CN109490685A - A kind of transformer early defect method for early warning based on oil dissolved gas on-line monitoring - Google Patents
A kind of transformer early defect method for early warning based on oil dissolved gas on-line monitoring Download PDFInfo
- Publication number
- CN109490685A CN109490685A CN201811275492.XA CN201811275492A CN109490685A CN 109490685 A CN109490685 A CN 109490685A CN 201811275492 A CN201811275492 A CN 201811275492A CN 109490685 A CN109490685 A CN 109490685A
- Authority
- CN
- China
- Prior art keywords
- key
- node
- network
- transformer
- period
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/50—Testing of electric apparatus, lines, cables or components for short-circuits, continuity, leakage current or incorrect line connections
- G01R31/62—Testing of transformers
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N30/00—Investigating or analysing materials by separation into components using adsorption, absorption or similar phenomena or using ion-exchange, e.g. chromatography or field flow fractionation
- G01N30/02—Column chromatography
Abstract
Transformer early defect method for early warning based on oil dissolved gas on-line monitoring, comprising: step I. is based on Gases Dissolved in Transformer Oil on-Line Monitor Device, establishes the complex network model that can describe transformer Dynamic Evolutional characteristic；Each node data of complex network model is normalized in Real-time Monitoring Data of the step II. based on Gases Dissolved in Transformer Oil on-Line Monitor Device, and dynamic screens the key node in current transformer complex network model；The key node screened is formed key network by step III., and the average for calculating separately key network under day part (i) is poorAverage Pearson correlation coefficient between each nodeAnd the average Pearson correlation coefficient between the node and other non-key nodes in key networkAnd the quantized value I of the key network marker under present period (i) is obtained according to calculated resulti: step IV. is as (Ii‑Ii‑1) > 0.05 and (Ii‑Ii+1) > 0.05 when, issue warning signal at this time.
Description
Technical field
The present invention relates to the early warning of power transformer defect and fault diagnosis technology field, more particularly to one kind based in oil
The transformer early defect method for early warning of dissolved gas on-line monitoring.
Background technique
Power transformer is grid equipment most basic and important in electric system, while it is also that accident easily occurs
A kind of equipment, carries out defect early warning to transformer and fault diagnosis is the important content for guaranteeing power system stability operation, wherein
Monitoring and analysis oil dissolved gas (DGA) have important role to guarantee transformer safety stable operation.
Currently, having with relatively broad the problem of being applied at the scene based on IEC three-ratio method: 1) this method master
Gas concentration data are monitored with analysis, is a kind of static analysis method, can not embody the advantage of on-line monitoring system；2) should
The case where method is mainly threshold alarm, and volume fraction does not reach demand value is not applicable, and the setting of alarm threshold is difficult, lacks and fills
Divide foundation；3) there are scarce encoded questions, and encoded boundary is excessively absolute, easily judge by accident.For the deficiency of traditional diagnosis method,
The introducing of the intelligent algorithm based on machine learning such as support vector machines, artificial neural network, integrated intelligent algorithm improves transformation
The accuracy of device fault diagnosis, these methods are based primarily upon the oil dissolved gas sample under collected transformer typicalness
Data are trained built intelligent diagnostics model, then the intelligent diagnostics model established can effectively be known after training successfully
Not various typical defects, rate of correct diagnosis and robustness are superior to traditional IEC three-ratio method, therefore have received widespread attention.
But in the diagnostic model establishment process based on machine learning, it is desirable to provide the sample under diagnosis object typicalness
Notebook data.Although there is certain relevance between the monitoring data of different transformers, it is contemplated that running environment and transformation
Influence of device state (factors such as model, producer, enlistment age) for oil dissolved gas ingredient and concentration itself, Accurate Diagnosis
One running state of transformer needs enough typicalness sample datas based on individual, and which increase the difficulties of model foundation
Degree, also the popularization for the transformer DGA method for diagnosing faults based on machine learning brings difficulty.
