CN106908752A - The electric energy metrical abnormality diagnostic method and system of a kind of feature based packet - Google Patents
The electric energy metrical abnormality diagnostic method and system of a kind of feature based packet Download PDFInfo
- Publication number
- CN106908752A CN106908752A CN201710161079.XA CN201710161079A CN106908752A CN 106908752 A CN106908752 A CN 106908752A CN 201710161079 A CN201710161079 A CN 201710161079A CN 106908752 A CN106908752 A CN 106908752A
- Authority
- CN
- China
- Prior art keywords
- electric energy
- energy metrical
- index
- data
- exception type
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R35/00—Testing or calibrating of apparatus covered by the other groups of this subclass
- G01R35/04—Testing or calibrating of apparatus covered by the other groups of this subclass of instruments for measuring time integral of power or current
Abstract
The embodiment of the invention discloses the electric energy metrical abnormality diagnostic method and system of a kind of packet of feature based, judgement for solving failure exception type at present for electric energy measuring equipment mainly distinguishes determination by the experience of the technical staff of on-site maintenance treatment, judged result easily receives subjective impact, and lack theoretical property foundation, judge recognition time technical problem also more long.Present invention method includes：Initial data base is built according to the electric energy metrical Exception Type and the criterion of electric energy metrical Exception Type that get and corresponding data source；Data source data to initial data base carries out data conversion according to the criterion of electric energy metrical Exception Type, forms original measurement abnormal index collection；Index feature packet is carried out to original measurement abnormal index collection, multiple index groups are formed；Multi-categorizer training is carried out to index group, several different disaggregated models are obtained；Several different disaggregated models are merged, electric energy metrical anomalous identification result is obtained.
Description
Technical field
The present invention relates to electric energy metrical anomalous identification field, more particularly to the electric energy metrical exception that a kind of feature based is grouped
Diagnostic method and system.
Background technology
Power System Intelligent is the common direction of power industry development at this stage, and power information acquisition system is to ensure
The smooth important foundation stone of this action.Electricity consumption data is by electric power energy course of conveying by electric power meter
The data acquisition of middle generation gets.With the continuous application of the metering device, by various factors such as device aging, artificial destructions
Influence, usually cause that different degrees of abnormal conditions occurs in device.The abnormal situation of these meterings is either large or small all can be to electricity
Power enterprise brings certain economic loss, seriously also affects electrical management work.
The abnormal situation species of electric energy metrical is various, generally comprises abnormal gauge table, transformer exception, terminal box exception
Deng.Every kind of equipment can specifically segment out abnormal subclass again, such as：Electric sampling open-phase, voltage phase shortage, electric current overload etc..However,
The warp of the technical staff for mainly being processed by on-site maintenance for the judgement of the failure exception type of electric energy measuring equipment at present
Test to distinguish determination, its judged result easily receives subjective impact, and lacks theoretical property foundation, judges that recognition time is also more long, greatly
The big efficiency that have impact on field accident treatment.
The content of the invention
The electric energy metrical abnormality diagnostic method and system of a kind of feature based packet are the embodiment of the invention provides, is solved
The warp of the technical staff for mainly being processed by on-site maintenance for the judgement of the failure exception type of electric energy measuring equipment at present
Test to distinguish determination, its judged result easily receives subjective impact, and lacks theoretical property foundation, judges that recognition time is also more long, greatly
The big efficiency that have impact on field accident treatment.
A kind of electric energy metrical abnormality diagnostic method of feature based packet provided in an embodiment of the present invention, including：
According to the electric energy metrical Exception Type and the criterion of electric energy metrical Exception Type that get, electric energy abnormal data is formed
Collection, and initial data base is built according to the corresponding data source of criterion and electric energy abnormal data set of electric energy metrical Exception Type；
Data source data to initial data base carries out data conversion according to the criterion of electric energy metrical Exception Type, is formed just
Begin metering abnormal index collection；
Index feature packet is carried out to original measurement abnormal index collection, multiple index groups are formed；
Multi-categorizer training is carried out to index group, several different disaggregated models are obtained；
Several different disaggregated models are merged, electric energy metrical anomalous identification result is obtained.
Preferably, the data source data to initial data base carries out data change according to the criterion of electric energy metrical Exception Type
Change, forming original measurement abnormal index collection includes：
Data prediction is carried out to the data source data of initial data base and the criterion according to electric energy metrical Exception Type is entered
Row data are converted, and form original measurement abnormal index collection, and data prediction includes correcting or suppressing exception data, interpolation missing number
According to.
Preferably, index feature packet is carried out to original measurement abnormal index collection, forming multiple index groups includes：
The every kind of electric energy metrical Exception Type and every kind of electric energy metrical Exception Type pair concentrated to original measurement abnormal index
The index answered carries out matching arrangement；
Index feature packet is carried out according to the characteristic relation between index and electric energy metrical Exception Type, multiple indexs are formed
Group.
Preferably, multi-categorizer training is carried out to index group, obtaining several different disaggregated models includes：
According to different electric energy metrical Exception Types, the multiple graders of index set selection to each index group are instructed
Practice, obtain several different sorter models for training, the sorter model for training can fix the corresponding electric energy of identification
Metering Exception Type.
Preferably, several different disaggregated models are merged, obtaining electric energy metrical anomalous identification result includes：
The electric energy metrical Exception Type of the disaggregated model correspondence identification that will be trained is collected, and is realized to all electric energy meters
Measure the differentiation of Exception Type.
A kind of electric energy metrical abnormity diagnostic system of feature based packet provided in an embodiment of the present invention, including：
Module is built, for according to the electric energy metrical Exception Type and the criterion of electric energy metrical Exception Type for getting, shape
Into electric energy abnormal data set, and built according to the corresponding data source of criterion and electric energy abnormal data set of electric energy metrical Exception Type
Initial data base；
Conversion module, line number is entered for the data source data to initial data base according to the criterion of electric energy metrical Exception Type
According to conversion, original measurement abnormal index collection is formed；
Grouping module, for carrying out index feature packet to original measurement abnormal index collection, forms multiple index groups；
Training module, for carrying out multi-categorizer training to index group, obtains several different disaggregated models；
Fusion Module, for several different disaggregated models to be merged, obtains electric energy metrical anomalous identification result.
Preferably, conversion module includes：
Data variation unit, data prediction is carried out and according to electric energy metrical for the data source data to initial data base
The criterion of Exception Type carries out data conversion, forms original measurement abnormal index collection, and data prediction includes correcting or deletes different
Regular data, interpolation missing data.
Preferably, grouping module includes：
Matching unit, based on the every kind of electric energy metrical Exception Type concentrated to original measurement abnormal index and every kind of electric energy
The corresponding index of amount Exception Type carries out matching arrangement；
Grouped element, for carrying out index feature point according to the characteristic relation between index and electric energy metrical Exception Type
Group, forms multiple index groups.
Preferably, training module includes：
Training unit, for according to different electric energy metrical Exception Types, the index set selection to each index group to be more
Individual grader is trained, and obtains several different sorter models for training, and the sorter model for training can be fixed
Recognize corresponding electric energy metrical Exception Type.
Preferably, Fusion Module includes：
Collection unit, the electric energy metrical Exception Type of the disaggregated model correspondence identification for that will train is collected, real
Now to the differentiation of all electric energy metrical Exception Types.
As can be seen from the above technical solutions, the embodiment of the present invention has advantages below：
The electric energy metrical abnormality diagnostic method and system of a kind of feature based packet are the embodiment of the invention provides, including：
According to the electric energy metrical Exception Type and the criterion of electric energy metrical Exception Type that get, electric energy abnormal data set, and root are formed
Initial data base is built according to the corresponding data source of the criterion of electric energy metrical Exception Type and electric energy abnormal data set；To primary data
The data source data in storehouse carries out data conversion according to the criterion of electric energy metrical Exception Type, forms original measurement abnormal index collection；
Index feature packet is carried out to original measurement abnormal index collection, multiple index groups are formed；Multi-categorizer is carried out to index group
Training, obtains several different disaggregated models；Several different disaggregated models are merged, electric energy metrical exception is obtained
Recognition result, by abnormal each of the electric energy metrical that will occur in power supply enterprise's actual power application process in the embodiment of the present invention
Type and basis for estimation are integrated, and propose the Multi-classifers integrated model of feature based packet, and the application not only can be with
Whether all-around intelligentization ground " considering " research object there is metering exception, accurately judge Exception Type；Meanwhile, feature based point
Group method, Data duplication rate, the optimization of implementation model performance can also be greatly reduced, it is ensured that Model Identification it is accurate
Property, solve the technology that the judgement of failure exception type at present for electric energy measuring equipment is mainly processed by on-site maintenance
The experience of personnel distinguishes determination, and its judged result easily receives subjective impact, and lacks theoretical property foundation, judges recognition time
It is more long, leverage the technical problem of the efficiency of field accident treatment.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
The accompanying drawing to be used needed for having technology description is briefly described, it should be apparent that, drawings in the following description are only this
Some embodiments of invention, for those of ordinary skill in the art, without having to pay creative labor, may be used also
Other accompanying drawings are obtained with according to these accompanying drawings.
Fig. 1 is an a kind of reality of the electric energy metrical abnormality diagnostic method of feature based packet provided in an embodiment of the present invention
Apply the schematic flow sheet of example；
Fig. 2 is another of the electric energy metrical abnormality diagnostic method of a kind of feature based packet provided in an embodiment of the present invention
The schematic flow sheet of embodiment；
Fig. 3 is that a kind of structure of the electric energy metrical abnormity diagnostic system of feature based packet provided in an embodiment of the present invention is shown
It is intended to.
Specific embodiment
The electric energy metrical abnormality diagnostic method and system of a kind of feature based packet are the embodiment of the invention provides, for solving
The technical staff's for certainly mainly being processed by on-site maintenance for the judgement of the failure exception type of electric energy measuring equipment at present
Experience distinguishes determination, and its judged result easily receives subjective impact, and lacks theoretical property foundation, judges that recognition time is also more long,
Leverage the technical problem of the efficiency of field accident treatment.
To enable that goal of the invention of the invention, feature, advantage are more obvious and understandable, below in conjunction with the present invention
Accompanying drawing in embodiment, is clearly and completely described, it is clear that disclosed below to the technical scheme in the embodiment of the present invention
Embodiment be only a part of embodiment of the invention, and not all embodiment.Based on the embodiment in the present invention, this area
All other embodiment that those of ordinary skill is obtained under the premise of creative work is not made, belongs to protection of the present invention
Scope.
