CN106355041B - The method of oil-immersed transformer online monitoring data and the correction of live detection data fusion - Google Patents
The method of oil-immersed transformer online monitoring data and the correction of live detection data fusion Download PDFInfo
- Publication number
- CN106355041B CN106355041B CN201610971555.XA CN201610971555A CN106355041B CN 106355041 B CN106355041 B CN 106355041B CN 201610971555 A CN201610971555 A CN 201610971555A CN 106355041 B CN106355041 B CN 106355041B
- Authority
- CN
- China
- Prior art keywords
- data
- live detection
- online monitoring
- oil
- detection data
- 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.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 70
- 238000012544 monitoring process Methods 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 title claims abstract description 37
- 230000004927 fusion Effects 0.000 title claims abstract description 32
- 238000012937 correction Methods 0.000 title claims abstract description 27
- 239000013598 vector Substances 0.000 claims abstract description 48
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 12
- 239000007789 gas Substances 0.000 claims description 21
- 238000007500 overflow downdraw method Methods 0.000 claims description 6
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 claims description 5
- 229910052739 hydrogen Inorganic materials 0.000 claims description 4
- 239000001257 hydrogen Substances 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 238000007476 Maximum Likelihood Methods 0.000 claims description 3
- 230000001419 dependent effect Effects 0.000 claims description 3
- 238000009795 derivation Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- VGGSQFUCUMXWEO-UHFFFAOYSA-N Ethene Chemical compound C=C VGGSQFUCUMXWEO-UHFFFAOYSA-N 0.000 claims description 2
- 239000005977 Ethylene Substances 0.000 claims description 2
- 229910002092 carbon dioxide Inorganic materials 0.000 claims description 2
- 239000001569 carbon dioxide Substances 0.000 claims description 2
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical compound [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 claims 1
- 229910002091 carbon monoxide Inorganic materials 0.000 claims 1
- 125000004435 hydrogen atom Chemical class [H]* 0.000 claims 1
- 230000000295 complement effect Effects 0.000 abstract description 5
- 239000000729 antidote Substances 0.000 description 7
- 238000001556 precipitation Methods 0.000 description 7
- 238000012545 processing Methods 0.000 description 6
- 230000002776 aggregation Effects 0.000 description 5
- 238000004220 aggregation Methods 0.000 description 5
- 238000005259 measurement Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 241001269238 Data Species 0.000 description 2
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 229960004424 carbon dioxide Drugs 0.000 description 2
- 238000013499 data model Methods 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 239000004215 Carbon black (E152) Substances 0.000 description 1
- OTMSDBZUPAUEDD-UHFFFAOYSA-N Ethane Chemical compound CC OTMSDBZUPAUEDD-UHFFFAOYSA-N 0.000 description 1
- ZVNYJIZDIRKMBF-UHFFFAOYSA-N Vesnarinone Chemical compound C1=C(OC)C(OC)=CC=C1C(=O)N1CCN(C=2C=C3CCC(=O)NC3=CC=2)CC1 ZVNYJIZDIRKMBF-UHFFFAOYSA-N 0.000 description 1
- HSFWRNGVRCDJHI-UHFFFAOYSA-N alpha-acetylene Natural products C#C HSFWRNGVRCDJHI-UHFFFAOYSA-N 0.000 description 1
- 229910002090 carbon oxide Inorganic materials 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 125000002534 ethynyl group Chemical group [H]C#C* 0.000 description 1
- 229930195733 hydrocarbon Natural products 0.000 description 1
- 150000002430 hydrocarbons Chemical class 0.000 description 1
- 150000002431 hydrogen Chemical class 0.000 description 1
- 238000011065 in-situ storage Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000000465 moulding Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 235000015170 shellfish Nutrition 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- JLYXXMFPNIAWKQ-UHFFFAOYSA-N γ Benzene hexachloride Chemical compound ClC1C(Cl)C(Cl)C(Cl)C(Cl)C1Cl JLYXXMFPNIAWKQ-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Z—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
- G16Z99/00—Subject matter not provided for in other main groups of this subclass
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Complex Calculations (AREA)
Abstract
The invention discloses a kind of methods of oil-immersed transformer online monitoring data and the correction of live detection data fusion, and first online monitoring data and live detection data are pre-processed to obtain with time interval is identical and time point mutual corresponding two groups of characteristic gas content-time series;Bayes's hierarchical mode is established further according to on-line monitoring characteristic gas content-time series and live detection characteristic gas content-time series, and Bayes's hierarchical mode is estimated using EM algorithm, obtains hyper parameterAnd parameter vector β=[βo,βf] distribution situation, then β is obtained into Z β multiplied by orthogonal vectors ZfMean value and as missing data replace value, fill into obtained in live detection data fusion correction after transformer station high-voltage side bus data.The transformer station high-voltage side bus data and live detection data estimate that level is roughly the same, and farthest remain the variation tendency of online monitoring data, have accomplished have complementary advantages, have broken online data and the respective limitation of charged data.
Description
Technical field
The present invention relates to statistical analysis fields, and in particular to a kind of oil-immersed transformer online monitoring data and live detection
The method of data fusion correction.
Background technique
Currently, the monitoring means of oil-immersed transformer are divided into two kinds, one is on-line monitoring is used, i.e., composed using infrared survey
Instrument remote online monitoring equipment;One is live detection, i.e. professional is sampled detection to transformer oil to scene.Online
The monitoring interval of data is short (usually 1 day), and data distribution is more dense, but the problem of due to on-Line Monitor Device itself,
Accuracy is still to be tested.Sample detection of the live detection data from professional, accuracy is very high, but the mode of artificial detection
Higher cost, therefore detect and be spaced longer (usually at 1 month or more).Online monitoring data can make up for it between live detection data
Every length, defect at high cost, live detection data then can the accuracy to online monitoring data carry out certain correction.Cause
This, the sequence of a fusion live detection data and online monitoring data can reflect the operation conditions of transformer more in time.
Currently, the integration technology for oil-immersed transformer online monitoring data and live detection data still belongs to academic empty
White, coherent signal processing can not use for reference technical experience with DATA REASONING field, we can only be by the correlation of other field
Method carries out certain analysis and processing to the problem.Data fusion technique needs have complementary advantages to different data, beat
The limitation of broken data itself, such technology is in other field using very extensive.For example, satellite data and ground data merger
Technology is exactly a kind of typical data fusion method.So-called satellite data refers to covering satellite with ground data merging techniques
The statistics side that the wider real-time observed data of the lid range manual measurement data very sparse with ground coverage mutually merge
Method, density and the sparsity of manual measurement data and the characteristic of transformer monitoring data of satellite data are extremely agreed with, therefore,
We can use for reference such method, merge to oil-immersed transformer online monitoring data with live detection data.Merge number
It can be transformer actual motion according to that can embody online data trend characteristic and embody the accuracy of charged data
The monitoring of situation provides accurate data basis in real time.
For the processing technique of existing satellite data and ground data based on merger, many scholars utilize the side of aggregation of data
Method integrates two kinds of different data that observation scope greatly differs from each other, as Huffman (Huffman G J, Adler R F,
Rudolf B,et al.Global precipitation estimates based on a technique for
combining satellite-based estimates,rain gauge analysis,and NWP model
Precipitation information [J] .Journal of Climate, 1995,8 (5): 1284-1295.), Arkin
(Xie P,Arkin P A.Global precipitation:A17-year monthly analysis based on
gauge observations,satellite estimates,and numerical model outputs[J]
.Bulletin of the American Meteorological Society, 1997,78 (11): 2539.), Adler
(Adler R F,Huffman G J,Chang A,et al.The version-2global precipitation
climatology project(GPCP)monthly precipitation analysis(1979-present)[J]
.Journal of hydrometeorology, 2003,4 (6): 1147-1167.), Lin (Lin A, Wang X L.An
algorithm for blending multiple satellite precipitation estimates with in
situ precipitation measurements in Canada[J].Journal of Geophysical Research:
Atmospheres, 2011,116 (D21)) etc..In these methods, the method for the propositions such as Lin (2011) is current more forward position
And a kind of higher aggregation of data method of technical maturity, there is stronger reference, processing step can be simply general
It includes are as follows:
(1) set ground data observation sequence as O, satellite data observation sequence is S, satellite data with ground data mutual
The data sequence matched is SO, calculate system deviation ratio R=S of satellite dataO/O。
(2) interpolation is carried out to resulting system deviation ratio, obtains the system deviation ratio R for covering whole observation scopesS,
And utilize RSSatellite data is modified, revised satellite data S ' is obtained.