To " critical phase transformation " theory in complex systematic dynamics studies have shown that the parameter of complication system is being changed to certain
It will lead to system stability when one " critical point " to change, and this behavior of critical slowing can be shown near the point, and
Generate three possible pre-warning signals: disturbance restores relatively slow, autocorrelation increases, variance increases, therefore critical slow by detecting
Changing signal can be with the state change of forecasting system.Transformer fault generating process is a dynamic process, it from health status to
There is also a critical states during malfunction changes, when the characteristic gas content dissolved in oil increases to certain journey
Degree, when reaching critical point, system will be converted to suddenly unstable state from stable state, critical transitions phenomenon occur.
Therefore, the dynamic change of oil dissolved gas is monitored by oil chromatography on-line monitoring system, detects transformer state
The pre-warning signal occurred in transformation helps to carry out early defect early warning and identification to transformer.
Summary of the invention
It is a primary object of the present invention to overcome drawbacks described above in the prior art, propose a kind of based on oil dissolved gas
The transformer early defect method for early warning of on-line monitoring.
The present invention adopts the following technical scheme:
Transformer early defect method for early warning based on oil dissolved gas on-line monitoring, which is characterized in that including following
Step:
Step I. is based on Gases Dissolved in Transformer Oil on-Line Monitor Device, and foundation can describe transformer Dynamic Evolutional
The complex network model of characteristic；
Real-time Monitoring Data of the step II. based on Gases Dissolved in Transformer Oil on-Line Monitor Device, setting are reasonably adopted
Sample time interval is a time cycle, and each node data of complex network model is normalized, and dynamic is screened
Key node in current transformer complex network model, node data standard deviation normalize formula are as follows:
Wherein XijIndicate the normalization expression formula data of node in (j) a sampling time point in period (i), xijWhen being
In section (i) in (j) a sampling time point node value, xiRepresent the node data of period (i), and mean (xi) and SD (xi)
Respectively indicate the average and standard deviation of node in all sampling time points in the period (i)；
The key node screened is formed key network by step III., calculates separately crucial net under day part (i)
The average of network is poorAverage Pearson correlation coefficient in key network between each nodeAnd it is crucial
The average Pearson correlation coefficient between node and other non-key nodes in networkTo judge the key network
Whether meet the critical characteristic of transformer state transformation, and obtains the key network mark under present period (i) according to calculated result
The quantized value I of will objecti:
ε is one for avoiding the small normal number that denominator is zero in formula, if denominator, which is not zero, to be saved；
Step IV. calculates the quantized value I of the key network marker under the key network day part (i)i, as (Ii-Ii-1)
> 0.05 and (Ii-Ii+1) > 0.05 when, it is meant that complex network model corresponding to transformer has occurred in period (i) state
Threshold variations, by stable state Ii-1Through critical state IiIt is converted into defect state Ii+1, issue warning signal at this time.
The step I specifically:
Every kind of representative gases that Gases Dissolved in Transformer Oil on-Line Monitor Device is monitored are mapped as complex network mould
A node in type takes full mutual contact mode between node.
The step II specifically:
Step II-1. is established based on historical data sequence acquired in Gases Dissolved in Transformer Oil on-Line Monitor Device
The prediction model of each type gas；
The prediction data that step II-2. obtains currently practical monitoring data and the prediction model established based on historical data
Carry out significance difference analysis；
If certain gas measured value of step II-3. deviates from expected dynamic trajectory and has conspicuousness poor with desired value
It is different, then it is assumed that the variation of the gas may influence the state of transformer complex network model, which is key node.
The step II-2 specifically:
Step II-2-1. sets the period (i) and shares the measured data of 7 groups of representative gases and the prediction data of 7 groups of representative gases,
Gas data including each sampled point under the period；Variance analysis is carried out to each gas respectively using t method of inspection, if conspicuousness water
Flat α=0.05, it is (1) p, p (2), p (3), p (4), p (5), p (6), p respectively that the corresponding p value of each gas arranges from small to large
(7), wherein p value reflects measured data and the prediction data probability that there was no significant difference, and two groups of data differences are without aobvious if p > α
Meaning, otherwise two groups of data differences have significant meaning；
Step II-2-2. combination false discovery rate carries out FDR correction to p value, judges whether the p value of gas meets:
M=1,2 ..., 7 indicates 7 kinds of gases, tentatively assert that the gas is key node if meeting, wherein p (m) is represented
The corresponding p value of gas；
Step II-2-3. combines two multiple method of changing further to screen the tentatively selected gas variable being changed significantly,
If meeting:
Then think that the gas measured value deviates from expected dynamic trajectory and has significant difference, corresponding section with desired value
Point is key node；Wherein A1i, A2iGas A is represented in the measured value and predicted value of period (i), SD (A1i) be the gas when
The standard deviation of section (i) measured value, SD (A2i) it is standard deviation of the gas in period (i) predicted value.
