CN111310854A - Low false alarm rate electricity stealing detection secondary screening method based on electricity load fingerprint - Google Patents
Low false alarm rate electricity stealing detection secondary screening method based on electricity load fingerprint Download PDFInfo
- Publication number
- CN111310854A CN111310854A CN202010160079.XA CN202010160079A CN111310854A CN 111310854 A CN111310854 A CN 111310854A CN 202010160079 A CN202010160079 A CN 202010160079A CN 111310854 A CN111310854 A CN 111310854A
- Authority
- CN
- China
- Prior art keywords
- electricity
- power
- load
- electricity stealing
- power consumption
- 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 200
- 238000001514 detection method Methods 0.000 title claims abstract description 32
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000012216 screening Methods 0.000 title claims abstract description 17
- 238000004458 analytical method Methods 0.000 claims description 11
- 238000004519 manufacturing process Methods 0.000 abstract description 43
- 230000008859 change Effects 0.000 abstract description 15
- 238000007689 inspection Methods 0.000 abstract description 13
- 230000007613 environmental effect Effects 0.000 abstract description 3
- 238000004422 calculation algorithm Methods 0.000 description 13
- 230000035772 mutation Effects 0.000 description 9
- 230000002354 daily effect Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 239000004753 textile Substances 0.000 description 4
- 230000009467 reduction Effects 0.000 description 3
- 230000005856 abnormality Effects 0.000 description 2
- 238000007796 conventional method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000003203 everyday effect Effects 0.000 description 2
- 238000012804 iterative process Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000001052 transient effect Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000010224 classification analysis Methods 0.000 description 1
- 238000007621 cluster analysis Methods 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000009191 jumping Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008450 motivation Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000004043 responsiveness Effects 0.000 description 1
- 230000008685 targeting Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000005612 types of electricity Effects 0.000 description 1
- 238000004804 winding Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24143—Distances to neighbourhood prototypes, e.g. restricted Coulomb energy networks [RCEN]
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (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
A low false alarm rate electricity stealing detection secondary screening method based on electricity load fingerprints is used for carrying out secondary screening on electricity stealing users (namely suspected electricity stealing users) detected by the existing electricity stealing detection. Clustering power consumption of suspected electricity stealing users or power consumption load fingerprints before and after sudden change of line loss of the suspected electricity stealing users connected to a distribution line/a distribution transformer, then comparing whether the power consumption mode corresponding to the clustering class of the power consumption load fingerprints before and after the sudden change time point changes, and if the sudden change power consumption mode is compared with a new power consumption mode before the sudden change, judging that the suspected electricity stealing users are electricity stealing users; otherwise, the electricity consumption is normal, and the sudden change of the electricity consumption is caused by the fact that the user enters a normal production operation state with low electricity consumption. The method can identify sudden change of the power consumption of the production stop user caused by reasons of power limitation, environmental protection inspection, safety inspection and the like, can reduce the false alarm rate of judgment of the conventional electricity stealing detection method, and promotes the practical engineering application of electricity stealing detection.
Description
Technical Field
The invention belongs to the field of power utilization inspection of an electric power system, relates to a secondary screening method for power-stealing users detected by the conventional power-stealing detection, in particular to a secondary screening method for low false alarm rate power-stealing detection based on power consumption load fingerprints, and is mainly used for power consumption abnormity detection of large-scale industrial and commercial users.
Background
Traditionally, detecting electricity stealing behavior requires technicians of power supply enterprises to step on the door to conduct electricity inspection, and manpower and material resource consumption is huge. The intelligent electric meter can accurately record and report the electricity consumption data of the user at a high frequency, and the electricity stealing behavior can be identified through the electricity consumption data recorded by the metering system, so that a key foundation is laid for accurately identifying the electricity stealing behavior by adopting a data driving technology.
The existing method for detecting and identifying electricity stealing based on the metering data of the intelligent electric meter can be divided into two categories, namely a simple physical rule and a clustering/classification analysis-based category:
the electricity stealing detection method based on the simple physical rule adopts the simple rule to judge the electricity stealing behavior according to the characteristics of the user electric line structure and the electric meter wiring. For example, when zero sequence current, user current and power reversal occur in a low-voltage single-phase user, the low-voltage single-phase user can be determined as a power stealing user;
regardless of how the user steals electricity, the motivation is to reduce the electricity consumption to earn the difference in electricity consumption, which is shown as a significant reduction in electricity consumption on the load curve. The clustering-based method is based on the obvious difference between the load curve form of the electricity stealing users and normal users, and defines the clustering index by combining the characteristic that the electricity consumption of the electricity stealing users has mutation so as to identify the electricity stealing users. Similarly, classification-based detection of electricity stealing is also based on comparing detected load curves of electricity stealing users as templates to identify electricity stealing.