(3) interpolation is carried out to ground data, obtains sequence O ' identical with satellite data coverage area.O ', to S ' and O '
It is weighted, the fused data after merger can be obtained.
The method that we can use for reference Lin (2011) is identical as live detection data carry out to transformer online monitoring data
Merger processing, but there are following limitations under transformer data scenarios for its method: (1) stability of on-Line Monitor Device is owed
Good, live detection data then belong to manual measurement, and the two has certain, random nature measurement error, above-mentioned merger side
There is no consider the randomness of data wherein for method.(2) essence of the above method is a kind of interpolation method, and interpolation object is ground
Data, thus after merger data trend based on the trend of ground data.And under transformer data scenarios, we are more
Wish that fused data can embody the variation tendency of online data, accuracy is in the controlled range of charged data.
In conclusion there are still a certain distance for the actual demand of the method for Lin (2011) and transformer data fusion.
Summary of the invention
The present invention provides for oil-immersed transformer online monitoring data and live detection data fusion antidote, melt
Close obtained transformer station high-voltage side bus data and live detection data after correction to estimate level roughly the same, and farthest retain
The variation tendency of online monitoring data has accomplished have complementary advantages, has broken online data and the respective limitation of charged data.
The present invention provides a kind of oil-immersed transformer online monitoring data and live detection data fusion correction method,
Include the following steps:
(1), the content that characteristic gas in oil-immersed transformer oil is obtained by remote online monitoring equipment, is denoted as online prison
Measured data;The content that characteristic gas in oil-immersed transformer oil is obtained by manual sampling, is denoted as live detection data;It is right respectively
Online monitoring data and live detection data are pre-processed, obtain time point mutual corresponding two groups of characteristic gas content-when
Between sequence;
(2), according to on-line monitoring characteristic gas content-time series and live detection characteristic gas content-time series
Bayes's hierarchical mode is established, shown in concrete form such as following formula (I):
In formula, Y is the dependent variable of model, if online monitoring data is the random vector Y with complete observationo, electrification
Detection data is the random vector Y with a large amount of missing valuesf, random vector Y=[Yo,Yf] obey mean value be Z β, variance isNormal distribution, wherein Z be Fourier transformation parameter, for portraying the fluctuation of fused data;Parameter vector β=
[βo,βf], obedience mean value is β0, variance isNormal distribution;It is Θ that parameter vector Σ, which obeys mean value, and variance is δ's
GIW distribution;β0, F-1, Θ, δ are hyper parameter;
(3) Bayes's hierarchical mode is estimated using EM algorithm, obtains hyper parameterTo which parameter vector β=[β be calculatedo,βf] distribution situation, then by β multiplied by orthogonal vectors
Z obtains Z βfMean value, and select Z βfMean value as missing data replace value, fill into live detection data, thus
Transformer station high-voltage side bus data to after online monitoring data and the correction of live detection data fusion.
Preferably, the characteristic gas includes hydrogen, methane, ethane, ethylene, acetylene, total hydrocarbon, one in step (1)
Carbonoxide or carbon dioxide;
The content of the characteristic gas is characterized the mass concentration or volumetric concentration of gas.
Preferably, the pretreatment in step (1), to online monitoring data progress are as follows:
With one day for time interval, there is being averaged for multiple data in one day, data are not replaced with linear interpolation;
The pretreatment that live detection data are carried out are as follows:
Also with one day for time interval, there is being averaged for multiple data in one day, not data without being filled up.
On-line monitoring characteristic gas content-time series with complete observation is formd after pretreatment and is existed big
Measure live detection characteristic gas content-time series of missing values, live detection characteristic gas content-time series missing values
It is replaced with null value.
In step (2), parameter vector Σ, decomposed form is as follows:
Wherein, footmark o represents on-line monitoring data, and footmark f represents live detection data;
It is Θ that parameter vector Σ, which obeys mean value, and the GIW that variance is δ is distributed, and the expanded form of GIW distribution is as follows:
τf=(Σ[o,o])-1Σ[o,f]
Γf=Σ[f,f]-Σ[f,o](Σ[o,o])-1Σ[o,f]
Γo=Σ[o,o]
Wherein, random vector τf, ΓfWith ΓoDistribution situation it is as follows:
If Θ={ Ω, τ0,H0,Λf,Λo, δ={ δf,δo, then β0, F-1, Θ, δ are hyper parameter, are denoted asAbove-mentioned random vector and parameter vector constitute online monitoring data and melt with live detection data
The major architectural of molding type.
After establishing corresponding data model, it would be desirable to estimate the parameter of model.Due in model at random to
Measure Yf(live detection sequence) there are a large amount of missing datas, we estimate model using EM algorithm accordingly.EM
(Expectation-Maximization) algorithm is one for estimating the common method of missing data and latent variable, is divided into E-
Step and M-Step two parts, detailed process is as follows:
(a)E-Step
According to maximum likelihood estimate, the likelihood function Q of model is
When known to the observation of Y, above-mentioned likelihood function can be maximized, to acquire hyper parameterValue, however it is existing
(online monitoring data Y known to Y portion in real situationoIt is known that live detection data YfKnown to part), therefore, by known part
(the on-line monitoring whole observations of characteristic gas-time series and the observation of known live detection characteristic gas-time series
Value) it is set as D, and assign hyper parameter vectorOne value,According to given data D and parameter vector, we can count
Calculation obtains the desired value of unknown data, and constructs new likelihood function using obtained desired value
(b)M-Step
To new likelihood function derivation, maximizeObtain hyper parameter valueIt willAs new
E-Step is substituted into, until likelihood function value Q restrains to get the hyper parameter in the case of there is the charged data largely lacked
In step (3), hyper parameter is utilizedParameter vector β=[β is calculatedo,βf] distribution situation, specifically:
Hyper parameter is obtained after being computedTo obtain βo, i.e. parameter vector β=[βo,βf]
Mean value, so that parameter vector β=[β be calculatedo,βf] distribution situation.
Compared with prior art, this method is using Bayes's hierarchical mode and EM algorithm (EM algorithm) to online prison
Measured data and live detection data carry out fusion treatment, it is advantageous that:
(1) Bayes's hierarchical mode has very strong advantage based on probability statistics in terms of portraying data randomness.
(2) EM algorithm has innate advantage in terms of filling up missing data, meanwhile, this method becomes by evidence of online data
Amount, resulting fused data can preferably embody the variation tendency of online data.
Therefore, fusion antidote provided by the invention provides one more for transformer online data and charged data
Science is reliable, meets actual fusion correction solution, the transformer station high-voltage side bus data and live detection data obtained after fusion correction
Estimate horizontal roughly the same, and farthest remain the variation tendency of online monitoring data, accomplished have complementary advantages, beaten
Online data and the respective limitation of charged data are broken.
Detailed description of the invention
Density of hydrogen-the time graph, live detection monitored on-line in certain oil-immersed transformer are set forth in Fig. 1
Density of hydrogen-the time graph obtained after density of hydrogen-time graph, and fusion correction;
Density of hydrogen-the time graph, live detection monitored on-line in certain oil-immersed transformer are set forth in Fig. 2
Density of hydrogen-the time graph obtained after density of hydrogen-time graph, and fusion correction;
Density of hydrogen-the time graph, live detection monitored on-line in certain oil-immersed transformer are set forth in Fig. 3
Density of hydrogen-the time graph obtained after density of hydrogen-time graph, and fusion correction;
Density of hydrogen-the time graph, live detection monitored on-line in certain oil-immersed transformer are set forth in Fig. 4
Density of hydrogen-time graph, and using the density of hydrogen-time graph obtained after distinct methods fusion correction;Wherein, upper figure
The middle fusion antidote using in the present invention, existing aggregation of data method is used in the following figure.
Specific embodiment
Embodiment 1
We are with speaking frankly bright online monitoring data proposed by the present invention and live detection data fusion for following example
The effect of antidote.
Firstly, obtaining the concentration of hydrogen in certain oil-immersed transformer oil by remote online monitoring equipment, it is denoted as online prison
Measured data;The concentration that hydrogen in oil-immersed transformer oil is obtained by manual sampling, is denoted as live detection data;To on-line monitoring
Data are pre-processed with live detection data, keep the time interval of online monitoring data identical, are 1 day, monitor number on-line
It is corresponded according to the time point of live detection data.The case where for online monitoring data missing, take the side of linear interpolation
Method fills up online monitoring data, ultimately forms on-line monitoring characteristic gas content-time with complete observation
Sequence and live detection characteristic gas content-time series with a large amount of missing values.
After completing corresponding data prediction, it would be desirable to establish the shellfish for being directed to online monitoring data and live detection data
This hierarchical mode of leaf, concrete form are as follows:
In above formula, Y is the dependent variable of model, if online monitoring data is the random vector Y with complete observationo, band
Electro-detection number is the random vector Y with a large amount of missing valuesf, random vector Y=[Yo,Yf] obey mean value be Z β, variance isNormal distribution, wherein Z is Fourier transformation parameter, for portraying the fluctuation of fused data, parameter vector β=
[βo,βf, obedience mean value is β0, variance isNormal distribution.
For parameter vector Σ, decomposed form is as follows:
It is Θ that parameter vector Σ, which obeys mean value, and the GIW that variance is δ is distributed, and the expanded form of GIW distribution is as follows:
Definition:
τf=(Σ[o,o])-1Σ[o, f ]
Γf=Σ[ f, f ]-ΣF, o](Σ[o,o])-1Σ[o,f]
Γo=Σ[o,o]
Wherein, random vector τf, ΓfWith ΓoDistribution situation it is as follows:
If Θ={ Ω, τ0,H0,Λf,Λo, δ={ δf, δ o }, then β0, F-1, Θ, δ are hyper parameter, are denoted asAbove-mentioned random vector and parameter vector constitute online data and charged data Fusion Model
Major architectural.
After establishing corresponding data model, it would be desirable to estimate the parameter of model.Due in model at random to
Measure Yf(live detection sequence) there are a large amount of missing datas, we estimate model using EM algorithm accordingly.EM
(Expectation-Maximization) algorithm is one for estimating the common method of missing data and latent variable, is divided into E-
Step and M-Step two parts, detailed process is as follows:
(1)E-Step(Expectation)
According to maximum likelihood estimate, the likelihood function Q of model is
When known to the observation of Y, above-mentioned likelihood function can be maximized, to acquire hyper parameterValue, however it is existing
(online monitoring data Y known to Y portion in real situationoIt is known that live detection data YfKnown to part), therefore, by known part
(the on-line monitoring whole observations of characteristic gas-time series and the observation of known live detection characteristic gas-time series
Value) it is set as D, and assign hyper parameter vectorOne value,According to given data D and parameter vectorWe can count
Calculation obtains the desired value of unknown data, and constructs new likelihood function using obtained desired value
(2)M-Step(Maximization)
To new likelihood function derivation, maximizeObtain hyper parameter valueIt willAs new
E-Step is substituted into, until likelihood function value Q restrains to get the hyper parameter in the case of there is the charged data largely lacked
Estimate the hyper parameter of Bayes's hierarchical modeAfterwards, parameter vector β=[β that we obtaino,βf] distribution
Situation, meanwhile, in order to preferably portray the fluctuation situation of data, we carry out corresponding orthogonal transformation to β, multiplied by orthogonal vectors
Machine Z, finally having estimated mean value is Z βfThe charged data random vector Y containing a large amount of missing valuesf, wherein parameter betafContain
On-line checking data YoPartial information.Under normal conditions, our selection parameter Z βfMean value as missing data replace value,
Transformer station high-voltage side bus data after foring online monitoring data and the correction of live detection data fusion.
As shown in fig. 1, fused correction data and the level of charged data are close, ensure that the accuracy of data,
It is very close with the variation tendency of online data simultaneously, the advantages of having merged the two.
Embodiment 2~3
Using fusion antidote in the same manner as in Example 1, density of hydrogen-time graph for being obtained after fusion correction and
Density of hydrogen-time graph of on-line monitoring and density of hydrogen-time graph of live detection are shown in Fig. 2, Fig. 3 respectively.
Embodiment 4
Using fusion antidote in the same manner as in Example 1, density of hydrogen-time graph for being obtained after fusion correction and
Density of hydrogen-time graph of on-line monitoring and density of hydrogen-time graph of live detection are shown in upper figure in Fig. 4.
Comparative example
Based on online monitoring data and live detection data in embodiment 4, by Bayes's hierarchical mode with it is existing
Aggregation of data method (with reference to the method in Lin (2011)), which combines, obtains density of hydrogen-time graph of one group of fusion correction, sees
The following figure in Fig. 4.
Figure 4, it is seen that existing aggregation of data method is similar to the interpolation processing to charged data, number after fusion
According to trend tightened around charged data, the variation tendencies for not embodying online data well.Using in the present invention
Fusion antidote make fused data and charged data to estimate level roughly the same, and farthest remain
The variation tendency (fused data and the trend of online data are very consistent) of online data has really been accomplished have complementary advantages, has been beaten
Online data and the respective limitation of charged data are broken.
Claims (6)
1. a kind of method of oil-immersed transformer online monitoring data and the correction of live detection data fusion, which is characterized in that packet
Include following steps:
(1), the content that characteristic gas in oil-immersed transformer oil is obtained by remote online monitoring equipment, is denoted as on-line monitoring number
According to;The content that characteristic gas in oil-immersed transformer oil is obtained by manual sampling, is denoted as live detection data;Respectively to online
Monitoring data and live detection data are pre-processed, and time point mutual corresponding two groups of characteristic gas content-time sequence is obtained
Column;
(2), it is established according to on-line monitoring characteristic gas content-time series and live detection characteristic gas content-time series
Bayes's hierarchical mode, shown in concrete form such as following formula (I):
In formula,For the dependent variable of model, if online monitoring data is the random vector with complete observation, live detection
Data are the random vector with a large amount of missing values, random vectorObeying mean value is, variance be's
Normal distribution, whereinFor Fourier transformation parameter, for portraying the fluctuation of fused data;Parameter vector, clothes
It is from mean value, variance isNormal distribution;Parameter vectorObeying mean value is, variance isGIW distribution;,,,It is hyper parameter;
(3) Bayes's hierarchical mode is estimated using EM algorithm, obtains hyper parameter, thus
Parameter vector is calculatedDistribution situation, then willMultiplied by orthogonal vectors, obtainMean value, and selectMean value as missing data replace value, fill into live detection data, thus obtain online monitoring data with electric-examination
Transformer station high-voltage side bus data after measured data fusion correction.
2. the side of oil-immersed transformer online monitoring data according to claim 1 and the correction of live detection data fusion
Method, which is characterized in that in step (1), the characteristic gas includes hydrogen, ethylene, carbon monoxide or carbon dioxide;
The content of the characteristic gas includes the mass concentration or volumetric concentration of characteristic gas.
3. the side of oil-immersed transformer online monitoring data according to claim 1 and the correction of live detection data fusion
Method, which is characterized in that the pretreatment in step (1), to online monitoring data progress are as follows:
With one day for time interval, there is being averaged for multiple data in one day, data are not replaced with linear interpolation;
The pretreatment that live detection data are carried out are as follows:
Also with one day for time interval, there is being averaged for multiple data in one day, not data without being filled up.
4. the side of oil-immersed transformer online monitoring data according to claim 1 and the correction of live detection data fusion
Method, which is characterized in that in step (2), parameter vector, decomposed form is as follows:
Wherein, footmark o represents on-line monitoring data, and footmark f represents live detection data;
Parameter vectorObeying mean value is, variance isGIW distribution, GIW distribution expanded form it is as follows:
Wherein, random vector,WithDistribution situation it is as follows:
If,, then,,,It is hyper parameter, is denoted as。
5. the side of oil-immersed transformer online monitoring data according to claim 1 and the correction of live detection data fusion
Method, which is characterized in that in step (3), using EM algorithm to Bayes's hierarchical mode carry out estimation be divided into E-Step and
M-Step two parts, detailed process is as follows:
(a) E-Step
According to maximum likelihood estimate, the likelihood function Q of model is
Known part is set as, and assign hyper parameter vectorOne value,;According to given dataAnd parameter vector
, the desired value of unknown data can be calculated in we, and constructs new likelihood function using obtained desired value
(b) M-Step
To new likelihood function derivation, maximize, obtain hyper parameter value, willAs newSubstitute into E-
Step, until likelihood function value Q restrains to get the hyper parameter in the case of there is the charged data largely lacked
6. the side of oil-immersed transformer online monitoring data according to claim 1 and the correction of live detection data fusion
Method, which is characterized in that in step (3), utilize hyper parameterParameter vector is calculatedDistribution situation, specifically
Are as follows:
Hyper parameter is obtained after being computed, to obtain, i.e. parameter vectorMean value, from
And parameter vector is calculatedDistribution situation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610971555.XA CN106355041B (en) | 2016-11-04 | 2016-11-04 | The method of oil-immersed transformer online monitoring data and the correction of live detection data fusion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610971555.XA CN106355041B (en) | 2016-11-04 | 2016-11-04 | The method of oil-immersed transformer online monitoring data and the correction of live detection data fusion |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106355041A CN106355041A (en) | 2017-01-25 |
CN106355041B true CN106355041B (en) | 2019-02-05 |
Family
ID=57864882
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610971555.XA Active CN106355041B (en) | 2016-11-04 | 2016-11-04 | The method of oil-immersed transformer online monitoring data and the correction of live detection data fusion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106355041B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107145624B (en) * | 2017-03-28 | 2019-07-30 | 浙江大学 | Gases Dissolved in Transformer Oil online monitoring data antidote based on artificial neural network |
CN106896219B (en) * | 2017-03-28 | 2019-01-29 | 浙江大学 | The identification of transformer sub-health state and average remaining lifetime estimation method based on Gases Dissolved in Transformer Oil data |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN202141720U (en) * | 2011-07-01 | 2012-02-08 | 昆明益通美尔防雷工程有限公司 | Device for intelligently monitoring dissolved gas in transformer oil on line |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10289954B2 (en) * | 2015-01-06 | 2019-05-14 | Accenture Global Services Limited | Power distribution transformer load prediction analysis system |
-
2016
- 2016-11-04 CN CN201610971555.XA patent/CN106355041B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN202141720U (en) * | 2011-07-01 | 2012-02-08 | 昆明益通美尔防雷工程有限公司 | Device for intelligently monitoring dissolved gas in transformer oil on line |
Non-Patent Citations (2)
Title |
---|
变压器典型缺陷局部特性及其带电检测技术研究;钟理鹏等;《高压电器》;20150304;第15-21页 |
油浸式变压器绝缘在线监测系统研究;宋天斌;《中国优秀硕士学位论文全文数据库》;20110415;全文 |
Also Published As
Publication number | Publication date |
---|---|
CN106355041A (en) | 2017-01-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Karra et al. | Global land use/land cover with Sentinel 2 and deep learning | |
CN106815643B (en) | Infrared spectroscopy Model Transfer method based on random forest transfer learning | |
CN112749627A (en) | Method and device for dynamically monitoring tobacco based on multi-source remote sensing image | |
Carouge et al. | What can we learn from European continuous atmospheric CO 2 measurements to quantify regional fluxes–Part 2: Sensitivity of flux accuracy to inverse setup | |
CN103699809B (en) | Water and soil loss space monitoring method based on Kriging interpolation equations | |
CN107918166A (en) | More satellite fusion precipitation methods and system | |
CN106355041B (en) | The method of oil-immersed transformer online monitoring data and the correction of live detection data fusion | |
CN111680870A (en) | Comprehensive evaluation method for target motion trajectory quality | |
CN115000947A (en) | Power distribution network topological structure and line parameter identification method based on intelligent electric meter measurement | |
CN112434814A (en) | Method for analyzing shipping economic potential based on multi-source heterogeneous information fusion algorithm | |
Li et al. | Prediction results of different modeling methods in soil nutrient concentrations based on spectral technology | |
CN108154170A (en) | A kind of NDVI data product fusion methods of long-term sequence | |
CN104424373B (en) | A kind of fine expression of space variable correlation | |
CN107492129B (en) | Non-convex compressive sensing optimization reconstruction method based on sketch representation and structured clustering | |
CN117271979A (en) | Deep learning-based equatorial Indian ocean surface ocean current velocity prediction method | |
CN105092509B (en) | A kind of sample component assay method of PCR-based ELM algorithms | |
CN116183868A (en) | Remote sensing estimation method and system for organic carbon in soil of complex ecological system | |
CN110348094A (en) | Petroleum pipeline leakage detection method and system based on influence network | |
CN109800690A (en) | A kind of non-linear Hyperspectral imaging mixed pixel decomposition method and device | |
CN109444055A (en) | A kind of EO-1 hyperion calculation method of alkaline land soil salt content | |
CN115359197A (en) | Geological curved surface reconstruction method based on spatial autocorrelation neural network | |
Tan et al. | Research on measurement model of water content of oil well based on data fusion | |
Trifunov et al. | A data-driven approach to partitioning net ecosystem exchange using a deep state space model | |
Shoshan et al. | Synthetic data for model selection | |
CN112464848A (en) | Information flow abnormal data monitoring method and device based on density space clustering |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |