnormity detection method and system |
Also Published As
Publication number | Publication date |
---|---|
CN107966600B (en) | 2021-03-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107966600A (en) | A kind of electricity anti-theft system and its electricity anti-theft method based on deep learning algorithm | |
CN103324980B (en) | A kind of method for forecasting | |
CN102722759B (en) | Method for predicting power supply reliability of power grid based on BP neural network | |
CN104280612B (en) | Distributed harmonic source identification method based on single-frequency current transmission characteristics | |
CN107423839A (en) | A kind of method of the intelligent building microgrid load prediction based on deep learning | |
CN103106535B (en) | Method for solving collaborative filtering recommendation data sparsity based on neural network | |
CN106529719A (en) | Method of predicting wind power of wind speed fusion based on particle swarm optimization algorithm | |
CN103646354A (en) | Effective index FCM and RBF neural network-based substation load characteristic categorization method | |
CN106022614A (en) | Data mining method of neural network based on nearest neighbor clustering | |
CN103093285A (en) | Short-term load forecast method based on artificial neural network | |
Liang et al. | A combined model based on CEEMDAN, permutation entropy, gated recurrent unit network, and an improved bat algorithm for wind speed forecasting | |
CN103020740A (en) | Micrometeorological data based electric power circuit icing thickness prediction method | |
CN109376950A (en) | A kind of polynary Load Forecasting based on BP neural network | |
CN105844294A (en) | Electricity usage behavior analysis method based on FCM cluster algorithm | |
CN104933629A (en) | Power user equipment evaluation method based on interval level analysis and interval entropy combination | |
CN104573857A (en) | Power grid load rate prediction method based on intelligent algorithm optimization and combination | |
CN104050547A (en) | Non-linear optimization decision-making method of planning schemes for oilfield development | |
CN109066819A (en) | A kind of idle work optimization method of the power distribution network based on case reasoning | |
Nourani et al. | Classification of groundwater level data using SOM to develop ANN-based forecasting model | |
Zhang et al. | Wind power prediction based on CNN-LSTM | |
CN106372440A (en) | Method and device for estimating self-adaptive robust state of distribution network through parallel computation | |
Zhangang et al. | Genetic algorithm-based RBF neural network load forecasting model | |
CN104182854A (en) | Mixed energy consumption measuring method for enterprise energy management system | |
Liu et al. | Theoretical line loss calculation method for low-voltage distribution network via matrix completion and ReliefF-CNN | |
CN107730132A (en) | A kind of village land intensive and economical use appraisal procedure and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |