CN106384308A - Method and device for processing electric load data of power system - Google Patents
Method and device for processing electric load data of power system Download PDFInfo
- Publication number
- CN106384308A CN106384308A CN201611036325.0A CN201611036325A CN106384308A CN 106384308 A CN106384308 A CN 106384308A CN 201611036325 A CN201611036325 A CN 201611036325A CN 106384308 A CN106384308 A CN 106384308A
- Authority
- CN
- China
- Prior art keywords
- attribute set
- attribute
- power load
- power system
- score
- 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
- 238000000034 method Methods 0.000 title claims abstract description 23
- 238000012545 processing Methods 0.000 title claims abstract description 14
- 238000004458 analytical method Methods 0.000 claims abstract description 28
- 238000010219 correlation analysis Methods 0.000 claims abstract description 17
- 238000007637 random forest analysis Methods 0.000 claims description 20
- 238000013519 translation Methods 0.000 claims description 19
- 230000005611 electricity Effects 0.000 claims description 16
- 238000003672 processing method Methods 0.000 claims description 14
- 238000013459 approach Methods 0.000 claims description 13
- 238000000556 factor analysis Methods 0.000 claims description 13
- 238000001514 detection method Methods 0.000 claims description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 10
- 239000012141 concentrate Substances 0.000 claims description 9
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 239000000284 extract Substances 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 4
- 238000000513 principal component analysis Methods 0.000 claims description 3
- 230000003362 replicative effect Effects 0.000 claims description 2
- 238000012216 screening Methods 0.000 claims description 2
- 239000013598 vector Substances 0.000 abstract description 3
- 230000001131 transforming effect Effects 0.000 abstract 1
- 230000002159 abnormal effect Effects 0.000 description 11
- 238000009826 distribution Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 2
- 240000002853 Nelumbo nucifera Species 0.000 description 2
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005265 energy consumption Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000018199 S phase Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002146 bilateral effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (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)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention relates to a method and a device for processing electric load data of an electric power system, wherein the method comprises the following steps: detecting the power load of a power system, obtaining a power load curve, and extracting a plurality of attribute subsets according to the load curve; calculating the importance of each attribute subset, and selecting an effective attribute subset according to the importance of each attribute subset; and transforming the effective attribute subset through a correlation analysis method to obtain the characteristic data of the power load curve. According to the invention, the low-dimensional power load curve characteristic data can be obtained, and the power system is analyzed based on the low-dimensional power load curve characteristic data, so that the storage space of data can be reduced, the time overhead of calculating the Euclidean distance between vectors can be reduced, and the analysis efficiency is obviously improved.
Description
Technical field
The present invention relates to electric power system data digging technology field, more particularly to a kind of power system power load data
Treating method and apparatus.
Background technology
Continuous improvement with the electric power information degree and electrical network active demand to business model novelty, sends out
Each electrical energy production such as electricity, transmission of electricity, distribution, electricity consumption and the data that all can produce and record magnanimity using link, how reasonably
Using these data and therefrom extract the major issue that valuable information is that current power system faces.
Adapted electricity side is the key data source of power system, in particular with the fast development of intelligent power grid technology, various
Advanced detection means and measuring equipment achieve in power distribution network and are widely applied.Information integrated platform is joined by the battalion of grid company
Achieve distribution generalized information system, marketing system, metering automation system, distribution automation system, distribution production system, distribution work
Multiple different business systems such as journey system are integrated, carry out effective data mining to the multi-source mass data of data integrated system and are
The inexorable trend of intelligent grid development.
In intelligent adapted electrical domain, if the intelligent electric meter in the whole nation all beams back number with every 15 minutes frequencies once at present
According to the daily data total amount producing just can reach PB level, exceedes the data storage of many electronic business enterprise.Intelligent grid
All expect to excavate by magnanimity adapted electricity data value with electricity market, realize the participation electricity consumption interaction of user's depth and price is rung
Should, but up to the present, " data magnanimity, absence of information " is still the major issue that electric power enterprise faces.It is directed to adapted electricity
Data has magnanimity, the feature such as high-dimensional it is necessary to provide a kind of data processing method, from substantial amounts of, fuzzy, noisy
The information of potentially useful is extracted in real data.
Content of the invention
Based on this, the present invention provides a kind for the treatment of method and apparatus of power system power load data, can be to high-dimensional
Power load curvilinear characteristic data carry out dimensionality reduction, from mass data extract potentially useful characteristic.
For achieving the above object, the embodiment of the present invention employs the following technical solutions：
A kind of processing method of power system power load data, comprises the steps：
The power load of detection power system, obtains power load curve, and extracts multiple genus according to described load curve
Temper collection；
Calculate the importance degree of each attribute set, and choose effective attribute set according to the importance degree of each attribute set；
By correlation analysis, described effective attribute set is entered with line translation, obtain power load curvilinear characteristic number
According to.
A kind of processing meanss of power system power load data, including：
Detection module, for detecting the power load of power system, obtains power load curve；
Extraction module, for extracting multiple attribute sets according to described load curve；
Selecting module, for calculating the importance degree of each attribute set, and chooses according to the importance degree of each attribute set
Effectively attribute set；
Conversion module, for described effective attribute set being entered with line translation by correlation analysis, obtains electricity consumption and bears
Lotus curvilinear characteristic data.
Based on the technique scheme of the embodiment of the present invention, multiple attribute sets are extracted according to power load curve, and
Therefrom select effective attribute set, carry out dimension stipulations after this further, correlation is carried out to effective attribute set and divides
Analysis, represents that with the new attribute of negligible amounts belonging to temper originally concentrates information as much as possible, thus the electricity consumption obtaining low dimensional is born
Lotus curvilinear characteristic data.Based on the power load curvilinear characteristic data of low dimensional, power system is analyzed, both can reduce
The memory space of data, and the time overhead calculating Euclidean distance between vector can be reduced, significantly improve analysis efficiency.
Brief description
Fig. 1 is that the processing method flow process in one embodiment of the power system power load data of the present invention is illustrated
Figure；
Calculate the importance degree of attribute set in Fig. 2 embodiment of the present invention and choose effectively according to the importance degree of attribute set
The schematic flow sheet of attribute set；
Fig. 3 is the schematic flow sheet carrying out dimension stipulations in the embodiment of the present invention using principal component analytical method；
Fig. 4 is the schematic flow sheet carrying out dimension stipulations in the embodiment of the present invention using factor-analysis approach；
Fig. 5 is the box traction substation of the tried to achieve Z score of each attribute set successive ignition in the embodiment of the present invention；
Fig. 6 is aimed at the box traction substation of attribute set V3 normal users and abnormal user property value；
Fig. 7 is aimed at the box traction substation of attribute set V10 normal users and abnormal user property value；
Fig. 8 is aimed at the box traction substation of attribute set V2 normal users and abnormal user property value；
Fig. 9 is aimed at the box traction substation of attribute set V14 normal users and abnormal user property value；
Figure 10 is the correlation matrix figure of the attribute set V1-V14 being obtained by correlation analysis；
Figure 11 is the information interpretation ability variance-rate figure of the attribute set being obtained by principal component analytical method；
Figure 12 is the information interpretation ability variance-rate figure of the attribute set being obtained by factor-analysis approach；
Figure 13 is the processing meanss structural representation in one embodiment of the power system power load data of the present invention
Figure；
Figure 14 is that the processing meanss structure in another embodiment of the power system power load data of the present invention is shown
It is intended to.
Specific embodiment
Below in conjunction with preferred embodiment and accompanying drawing, present disclosure is described in further detail.Obviously, hereafter institute
The embodiment of description is only used for explaining the present invention, rather than limitation of the invention.Based on the embodiment in the present invention, this area is general
The every other embodiment that logical technical staff is obtained under the premise of not making creative work, broadly falls into present invention protection
Scope.Although it should be appreciated that hereinafter adopting term " first ", " second " etc. to describe various information, these letters
Breath should not necessarily be limited by these terms, and these terms are only used for same type of information is distinguished from each other out.For example, without departing from this
In the case of bright scope, " first " information can also be referred to as " second " information, similar, and " second " information can also be referred to as
" first " information.It also should be noted that, for the ease of description, illustrate only part related to the present invention in accompanying drawing and
Not all content.
Fig. 1 is that the processing method flow process in one embodiment of the power system power load data of the present invention is illustrated
Figure, as shown in figure 1, the processing method of the power system power load data in the present embodiment comprises the following steps：
Step S110, the power load of detection power system, obtain load curve, and extracted according to described load curve many
Individual attribute set；
Power load refers to the summation of the electrical power that the electrical equipment of power consumer is at a time taken to power system.
Power load examinations to each power consumer in power system, thus power load curve can be obtained, in order to obtain relatively
Adequately data, detection frequency ratio is higher, and detection duration is typically also long, and the dimension of therefore power load curve is very big.
In order that power load curve is suitable to electrical energy consumption analysis, in the present embodiment, dimensionality reduction is carried out to power load curvilinear characteristic, on the one hand may be used
To reduce the memory space of data, the efficiency of electrical energy consumption analysis on the other hand can be improved.
In order to realize dimensionality reduction, in the present embodiment, the physical characteristic according to power load curve extracts multiple attribute sets,
The ascendant trend index of such as power load curve, downward trend index, in front and back difference of the average load of 1 month etc..
Step S120, calculates the importance degree of each attribute set, and chooses effectively according to the importance degree of each attribute set
Attribute set；
After extracting multiple attribute sets, calculate the importance degree of each attribute set by present algorithm, then
Importance degree according to attribute set picks out effective attribute set.
In another embodiment, each attribute set is calculated by the Boruta algorithm based on random forest grader
Importance degree, now the importance degree of attribute set is Z score (Z-score), specifically, with reference to shown in Fig. 2, by based on gloomy at random
The Boruta algorithm of woods grader calculates the importance degree of each attribute set, and has according to the importance degree selection of each attribute set
The process of effect attribute set includes：
Step S121, replicates each described attribute set, and the value of each element in attribute set each described is carried out
Random alignment, obtains the shadow attribute of each described attribute set；
Step S122, the set of described attribute set and the set of described shadow attribute is merged, be expanded genus
Property collection；
Step S123, sets up Random Forest model according to the set of described attribute set, calculates each described attribute set
Z score；
Step S124, sets up Random Forest model according to described extended attribute collection, and the Z calculating each described shadow attribute divides
Number；
Step S125, if the Z score of current attribute set is more than MSZA, using this attribute set as valid genus temper
Collection；Wherein, MSZA is the maximum in the Z score of described shadow attribute.
Assume that attribute set X1 is the ordered series of numbers of a 1*n, then the value of this n element is carried out random alignment, obtains
One new ordered series of numbers is exactly the shadow attribute of X1.After obtaining the shadow attribute of each attribute set, by the set of attribute set with
The set of shadow attribute merges, and be expanded property set.
Random Forest model is a kind of multi-functional machine learning algorithm, in Random Forest model, will generate multiple determining
Plan tree, each decision-making when carrying out discriminant classification in the object new to based on some attributes, in Random Forest model
Tree all can provide the categorizing selection of oneself, and thus carries out " ballot ", and the overall output result of forest would is that poll is most
Classification options.Random Forest model can process High Dimensional Data Set, determine most important variable, and be capable of the weight of output variable
The property wanted degree.Therefore, the present embodiment sets up Random Forest model on extended attribute collection, and calculates the Z of each shadow attribute
Fraction, closes in the collection of attribute set and equally sets up Random Forest model, calculates the Z score of each attribute set, then finds out
Maximum MSZA in the Z score of shadow attribute.For current attribute set Xk (1≤k≤n), judge whether its Z score is big
In MSZA, if it is believed that this attribute set is critically important, retain as effective attribute set；If it is not, then carrying out bilateral hypothesis
Inspection, if its Z score significantly less than MSZA then it is assumed that this attribute set is inessential, by its from extended attribute concentrate remove, such as
Really its Z score be noticeably greater than MSZA then it is assumed that this attribute set important and as effective attribute set retain.By above-mentioned mistake
Journey, loop iteration detects the importance degree of each attribute set and updates extended attribute collection, until all of attribute set is all tested
Survey or reach default maximum iteration time, thus obtaining importance degree to meet the effective attribute set requiring.
Step S130, enters line translation by correlation analysis to described effective attribute set, obtains power load bent
Line characteristic.
After selecting effective attribute set, carry out dimension stipulations further, correlation is carried out to effective attribute set
Analysis, represents that with the new attribute of negligible amounts belonging to temper originally concentrates information as much as possible, thus obtaining the electricity consumption of low dimensional
Load curve characteristic, is hereafter analyzed to power system using the power load curvilinear characteristic data of low dimensional, both may be used
To reduce the memory space of data, and the time overhead calculating Euclidean distance between vector can be reduced, significantly improve analysis efficiency.
In another embodiment, dimension stipulations are carried out using principal component analytical method, with reference to shown in Fig. 3, by correlation
Property analysis method described effective attribute set is entered with line translation, the process obtaining power load curvilinear characteristic data includes：
Step S311, enters line translation by principal component analytical method to described effective attribute set, obtains principal component subset；
Step S312, chooses the principal component subset of predetermined number, concentrates, with from described valid genus temper, attribute selected
Collection merges, and obtains power load curvilinear characteristic data.
Principal component analytical method is a kind of statistical method.By orthogonal transformation, the variable that a group there may be correlation is turned
It is changed to one group of linearly incoherent variable, after conversion, obtain main variables.In the present embodiment, using principal component analytical method
Line translation is entered to effective attribute set, it is possible to obtain multiple principal component subsets, then therefrom choose predetermined number (as 4) main one-tenth
Molecule Set, itself and valid genus temper is concentrated the attribute set selected to merge, such as with power consumer credit grade record
Merge, thus obtaining the power load curvilinear characteristic data of low dimensional.
Above-mentioned power consumer credit grade record, as an attribute set, can pay note according to the power consumer electricity charge
Record or the mode such as ammeter inspection situation are determining.Optionally, power consumer credit grade have respectively from high to low five discrete etc.
Level.
In another embodiment, dimension stipulations are carried out using factor-analysis approach, with reference to shown in Fig. 4, described by phase
Closing property analysis method enters line translation to described effective attribute set, and the process obtaining power load curvilinear characteristic data includes：
Step S321, enters line translation by factor-analysis approach to described effective attribute set, obtains factor subset；
Step S322, chooses the factor subset of predetermined number, concentrates, with from described valid genus temper, the attribute set selected
Merge, obtain power load curvilinear characteristic data.
The basic object of factor-analysis approach is exactly to go to describe the connection between many indexs or factor with a few factor
System, will return in same class by the closer several variables of correlation ratio, each class variable just becomes a factor, with less several
The individual factor reflects the most information of raw data.By factor-analysis approach, effective attribute set is become in the present embodiment
Change, obtain multiple factor subsets, then therefrom choose predetermined number (as 4) principal component subset, by itself and effective attribute set
In select attribute set merge, for example merge with power consumer credit grade record, thus obtaining low dimensional
Power load curvilinear characteristic data.
The processing method of power system power load data of the present invention is carried out in conjunction with a specific example below
Explanation.
Assume that power system has 6200 power consumers, detect 6200 power consumers power load of continuous 18 months,
Sample frequency takes 30 minutes, and the research duration of every power load curved needle pair takes 1 month, then can get after detection
Article 111600, power load curve, every power load curve has 1440 sampled points, and data dimension is 1440.As can be seen here,
The data dimension that direct detection is obtained is too high, is unfavorable for follow-up researching and analysing, therefore needs to carry out dimension-reduction treatment to it.
According to the physical characteristic of power load curve, extract 14 attribute sets based on power load curve first,
Including：
(1) ascendant trend index V1；
(2) downward trend index V2；
(3) difference V3 of the average load of 1 month before and after
(4) difference V4 of the average load of 3 months before and after
(5) difference V5 of the average load of 6 months before and after；
(6) standard deviation V6 of all monthly average load sequences；
(7) standard deviation V7 of front 6 monthly average load sequences；
(8) standard deviation V8 of rear 6 monthly average load sequences；
(9) ratio V9 of latter 3 months load averages and all months load averages；
(10) ratio V10 of latter 6 months load averages and all months load averages；
(11) ratio V11 of latter 9 months load averages and all months load averages；
(12) the slope V12 of average load data linear fit；
(13) the mould V13 of 6 months Fourier coefficient sequence of differences before and after；
(14) coefficient correlation V14 of each customer charge sequence and all customer charge median sequences；
In addition, being also extracted attribute set V15：Power consumer credit grade record.This attribute set can be according to its electricity
Expense is paid record and ammeter inspection situation to determine, it has five discrete level of A, B, C, D, E from high to low respectively.
After determining 15 attribute sets, by each attribute is drawn based on the Boruta algorithm of Random Forest model
The importance degree (Z score) of collection, and be ranked up, choose the high attribute set of importance degree as effective attribute set.Fig. 5 is each genus
The box traction substation of the tried to achieve Z score of temper collection successive ignition, in Fig. 5, shmin, shmean, shmax represent shadow Attribute Significance
Minimum of a value, mean value and maximum box traction substation, the case chart that importance degree is more than shadow Attribute Significance maximum shMax is shown with effect
Attribute set, the case figure that importance degree is less than shadow Attribute Significance minimum of a value shMin represents invalid attribute set.It can be seen that and classification
The larger attribute set of results relevance has V2-V15, totally 14 attribute sets, has efficiently identified out 5 nothings in calculating process
Effect attribute set, including 4 random attribute V16-V19 of ascendant trend index V1 and interpolation.
With the data of known normal users and abnormal user, the effective attribute set chosen is verified below.Fig. 6 is
It is directed to attribute set V3 (difference of the average load of 1 month in front and back), the box traction substation of normal users and abnormal user property value；Figure
7 are aimed at attribute set V10 (ratio of latter 6 months load averages and all months load averages) normal users and abnormal user
The box traction substation of property value.As shown in Figure 6, Figure 7, attribute set V3, attribute set V10 be two on classification results affect larger
Attribute set, be directed to this two attribute sets, the box traction substation of normal users and abnormal user has very big difference.
Fig. 8 is aimed at the box traction substation of attribute set V2 (downward trend index) normal users and abnormal user property value；
It is normal that Fig. 9 is aimed at attribute set V14 (coefficient correlations of each customer charge sequence and all customer charge median sequences)
User and the box traction substation of abnormal user property value, as shown in Figure 8, Figure 9, attribute set V2, V14 are two to be affected on classification results
Less attribute set, for this two attribute sets, the box traction substation difference of normal users and abnormal user is less.Thus may be used
See, after the Boruta algorithm based on Random Forest model filters out effective attribute set, wherein also there is part attribute set
Image for classification results is less, therefore can carry out dimension stipulations by correlation analysis further.
Figure 10 is the correlation matrix figure of the attribute set V1-V14 being obtained by correlation analysis, as shown in Figure 10, can
Intuitively to find out the coefficient correlation between each attribute set of extraction.The degree of correlation of visible part attribute set is higher,
It is that these attribute sets comprise overlay information.Figure 11 is the information interpretation energy of the attribute set being obtained by principal component analytical method
All connection attribute subsets as shown in figure 11, are carried out principal component analysis, obtain principal component subset by power variance-rate figure, main
Become in Molecule Set and contain the most information belonging to temper concentration originally, take wherein 4 principal component subsets, along with attribute set
V15, you can constitute new property set, this property set is as power load curvilinear characteristic data so that carrying out to power system
During analysis, the dimension of data falls below 5 from 1440, greatly reduces data on the premise of ensureing to reflect real information as far as possible
Memory space, also improves analysis efficiency.
Figure 12 is the information interpretation ability variance-rate figure of the attribute set being obtained by factor-analysis approach, such as Figure 12
Shown, respectively all connection attribute subsets are carried out with factorial analysis, obtains factor subset, contain in factor subset and belong to temper originally
The most information concentrated, in the same manner, chooses 4 factor subsets, along with attribute set V15, after constituting new property set, equally
The dimension of analyze data can be reduced.In Figure 12, factor subset 1 mainly reflects the difference of before and after's phase power mode, factor subset 2
Mainly reflect the correlation of user power utilization pattern, factor subset 3 mainly reflects the trend of electricity consumption, factor subset 4 mainly reflects
The fluctuation of mould power mode.
It should be noted that for aforesaid each method embodiment, for easy description, it is all expressed as a series of
Combination of actions, but those skilled in the art should know, and the present invention is not limited by described sequence of movement, because according to
According to the present invention, some steps can be carried out using other orders or simultaneously.
The processing method of the power system power load data according to the invention described above, the present invention also provides a kind of power train
The processing meanss of system power load data, below in conjunction with the accompanying drawings and the power system power load number to the present invention for the preferred embodiment
According to processing meanss be described in detail.
Figure 13 is the processing meanss structural representation in one embodiment of the power system power load data of the present invention
Figure.As shown in figure 13, the processing meanss of the power system power load data in this embodiment include：
Detection module 100, for detecting the power load of power system, obtains power load curve；
Extraction module 200, for extracting multiple attribute sets according to described load curve；
Selecting module 300, for calculating the importance degree of each attribute set, and the importance degree choosing according to each attribute set
Take effective attribute set；
Conversion module 400, for described effective attribute set being entered with line translation by correlation analysis, obtains electricity consumption
Load curve characteristic.
In a kind of optional embodiment, selecting module 300 is by the Boruta algorithm based on random forest grader
Calculate the importance degree of each attribute set, described importance degree is Z score；Selecting module 300 includes：
Shadow attribute acquisition module 301, for replicating each described attribute set, and to each in attribute set each described
The value of element carries out random alignment, obtains the shadow attribute of each described attribute set；
Expansion module 302, for merging the set of described attribute set and the set of described shadow attribute, obtains
Extended attribute collection；
Attribute set Z score computing module 303, for Random Forest model is set up according to the set of described attribute set,
Calculate the Z score of each described attribute set；
Shadow attribute Z score computing module 304, sets up Random Forest model according to described extended attribute collection, calculates each
The Z score of described shadow attribute；
Screening module 305, for current attribute set Z score be more than MSZA when, then using this attribute set as
Effectively attribute set；Wherein, MSZA is the maximum in the Z score of described shadow attribute.
In one embodiment, described correlation analysis include principal component analytical method, with reference to shown in Figure 13, convert
Module 400 includes：
Principal component analysis module 411, for entering line translation by principal component analytical method to described effective attribute set, obtains
Obtain principal component subset；
Fisrt feature determining module 412, for choosing the principal component subset of predetermined number, and from described effective attribute set
In select attribute set merge, obtain power load curvilinear characteristic data.
In another embodiment, described correlation analysis include factor-analysis approach, with reference to shown in Figure 14, convert
Module 400 includes：
Factorial analysis module 421, for entering line translation by factor-analysis approach to described effective attribute set, obtains the factor
Subset；
Second feature determining module 422, for choosing the factor subset of predetermined number, concentrates with from described valid genus temper
Selected attribute set merges, and obtains power load curvilinear characteristic data.
The processing meanss of above-mentioned power system power load data can perform the power system that the embodiment of the present invention is provided
The processing method of power load data, possesses the corresponding functional module of execution method and beneficial effect.
Each technical characteristic of embodiment described above can arbitrarily be combined, for making description succinct, not to above-mentioned reality
The all possible combination of each technical characteristic applied in example is all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, all it is considered to be the scope of this specification record.
Embodiment described above only have expressed the several embodiments of the present invention, and its description is more concrete and detailed, but simultaneously
Can not therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
Say, without departing from the inventive concept of the premise, some deformation can also be made and improve, these broadly fall into the protection of the present invention
Scope.Therefore, the protection domain of patent of the present invention should be defined by claims.
Claims (10)
1. a kind of processing method of power system power load data is it is characterised in that comprise the steps：
The power load of detection power system, obtains power load curve, and extracts multiple attribute according to described load curve
Collection；
Calculate the importance degree of each attribute set, and choose effective attribute set according to the importance degree of each attribute set；
By correlation analysis, described effective attribute set is entered with line translation, obtain power load curvilinear characteristic data.
2. power system power load data according to claim 1 processing method it is characterised in that by based on
The Boruta algorithm of machine forest classified device calculates the importance degree of each attribute set.
3. the processing method of power system power load data according to claim 2 is it is characterised in that described importance degree
For Z score, by calculating the importance degree of each attribute set based on the Boruta algorithm of random forest grader, and according to each
The process that the importance degree of attribute set chooses effective attribute set includes：
Replicate each described attribute set, and random alignment is carried out to the value of each element in attribute set each described, obtain
The shadow attribute of each described attribute set；
The set of described attribute set and the set of described shadow attribute are merged, be expanded property set；
Random Forest model is set up according to the set of described attribute set, calculates the Z score of each described attribute set；
Random Forest model is set up according to described extended attribute collection, calculates the Z score of each described shadow attribute；
If the Z score of current attribute set is more than MSZA, using this attribute set as effective attribute set；Wherein, MSZA
For the maximum in the Z score of described shadow attribute.
4. the processing method of power system power load data according to claim 1 is it is characterised in that described correlation
Analysis method includes principal component analytical method, described by correlation analysis, line translation is entered to described effective attribute set,
The process obtaining power load curvilinear characteristic data includes：
Line translation is entered by principal component analytical method to described effective attribute set, obtains principal component subset；
Choose the principal component subset of predetermined number, concentrate the attribute set selected to merge with from described valid genus temper, obtain
Obtain power load curvilinear characteristic data.
5. the processing method of power system power load data according to claim 1 is it is characterised in that described correlation
Analysis method includes factor-analysis approach, described by correlation analysis, line translation is entered to described effective attribute set, obtain
The process obtaining power load curvilinear characteristic data includes：
Line translation is entered by factor-analysis approach to described effective attribute set, obtains factor subset；
Choose the factor subset of predetermined number, concentrate the attribute set selected to merge with from described valid genus temper, obtain electricity consumption
Load curve characteristic.
6. the processing method of the power system power load data according to claim 4 or 5 is it is characterised in that from described
It is power consumer credit grade record that valid genus temper concentrates the attribute set selected.
7. a kind of processing meanss of power system power load data are it is characterised in that include：
Detection module, for detecting the power load of power system, obtains power load curve；
Extraction module, for extracting multiple attribute sets according to described load curve；
Selecting module, for calculating the importance degree of each attribute set, and chooses effectively according to the importance degree of each attribute set
Attribute set；
Conversion module, for described effective attribute set being entered with line translation by correlation analysis, obtains power load bent
Line characteristic.
8. the processing meanss of power system power load data according to claim 7 are it is characterised in that described selection mould
Block is divided for Z by being calculated the importance degree of each attribute set, described importance degree based on the Boruta algorithm of random forest grader
Number；Described selecting module includes：
Shadow attribute acquisition module, for replicating each described attribute set, and to each element in attribute set each described
Value carries out random alignment, obtains the shadow attribute of each described attribute set；
Expansion module, for merging the set of described attribute set and the set of described shadow attribute, be expanded genus
Property collection；
Attribute set Z score computing module, for setting up Random Forest model according to the set of described attribute set, calculates each
The Z score of described attribute set；
Shadow attribute Z score computing module, sets up Random Forest model according to described extended attribute collection, calculates each described shadow
The Z score of attribute；
Screening module, for when the Z score of current attribute set is more than MSZA, then using this attribute set as effective attribute
Subset；Wherein, MSZA is the maximum in the Z score of described shadow attribute.
9. the processing meanss of power system power load data according to claim 7 are it is characterised in that described correlation
Analysis method includes principal component analytical method, and described conversion module includes：
Principal component analysis module, for described effective attribute set being entered with line translation by principal component analytical method, obtains main one-tenth
Molecule Set；
Fisrt feature determining module, for choosing the principal component subset of predetermined number, selectes with concentrating from described valid genus temper
Attribute set merge, obtain power load curvilinear characteristic data.
10. the processing meanss of power system power load data according to claim 7 are it is characterised in that described correlation
Property analysis method includes factor-analysis approach, and described conversion module includes：
Factorial analysis module, for entering line translation by factor-analysis approach to described effective attribute set, obtains factor subset；
Second feature determining module, for choosing the factor subset of predetermined number, selected with from described valid genus temper concentration
Attribute set merges, and obtains power load curvilinear characteristic data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611036325.0A CN106384308A (en) | 2016-11-16 | 2016-11-16 | Method and device for processing electric load data of power system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611036325.0A CN106384308A (en) | 2016-11-16 | 2016-11-16 | Method and device for processing electric load data of power system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106384308A true CN106384308A (en) | 2017-02-08 |
Family
ID=57959239
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611036325.0A Pending CN106384308A (en) | 2016-11-16 | 2016-11-16 | Method and device for processing electric load data of power system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106384308A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108011367A (en) * | 2017-12-04 | 2018-05-08 | 贵州电网有限责任公司电力科学研究院 | A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms |
| CN108954680A (en) * | 2018-07-13 | 2018-12-07 | 电子科技大学 | A kind of air-conditioning energy consumption prediction technique based on operation data |
| CN118296296A (en) * | 2024-06-06 | 2024-07-05 | 湖北华中电力科技开发有限责任公司 | Intelligent power distribution method based on electric power big data |
-
2016
- 2016-11-16 CN CN201611036325.0A patent/CN106384308A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108011367A (en) * | 2017-12-04 | 2018-05-08 | 贵州电网有限责任公司电力科学研究院 | A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms |
| CN108011367B (en) * | 2017-12-04 | 2020-12-18 | 贵州电网有限责任公司电力科学研究院 | Power load characteristic mining method based on depth decision tree algorithm |
| CN108954680A (en) * | 2018-07-13 | 2018-12-07 | 电子科技大学 | A kind of air-conditioning energy consumption prediction technique based on operation data |
| CN118296296A (en) * | 2024-06-06 | 2024-07-05 | 湖北华中电力科技开发有限责任公司 | Intelligent power distribution method based on electric power big data |
| CN118296296B (en) * | 2024-06-06 | 2024-08-20 | 湖北华中电力科技开发有限责任公司 | Intelligent power distribution method based on electric power big data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109492667A (en) | A kind of feature selecting discrimination method for non-intrusive electrical load monitoring | |
| CN108960833A (en) | A kind of abnormal transaction identification method based on isomery finance feature, equipment and storage medium | |
| Feng et al. | Data mining for abnormal power consumption pattern detection based on local matrix reconstruction | |
| CN109767054A (en) | Efficiency cloud appraisal procedure and edge efficiency gateway based on deep neural network algorithm | |
| Zhu et al. | Hybrid-order representation learning for electricity theft detection | |
| CN110321934A (en) | Method and system for detecting abnormal data of user electricity consumption | |
| CN106384308A (en) | Method and device for processing electric load data of power system | |
| CN112308341B (en) | Power data processing method and device | |
| Dylewsky et al. | Stochastically forced ensemble dynamic mode decomposition for forecasting and analysis of near-periodic systems | |
| Sun et al. | PF-FEDG: An open-source data generator for frequency disturbance event detection with deep-learning reference classifiers | |
| Grigoras et al. | Processing of smart meters data for peak load estimation of consumers | |
| Tayeb et al. | Time series database preprocessing for data mining using Python | |
| Yao et al. | SVD-AE: An asymmetric autoencoder with SVD regularization for multivariate time series anomaly detection | |
| CN111199014B (en) | Time sequence based seq2point NILM method and device | |
| Wu et al. | Forecast of short‐term electricity price based on data analysis | |
| Naderian | A novel hybrid deep learning approach for non-intrusive load monitoring of residential appliance based on long short term memory and convolutional neural networks | |
| CN110457572B (en) | Commodity information recommendation method based on graph network and electronic equipment | |
| CN114117210A (en) | Intelligent financial product recommendation method and device based on federal learning | |
| CN112614005A (en) | Enterprise rework state processing method and device | |
| Aravind et al. | DEVELOPMENT OF AN EFFICIENT DEEP LEARNING SYSTEM FOR AUTOMATIC PREDICTION OF POWER DEMAND BASED ON THE FORECASTING OF POWER DISTRIBUTION. | |
| Reddy et al. | Detection of Electricity Theft Cyber-Attacks In Renewable Distributed Generation for Future IoT-based Smart Electric Meters | |
| CN110288114A (en) | Violation electricity consumption behavior prediction method based on power marketing data | |
| Mhaske et al. | An efficient electricity theft detection using xg boost | |
| Weng et al. | Research on the Construction of a Risk Assessment Indicator System for Transportation Infrastructure Investment under Public–Private Partnership Model | |
| Zhang et al. | Topology identification method of low voltage aea based on topological data analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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: 20170208 |