In the two types of electricity stealing detection methods, the former has an exact physical concept and can accurately identify and position electricity stealing users, but the application range is limited, and a considerable proportion of electricity stealing (such as electricity connection by winding a meter, private change of transformer transformation ratio and the like) cannot be directly judged according to simple physical rules. The core idea of the clustering/classification-based electricity stealing detection is that the electricity load curve of an electricity stealing user drops suddenly, but the sudden change of the user electricity consumption can be caused by the reasons of equipment overhaul, production halt reconstruction, fire control modification, environmental protection inspection, safety inspection and the like in actual production, and the sudden drop of the user electricity consumption caused by electricity stealing is easily confused with the sudden drop of the electricity consumption. Therefore, the method for detecting electricity stealing based on the electricity load mutation criterion generally has the problem of high misjudgment rate in engineering application. When the actual power grid utilizes the clustering/classifying algorithm to carry out electricity stealing detection, the misjudgment rate is even higher than 80%, the popularization and application of electricity stealing detection based on the clustering/classifying algorithm are seriously hindered by the problem that electricity stealing criterion adopted by the electricity stealing detection algorithm is easy to misjudge, and a method for carrying out secondary screening on electricity stealing users detected by electricity stealing detection to eliminate normal users and reduce the misjudgment rate of electricity stealing detection is urgently needed so as to improve the practicability level of intelligent electricity stealing detection.
In addition, because the electricity stealing users directly access the line loss electricity of the distribution transformer/distribution line corresponding to the electricity stealing users, research is also carried out to take all the users accessing the distribution transformer/distribution line as the electricity stealing users according to the obvious increase of the line loss electricity of the distribution transformer/distribution line, and the detection target at the moment is to judge whether the users are the electricity stealing users according to the load data of the users before and after the line loss electricity of the distribution transformer/distribution line changes suddenly.
The non-invasive load analysis is a research hotspot in recent years, and is a method for identifying the power load constitution and the consumption behavior habit according to the user power consumption data recorded by the intelligent electric meter. The non-invasive load analysis method can analyze the electric load composition of a user according to transient load fingerprints (jumping amplitude steepness, starting transient harmonic wave, pulse characteristics and the like) and steady load fingerprints (power factor, harmonic wave content, load curve form, duration, power magnitude and the like), and further identify deep information such as electric consumption habits of the user. At present, a new type of non-invasive load analyzing electricity meter has appeared which can identify the user load composition from the metering data and measure the power consumption of each load.
Disclosure of Invention
Referring to the idea of non-invasive load analysis, it is possible to use the three-phase load, three-phase power consumption, three-phase power factor, three-phase harmonic content, three-phase current and the like of the power consumer as the power load fingerprint, and identify the production and operation state of the power consumer according to the power load fingerprint. For large industry and general industrial and commercial users, the electric load composition, production operation and the production electric equipment which is started under different production operation states are basically determined. The production management scheduling plan is generally made by daily business users, and the electric equipment used in a production management state is basically fixed, so that indexes such as three-phase power, three-phase electricity consumption, three-phase current, three-phase power factor, harmonic content and the like recorded at intervals of different time periods (such as 15 minutes, 30 minutes or 60 minutes) in a day can be used as electricity load fingerprints to represent the production management state of the users on the same day, for example, the production management state of the users can be represented according to a single electricity load fingerprint (such as 96 three-phase power data recorded at intervals of 15 minutes every day) or according to a combination of a plurality of electricity load fingerprints (such as three-phase power and three-phase power factor recorded at intervals of 15 minutes). Because the power consumption load fingerprints corresponding to the power consumption modes under the same production and operation state are similar, clustering can be performed according to the power consumption load fingerprints of the user every day in a period of time (such as one year), and the power consumption modes and the corresponding production and operation state of the user can be judged. The normal users are always switched between limited operation states, and at the moment, even if the power consumption is suddenly reduced, the production operation state with large power consumption is switched to the production operation state with low power consumption from the view of load fingerprints. However, electricity stealing not only causes the reduction of electricity consumption, but also causes the sudden change of load fingerprint structure, so that a new category different from that in the normal production and operation state is formed when the power load fingerprint clustering is carried out.
For a single user, the number of the production operation states of the user is always limited, and the normal production operation states of different modes such as normal production, partial shutdown and maintenance, vacation shutdown, rush season driving, slack season half vacation and the like can be traversed all the time. In each production and operation state, relatively fixed electric equipment is put into use and combined to form a corresponding power utilization mode. Under the production and operation state of long-term false shutdown, only low-power single-phase load is used throughout the day, and at the moment, the harmonic content and the power level are not high, but the load can be concentrated on a certain phase, so that the load has high three-phase load asymmetry; under the normal production and operation state, a large amount of high-power three-phase symmetrical electrical equipment is probably put into use in the daytime, only low-power single-phase lighting load is reserved at night, the symmetry, the load level and the harmonic content of three-phase power corresponding to the daytime period are higher, and the symmetry, the load level and the harmonic content of the three-phase power are lower at night; under the condition of high-season production and operation, the symmetry, the load level and the harmonic content of three-phase power possibly all day long are higher. In addition, the user can mark the production operation state of the user by using equipment under different production operation states and three-phase power factors of different time periods (such as working time periods and night rest time periods) in one day determined by reactive compensation switching. When the user switches between different production and operation states, the daily electric quantity is reduced, and the actual situation is reflected on the corresponding power consumption load fingerprint. According to the clustering of the power load fingerprints, the production operation state of the user can be identified.
When the user steals electricity, part or all of the electricity load electricity is not metered by the ammeter. On one hand, the measured electric quantity is reduced; on the other hand, indexes such as three-phase load/current level, power factor, harmonic content and the like reflected by the metering data of the electric meter obviously deviate from the original normal production operation state in the state space. Therefore, the power consumption load fingerprints of the power consumption users detected by the existing power consumption detection can be clustered, and the power consumption load fingerprints are clustered into categories for identifying different power consumption modes through clustering analysis, so that the difference between the production operation state of the users and the production operation state of the users in the normal power consumption period when the power consumption drops abnormally is diagnosed. If the category of the power consumption load fingerprint clustering in the abnormal power consumption period is included in the category of the power consumption load fingerprint clustering in the normal power consumption period, the sudden drop of the power consumption is considered to be caused by the normal production and operation state change of the user, and no alarm and power stealing inspection are needed; if the power load fingerprint after the sudden drop of the power consumption forms a new cluster type, the user is identified as a power stealing user, so that the sudden drop of the power consumption caused by external interference is identified, and the false alarm rate of the power stealing alarm is reduced.
Therefore, the invention aims to provide a secondary screening method for low false alarm rate electricity stealing detection based on electricity load fingerprints, which is used for carrying out secondary electricity stealing screening on suspected electricity stealing users through cluster analysis of the electricity load fingerprints so as to identify the change of the production and operation states of the users, thereby identifying the real electricity stealing users more accurately.
Therefore, the technical scheme adopted by the invention is as follows: a low false alarm rate electricity stealing detection secondary screening method based on electricity load fingerprints comprises the following steps:
a. collecting suspected electricity stealing users, carrying out interval recording according to the same time period and containing electricity utilization load fingerprint data of the suspected electricity stealing time period;
the suspected electricity stealing users mentioned herein are detected by the existing electricity stealing detection methods (mentioned in the background of the existing electricity stealing detection methods) based on a significant drop in electricity usage by the user or based on a significant rise in line loss of the accessed distribution line/distribution transformer. A significant change (a significant drop or a significant increase) referred to herein is an abrupt change, as a user stealing electricity may cause a significant drop in his or her own power usage or a significant increase in line loss to the distribution line/distribution transformer. And the suspected electricity stealing period is a period when the electricity consumption of the user or the line loss of the connected distribution line/distribution transformer changes significantly.
The fingerprint of the electricity load of the suspected electricity stealing user is one or a combination of a plurality of three-phase power, three-phase electricity consumption, three-phase power factor, three-phase current or three-phase harmonic content, the fingerprint data of the electricity load is collected from the electricity consumption data of the user recorded by the intelligent ammeter, and preferably, the data of one year is collected.
Preferably, the above-mentioned time period in interval recording at the same time period is 15 minutes, 30 minutes, or 60 minutes, that is, the sampling recording is performed every 15 minutes, 30 minutes, or 60 minutes.
b. Taking daily power consumption load fingerprint data of suspected power stealing users as a clustering index, carrying out clustering analysis on power consumption load fingerprints of the suspected power stealing users before and after the power consumption is remarkably reduced, and determining power consumption mode types of the suspected power stealing users before and after the power consumption is remarkably reduced; or clustering and analyzing the power consumption load fingerprints of suspected power stealing users before and after the line loss power accessed to the distribution line/distribution transformer is remarkably increased, and determining the power consumption mode types of the suspected power stealing users before and after the line loss power accessed to the distribution line/distribution transformer is remarkably increased;
the above mentioned method of clustering power consumption load fingerprints before and after the power consumption is significantly reduced or the line loss of the power distribution line/distribution transformer is significantly increased is a conventional technique in the art, and any clustering method not requiring a specified clustering number is applicable to the present invention.
Preferably, before clustering, the collected power consumption load fingerprint data can be subjected to unified dimensionless processing by adopting Min-Max standardization. If the load fingerprint is a three-phase load, the maximum value and the minimum value are taken as the maximum value and the minimum value of a single-phase load in all the load fingerprints; when the load fingerprints are three-phase power consumption, taking the maximum value and the minimum value as the maximum value and the minimum value of single-phase power consumption in all the load fingerprints; when the load fingerprints are three-phase power factors, taking the maximum value and the minimum value as the maximum value and the minimum value of single-phase power factors in all the load fingerprints; and when the load fingerprints are three-phase harmonic content, taking the maximum value and the minimum value as the maximum value and the minimum value of the single-phase harmonic content in all the load fingerprints. The Min-Max standardization is the prior art in the field, and the Min-Max standardization is adopted to improve the clustering precision and enable the clustering program to be converged faster during operation.
c. Comparing the electricity consumption mode categories to which the electricity consumption load fingerprints belong before and after the electricity consumption of the suspected electricity stealing users is remarkably reduced or the electricity consumption mode categories to which the electricity consumption load fingerprints belong before and after the line loss electricity accessed to the distribution line/distribution transformer is remarkably increased, and if the electricity consumption mode category to which the electricity consumption load fingerprints belong after the electricity consumption of the suspected electricity stealing users is remarkably reduced is contained in the electricity consumption mode categories before the electricity consumption is remarkably reduced or the electricity consumption mode category to which the electricity consumption load fingerprints belong after the line loss electricity accessed to the distribution line/distribution transformer is remarkably increased is contained in the electricity consumption mode categories before the electricity consumption is remarkably increased, the suspected electricity stealing users are normal users, otherwise, the suspected electricity stealing users are electricity stealing users.
The method comprises the steps that with the help of a power load fingerprint clustering analysis mode, power consumption mode mutation detection based on index representation such as three-phase balance, power factors and harmonic content is carried out on user power consumption data which are identified to contain power consumption mutation or line loss power mutation of a power distribution line/a power distribution transformer, when the type of the power consumption mode of a power consumption remarkably reduced period or a line loss power remarkably increased period of the power distribution line/the power distribution transformer is contained in the existing power consumption mode type (such as holidays like spring festival) of a user, a new clustering type is not formed, normal load mutation can be identified, and the method is caused when the user enters a normal low power consumption production operation state; when the electricity consumption is remarkably reduced or the line loss electricity is remarkably increased, the type of the electricity consumption mode is obviously different from the type of the existing electricity consumption mode, and a new cluster type is formed, the electricity stealing can be identified, an electricity stealing warning is sent out, and field inspection and confirmation are carried out. The method can identify sudden changes of the power consumption of the production stop users caused by power limiting, environmental protection inspection, safety inspection and the like, improves the targeting of electricity stealing inspection by reducing the false alarm rate of electricity stealing, reduces non-technical line loss, and has important theoretical significance and practical application value.
Drawings
FIG. 1 is a block flow diagram of the method of the present invention.
FIG. 2 is a user historical load curve according to an embodiment of the present invention.
FIG. 3 is a three-phase power curve corresponding to each clustering center of a user in the embodiment of the present invention;
wherein: a-class center 1, b-class center 2, c-class center 3, d-class center 4, e-class center 5, f-class center 6.
FIG. 4 is a per-unit three-phase power scatter cloud diagram corresponding to each cluster center of a user in the embodiment of the present invention;
wherein: a-class center 1, b-class center 2, c-class center 3, d-class center 4, e-class center 5, f-class center 6.
Detailed Description
Referring to fig. 1 in combination, the invention relates to a low false alarm rate electricity stealing detection secondary screening method based on electricity load fingerprints, which comprises the following specific steps:
a. collecting suspected electricity stealing users, carrying out interval recording according to the same time period and containing electricity utilization load fingerprint data of the suspected electricity stealing time period;
the electrical load fingerprint mentioned above is one or a combination of several of three-phase power, three-phase power consumption, three-phase power factor, three-phase current or three-phase harmonic content.
Preferably, the aforementioned time period is 15 minutes, 30 minutes, or 60 minutes, i.e., recorded at intervals of 15 minutes, 30 minutes, or 60 minutes.
b. Taking daily power consumption load fingerprint data of suspected power stealing users as a clustering index, carrying out clustering analysis on power consumption load fingerprints of the suspected power stealing users before and after the power consumption is remarkably reduced or power consumption load fingerprints of the suspected power stealing users before and after the line loss power of the power distribution line/distribution transformer is remarkably increased, and determining power consumption mode types obtained by clustering the power consumption load fingerprints of the suspected power stealing users before and after the power consumption is remarkably reduced or the line loss power of the power distribution line/distribution transformer is remarkably increased;
the above-mentioned method of clustering power consumption load fingerprints before and after a significant drop in power consumption or line loss is a conventional technique in the art, and any clustering method that does not require a specified number of clusters is applicable to the present invention. Here, the specific description is made by using an Affinity Propagation (AP) cluster.
The basic idea of the neighbor propagation clustering algorithm is as follows: is to put all initial samples X ═ XiI ═ 1, 2, … M are regarded as nodes of the network, and then the cluster center (Exemplar) of each sample is calculated through message passing of each edge in the network. In the clustering process, two messages are transmitted among nodes, namely an attraction degree r (responsiveness) and an attribution degree a (availabilitity). And continuously updating the attraction degree and the attribution degree value of each point by the neighbor propagation algorithm through an iterative process until m high-quality clustering centers are generated, and simultaneously distributing all the rest sample points to the corresponding clusters. The specific algorithm is as follows:
s1: initializing an algorithm, and calculating an initial similarity matrix s; s (x)i，xk) Is a similarity matrix between data points i and k, where the Euclidean distance is used for similarity testing as follows:
the similarity values of the points are all negative, that is, the larger the similarity value is, the closer the distance between the points is.
S2: selecting a deviation parameter P to represent the tendency of a data point to be selected as a clustering center, wherein the value of P determines the output category of the neighbor propagation algorithm, and the P is set as the mean value of elements in the similarity matrix s because all data points are taken as potential category representatives without prior knowledge.
S3: calculating an attraction degree r and an attribution value a between the sample points, wherein the attraction degree r (i, k) represents the attraction degree of the data point k suitable for being used as a class representation of the data point i; the attribution degree a (i, k) represents the attribution degree of the data point i by selecting the point k as the class representation thereof, and the calculation formula is as follows:
s4: r and a are updated, and subscripts t and t +1 represent the final results of the last and present update messages, respectively. λ is an attenuation coefficient, λ is more than or equal to 0 and less than or equal to 1, and is used for adjusting the convergence speed of the algorithm and the stability of the iterative process, where λ is set to 0.5, and the maximum iteration number T is set according to the attenuation coefficient λ through iterative calculation of the following two formulas:
rt+1(i，k)＝λ*rt(i，k)+(1-λ)*rt(i，k)
at+1(i，k)＝λ*at(i，k)+(1-λ)*at(i，k)；
s5: and summing the attribution degree and the attraction degree of each data point to determine a clustering center. When a (i, k) + r (i, k) takes the maximum value, if i ═ k, it can be determined that i is the cluster center; if i ≠ k, it can be determined that k is the cluster center of i.
S6: if the iteration times exceed the maximum iteration times T or when the clustering center does not change in a plurality of iterations, stopping calculation, and determining the clustering center and various sample points; otherwise, return to S3 to continue calculation.
Different clustering centers obtained through the algorithm respectively represent different power utilization modes, and the power utilization mode category is a clustering category.
c. And comparing and judging whether the type of the electricity consumption mode to which the electricity consumption load fingerprint belongs after the electricity consumption of the suspected electricity stealing users is remarkably reduced or the type of the electricity consumption mode to which the electricity consumption load fingerprint belongs after the line loss electricity accessed to the distribution line/distribution transformer is remarkably increased is contained in the type of the electricity consumption mode before the electricity consumption of the suspected electricity stealing users is remarkably reduced or the type of the electricity consumption mode before the line loss electricity accessed to the distribution line/distribution transformer is remarkably increased, if so, the suspected electricity stealing users are normal users, and if not, the suspected electricity stealing users are electricity stealing users.
The method comprises the steps of clustering power load fingerprints such as three-phase power, three-phase power consumption, three-phase current, three-phase power factors or three-phase harmonic content, which are recorded at intervals in the same time period in one day through a clustering technology, drawing user power load fingerprints after the power consumption of a user is remarkably reduced or the line loss power connected to a distribution line/a distribution transformer is remarkably increased in a state space, and judging whether misjudgment electricity stealing is caused by normal switching of a production operation state or not according to whether the user power load fingerprints are included in the clustering categories of the existing power load fingerprints before mutation. If the power consumption mode category of the user load fingerprint after the power consumption is remarkably reduced or the line loss is remarkably increased is contained in the power consumption mode category of the power consumption load fingerprint before mutation, and no new category is formed, the power stealing is judged to be misjudged, namely the suspected power stealing user is a normal user, and the power consumption mutation is caused by the change of the normal production operation state; if the fingerprint of part of the electricity load forms a new electricity utilization mode category different from the existing electricity utilization mode category after the electricity utilization amount is remarkably reduced or the line loss amount is remarkably increased, the electricity utilization is really an electricity stealing user.
Example 1
The flow and effect of screening the power utilization abnormality of the user by the method are concretely explained by taking a user of a single-phase power-stealing textile mill as an example. The instantaneous active power and three-phase power data of the user in 2019 year all the year are selected from the intelligent electric meter for specific analysis, data samples comprise suspected electricity stealing time periods and normal electricity using time periods, sampling intervals are 15 minutes, and the electricity load curve of the user all the year is shown in fig. 2. In fig. 2, the horizontal axis represents date and the vertical axis represents daily power consumption. Wherein, the first shadow section is suspected to steal electricity (8 months and 9 days to 9 months and 2 days), and the second shadow section is national celebration and long-time faking.
As can be seen from fig. 2:
● the textile mill electrical load has definite periodicity, and the electrical load on weekends is obviously lower than that on working days;
● 2 in late spring festival, the user is in complete halt state, and the electric load gradually falls and recovers with the return of the person to the country and back course before and after the spring festival.
● 8 in the middle and late months, the power consumption of the user is reduced obviously. Through field inspection, the fact that a user steals electricity by adopting a current transformer B phase shunt mode is confirmed, and the phase metering load is only about 20% of the real load. The electricity stealing time period is 8 months and 9 days to 9 months and 2 days.
● the first day of quinqueen faking and the first 4 days of eleven faking, the user approaches a production stop. The power load of Qingming province is also obviously reduced.
Adopting a neighbor propagation clustering algorithm to perform clustering analysis on the three-phase power of the electricity stealing period and the time duration before and after the electricity stealing period, wherein the parameters of the neighbor propagation clustering algorithm are set as follows: the attenuation coefficient λ is set to 0.5; the maximum iteration time T is set to be 500 times, the time T of the maximum iteration of the clustering center which is not changed is set to be 50 times, and the reference degree is set to be the median of all values in the similarity matrix. The algorithm groups the data of the electricity stealing time period and the previous 9 months into 6 categories, and the category dates form, and the corresponding production and operation states and the category centers are listed in the following table 1. For convenience of comparison and explanation, daily load curves and three-phase instantaneous power scatter cloud charts of the centers of all the classes are also drawn as shown in fig. 3 and 4. Each sub-graph in fig. 3 is the center of a class, with the horizontal axis at 24 hours of the day and the vertical axis at power level, and the three curves identify A, B, C three-phase power every 15 minutes of the day. Each sub-graph in fig. 4 is the class center of the sub-graph corresponding to the same reference number in fig. 3, the three coordinate axes are A, B, C three-phase power per unit, and 96 data points in each sub-graph correspond to the three-phase power recorded every 15 minutes.
TABLE 1 Cluster class Specification
The detailed analysis of each category, in conjunction with table 1 and fig. 3-4, is illustrated below:
● the category 1 marks the production and operation state of the user in spring festival, five one law fixed length and the like, at this time, the worker has vacation and the machine has shutdown, only has basic electricity consumption such as illumination and the like, the electricity consumption is very little, and the three-phase power scatter diagram is concentrated in the positions of zero power at the lower right corner and lower bottom power.
● Category 2 identifies the production operating status of the user on a normal work day. The textile mill is in a shift production mode, equipment runs without stopping, the load is distributed uniformly all day, and the daytime is slightly larger than night; because the load is all great all day long, all power scatter the bottom far away from the lower right corner of zero power and lower bottom of power, upper portion is along diagonal scatter distribution in the scatter diagram.
● type 3 indicates the period of time before and after spring festival when workers return to countryside/rework, during which only part of equipment is put into production in the daytime due to insufficient start-up, and the load is slightly less than normal working day; the three-phase power scatter diagram is divided into two parts, the metering data in the daytime are distributed in the scatter diagram along the diagonal line at the upper part similarly to the category 2, and the metering data in the nighttime are concentrated at the lower right corner of zero power.
● the class 4 marks the normal holiday and the Qingming festival, the holiday of the textile mill is set at Friday, the load level of day and night is obviously less than the normal working day marked by the class 2, and the three-phase power scatter cloud picture is distributed near the lower part.
● Category 5 and Category 6 identify normal weekdays and weekends for electricity stealing, respectively, since the fractional flow of phase B is nearly 80%, and is significantly lower on the daily load curve of the category center day than the other two phases; both are clearly off diagonal on the scatter cloud plot, and are distributed near the B-phase low power region on the right side of the plot. Where category 6 is tightly clustered in the bottom right corner of the scatter plot at zero power because of light daytime loads. Therefore, the used amount decrease period forms two new categories that are different from the previous categories, respectively.
● because the load data of the time section after the power consumption is reduced and the load data before are clustered to form different categories, the suspected power consumption abnormality of the power consumption reduction can be attributed to electricity stealing, and the field inspection needs to be performed by alarming.
Claims (5)
1. A low false alarm rate electricity stealing detection secondary screening method based on electricity load fingerprints is characterized in that: the method comprises the following steps:
a. collecting suspected electricity stealing users, carrying out interval recording according to the same time period and containing electricity utilization load fingerprint data of the suspected electricity stealing time period;
b. taking daily power consumption load fingerprint data of suspected power stealing users as a clustering index, carrying out clustering analysis on power consumption load fingerprints of the suspected power stealing users before and after the power consumption is remarkably reduced or before and after the line loss power connected to the distribution line/distribution transformer is remarkably increased, and determining power consumption mode types obtained by clustering the power consumption load fingerprints of the suspected power stealing users before and after the power consumption is remarkably reduced or the power consumption load fingerprints before and after the line loss power connected to the distribution line/distribution transformer is remarkably increased;
c. and comparing and judging whether the electricity consumption mode category to which the electricity load fingerprint belongs after the electricity consumption of the suspected electricity stealing user is remarkably reduced or the line loss electricity accessed to the distribution line/distribution transformer is remarkably increased is contained in the electricity consumption mode category before the electricity consumption of the suspected electricity stealing user is remarkably reduced or the line loss electricity accessed to the distribution line/distribution transformer is remarkably increased, if so, the suspected electricity stealing user is a normal user, and if not, the suspected electricity stealing user is an electricity stealing user.
2. The secondary screening method for low false alarm rate electricity stealing detection based on electricity load fingerprint as claimed in claim 1, wherein the electricity load fingerprint in step a is one or more of three-phase power, three-phase electricity consumption, three-phase power factor, three-phase current or three-phase harmonic content.
3. The secondary screening method for detecting electricity stealing with low false alarm rate based on fingerprint of electricity load as claimed in claim 1 or 2, wherein the recording of intervals in the same period of time in step a means recording intervals in 15 minutes, 30 minutes or 60 minutes.
4. The secondary screening method for low false alarm rate electricity stealing detection based on electricity load fingerprint as claimed in claim 1, wherein the suspected electricity stealing users in step a are detected according to the existing electricity stealing detection method based on sudden changes of electricity consumption of users or on sudden changes of line loss of power accessed to the distribution line/distribution transformer.
5. The secondary screening method for electricity stealing detection based on electricity load fingerprint as claimed in claim 1, wherein the electricity load fingerprint data of suspected electricity stealing users in step a is collected from the user electricity data recorded by the smart meter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010160079.XA CN111310854A (en) | 2020-03-10 | 2020-03-10 | Low false alarm rate electricity stealing detection secondary screening method based on electricity load fingerprint |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010160079.XA CN111310854A (en) | 2020-03-10 | 2020-03-10 | Low false alarm rate electricity stealing detection secondary screening method based on electricity load fingerprint |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111310854A true CN111310854A (en) | 2020-06-19 |
Family
ID=71145621
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010160079.XA Pending CN111310854A (en) | 2020-03-10 | 2020-03-10 | Low false alarm rate electricity stealing detection secondary screening method based on electricity load fingerprint |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111310854A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112162136A (en) * | 2020-09-14 | 2021-01-01 | 佰聆数据股份有限公司 | Electricity larceny prevention method and system applied to power distribution network fusion terminal |
| CN112329895A (en) * | 2021-01-05 | 2021-02-05 | 国网江西综合能源服务有限公司 | Method and device for identifying user with suspicion of electricity stealing |
| CN114254854A (en) * | 2021-10-28 | 2022-03-29 | 国网山东省电力公司安丘市供电公司 | Distribution network line loss abnormity judgment method and system based on outlier monitoring |
| CN116070162A (en) * | 2023-03-06 | 2023-05-05 | 国网浙江省电力有限公司宁波供电公司 | Anti-electricity-stealing monitoring method and system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012111377A1 (en) * | 2011-02-16 | 2012-08-23 | パナソニック株式会社 | Electricity use monitoring device, electricity use monitoring system |
| CN104751260A (en) * | 2013-12-30 | 2015-07-01 | 通用电气公司 | System and method for identifying abnormal reading of meter |
| CN106682079A (en) * | 2016-11-21 | 2017-05-17 | 云南电网有限责任公司电力科学研究院 | Detection method of user's electricity consumption behavior of user based on clustering analysis |
| CN108593990A (en) * | 2018-06-04 | 2018-09-28 | 国网天津市电力公司 | A kind of stealing detection method and application based on electric power users electricity consumption behavior pattern |
| CN110568238A (en) * | 2019-10-15 | 2019-12-13 | 长沙理工大学 | Street lamp special transformer electricity stealing detection analysis method based on electricity consumption behavior characteristics |
-
2020
- 2020-03-10 CN CN202010160079.XA patent/CN111310854A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012111377A1 (en) * | 2011-02-16 | 2012-08-23 | パナソニック株式会社 | Electricity use monitoring device, electricity use monitoring system |
| CN104751260A (en) * | 2013-12-30 | 2015-07-01 | 通用电气公司 | System and method for identifying abnormal reading of meter |
| CN106682079A (en) * | 2016-11-21 | 2017-05-17 | 云南电网有限责任公司电力科学研究院 | Detection method of user's electricity consumption behavior of user based on clustering analysis |
| CN108593990A (en) * | 2018-06-04 | 2018-09-28 | 国网天津市电力公司 | A kind of stealing detection method and application based on electric power users electricity consumption behavior pattern |
| CN110568238A (en) * | 2019-10-15 | 2019-12-13 | 长沙理工大学 | Street lamp special transformer electricity stealing detection analysis method based on electricity consumption behavior characteristics |
Non-Patent Citations (1)
| Title |
|---|
| 王桂兰等: ""大规模用电数据流的快速聚类和异常检测技术"", 《电力系统自动化》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112162136A (en) * | 2020-09-14 | 2021-01-01 | 佰聆数据股份有限公司 | Electricity larceny prevention method and system applied to power distribution network fusion terminal |
| CN112329895A (en) * | 2021-01-05 | 2021-02-05 | 国网江西综合能源服务有限公司 | Method and device for identifying user with suspicion of electricity stealing |
| CN114254854A (en) * | 2021-10-28 | 2022-03-29 | 国网山东省电力公司安丘市供电公司 | Distribution network line loss abnormity judgment method and system based on outlier monitoring |
| CN116070162A (en) * | 2023-03-06 | 2023-05-05 | 国网浙江省电力有限公司宁波供电公司 | Anti-electricity-stealing monitoring method and system |
| CN116070162B (en) * | 2023-03-06 | 2023-08-29 | 国网浙江省电力有限公司宁波供电公司 | Anti-electricity-stealing monitoring method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111310854A (en) | Low false alarm rate electricity stealing detection secondary screening method based on electricity load fingerprint | |
| CN110097297A (en) | A kind of various dimensions stealing situation Intellisense method, system, equipment and medium | |
| EP1309062A2 (en) | Apparatus and method for determining days of the week with similar utility consumption profiles | |
| CN112288303B (en) | Method and device for determining line loss rate | |
| Hartmann et al. | Suspicious electric consumption detection based on multi-profiling using live machine learning | |
| CN116522746A (en) | Power distribution hosting method for high-energy-consumption enterprises | |
| CN114878934A (en) | Electric energy consumption data abnormity early warning method | |
| CN111160993A (en) | User practical and stage release coefficient analysis method based on user load application | |
| CN117610214B (en) | Intelligent power distribution network wiring planning method based on dynamic geographic features | |
| CN110298765A (en) | A kind of power distribution network electric power method for detecting abnormality based on objective correlative factor | |
| CN110750760A (en) | Abnormal theoretical line loss detection method based on situation awareness and control chart | |
| Lin et al. | The data mining application in the power quality monitoring data analysis | |
| CN112598257A (en) | Power failure analysis method and system based on big data feature mining | |
| CN109245109B (en) | Power supply low-voltage early warning system and method | |
| CN113723671B (en) | Data clustering analysis method based on electricity consumption condition big data | |
| Wu et al. | Survey on negative line loss rate of transformer region: rectification measures and challenges | |
| Ya’An | Application of artificial intelligence in computer network technology in the era of big data | |
| Dashti et al. | Decisions, Decisions: An Asset Management—Based Distribution System Framework | |
| CN114839462A (en) | Intelligent anti-electricity-stealing monitoring method and system | |
| CN114168662A (en) | Power distribution network problem combing and analyzing method and system based on multiple data sources | |
| CN114638537A (en) | Digital smart power grid region management method, system and storage medium | |
| Li et al. | Research and design of power Visualization intelligent management platform based on big data technology | |
| Grigoras et al. | Knowlegde extraction from Smart Meters for consumer classification | |
| Li et al. | Research on Key Technology of State Evaluation and Replacement Acquisition System of Intelligent Electric Energy Meter Based on Smart Grid | |
| Shikhin et al. | Significant Events Detection and Identification through Electrical Grid Load Profile |
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 |
Application publication date: 20200619 |
|
| RJ01 | Rejection of invention patent application after publication |