CN103208091A  Electric larceny preventing method based on data mining of electric load management system  Google Patents
Electric larceny preventing method based on data mining of electric load management system Download PDFInfo
 Publication number
 CN103208091A CN103208091A CN2013101481057A CN201310148105A CN103208091A CN 103208091 A CN103208091 A CN 103208091A CN 2013101481057 A CN2013101481057 A CN 2013101481057A CN 201310148105 A CN201310148105 A CN 201310148105A CN 103208091 A CN103208091 A CN 103208091A
 Authority
 CN
 China
 Prior art keywords
 value
 stealing
 centerdot
 management system
 load
 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
Images
Landscapes
 Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to an electric larceny preventing method based on data mining of an electric load management system. The method comprises the following steps of: 1) acquiring user's historical load data in an electric load management system database, and performing attribute supplement on missing data through a Newton interpolation method; 2) arranging attribute supplemented historical load data samples according to the time rule, establishing an autoregressive moving average model through model identification and parameter estimation, acquiring a shortterm load prediction equation, and calculating a shortterm load prediction value; and 3) comparing the load prediction value with the actually acquired value, judging whether a current user has electric larceny possibility behavior or not, if so, sending an alarm of 'electric larceny possibility', otherwise, storing the actually acquired value in the electric load management system database, and returning to the step 1). Compared with the prior art, the method has the advantages of high prediction precision, high electric larceny suspect screening accuracy and the like.
Description
Technical field
The present invention relates to a kind of method of opposing electricitystealing, especially relate to a kind of method of excavating based on the power load Management System Data of opposing electricitystealing.
Background technology
Electricity filching behavior is the major issue that electric power enterprise runs into aspect marketing always, and it has also caused enormous economic loss to electric power enterprise when upsetting for the electricity consumption order.The means of opposing electricitystealing of the current overwhelming majority all concentrate in measurement loop and the electric energy meter meter prophylacticotherapeutic measures are set, or compare by the rough criterion of empirical data, lack stricter believable mathematical model as the basis.And, the user that power supply enterprise faces is numerous, the electricity filching behavior of Pan Dinging must be investigated further in practice just and can determine in theory, therefore, if can dwindle stealing user's suspicion list based on the data of current power load management system significantly, will bring great advantage to the pressure that alleviates the work of opposing electricitystealing.Therefore the requirement to the method for opposing electricitystealing is: can carry out the stealing suspicion judgement of power consumer based on the data of power load management system 1.; 2. can filter out stealing suspicion list rapidly and accurately and propose alarm; 3. can be by to having the stealing suspicious data to judge to improve the confidence level of power load Management System Data.
The final purpose of stealing is to make electric energy meter metering numeral less than actual power consumption by the whole bag of tricks, thereby reduces the electricity charge that institute should pay, and obtains the economic interests of reality, and different stealing modes has certain difference for the variation of customer charge or electric weight.At present, in order to reduce the metering numeral of electric energy meter, more common stealing electricity method mainly contains following several.
1) make that measurement loop opens circuit, short circuit, short circuit.
Such as a wherein phase of cutting off in 3 phase voltages, thereby will reduce by 1/3 actual power consumption.
2) change secondary connection.
Such as the position that changes in secondary voltage electric current access voltage transformer (VT) (PT), the current transformer (CT), it is reverse to cause calculating the gained electric weight, thereby has reduced total power consumption.
3) people is for making the table slow step, stopping walking.
Insert little pin such as artificially walking the circle place at electric energy meter, make that rotating speed is slackoff even stop, thereby reduce the electric weight counting.
4) make the table reversing with reduction table indicated number etc. for the CT reverse irrigated current.
That all electricity filching behaviors show or make electric energy meter not respond to electric power, or make electric power that electric energy meter senses and this load down the electric power during normal metering compare, produce the transition of a decline or the table indicated number descended unusually.In above commonplace stealing electricity method, some will cause the load of fixed proportion to reduce.The most common first method wherein, often causing loads be reduced to 0 or actual load 2/3.
Also there is the unusual reason of some meterings in addition, such as voltage transformer circuit fuse blows, loosening, the broken string of terminal screw, perhaps table is counted internal fault etc., though be not the subjective electricity filching behavior of user, but the actual result who produces also is the reduction of load, still can cause economically loss to electric power enterprise.
At present, the method of electric power enterprise investigation electricity filching behavior is more limited, have only when the user power utilization load has fairly obvious variation, electric power enterprise could come electricity filching behavior is judged and investigated and prosecuted according to corresponding experience and actual prospecting, but still have many electricity filching behaviors to escape punishment, thereby caused economic loss to Utilities Electric Co..Method more common and that actual effect is arranged has only following several at present.
1) method that adopts measurement loop and analog loopback to cooperatively interact.
For the user who always is subjected to, amounts to, beyond total metering cabinet, also have one and always be subjected to cabinet, from always being subjected to discretely the cabinet to go out one road analogue measurement loop, cooperatively interact with measurement loop, monitor mutually, by relatively obtaining the electric weight mistake, thereby obtain stealing information.But this method needs the extra measurement loop that increases, and is not suitable for the user who has only total, such as the Shanghai Power Network present case.And when the electricity filching behavior of the power circuit structure that changes, this method will lose efficacy.