Refer to Fig. 1, a kind of electric energy metrical abnormality diagnostic method bag of feature based packet provided in an embodiment of the present invention
Include：
101st, according to the electric energy metrical Exception Type and the criterion of electric energy metrical Exception Type for getting, electric energy exception is formed
Data set, and primary data is built according to the corresponding data source of criterion and electric energy abnormal data set of electric energy metrical Exception Type
Storehouse；
First, to by the end of the current electric energy metrical Exception Type for occurring and based on experience or data of literatures institute
The basis for estimation of the various metering Exception Types put in order is collected.According to the electric energy metrical Exception Type and electric energy that get
The criterion of Exception Type is measured, electric energy abnormal data set is formed, and according to the corresponding data of criterion of electric energy metrical Exception Type
Source and electric energy abnormal data set build initial data base.
102nd, the data source data to initial data base carries out data conversion, shape according to the criterion of electric energy metrical Exception Type
Into original measurement abnormal index collection；
According to the electric energy metrical Exception Type and the criterion of electric energy metrical Exception Type for getting, electric energy exception number is being formed
According to collection, and according to the corresponding data source of criterion and electric energy abnormal data set of electric energy metrical Exception Type build initial data base it
Afterwards, the data source data to initial data base carries out data conversion according to the criterion of electric energy metrical Exception Type, forms initial meter
Amount abnormal index collection.
103rd, index feature packet is carried out to original measurement abnormal index collection, multiple index groups are formed；
Data conversion is carried out according to the criterion of electric energy metrical Exception Type in the data source data to initial data base, is formed
After original measurement abnormal index collection, index feature packet is carried out to original measurement abnormal index collection, form multiple index groups.
104th, multi-categorizer training is carried out to index group, obtains several different disaggregated models；
Index feature packet is being carried out to original measurement abnormal index collection, after the multiple index groups of formation, to index group
Multi-categorizer training is not carried out, obtains several different disaggregated models.
105th, several different disaggregated models are merged, is obtained electric energy metrical anomalous identification result.
Carrying out multi-categorizer training to index group, obtain after several different disaggregated models, it is necessary to will be some
Individual different disaggregated model is merged, and obtains electric energy metrical anomalous identification result.
It is more than to a kind of electric energy metrical abnormality diagnostic method of feature based packet provided in an embodiment of the present invention
The detailed description of individual embodiment, will extremely examine a kind of electric energy metrical of feature based packet provided in an embodiment of the present invention below
Another embodiment of disconnected method is described in detail.
Fig. 2 is referred to, a kind of electric energy metrical abnormality diagnostic method of feature based packet provided in an embodiment of the present invention
Another embodiment, including：
201st, according to the electric energy metrical Exception Type and the criterion of electric energy metrical Exception Type for getting, electric energy exception is formed
Data set, and primary data is built according to the corresponding data source of criterion and electric energy abnormal data set of electric energy metrical Exception Type
Storehouse；
First, to by the end of the current electric energy metrical Exception Type for occurring and based on experience or data of literatures institute
The basis for estimation of the various metering Exception Types put in order is collected.According to the electric energy metrical Exception Type and electric energy that get
The criterion of Exception Type is measured, electric energy abnormal data set is formed, and according to the corresponding data of criterion of electric energy metrical Exception Type
Source and electric energy abnormal data set build initial data base.
202nd, the data source data to initial data base carries out data prediction and sentencing according to electric energy metrical Exception Type
According to data conversion is carried out, original measurement abnormal index collection is formed, data prediction includes correcting or suppressing exception data, interpolation lack
Lose data；
Then, the data source data to initial data base carries out data prediction and sentencing according to electric energy metrical Exception Type
According to carrying out data conversion, and follow the criterion referred to during electric energy metrical judges extremely, the data after conversion are mapped or
The method of conversion forms original measurement abnormal index collection, wherein, data prediction includes correcting or suppressing exception data, interpolation lack
Lose data.
203rd, the every kind of electric energy metrical Exception Type and every kind of electric energy metrical exception class concentrated to original measurement abnormal index
The corresponding index of type carries out matching arrangement；
Data prediction is carried out in the data source data to initial data base and according to the criterion of electric energy metrical Exception Type
Data conversion is carried out, after formation original measurement abnormal index collection, the every kind of electric energy metrical concentrated to original measurement abnormal index
Exception Type and the corresponding index of every kind of electric energy metrical Exception Type carry out matching arrangement.Wherein, a kind of electric energy metrical exception class
Type includes one or more indexs；Can there is independent, intersection, bag between index involved by two kinds of different metering Exception Types
A kind of relation in containing.
204th, index feature packet is carried out according to the characteristic relation between index and electric energy metrical Exception Type, forms multiple
Index group；
In the every kind of electric energy metrical Exception Type and every kind of electric energy metrical Exception Type concentrated to original measurement abnormal index
Corresponding index is carried out after matching arrangement, it is necessary to be referred to according to the characteristic relation between index and electric energy metrical Exception Type
Mark feature packet, forms multiple index groups.Using every kind of index for being related to of metering Exception Type as research object, when occur with
During any one in lower 3 kinds of situations, its corresponding Exception Type and index can be returned into single one group, for overlapping
Index only take once；Otherwise, every kind of Exception Type and its index are independent into one group.
In above-mentioned three kinds of situations, Ua、Ub、UcExpression represents the index involved by electric energy metrical Exception Type a, b, c respectively
Collection.After the completion of index packet, every group of index comprises at least a kind of corresponding all indexs of metering Exception Type.
205th, according to different electric energy metrical Exception Types, the multiple graders of index set selection to each index group enter
Row training, obtains several different sorter models for training, and it is corresponding that the sorter model for training can fix identification
Electric energy metrical Exception Type；
Index feature packet is being carried out according to the characteristic relation between index and electric energy metrical Exception Type, multiple is being formed and is referred to
After mark group, according to different metering Exception Types, the multiple graders of index set selection to each group are trained, obtain
To several sorter models, wherein, grader selection has the sorter model of supervision.
In several resulting sorter models, the sorter model that each index group is trained can regularly be known
Metering Exception Type not corresponding to it.
The electric energy metrical Exception Type of the disaggregated model correspondence identification that the 206th, will be trained is collected, and is realized to all electricity
The differentiation of Exception Type can be measured.
In several sorter models for training, the current all metering Exception Types being related to have been enumerated.Most
Then the metering Exception Type that each sorter model is recognized is collected afterwards, is realized all fault types of each research object
Differentiate.
It is more than to a kind of the another of electric energy metrical abnormality diagnostic method of feature based packet provided in an embodiment of the present invention
The detailed description of one embodiment, for ease of understanding, the one kind that will be provided the embodiment of the present invention with specific example below
The electric energy metrical abnormality diagnostic method of feature based packet is described in detail.
The first step, according to the real needs of Power Project, obtains the relevant information of electric energy metrical Exception Type and sorts out
Related data source.
The electric energy metrical Exception Type information of table 1
As shown in table 1, metering Exception Type contains common and urgently to be resolved hurrily metering Exception Type and its corresponding sentences
According to two fields.In view of the data that the Power Project is related to are excessive, voltage phase shortage, Voltage unbalance, electricity are simply listed here
Sampling open-phase, electric current overload, electric current reversed polarity, current imbalance, 7 kinds of metering abnormal conditions of defluidization carry out combing.
The criterion of voltage phase shortage mainly has three：1st, this phase voltage<80%Un；2nd, this phase current>1%In；When the 3rd, continuing
Between be more than 1 hour.The data source being previously mentioned according to these three criterions should include：Three-phase three (three-phase four) voltage, rated voltage,
Three-phase three (three-phase four) electric current, rated current, time data.By that analogy, you can find all metering Exception Type criterions
Data source.Load data of the data source from Utilities Electric Co.'s major network data, row degrees of data, archive information.
Second step, the data source data to initial data base carries out data change according to the criterion of electric energy metrical Exception Type
Change, form original measurement abnormal index collection.
The main target of this step is to realize the construction of index set, and Main Basiss are the abnormal criterion of metering.Why not
Directly directly judge whether that it is that the determination of criterion is given by expert that the reason for metering is abnormal occurs using the abnormal criterion of metering,
The determination of many threshold values is too artificial subjective, it is easier to the situation of the erroneous judgement that occurs failing to judge.So that construction index, by machine
Recognition rule is removed in study.
The data of reality often occur a large amount of irrational data, need to enter source data first before index construction is done
Row pretreatment, including correction or deletion to abnormal data, missing data selects appropriate method interpolation, so as to obtain correction number
According to.After data prediction, it then follows the criterion that electric energy metrical is referred in judging extremely, to correcting data mapping or converting
Method construct original measurement abnormal index collection.
The index set of table 2 is constructed
By taking voltage phase shortage as an example, there are two criterions to be related to two comparings of data in three criterions of voltage phase shortage, this
The data of sample are generally converted into index in the way of ratio.So the index of voltage phase shortage construction is for voltagerating ratio, and (certain is mutually electric
Pressure Ux/ rated voltage Un*100), current rating ratio (certain phase current Ix/ rated current In*100), state duration it is (small
When).
By taking voltage phase shortage as an example, duration index builds：
" duration is more than 1 hour " mentioned in criterion is understood that, refers to meeting " this phase voltage at the same time<
80%Un " and " this phase current>It is sustained for longer than under 1%In " states 1 hour.
When constructing the index, need to calculate the separate state duration of other indexs in addition to time index first, respectively
There is voltagerating to compare state duration than state duration, current rating.
Voltagerating compares state duration：Note voltagerating is α, criterion " this phase voltage than desired value<80%Un ", this
In to need duration of record should be time span when front and rear adjacent record falls into [0, α].
Current rating compares state duration：Note current rating is β, criterion " this phase current than desired value>1%In ", this
In to need duration of record should be time span when front and rear adjacent record falls into [β, 100].
The state duration of so voltage phase shortage is:
Min (, than state duration, current rating is than state duration for voltagerating).
By that analogy, the index set construction involved by electric energy metrical Exception Type is completed.
4th step, index feature packet is carried out to original measurement abnormal index collection, forms multiple index groups.
Feature packet to original measurement fault indices collection includes following sub-step：
S3.1：Integrate the abnormal and corresponding index inventory of metering.
Index involved by every kind of metering Exception Type is carried out into induction-arrangement.As shown in table 3, voltage phase shortage, voltage is not
The index set that balance, electric sampling open-phase, electric current overload, electric current reversed polarity, current imbalance, defluidization are related separately to is in index field
In can find.
The Exception Type index set of table 3
Each of the above metering Exception Type is directed to 2~3 indexs.
In this project, when dividing the relation between the different abnormal corresponding index sets of metering, first by state duration
Ignore subdivided, such as in addition to index state duration, belong between the corresponding index set of voltage phase shortage and electric sampling open-phase
In comprising (coincidence) relation, i.e.,OrSimilarly divide other difference meterings abnormal
Relation between type correspondence index set.
S3.2：Index feature is grouped.
It is any one in 3 kinds of situations below occur using every kind of index for being related to of metering Exception Type as research object
When planting, its corresponding Exception Type and index can be returned into single one group, the index for overlapping only takes once；It is no
Then, every kind of Exception Type and its index are independent into one group.
In above formula, Ua、Ub、UcExpression represents the index set involved by electric energy metrical Exception Type a, b, c respectively.
The index set packet obtained according to above-mentioned criterion is as shown in table 4.
Table 4 is grouped index
Voltage phase shortage, electric sampling open-phase, the corresponding index set of electric current overload have been divided into first group, corresponding New Set
Collect and be：Voltagerating ratio, current rating ratio, state duration 1, state duration 3, state duration 4.First group refers to
Mark collection can simultaneously be completed voltage phase shortage, electric sampling open-phase, three kinds of metering Exception Types of electric current overload by model training and be examined
It is disconnected, while reducing diagnostic model, index redundancy can be well reduced again.The index set of other Exception Types is all mutual
Independent relation, therefore individually turn into one group.
Multiple index groups are carried out multi-categorizer training by the 4th step, draw different classifications model.
According to different metering Exception Types, the multiple graders of index set selection to each group are trained, obtain
Several sorter models.The sorter model that the embodiment of the present invention is used to train has used neutral net, decision tree, simple shellfish
Ye Si, SVMs.
There are 12147 users in the test small sample storehouse of the embodiment of the present invention, wherein being asked comprising normal users, single metering
Topic user, multiple metering abnormal problem clients.Random selection 20% is remaining as training sample as test sample.In instruction
Elimination dimension treatment is first done before practicing model to every group of index, the method for using is standardized for extreme difference.
Training set each group index model accuracy rate (%) of table 5
Table 5 is four sorter models training accuracy rate of training set each group index, and accuracy rate is higher, illustrates model training
Effect is better.Every group of index correspondence accuracy rate highest grader is selected to be shown in Table 6 by the table.
The accuracy rate highest grader of table 6
| Group | Grader |
| First group | Neutral net |
| Second group | Neutral net |
| 3rd group | SVMs |
| 4th group | SVMs |
| 5th group | Naive Bayesian |
In combination with the another two evaluation result of model, can obtain the normal users of table 7 be mistaken for abnormal user ratio,
The abnormal user of table 8 is mistaken for the ratio of normal users, and table 7 and table 8 are specific as follows:
The normal users of table 7 are mistaken for the ratio (%) of abnormal user
| Neutral net | Decision tree | Naive Bayesian | SVMs | |
| First group | 1.95 | 5.85 | 5.05 | 1.89 |
| Second group | 0.63 | 4.46 | 1.35 | 7.40 |
| 3rd group | 6.85 | 5.05 | 7.34 | 4.77 |
| 4th group | 3.64 | 6.80 | 8.23 | 2.90 |
| 5th group | 5.99 | 3.71 | 1.34 | 5.49 |
The abnormal user of table 8 is mistaken for the ratio (%) of normal users
| Neutral net | Decision tree | Naive Bayesian | SVMs | |
| First group | 28.6 | 22.86 | 38.90 | 23.77 |
| Second group | 19.41 | 32.25 | 26.31 | 20.00 |
| 3rd group | 27.16 | 29.53 | 35.31 | 24.80 |
| 4th group | 20.07 | 22.98 | 27.38 | 25.77 |
| 5th group | 30.12 | 28.73 | 29.01 | 27.96 |
The neural network classifier training result of first group of index shows that its accuracy rate highest, normal users misjudge ratio
It is minimum, although abnormal user ratio of failing to judge is not minimum, but is more or less the same with 22.86%, is considered, by first group of index
Sorter model is chosen to be neutral net.Similarly determine the grader of other groups, second group：Neutral net；3rd group：Support
Vector machine；4th group：SVMs；5th group：Naive Bayesian.
5th step, multi-categorizer training result is merged, and obtains final electric energy metrical anomalous identification result.
After determining multi-categorizer, race, five index groups are surveyed during 20% test sample is substituted into five sorter models
Other accuracy rate is as follows：
The test data grader accuracy rate of table 9
| Group | Accuracy rate (%) |
| First group | 92.04 |
| Second group | 96.50 |
| 3rd group | 93.02 |
| 4th group | 98.73 |
| 5th group | 95.77 |
The survey of the display grader of table 9 runs effect very well, and all more than 90%, highest can reach 98.73% to accuracy rate.
Final step is to realize collecting for the various metering abnormity diagnosis results of user, and effect is as follows：
The user of table 10 measures abnormity diagnosis result
| ID | Metering is abnormal | Time point |
| 09297 | Nothing | 2016-12-0700：00：00 |
| 11532 | Voltage phase shortage | 2016-12-0104：00：00 |
| 21200 | Defluidization, current imbalance | 2016-12-0512：00：00 |
| 09349 | Nothing | 2016-12-0706：00：00 |
| 97008 | Defluidization | 2016-12-0421：00：00 |
As shown in table 10, the Multi-classifers integrated model of feature based packet can recognize the various of synchronization generation
Metering is abnormal, and multiple metering exception is exported together.
It is more than to a kind of tool of the electric energy metrical abnormality diagnostic method of feature based packet provided in an embodiment of the present invention
The detailed description of style, below by a kind of electric energy metrical abnormity diagnosis of feature based packet provided in an embodiment of the present invention
System is described in detail.
Refer to Fig. 3, a kind of electric energy metrical abnormity diagnostic system bag of feature based packet provided in an embodiment of the present invention
Include：
Module 301 is built, the electric energy metrical Exception Type and the criterion of electric energy metrical Exception Type got for basis,
Electric energy abnormal data set is formed, and according to the corresponding data source of criterion and electric energy abnormal data set structure of electric energy metrical Exception Type
Build initial data base；
Conversion module 302, enters for the data source data to initial data base according to the criterion of electric energy metrical Exception Type
Row data are converted, and form original measurement abnormal index collection；Conversion module 302 includes：
Data variation unit 3021, data prediction is carried out and according to electric energy for the data source data to initial data base
Measuring the criterion of Exception Type carries out data conversion, forms original measurement abnormal index collection, and data prediction includes correcting or deleting
Except abnormal data, interpolation missing data.
Grouping module 303, for carrying out index feature packet to original measurement abnormal index collection, forms multiple index groups
Not；Grouping module 303 includes：
Matching unit 3031, for the every kind of electric energy metrical Exception Type and every kind of electricity concentrated to original measurement abnormal index
The corresponding index of Exception Type can be measured carries out matching arrangement；
Grouped element 3032, for carrying out index feature according to the characteristic relation between index and electric energy metrical Exception Type
Packet, forms multiple index groups.
Training module 304, for carrying out multi-categorizer training to index group, obtains several different disaggregated models；
Training module 304 includes：
Training unit 3041, for according to different electric energy metrical Exception Types, the index set choosing to each index group
Select multiple graders to be trained, obtain several different sorter models for training, the sorter model for training can
It is fixed to recognize corresponding electric energy metrical Exception Type.
Fusion Module 305, for several different disaggregated models to be merged, obtains electric energy metrical anomalous identification knot
Really；Fusion Module 305 includes：
Collection unit 3051, the electric energy metrical Exception Type of the disaggregated model correspondence identification for that will train is converged
Always, the differentiation to all electric energy metrical Exception Types is realized.
It is apparent to those skilled in the art that, for convenience and simplicity of description, the system of foregoing description,
The specific work process of device and unit, may be referred to the corresponding process in preceding method embodiment, will not be repeated here.
In several embodiments provided herein, it should be understood that disclosed system, apparatus and method can be with
Realize by another way.For example, device embodiment described above is only schematical, for example, the unit
Divide, only a kind of division of logic function there can be other dividing mode when actually realizing, for example multiple units or component
Can combine or be desirably integrated into another system, or some features can be ignored, or do not perform.It is another, it is shown or
The coupling each other for discussing or direct-coupling or communication connection can be the indirect couplings of device or unit by some interfaces
Close or communicate to connect, can be electrical, mechanical or other forms.
The unit that is illustrated as separating component can be or may not be it is physically separate, it is aobvious as unit
The part for showing can be or may not be physical location, you can with positioned at a place, or can also be distributed to multiple
On NE.Some or all of unit therein can be according to the actual needs selected to realize the mesh of this embodiment scheme
's.
In addition, during each functional unit in each embodiment of the invention can be integrated in a processing unit, it is also possible to
It is that unit is individually physically present, it is also possible to which two or more units are integrated in a unit.Above-mentioned integrated list
Unit can both be realized in the form of hardware, it would however also be possible to employ the form of SFU software functional unit is realized.
If the integrated unit is to realize in the form of SFU software functional unit and as independent production marketing or use
When, can store in a computer read/write memory medium.Based on such understanding, technical scheme is substantially
The part for being contributed to prior art in other words or all or part of the technical scheme can be in the form of software products
Embody, the computer software product is stored in a storage medium, including some instructions are used to so that a computer
Equipment (can be personal computer, server, or network equipment etc.) performs the complete of each embodiment methods described of the invention
Portion or part steps.And foregoing storage medium includes：USB flash disk, mobile hard disk, read-only storage (ROM, Read-Only
Memory), random access memory (RAM, RandomAccess Memory), magnetic disc or CD etc. are various can store journey
The medium of sequence code.
The above, the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although with reference to preceding
Embodiment is stated to be described in detail the present invention, it will be understood by those within the art that：It still can be to preceding
State the technical scheme described in each embodiment to modify, or equivalent is carried out to which part technical characteristic；And these
Modification is replaced, and does not make the spirit and scope of the essence disengaging various embodiments of the present invention technical scheme of appropriate technical solution.
Claims (10)
1. the electric energy metrical abnormality diagnostic method that a kind of feature based is grouped, it is characterised in that including：
According to the electric energy metrical Exception Type for getting and the criterion of the electric energy metrical Exception Type, electric energy abnormal data is formed
Collection, and the corresponding data source of criterion and the electric energy abnormal data set structure initial number according to the electric energy metrical Exception Type
According to storehouse；
Data source data to initial data base carries out data conversion according to the criterion of the electric energy metrical Exception Type, is formed just
Begin metering abnormal index collection；
Index feature packet is carried out to the original measurement abnormal index collection, multiple index groups are formed；
Multi-categorizer training is carried out to the index group, several different disaggregated models are obtained；
Described several different disaggregated models are merged, electric energy metrical anomalous identification result is obtained.
2. the electric energy metrical abnormality diagnostic method that feature based according to claim 1 is grouped, it is characterised in that described right
The data source data of initial data base carries out data conversion according to the criterion of the electric energy metrical Exception Type, forms original measurement
Abnormal index collection includes：
Data prediction is carried out to the data source data of initial data base and the criterion according to the electric energy metrical Exception Type is entered
Row data are converted, and form original measurement abnormal index collection, and the data prediction includes correcting or suppressing exception data, interpolation lack
Lose data.
3. the electric energy metrical abnormality diagnostic method that feature based according to claim 2 is grouped, it is characterised in that described right
The original measurement abnormal index collection carries out index feature packet, and forming multiple index groups includes：
The every kind of electric energy metrical Exception Type and every kind of electric energy metrical exception class concentrated to the original measurement abnormal index
The corresponding index of type carries out matching arrangement；
Index feature packet is carried out according to the characteristic relation between the index and electric energy metrical Exception Type, multiple indexs are formed
Group.
4. the electric energy metrical abnormality diagnostic method that feature based according to claim 3 is grouped, it is characterised in that described right
The index group carries out multi-categorizer training, and obtaining several different disaggregated models includes：
According to different electric energy metrical Exception Types, the multiple graders of index set selection to index group each described are instructed
Practice, obtain several different sorter models for training, it is corresponding that the sorter model for training can fix identification
Electric energy metrical Exception Type.
5. the electric energy metrical abnormality diagnostic method that feature based according to claim 4 is grouped, it is characterised in that described to incite somebody to action
Described several different disaggregated models are merged, and obtaining electric energy metrical anomalous identification result includes：
The electric energy metrical Exception Type of the disaggregated model correspondence identification for training is collected, is realized to all electric energy meters
Measure the differentiation of Exception Type.
6. the electric energy metrical abnormity diagnostic system that a kind of feature based is grouped, it is characterised in that including：
Module is built, for according to the electric energy metrical Exception Type and the criterion of the electric energy metrical Exception Type for getting, shape
Into electric energy abnormal data set, and according to the corresponding data source of criterion of the electric energy metrical Exception Type and electric energy exception number
Initial data base is built according to collection；
Conversion module, line number is entered for the data source data to initial data base according to the criterion of the electric energy metrical Exception Type
According to conversion, original measurement abnormal index collection is formed；
Grouping module, for carrying out index feature packet to the original measurement abnormal index collection, forms multiple index groups；
Training module, for carrying out multi-categorizer training to the index group, obtains several different disaggregated models；
Fusion Module, for described several different disaggregated models to be merged, obtains electric energy metrical anomalous identification result.
7. the electric energy metrical abnormity diagnostic system that feature based according to claim 6 is grouped, it is characterised in that the change
Mold changing block includes：
Data variation unit, data prediction is carried out and according to the electric energy metrical for the data source data to initial data base
The criterion of Exception Type carries out data conversion, forms original measurement abnormal index collection, and the data prediction includes correcting or deleting
Except abnormal data, interpolation missing data.
8. the electric energy metrical abnormity diagnostic system that feature based according to claim 7 is grouped, it is characterised in that described point
Group module includes：
Matching unit, for the every kind of electric energy metrical Exception Type and every kind of electricity concentrated to the original measurement abnormal index
The corresponding index of Exception Type can be measured carries out matching arrangement；
Grouped element, for carrying out index feature point according to the characteristic relation between the index and electric energy metrical Exception Type
Group, forms multiple index groups.
9. the electric energy metrical abnormity diagnostic system that feature based according to claim 8 is grouped, it is characterised in that the instruction
Practicing module includes：
Training unit, for according to different electric energy metrical Exception Types, the index set selection to index group each described to be more
Individual grader is trained, and obtains several different sorter models for training, and the sorter model for training can
It is fixed to recognize corresponding electric energy metrical Exception Type.
10. the electric energy metrical abnormity diagnostic system that feature based according to claim 9 is grouped, it is characterised in that described
Fusion Module includes：
Collection unit, it is real for the electric energy metrical Exception Type of the disaggregated model correspondence identification for training to be collected
Now to the differentiation of all electric energy metrical Exception Types.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710161079.XA CN106908752A (en) | 2017-03-17 | 2017-03-17 | The electric energy metrical abnormality diagnostic method and system of a kind of feature based packet |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710161079.XA CN106908752A (en) | 2017-03-17 | 2017-03-17 | The electric energy metrical abnormality diagnostic method and system of a kind of feature based packet |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106908752A true CN106908752A (en) | 2017-06-30 |
Family
ID=59186483
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710161079.XA Pending CN106908752A (en) | 2017-03-17 | 2017-03-17 | The electric energy metrical abnormality diagnostic method and system of a kind of feature based packet |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106908752A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108227678A (en) * | 2018-01-04 | 2018-06-29 | 广东电网有限责任公司电力科学研究院 | A kind of power distribution network no-voltage fault diagnostic method and device based on metering automation system |
| CN111767982A (en) * | 2020-05-20 | 2020-10-13 | 北京大米科技有限公司 | Training method and device for user conversion prediction model, storage medium and electronic equipment |
| CN112084220A (en) * | 2020-09-21 | 2020-12-15 | 国网重庆市电力公司营销服务中心 | Method and device for diagnosing abnormity of electric energy metering device and readable storage medium |
| CN112098917A (en) * | 2020-08-31 | 2020-12-18 | 国网福建省电力有限公司莆田供电公司 | Low-voltage electric energy meter reverse-direction word-walking abnormity troubleshooting method based on electricity consumption data analysis |
| CN112510699A (en) * | 2020-11-25 | 2021-03-16 | 国网湖北省电力有限公司咸宁供电公司 | Transformer substation secondary equipment state analysis method and device based on big data |
| CN113347201A (en) * | 2021-06-25 | 2021-09-03 | 安徽容知日新科技股份有限公司 | Anomaly detection method and system and computing device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1908960A (en) * | 2005-08-02 | 2007-02-07 | 中国科学院计算技术研究所 | Feature classification based multiple classifiers combined people face recognition method |
| CN105652232A (en) * | 2015-12-30 | 2016-06-08 | 国家电网公司 | Stream processing-based electric energy metering device online abnormality diagnosis method and system |
-
2017
- 2017-03-17 CN CN201710161079.XA patent/CN106908752A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1908960A (en) * | 2005-08-02 | 2007-02-07 | 中国科学院计算技术研究所 | Feature classification based multiple classifiers combined people face recognition method |
| CN105652232A (en) * | 2015-12-30 | 2016-06-08 | 国家电网公司 | Stream processing-based electric energy metering device online abnormality diagnosis method and system |
Non-Patent Citations (2)
| Title |
|---|
| 刘毅: "电能计量装置异常状态检测系统总体设计", 《电力系统通信》 * |
| 范洁等: "基于用电信息采集系统的电能计量装置异常智能分析方法研究", 《电测与仪表》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108227678A (en) * | 2018-01-04 | 2018-06-29 | 广东电网有限责任公司电力科学研究院 | A kind of power distribution network no-voltage fault diagnostic method and device based on metering automation system |
| CN111767982A (en) * | 2020-05-20 | 2020-10-13 | 北京大米科技有限公司 | Training method and device for user conversion prediction model, storage medium and electronic equipment |
| CN112098917A (en) * | 2020-08-31 | 2020-12-18 | 国网福建省电力有限公司莆田供电公司 | Low-voltage electric energy meter reverse-direction word-walking abnormity troubleshooting method based on electricity consumption data analysis |
| CN112098917B (en) * | 2020-08-31 | 2022-09-06 | 国网福建省电力有限公司莆田供电公司 | Low-voltage electric energy meter reverse-direction word-walking abnormity troubleshooting method based on electricity consumption data analysis |
| CN112084220A (en) * | 2020-09-21 | 2020-12-15 | 国网重庆市电力公司营销服务中心 | Method and device for diagnosing abnormity of electric energy metering device and readable storage medium |
| CN112510699A (en) * | 2020-11-25 | 2021-03-16 | 国网湖北省电力有限公司咸宁供电公司 | Transformer substation secondary equipment state analysis method and device based on big data |
| CN113347201A (en) * | 2021-06-25 | 2021-09-03 | 安徽容知日新科技股份有限公司 | Anomaly detection method and system and computing device |
| CN113347201B (en) * | 2021-06-25 | 2023-08-18 | 安徽容知日新科技股份有限公司 | Abnormality detection method, abnormality detection system and computing device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106908752A (en) | The electric energy metrical abnormality diagnostic method and system of a kind of feature based packet | |
| CN102866321B (en) | Self-adaptive stealing-leakage prevention diagnosis method | |
| CN107340492A (en) | Electric power meter failure analysis methods with scene anticipation are excavated based on big data | |
| CN110516912B (en) | Method for identifying household transformer relation of distribution station | |
| CN109856299A (en) | A kind of transformer online monitoring differentiation threshold value dynamic setting method, system | |
| CN110609200B (en) | Power distribution network earth fault protection method based on fuzzy metric fusion criterion | |
| CN106154180B (en) | Energy-storage battery charge/discharge anomaly detection method and detection system | |
| CN106443556A (en) | Method for intelligently diagnosing electric energy meter | |
| CN110276511A (en) | A kind of line change relationship anomalous discrimination method based on electricity and line loss relevance | |
| CN111738462B (en) | Fault first-aid repair active service early warning method for electric power metering device | |
| CN105548949A (en) | Method and system for fault remote determination of intelligent ammeter | |
| CN106645934A (en) | Power utilization behavior electricity theft prevention diagnosis method and device based on dynamic grid outliers | |
| CN107991580B (en) | Electrical power distribution network fault location method based on associated weights discreteness multi-source information | |
| CN113267692A (en) | Low-voltage transformer area line loss intelligent diagnosis and analysis method and system | |
| CN106569030A (en) | Alarm threshold optimizing method and device in electric energy metering abnormity diagnosis | |
| CN113222036A (en) | Automatic defect identification method and device for high-voltage cable grounding system | |
| CN110311709A (en) | Power information acquisition system fault distinguishing method | |
| CN114519514B (en) | Low-voltage transformer area reasonable line loss value measuring and calculating method, system and computer equipment | |
| CN111612326A (en) | Comprehensive evaluation method for power supply reliability of distribution transformer | |
| CN110489783A (en) | A kind of low-voltage platform area line loss per unit zone of reasonableness estimation method based on QRNN | |
| CN112688428A (en) | 10kV power grid synchronous line loss treatment method based on Internet of things technology | |
| CN106843100A (en) | Substation secondary device running status level determination method and device | |
| CN111654392A (en) | Low-voltage distribution network topology identification method and system based on mutual information | |
| Benzerga et al. | Low-voltage network topology and impedance identification using smart meter measurements | |
| Ferreira et al. | A data-mining-based methodology for transmission expansion planning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170630 |