2) method of employing intelligent remote kilowatt meter readingout system.
This method is mainly based on novel intelligent electric meter and longdistance meterreading system, need the ammeter whether phase failure of testing circuit on the spot, PT whether short circuit, CT whether shunt etc., should be used for detecting more exactly electricity filching behavior based on senior on these ammeters.But the spent amount of money of this method is higher, changes if carry out ammeter, not only needs a large amount of fund inputs, also needs the backstage of load management system is transformed; And from now on along with the understanding of user to intelligent electric meter, may have the generation of new stealing mode again, thereby cause the inefficacy of the method.
3) based on the method for coulometric analysis.
This method mainly be the existing electric weight of user and last time are lacked or historical same period electric weight carry out corresponding comparison, obtain the abnormal electrical numerical quantity, detect electricity filching behavior.But the criterion that this method is chosen at present is comparatively coarse, needs artificially to judge, and can be subjected to the influence of some specific factors, and cause error in judgement.What this method was used is the artificial experience conjecture, and one comes criterion too coarse, two to the randomness of load without any consideration.
γ represents autocovariance function,
${\mathrm{\γ}}_{k}={\mathrm{\Σ}}_{t=1}^{Nk}Y\left(t\right)Y(t+k)/N;$
106) order
Covariance function be
Calculate the estimated value of running mean coefficient by moments method:
5. a kind of method of excavating based on the power load Management System Data of opposing electricitystealing according to claim 4 is characterized in that described shortterm load forecasting equation is:
6. a kind of method of excavating based on the power load Management System Data of opposing electricitystealing according to claim 5 is characterized in that in the described step 3), judge whether the active user exists the concrete steps of stealing probable behavior to be:
301) warning factor r and warning window length W are set;
302) according to the load prediction value
yCalculate alarm threshold value y
_{Th}: y
_{Th}=r
y
303) realized load curve of more same time period and alarm threshold value curve, if meet the following conditions from n actual load value y (t) constantly:
Judge that then there is the stealing probable behavior in the active user, send the alarm of " stealing possibility ".
7. a kind of method of excavating based on the power load Management System Data of opposing electricitystealing according to claim 6 is characterized in that described warning factor r is: 1/2≤r≤2/3.
Be autoregressive coefficient and running mean coefficient, i=1,2 ..., p, j=1,2 ..., q.
Described Model Distinguish and parameter estimation are specially:
101) the historical load data sample of arranging according to the time is carried out null transformation:
Y (i) is i sample value, and Y (i) is the value after the corresponding conversion,
Be the mean value of all samples, N is sample size and N 〉=50;
102) calculate autocorrelation function ρ
_{k}Estimated value:
K represents the item number of autocorrelation function, if satisfy condition
Frequency less than 95%, autocorrelation function ρ then
_{k}Has truncation, as if the increase along with k, ρ
_{k}Trend towards 0, then autocorrelation function ρ
_{k}Has hangover;
103) calculate partial correlation function a by following recursion formula
_{Kk}Valuation:
J=1,2 ..., k is if satisfy condition
Frequency less than 95%, partial correlation function a then
_{Kk}Has truncation, as if the increase along with k, a
_{Kk}Trend towards 0, then partial correlation function a
_{Kk}Has hangover;
104) calculate autoregression exponent number p and running mean exponent number q according to autocorrelation function and partial correlation function property:
If ρ
_{k}Have hangover property and a
_{Kk}Have truncation, then maximum satisfies
The k value of condition is exactly the value of the p of model, q=0;
If ρ
_{k}Have truncation and a
_{Kk}Have hangover, then maximum satisfies
The k value of condition is exactly the value of the q of model, p=0;
If ρ
_{k}, a
_{Kk}All have truncation, then maximum satisfies
The k value of condition is exactly the value of the p of model, and maximum satisfies
The k value of condition is exactly the value of the q of model;
105) pass through the estimated value that the YuleWalker equation calculates autoregressive coefficient:
γ represents autocovariance function,
${\mathrm{\γ}}_{k}={\mathrm{\Σ}}_{t=1}^{Nk}Y\left(t\right)Y(t+k)/N;$
106) order
Covariance function be
Calculate the estimated value of running mean coefficient by moments method:
The variance of expression white noise sequence.
Described shortterm load forecasting equation is:
In the described step 3), judge whether the active user exists the concrete steps of stealing probable behavior to be:
301) warning factor r and warning window length W are set;
302) according to the load prediction value
yCalculate alarm threshold value y
_{Th}: y
_{Th}=r
y
303) realized load curve of more same time period and alarm threshold value curve, if meet the following conditions from n actual load value y (t) constantly:
Judge that then there is the stealing probable behavior in the active user, send the alarm of " stealing possibility ".
Described warning factor r is: 1/2≤r≤2/3.
Compared with prior art, the present invention has the following advantages:
1, the present invention utilizes historical data to carry out the stealing monitoring, comprises historical data consolidation and shortterm load forecasting, improves the accuracy of stealing suspicion screening, realizes the accurate judgement to power consumer stealing suspicion;
2, the present invention carries out the science modeling by the method for shortterm load forecasting to user data, effectively utilized historical data base, has improved the degree of accuracy of prediction significantly, and lays a good foundation for the intellectuality of opposing electricitystealing;
3, time series forecasting method of the present invention is carried out shortterm load forecasting, calculate simple relatively and predict the outcome more accurate.
Description of drawings
Fig. 1 is schematic flow sheet of the present invention;
Fig. 2 is the 51st day power load of certain user prediction synoptic diagram among the embodiment;
Fig. 3 is the 52nd day power load of certain user prediction synoptic diagram among the embodiment.
Embodiment
The present invention is described in detail below in conjunction with the drawings and specific embodiments.Present embodiment is that prerequisite is implemented with the technical solution of the present invention, provided detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment.
Embodiment
As shown in Figure 1, a kind of method of excavating based on the power load Management System Data of opposing electricitystealing, this method may further comprise the steps:
1) obtains user's historical load data in the power load Management System Data storehouse, and missing data is carried out consolidation by Newton interpolating method replenish;
2) the historical load data sample after according to temporal regularity consolidation being replenished is arranged, and sets up autoregressive movingaverage model by Model Distinguish and parameter estimation, obtains the shortterm load forecasting equation, calculates the shortterm load forecasting value;
3) compare load prediction value and actual acquisition value, judge whether the active user exists the stealing probable behavior, if the alarm of then sending " stealing possibility " if not, then is saved to the actual acquisition value power load Management System Data storehouse, returns step 1).
The feature of power load management system is once to gather every the load of power consumer of set time t, in general, the t of a lot of power load management systems is 15 minutes, also gathered 4 points in namely one hour, gathered 96 points in one day, the data consolidation that then these is collected is to historical data base.All customer data in power load Management System Data storehouse, owing to the reasons such as communicating interrupt that may exist, incomplete situations such as some bad datas, empty data will be arranged, these data flaws can have influence on the accuracy of data mining to a certain extent, therefore need to adopt suitable numerical computation method that it is improved and replenishes.
In the step 1), carry out consolidation by Newton interpolating method and replenish and to be specially:
x
_{0}, x
_{1}..., x
_{n}Be known n sampling time point, f (x
_{0}), f (x
_{1}) ..., f (x
_{n}) be corresponding sampled value, lacked sampled value f (x) at time point x, then
f(x)＝f(x
_{0})+f[x
_{0}，x
_{1}](xx
_{0})+…+f[x
_{0}，x
_{1}，…x
_{n}](xx
_{0})(xx
_{1})…(xx
_{n})+P
_{n}(x)
Wherein, f[x
_{0}, x
_{1}... x
_{n}] be the n jump merchant of function f,
f[x
_{0}，x
_{1}，…x
_{n}]＝(f[x
_{0}，x
_{1}，…x
_{n1}]f[x
_{1}，…x
_{n}])/(x
_{0}x
_{n})
On the data of the power load management system after the consolidation, this method is by selected load forecasting method and model, the user is carried out shortterm load forecasting, predicted value and observed reading are compared, by certain comparison means, set up rational decision threshold and warning window size, user's stealing suspicion is rationally judged, the suspicion name monoenergetic that produces reduces artificial screening stealing user's workload significantly, and can improve the accuracy of screening by the data analysis of science.
Traditional load forecasting method mainly contains dynamic method and mathematical statistics method, and these method common features are subjective model of setting up data sequence earlier, calculates then and predicts.Method mainly contains following several: least square fitting method, regression analysis, expert system approach, grey method, Time Series Method.Least square fitting method and regression analysis also are not suitable for shortterm load forecasting; The artificial neural network method is too complicated, is difficult for producing canonical solution in unified system; Though expert system approach is comparatively accurate in shortterm forecasting, need expend too many manpower and materials, and can't realize robotization, be not suitable for being applied in a large amount of electricity consumption clients' the system that opposes electricitystealing; That grey method is fit to a large amount of random quantitys and value is not suitable for sole user's random quantity prediction.Therefore, the present invention has selected to calculate the time series forecasting method uncomplicated relatively, that shortterm forecasting is relatively accurate, and the historical data after excavating by the power load management system is carried out modeling and prediction to the load of unique user, and the suspicion of opposing electricitystealing is again judged.
Because the electricity consumption client has certain production cycle or work habits and customs, therefore the work of handling in the customer charge data is arranged historical data according to temporal regularity, and some specific production firms are divided into " load on working day " and " nonworkdays load ".
In the time series forecasting method, the model of data fit is AR (autoregression) model, MA (running mean) model generally speaking, perhaps ARMA (autoregressive moving average) model, and wherein AR and MA model all are the special cases of arma modeling.Typical autoregressive movingaverage model is:
Wherein, y (tp), y (tp+1) ..., y (t1), y (t) they are the load data values of continuous a period of time; α
_{t}, α
_{T1}..., α
_{0}Be white noise sequence (concrete probability distribution will be decided on data, generally is assumed to be white Gaussian noise,
P, q are respectively autoregression exponent number and running mean exponent number;
And θ
_{j}Be respectively autoregressive coefficient and running mean coefficient, i=1,2 ..., p, j=1,2 ..., q.
After with the data tranquilization, can its identification and parameter estimation of carrying out arma modeling be specially:
101) the historical load data sample of arranging according to the time is carried out null transformation:
Y (i) is i sample value, and Y (i) is the value after the corresponding conversion,
Be the mean value of all samples, N is sample size and N 〉=50;
102) calculate autocorrelation function ρ
_{k}Estimated value:
K represents the item number of autocorrelation function, if satisfy condition
Frequency less than 95%, autocorrelation function ρ then
_{k}Has truncation, as if the increase along with k, ρ
_{k}Trend towards 0, then autocorrelation function ρ
_{k}Has hangover;
103) calculate partial correlation function a by following recursion formula
_{Kk}Valuation:
J=1,2 ..., k is if satisfy condition
Frequency less than 95%, partial correlation function a then
_{Kk}Has truncation, as if the increase along with k, a
_{Kk}Trend towards 0, then partial correlation function a
_{Kk}Has hangover;
104) calculate autoregression exponent number p and running mean exponent number q according to autocorrelation function and partial correlation function property:
If ρ
_{k}Have hangover property and a
_{Kk}Have truncation, then maximum satisfies
The k value of condition is exactly the value of the p of model, q=0;
If ρ
_{k}Have truncation and a
_{Kk}Have hangover, then maximum satisfies
The k value of condition is exactly the value of the q of model, p=0;
If ρ
_{k}, a
_{Kk}All have truncation, then maximum satisfies
The k value of condition is exactly the value of the p of model, and maximum satisfies
The k value of condition is exactly the value of the q of model;
105) pass through the estimated value that the YuleWalker equation calculates autoregressive coefficient:
γ represents autocovariance function, for
${\mathrm{\γ}}_{k}={\mathrm{\Σ}}_{t=1}^{Nk}Y\left(t\right)Y(t+k)/N;$
106) order
Covariance function be
Calculate the estimated value of running mean coefficient by moments method:
So far finish the foundation of arma modeling, draw the equation of shortterm load forecasting, carry out the shortterm load forecasting of power consumer again by the historical data base of power load management system.
Described shortterm load forecasting equation is:
If the customer charge data can't satisfy " hangover " and " truncation " property of autocorrelation function and partial correlation function, then can carry out the periodicity difference to data, judge " hangover " and " truncation " property of autocorrelation function and the partial correlation function of data after the difference again.Differentiated data are observed " hangover " and " truncation " property of autocorrelation function and partial correlation function again, and then pass through the Time Series Method modeling, and finish the shortterm load forecasting of this power consumer.
The data sequence of supposing current collection is y (i), i ∈ [1, T+t], and the value that predict is that y (T+t+1) then is that the difference method of periodic term is as follows with T:
z(i)＝y(T+i)y(i)，i∈[1，t]
Z (i), i=1,2 ..., t is differentiated data, by time series z (i) is carried out modeling, dopes z (t+1), and then returns the predicted value of y (T+t+1) by following calculating:
y(T+t+1)＝y(t+1)+z(t+1)。
For the sole user, electricity filching behavior has two tangible indexs, first powersteeling quantity, and it two is stealing durations.Whether the described active user of judgement exists the concrete steps of stealing probable behavior to be:
301) warning factor r and warning window length W are set;
302) according to the load prediction value
yCalculate alarm threshold value y
_{Th}: y
_{Th}=r
y
303) realized load curve of more same time period and alarm threshold value curve, if meet the following conditions from n actual load value y (t) constantly:
Judge that then there is the stealing probable behavior in the active user, send the alarm of " stealing possibility ".In case the alarm of " stealing possibility " appears in system, then work as day data and will be considered to " insincere ", namely fict load data can not be entered into the database of predicting usefulness; Otherwise the data of " credible " then finish the database of back typing prediction usefulness in every day, are used for predicting the power load of next day.
According to the actual conditions of power consumer warning factor r is carried out one and reasonably arrange, as 1/2≤r≤2/3.The setting of warning window is because sole user's power load has bigger randomness, if each actual value is lower than the warning that alarm threshold value just carries out " stealing possibility ", may cause system frequently to report to the police, and reduces the stealing accuracy for predicting.In general, the stealing user is because the loaded down with trivial details property of stealing work, can the very short time of a stealing, the worker also can carry out a reasonably setting to warning window W according to actual conditions and the historical data of locality so oppose electricitystealing, W is too small, then the wrong forecast rate of opposing electricitystealing of system will reduce, and W is excessive, and then the Lou forecast rate of opposing electricitystealing of system will increase.
Based on the continuous 50 days actual power load of certain user in the power load management system, obtained the 51st day power load prediction as Fig. 2, and compare with the 51st day realized load curve, get r=2/3, W=8 (once sampled in per 15 minutes by this power load management system, gathered 96 points in one day altogether, sample window length is 8 namely to represent 2 hours), the prediction load data is comparatively close with the actual load data.The 51st day actual load data do not reach and cause the requirement of " stealing may " alarm, and to think the 51st day be believable in system accordingly, and the 51st day data are included into historical data concentrate.
History data set according to after upgrading carries out shortterm load forecasting again, obtains the 52nd day prediction load prediction curve, as shown in Figure 3.Just can find that by setting before preceding 24 sampled points in the realized load curve all are starkly lower than red alarm threshold curve, reach the requirement of " stealing possibility " alarm, 52 sampled datas afterwards are then approaching substantially with predicted value.Learn that by comparative analysis there is bigger stealing suspicion in this user in 6 hours of 0 o'clock to 6 o'clock morning of the 52nd day, so the data of this day are listed in " insincere ", are not included into historical data and concentrate.According to The numerical results, this user electricity consumption behavior during this period of time is locked as " stealing suspicion ".
Claims (7)
1. method of excavating based on the power load Management System Data of opposing electricitystealing is characterized in that this method may further comprise the steps:
1) obtains user's historical load data in the power load Management System Data storehouse, and missing data is carried out consolidation by Newton interpolating method replenish;
2) the historical load data sample after according to temporal regularity consolidation being replenished is arranged, and sets up autoregressive movingaverage model by Model Distinguish and parameter estimation, obtains the shortterm load forecasting equation, calculates the shortterm load forecasting value;
3) compare load prediction value and actual acquisition value, judge whether the active user exists the stealing probable behavior, if the alarm of then sending " stealing possibility " if not, then is saved to the actual acquisition value power load Management System Data storehouse, returns step 1).
2. a kind of method of excavating based on the power load Management System Data of opposing electricitystealing according to claim 1 is characterized in that, describedly carries out consolidation by Newton interpolating method and replenishes and be specially:
x
_{0}, x
_{1}..., x
_{n}Be known n sampling time point, f (x
_{0}), f (x
_{1}) ..., f (x
_{n}) be corresponding sampled value, lacked sampled value f (x) at time point x, then
f(x)＝f(x
_{0})+f[x
_{0}，x
_{1}](xx
_{0})+…+f[x
_{0}，x
_{1}，…x
_{n}](xx
_{0})(xx
_{1})…(xx
_{n})+P
_{n}(x)
Wherein, f[x
_{0}, x
_{1}... x
_{n}] be the n jump merchant of function f,
f[x
_{0}，x
_{1}，…x
_{n}]＝(f[x
_{0}，x
_{1}，…x
_{n1}]f[x
_{1}，…x
_{n}])/(x
_{0}x
_{n})
3. a kind of method of excavating based on the power load Management System Data of opposing electricitystealing according to claim 1 is characterized in that described autoregressive movingaverage model is specially:
Wherein, y (tp), y (tp+1) ..., y (t1), y (t) they are the load data values of continuous a period of time; α
_{t}, α
_{T1}..., α
_{0}It is white noise sequence; P, q are respectively autoregression exponent number and running mean exponent number;
And θ
_{j}Be respectively autoregressive coefficient and running mean coefficient, i=1,2 ..., p, j=1,2 ..., q.
4. a kind of method of excavating based on the power load Management System Data of opposing electricitystealing according to claim 3 is characterized in that described Model Distinguish and parameter estimation are specially:
101) the historical load data sample of arranging according to the time is carried out null transformation:
Y (i) is i sample value, and Y (i) is the value after the corresponding conversion,
Be the mean value of all samples, N is sample size and N 〉=50;
102) calculate autocorrelation function ρ
_{k}Estimated value:
K represents the item number of autocorrelation function, if satisfy condition
Frequency less than 95%, autocorrelation function ρ then
_{k}Has truncation, as if the increase along with k, ρ
_{k}Trend towards 0, then autocorrelation function ρ
_{k}Has hangover;
103) calculate partial correlation function a by following recursion formula
_{Kk}Valuation:
J=1,2 ..., k is if satisfy condition
Frequency less than 95%, partial correlation function a then
_{Kk}Has truncation, as if the increase along with k, a
_{Kk}Trend towards 0, then partial correlation function a
_{Kk}Has hangover;
104) calculate autoregression exponent number p and running mean exponent number q according to autocorrelation function and partial correlation function property:
If ρ
_{k}Have hangover property and a
_{Kk}Have truncation, then maximum satisfies
The k value of condition is exactly the value of the p of model, q=0;
If ρ
_{k}Have truncation and a
_{Kk}Have hangover, then maximum satisfies
The k value of condition is exactly the value of the q of model, p=0;
If ρ
_{k}, a
_{Kk}All have truncation, then maximum satisfies
The k value of condition is exactly the value of the p of model, and maximum satisfies
The k value of condition is exactly the value of the q of model;
105) pass through the estimated value that the YuleWalker equation calculates autoregressive coefficient:
γ represents autocovariance function,
${\mathrm{\γ}}_{k}={\mathrm{\Σ}}_{t=1}^{Nk}Y\left(t\right)Y(t+k)/N.$
106) order
Covariance function be
Calculate the estimated value of running mean coefficient by moments method:
6. a kind of method of excavating based on the power load Management System Data of opposing electricitystealing according to claim 5 is characterized in that in the described step 3), judge whether the active user exists the concrete steps of stealing probable behavior to be:
301) warning factor r and warning window length W are set;
302) according to the load prediction value
yCalculate alarm threshold value y
_{Th}: y
_{Th}=r
y
303) realized load curve of more same time period and alarm threshold value curve, if meet the following conditions from n actual load value y (t) constantly:
Judge that then there is the stealing probable behavior in the active user, send the alarm of " stealing possibility ".
7. a kind of method of excavating based on the power load Management System Data of opposing electricitystealing according to claim 6 is characterized in that described warning factor r is: 1/2≤r≤2/3.
Priority Applications (1)
Application Number  Priority Date  Filing Date  Title 

CN201310148105.7A CN103208091B (en)  20130425  20130425  A kind of method of opposing electricitystealing excavated based on power load Management System Data 
Applications Claiming Priority (1)
Application Number  Priority Date  Filing Date  Title 

CN201310148105.7A CN103208091B (en)  20130425  20130425  A kind of method of opposing electricitystealing excavated based on power load Management System Data 
Publications (2)
Publication Number  Publication Date 

CN103208091A true CN103208091A (en)  20130717 
CN103208091B CN103208091B (en)  20160127 
Family
ID=48755308
Family Applications (1)
Application Number  Title  Priority Date  Filing Date 

CN201310148105.7A Active CN103208091B (en)  20130425  20130425  A kind of method of opposing electricitystealing excavated based on power load Management System Data 
Country Status (1)
Country  Link 

CN (1)  CN103208091B (en) 
Cited By (27)
Publication number  Priority date  Publication date  Assignee  Title 

CN103455855A (en) *  20130912  20131218  国家电网公司  Intermittent electricity stealing prevention detecting method based on electricity consumption information data analysis 
CN103792420A (en) *  20140126  20140514  威胜集团有限公司  Electricity larceny preventing and electricity utilization monitoring method based on load curves 
CN104361233A (en) *  20141111  20150218  国家电网公司  Antielectric larceny management method under condition of access of distributed generation 
CN104765015A (en) *  20150416  20150708  国家电网公司  Method for rapidly testing measurement wiring correctness by means of data integration method 
CN105510656A (en) *  20160201  20160420  南安普敦咨询服务有限公司  Intelligent electricity meter 
CN105866511A (en) *  20160527  20160817  国电南瑞科技股份有限公司  Method for recognizing trailing currents based on adjacent sampling point specific values of current wave forms 
CN106019087A (en) *  20160720  20161012  国网上海市电力公司  Intermittent electricity stealing monitoring system 
CN106093707A (en) *  20160712  20161109  亿米特（上海）信息科技有限公司  Intelligent electricity antitheft analyzes the data processing method of system 
CN106127602A (en) *  20160615  20161116  国网江苏省电力公司电力科学研究院  A kind of stealing discrimination method based on yojan outlier algorithm and device 
CN106384298A (en) *  20160919  20170208  合肥工业大学  Method for correcting missing data in intelligent power utilization based on twostage interpolation model 
CN107196953A (en) *  20170614  20170922  上海丁牛信息科技有限公司  A kind of anomaly detection method based on user behavior analysis 
CN107305611A (en) *  20160422  20171031  腾讯科技（深圳）有限公司  The corresponding method for establishing model of malice account and device, the method and apparatus of malice account identification 
CN107464015A (en) *  20170721  20171212  南京林洋电力科技有限公司  A kind of short term and power predicating method that possess fusing mechanism based on electricity consumption behavior 
CN107807309A (en) *  20171027  20180316  广东电网有限责任公司中山供电局  A kind of transmission line malfunction method for early warning and system based on big data 
CN108197920A (en) *  20171228  20180622  福州信诺通信息技术有限公司  A kind of gas meter reading method for detecting abnormality and storage medium 
CN108416693A (en) *  20180201  20180817  东华大学  A kind of multiplexing electric abnormality Activity recognition method based on fuzzy neural network 
CN108898120A (en) *  20180704  20181127  四川泰立智汇科技有限公司  Water cooler method for diagnosing faults based on hybrid neural networks 
CN109345013A (en) *  20180925  20190215  广西电网有限责任公司电力科学研究院  A kind of distribution user power utilization behavior prediction method 
CN109409661A (en) *  20180919  20190301  平安科技（深圳）有限公司  Electronic device, the building Electrical Safety recognition methods based on cusum algorithm and storage medium 
CN109784388A (en) *  20181229  20190521  北京中电普华信息技术有限公司  Stealing user identification method and device 
CN111049195A (en) *  20200103  20200421  云南电网有限责任公司电力科学研究院  AGC parameter optimization method and device 
CN111178618A (en) *  20191225  20200519  国网天津市电力公司电力科学研究院  Intelligent power grid load prediction method 
CN111245926A (en) *  20200109  20200605  浙江中新电力工程建设有限公司自动化分公司  Working condition load analysis system based on ubiquitous power sensor Internet of things 
CN112215721A (en) *  20200904  20210112  国网青海省电力公司信息通信公司  Electric power electricity stealing accurate identification and electricity stealing feedback data analysis model based on big data 
CN113406558A (en) *  20210729  20210917  天目数据(福建)科技有限公司  Electric meter misalignment detection method and device based on linear regression and electronic equipment 
CN113514695A (en) *  20210304  20211019  华北电力大学  Detection system and detection method suitable for group fixed proportion electricity stealing behavior 
CN116933986A (en) *  20230919  20231024  国网湖北省电力有限公司信息通信公司  Electric power data safety management system based on deep learning 
Citations (5)
Publication number  Priority date  Publication date  Assignee  Title 

CN1442697A (en) *  20020305  20030917  刘汪  Device for anti stealing electricity 
CN200953031Y (en) *  20060922  20070926  刘志  Portable detecting instrument for indicating fraudulent use of electricity 
CN102708427A (en) *  20120629  20121003  山东电力集团公司电力科学研究院  System and method for analyzing and determining influence of largescale charging pile to community distribution system 
CN102866321A (en) *  20120813  20130109  广东电网公司电力科学研究院  Selfadaptive stealingleakage prevention diagnosis method 
CN202676777U (en) *  20120719  20130116  鹿波  Antielectricitytheft remote monitoring and management system 

2013
 20130425 CN CN201310148105.7A patent/CN103208091B/en active Active
Patent Citations (5)
Publication number  Priority date  Publication date  Assignee  Title 

CN1442697A (en) *  20020305  20030917  刘汪  Device for anti stealing electricity 
CN200953031Y (en) *  20060922  20070926  刘志  Portable detecting instrument for indicating fraudulent use of electricity 
CN102708427A (en) *  20120629  20121003  山东电力集团公司电力科学研究院  System and method for analyzing and determining influence of largescale charging pile to community distribution system 
CN202676777U (en) *  20120719  20130116  鹿波  Antielectricitytheft remote monitoring and management system 
CN102866321A (en) *  20120813  20130109  广东电网公司电力科学研究院  Selfadaptive stealingleakage prevention diagnosis method 
NonPatent Citations (1)
Title 

张伊辉等: "电力负荷管理终端防电流法窃电装置研制", 《电力自动化设备》 * 
Cited By (37)
Publication number  Priority date  Publication date  Assignee  Title 

CN103455855B (en) *  20130912  20160615  国家电网公司  A kind of intermittent electricity stealing prevention detection method based on power information data analysis 
CN103455855A (en) *  20130912  20131218  国家电网公司  Intermittent electricity stealing prevention detecting method based on electricity consumption information data analysis 
CN103792420A (en) *  20140126  20140514  威胜集团有限公司  Electricity larceny preventing and electricity utilization monitoring method based on load curves 
CN104361233A (en) *  20141111  20150218  国家电网公司  Antielectric larceny management method under condition of access of distributed generation 
CN104765015A (en) *  20150416  20150708  国家电网公司  Method for rapidly testing measurement wiring correctness by means of data integration method 
CN105510656A (en) *  20160201  20160420  南安普敦咨询服务有限公司  Intelligent electricity meter 
CN107305611A (en) *  20160422  20171031  腾讯科技（深圳）有限公司  The corresponding method for establishing model of malice account and device, the method and apparatus of malice account identification 
CN107305611B (en) *  20160422  20201103  腾讯科技（深圳）有限公司  Method and device for establishing model corresponding to malicious account and method and device for identifying malicious account 
CN105866511A (en) *  20160527  20160817  国电南瑞科技股份有限公司  Method for recognizing trailing currents based on adjacent sampling point specific values of current wave forms 
CN106127602B (en) *  20160615  20191224  国网江苏省电力公司电力科学研究院  Electricity stealing identification method and device based on reduction outlier algorithm 
CN106127602A (en) *  20160615  20161116  国网江苏省电力公司电力科学研究院  A kind of stealing discrimination method based on yojan outlier algorithm and device 
CN106093707B (en) *  20160712  20181130  亿米特（上海）信息科技有限公司  The data processing method of intelligent electricity antitheft analysis system 
CN106093707A (en) *  20160712  20161109  亿米特（上海）信息科技有限公司  Intelligent electricity antitheft analyzes the data processing method of system 
CN106019087A (en) *  20160720  20161012  国网上海市电力公司  Intermittent electricity stealing monitoring system 
CN106384298B (en) *  20160919  20190531  合肥工业大学  A kind of intelligent power missing data modification method based on two stages interpolation model 
CN106384298A (en) *  20160919  20170208  合肥工业大学  Method for correcting missing data in intelligent power utilization based on twostage interpolation model 
CN107196953B (en) *  20170614  20200508  上海境领信息科技有限公司  Abnormal behavior detection method based on user behavior analysis 
CN107196953A (en) *  20170614  20170922  上海丁牛信息科技有限公司  A kind of anomaly detection method based on user behavior analysis 
CN107464015A (en) *  20170721  20171212  南京林洋电力科技有限公司  A kind of short term and power predicating method that possess fusing mechanism based on electricity consumption behavior 
CN107807309A (en) *  20171027  20180316  广东电网有限责任公司中山供电局  A kind of transmission line malfunction method for early warning and system based on big data 
CN108197920A (en) *  20171228  20180622  福州信诺通信息技术有限公司  A kind of gas meter reading method for detecting abnormality and storage medium 
CN108416693A (en) *  20180201  20180817  东华大学  A kind of multiplexing electric abnormality Activity recognition method based on fuzzy neural network 
CN108898120A (en) *  20180704  20181127  四川泰立智汇科技有限公司  Water cooler method for diagnosing faults based on hybrid neural networks 
CN109409661A (en) *  20180919  20190301  平安科技（深圳）有限公司  Electronic device, the building Electrical Safety recognition methods based on cusum algorithm and storage medium 
CN109345013A (en) *  20180925  20190215  广西电网有限责任公司电力科学研究院  A kind of distribution user power utilization behavior prediction method 
CN109784388A (en) *  20181229  20190521  北京中电普华信息技术有限公司  Stealing user identification method and device 
CN111178618A (en) *  20191225  20200519  国网天津市电力公司电力科学研究院  Intelligent power grid load prediction method 
CN111049195A (en) *  20200103  20200421  云南电网有限责任公司电力科学研究院  AGC parameter optimization method and device 
CN111049195B (en) *  20200103  20231027  云南电网有限责任公司电力科学研究院  AGC parameter optimization method and device 
CN111245926A (en) *  20200109  20200605  浙江中新电力工程建设有限公司自动化分公司  Working condition load analysis system based on ubiquitous power sensor Internet of things 
CN112215721A (en) *  20200904  20210112  国网青海省电力公司信息通信公司  Electric power electricity stealing accurate identification and electricity stealing feedback data analysis model based on big data 
CN113514695A (en) *  20210304  20211019  华北电力大学  Detection system and detection method suitable for group fixed proportion electricity stealing behavior 
CN113514695B (en) *  20210304  20220506  华北电力大学  Detection system and detection method suitable for group fixed proportion electricity stealing behavior 
CN113406558A (en) *  20210729  20210917  天目数据(福建)科技有限公司  Electric meter misalignment detection method and device based on linear regression and electronic equipment 
CN113406558B (en) *  20210729  20240409  天目数据(福建)科技有限公司  Ammeter misalignment detection method and device based on linear regression and electronic equipment 
CN116933986A (en) *  20230919  20231024  国网湖北省电力有限公司信息通信公司  Electric power data safety management system based on deep learning 
CN116933986B (en) *  20230919  20240123  国网湖北省电力有限公司信息通信公司  Electric power data safety management system based on deep learning 
Also Published As
Publication number  Publication date 

CN103208091B (en)  20160127 
Similar Documents
Publication  Publication Date  Title 

CN103208091B (en)  A kind of method of opposing electricitystealing excavated based on power load Management System Data  
CN110097297B (en)  Multidimensional electricity stealing situation intelligent sensing method, system, equipment and medium  
CN110223196B (en)  Antielectricitystealing analysis method based on typical industry feature library and antielectricitystealing sample library  
CN107742127B (en)  Improved electricity stealing prevention intelligent early warning system and method  
CN106093652B (en)  A kind of nonintrusive electrical load monitoring System and method for having selflearning function  
Chicco et al.  Emergent electricity customer classification  
EP2529186B1 (en)  Robust automated determination of the hierarchical structure of utility monitoring systems  
Nizar et al.  Load profiling method in detecting nontechnical loss activities in a power utility  
CN104407268A (en)  Abnormal electricity utilization judgment method based on abnormal analysis of electric quantity, voltage and current  
CN108765004A (en)  A method of user's electricity stealing is identified based on data mining  
CN102866321A (en)  Selfadaptive stealingleakage prevention diagnosis method  
CN110749784B (en)  Line electricity stealing detection method based on electric power data wavelet analysis  
CN103886518A (en)  Early warning method for voltage sag based on electric energy quality data mining at monitoring point  
CN104239963A (en)  Method for finding abnormal electric energy meter based on gray GM (1, 1) model  
CN113673580A (en)  Electricity stealing suspicion judgment method based on big data modeling  
CN110244099A (en)  Stealing detection method based on user's voltage  
CN105844351A (en)  Prosperity index prediction method for electric power consumption market  
Hartmann et al.  Suspicious electric consumption detection based on multiprofiling using live machine learning  
CN116882804A (en)  Intelligent power monitoring method and system  
Jaramillo et al.  Nonintrusive load monitoring algorithm for PV identification in the residential sector  
EP2840402B1 (en)  Operating state estimating device for electrical apparatus, operating state estimating method for electrical apparatus, program, feature vector registration device for electrical apparatus, feature vector registration method for electrical apparatus, server apparatus, and operating state estimating system for electrical apparatus  
CN103208036A (en)  Power consumer databased shortterm load forecast method  
KR102332248B1 (en)  Performance evaluation system for Nonintrusive load monitoring according to multistate energy classification and method thereof  
CN111506636A (en)  System and method for analyzing residential electricity consumption behavior based on autoregressive and neighbor algorithm  
CN117093943A (en)  Power consumption monitoring and early warning method and device 
Legal Events
Date  Code  Title  Description 

C06  Publication  
PB01  Publication  
C10  Entry into substantive examination  
SE01  Entry into force of request for substantive examination  
C14  Grant of patent or utility model  
GR01  Patent grant 