The step III specifically:
The average that step III-1. calculates the key network of present period (i) is poor
M is key node number in network in formula, and N is sampling point number, XijIndicate network node the within the period (i)
(j) normalization data in a sampling time point,For in key network the gas data period (i) it is flat
Mean value；
Step III-2. calculates the average Pearson correlation coefficient in the key network of present period (i) between each node
R in formulamIndicate the related coefficient in key network between m-th of node and other interior nodes of network；
Step III-3. calculates the average skin between node and other non-key nodes in the key network of present period (i)
The inferior related coefficient of that
R in formulamIndicate the related coefficient in key network between m-th of node and other outer nodes of network；
Step III-4. is according to calculated result, if meeting following critical characteristic:
1) average of key network is poorWithCompared to dramatically increasing；
2) the average Pearson correlation coefficient in key network between each nodeWithCompared to increasing
Add；
3) the average Pearson correlation coefficient between the node in key network and other non-key nodesWithCompared to reduction；
Then complex network model corresponding to transformer may reach the Near The Critical Point of state transformation, and according to calculated result
Obtain the quantized value I of the key network marker under day part (i)i, by the dynamic change for monitoring the key network marker
Facilitate the pre-warning signal of detection of complex network model critical transitions.
Correlation coefficient r between the key network interior nodes between exterior node in key networkmAnd RmCalculation formula it is specific
Are as follows:
Pearson correlation coefficient between key network interior nodes are as follows:
Wherein m ≠ k, XimIndicate the normalization data of (m) a key node in key network in present period (i), Yik
Indicate the normalization data of (k) a key node in key network in period (i),WithIndicate corresponding node when current
Mean value in section (i) on all sampled points；
Pearson correlation coefficient calculation formula in key network between other non-key nodes are as follows:
Wherein XimIndicate the normalization data of (m) a key node in key network in present period (i), ZikIt indicates
The normalization data of (k) a non-key node in period (i),WithIndicate that corresponding node owns in present period (i)
Mean value on sampled point.
The calculating of first critical characteristic specifically:
Judge that average is poor using multiple method of changingWhether dramatically increase:
In formulaWithThe respectively key network standard deviation mean value of period (i) and period (i-1) is led to
Often using 2~3 times of differences as threshold value, work as FcThink that incrementss are smaller when < 2, as 2 < FcThink to increase when < 4, works as FcRecognize when > 4
It is larger for incrementss.
By the above-mentioned description of this invention it is found that compared with prior art, the invention has the following beneficial effects:
The present invention be model-free calculation method, according to during voltage transformer system critical transitions group dynamics effect and
The history and Real-time Monitoring Data of Gases Dissolved in Transformer Oil on-Line Monitor Device, to the pre-warning signal of transformer state transformation
It is detected, overcomes failure mistaken diagnosis caused by causing physical model unreliable because of sample size deficiency, leakage alarm and false alarm；
The present invention is a kind of dynamic diagnosis method, compared with traditional static analytic approach, is supervised to the dynamic of running state of transformer
Survey can reflect the dynamic process that system changes from normal operating condition to malfunction, hold the trend of transformer development and change；
The present invention is solved because of inside transformer complicated mechanism, and the complicated coupling relationship of oil colours modal data and fault type is led
The problem of causing data not to correspond to the mapping of failure, and the dynamic network established accurate can find in advance transformation
The critical transitions point of device failure, gives warning in advance to transformer defect.
Detailed description of the invention
Fig. 1 is flow chart of the present invention；
Fig. 2 is transformer complex network model established by the present invention and the corresponding key network of failure；
Fig. 3 is that the average of key network day part of the present invention is poorValue；
Fig. 4 is the average Pearson correlation coefficient of key network day part of the present inventionValue；
Fig. 5 is the average Pearson correlation coefficient of key network day part of the present inventionValue；
Fig. 6 is the quantized value I of the key network marker of key network day part of the present inventioni；
Specific embodiment
Below by way of specific embodiment, the invention will be further described.
For process referring to Fig. 1, transformer used by the present embodiment is Alstom, Britain (existing AREVA company) manufacture
3 × 277MVA 515/22KV single-phase transformer group, wiring group YNd1, outlet in the middle part of high and low pressure side, impedance voltage
16.6%.Main transformer is open type transformer, and conservator is interior without diaphragm and capsule.Transformer insulation oil is public using Sweden NYNAS
The Nytro 10GBN of department adds the light naphthenic mineral oil of hydrogen.It is established according to the variation of 1 different faults gas componant of table as shown in Figure 2
Transformer complex network model and failure correspond to key network, and (gas-monitoring project is as shown in table 2, does not measure in the present embodiment
CO2, therefore institute's establishing network is 6 meshed networks), wherein inclined stripe dot is that failure corresponds to key node, and soft dot is non-pass
Key node.Emulation explanation is carried out to the present invention using MATLAB as workbench, main transformer C phase chromatography is chosen and surveys level-one alarm condition
40 preceding sampled point chromatographic datas, every 5 sampled points are set as a time cycle, are divided into 8 periods, in this alarm condition
Before generation, each gas data variation is below demand value.
The variation of 1 different faults gas componant of table:
Each gas-monitoring project of table 2:
Based on the transformer complex network model established, the reality based on Gases Dissolved in Transformer Oil on-Line Monitor Device
When monitoring data, standard deviation normalization is carried out to node data each in complex network, formula is as follows:
Wherein XijIndicate the normalization expression formula data of node in (j) a sampling time point in period (i), xijWhen being
In section (i) in (j) a sampling time point node value, xiRepresent the node data of period (i), and mean (xi) and SD (xi)
Respectively indicate the average and standard deviation of node in all sampling time points in the period (i).
Based on historical data sequence acquired in Gases Dissolved in Transformer Oil on-Line Monitor Device, each type gas is established
The prediction model of body, the present embodiment are based on the regression forecasting mould that support vector machines (SVR) establishes associated gas monitoring project data
Type, and the prediction data obtained to currently practical monitoring data and the SVR prediction model established based on historical data carries out difference
Significance analysis, dynamic screen the key node in current transformer complex network.Detailed process are as follows:
If the period (i) shares the measured data of 6 groups of representative gases and the prediction data of 6 groups of representative gases, including the period
Under each sampled point gas data, data are also possible to 7 groups or other.Difference point is carried out to each gas respectively using t method of inspection
Analysis, if level of significance α=0.05, it is (1) p, p (2), p (3), p (4), p respectively that the corresponding p value of each gas arranges from small to large
(5), (6) p, wherein p value reflection measured data is with the prediction data probability that there was no significant difference, two groups of data differences if p > α
Without significant meaning, otherwise two groups of data differences have significant meaning.FDR correction is carried out to p value in conjunction with false discovery rate, judges gas
P value whether meet:
M=1,2 ..., 6 indicates 6 kinds of gases, tentatively assert that the gas is key node if meeting, wherein p (m) is represented
The corresponding p value of gas.Then two multiple method of changing are combined further to screen the tentatively selected gas variable being changed significantly, if
Meet:
Then think that the gas measured value deviates from expected dynamic trajectory and has significant difference with desired value, the gas
Variation may influence the state of transformer complex network, corresponding node is key node.Wherein A1i, A2iRepresent gas A when
The measured value and predicted value of section (i), SD (A1i) it is standard deviation of the gas in period (i) measured value, SD (A2i) it is that the gas exists
The standard deviation of period (i) predicted value.
The key node screened is formed into key network (key node that day part filters out is as shown in table 3),
The average for calculating separately key network under present period (i) is poorAverage Pierre in key network between each node
Inferior related coefficientAnd the average Pearson correlation coefficient between the node and other non-key nodes in key networkTo judge whether the key network meets the critical characteristic of transformer state transformation, detailed process are as follows:
The average for calculating the key network of present period (i) first is poor
M is key node number in network in formula, and N is sampling point number, XijIndicate network node the within the period (i)
(j) normalization data in a sampling time point,For in key network the gas data period (i) it is flat
Mean value.
Then the average Pearson correlation coefficient in the key network of present period (i) between each node is calculated
R in formulamIndicate the related coefficient in key network between m-th of node and other interior nodes of network.
Further calculate the average Pearson came between the node and other non-key nodes in the key network of present period (i)
Related coefficient
R in formulamIndicate the related coefficient in key network between m-th of node and other outer nodes of network.
Wherein, the correlation coefficient r between the key network interior nodes between exterior node in key networkmAnd RmCalculating it is public
Formula specifically:
Pearson correlation coefficient between key network interior nodes are as follows:
Wherein XimIndicate the normalization data of (m) a key node in key network in present period (i), YikIt indicates
In period (i) in key network (k) a key node normalization data,WithIndicate corresponding node in present period
(i) mean value on all sampled points.
Pearson correlation coefficient calculation formula in key network between other non-key nodes are as follows:
Wherein XimIndicate the normalization data of (m) a key node in key network in present period (i), ZikIt indicates
The normalization data of (k) a non-key node in period (i),WithIndicate that corresponding node owns in present period (i)
Mean value on sampled point.
Further, according to calculated result, if meeting following critical characteristic:
1) average of key network is poorWithCompared to dramatically increasing；
2) the average Pearson correlation coefficient in key network between each nodeWithCompared to increasing
Add；
3) the average Pearson correlation coefficient between the node in key network and other non-key nodesWithCompared to reduction.
Wherein, the calculating of first critical characteristic specifically:
Judge that average is poor using multiple method of changingWhether dramatically increase:
In formulaWithThe respectively key network standard deviation mean value of period (i) and period (i-1) is led to
Often using 2~3 times of differences as threshold value, work as FcThink that incrementss are smaller when < 2, as 2 < FcThink to increase when < 4, works as FcRecognize when > 4
It is larger for incrementss.The present embodiment chooses Fc=4 as the threshold value for judging that average difference dramatically increases, that is, work asWhen think that the average of key network is poorWithCompared to dramatically increasing.
Further, if meeting above three critical characteristic, illustrate that complex network corresponding to transformer may reach shape
The Near The Critical Point of state transformation, and obtain according to calculated result the quantized value I of the key network marker under present period (i)i,
Dynamic change by monitoring the key network marker facilitates the relevant early warning letter of detection of complex network critical transitions
Number.The quantized value I of key network markeriCalculation formula specifically:
ε is one for avoiding the small normal number that denominator is zero in formula, if denominator, which is not zero, to be saved.
Further, the quantized value I of the key network marker under the key network day part (i) is calculatedi, as (Ii-
Ii-1) > 0.05 and (Ii-Ii+1) > 0.05 when, it is meant that complex network corresponding to transformer has occurred in period (i) state
Threshold variations, by stable state Ii-1Through critical state IiIt is converted into defect state Ii+1, issue warning signal at this time.
The key node situation that 3 day part of table filters out:
The transformer from debugging to the operation phase, occur that neutral point connector overheat, air content be high and oil in hydrogen and total
The problems such as hydrocarbon content is higher is calculated according to the oil sample data of main transformer three-phase and belongs to cryogenic overheating fault coverage.
As seen in figures 3-6, the key network of period (3) each index under day part different degrees of calculated result occurs
Significant change.It can see from Fig. 3-6, the key network of period (3) is poor in the period (1) later averageWithCompared to sharply increasing, and reach peak value (Fig. 3) at period (2), while the average skin between the key network interior nodes
The inferior related coefficient of thatWithCompared to increase (Fig. 4), and between the node and non-key node in key network
Average Pearson correlation coefficientWithCompared to (Fig. 5) is then reduced, meet critical characteristic.Illustrate at this
Period key network forms the sub-network of a strong correlation, and issues warning signal, i.e. (I2-I1) > 0.05 and (I2-I3) >
0.05, show that system may reach the Near The Critical Point (Fig. 6) of state transformation, it means that complex network corresponding to transformer
In period (2), threshold variations are had occurred in state, by stable state I1Through critical state I2It is converted into defect state I3, issue at this time
Pre-warning signal.
The table of comparisons 3 it is found that the network of this 5 key nodes of Z1, Z2, Z3, Z4, Z6 composition is key network when period (3),
Respectively correspond H2、CO、CH4、C2H4、C2H6It is one strong to illustrate that 5 key nodes are formed later in the period (2) for this 5 kinds of characteristic gas
The sub-network of strong dynamical correlation, and this 5 key nodes are consistent with characteristic gas caused by oilpaper overheat, while and Fig. 2
The corresponding key network of oilpaper overheating fault is consistent in institute's establishing network, can speculate that the failure that transformer occurs may be oilpaper
Overheating fault.
Simulation result is consistent with oil sample data calculated result, it is seen that facing before transformer is converted to failure from normal condition
When boundary's state is converted to malfunction again, the quantized value of key network marker can also change therewith.Work as IiValue is constantly in low
It is worth steady state, illustrates that transformer is constantly in opposite normal operating condition；Work as IiValue sharply increases when reaching peak value, illustrates to become
Depressor is in the critical state before failure.Therefore the real-time monitoring number based on Gases Dissolved in Transformer Oil on-Line Monitor Device
According to based on the complex network model for describing transformer Dynamic Evolutional characteristic established, by voltage transformer system state turn
Different the established key networks that interact become between each node of Near The Critical Point can generate effective pre-warning signal, pass through
Observe IiThe dynamic change of value can carry out early defect early warning to transformer, and play certain reference to the diagnosis of failure
Effect.
As it can be seen that a kind of pre- police of transformer early defect based on oil dissolved gas on-line monitoring proposed by the present invention
Method, can more fully hold the dynamic evolution process of running state of transformer, and this method is independent of model and parameter, and
It is the Real-time Monitoring Data of historical data and on-Line Monitor Device based on Gases Dissolved in Transformer Oil, finally plays to transformation
Device carries out defect early warning and the effect of fault diagnosis.
The above is only a specific embodiment of the present invention, but the design concept of the present invention is not limited to this, all to utilize this
Design makes a non-material change to the present invention, and should all belong to behavior that violates the scope of protection of the present invention.
Claims (7)
1. the transformer early defect method for early warning based on oil dissolved gas on-line monitoring, which is characterized in that including following step
It is rapid:
Step I. is based on Gases Dissolved in Transformer Oil on-Line Monitor Device, and foundation can describe transformer Dynamic Evolutional characteristic
Complex network model；
Real-time Monitoring Data of the step II. based on Gases Dissolved in Transformer Oil on-Line Monitor Device, when setting reasonable sampling
Between between be divided into a time cycle, each node data of complex network model is normalized, and dynamic screen it is current
Key node in transformer complex network model, node data standard deviation normalize formula are as follows:
Wherein XijIndicate the normalization expression formula data of node in (j) a sampling time point in period (i), xijIt is the period (i)
The value of node, x in interior (j) a sampling time pointiRepresent the node data of period (i), and mean (xi) and SD (xi) difference table
Show the average and standard deviation of node in all sampling time points in the period (i)；
The key node screened is formed key network by step III., calculates separately key network under day part (i)
Average is poorAverage Pearson correlation coefficient in key network between each nodeAnd key network
Average Pearson correlation coefficient between interior node and other non-key nodesWhether to judge the key network
Meet the critical characteristic of transformer state transformation, and obtains the key network marker under present period (i) according to calculated result
Quantized value Ii:
ε is one for avoiding the small normal number that denominator is zero in formula, if denominator, which is not zero, to be saved；
Step IV. calculates the quantized value I of the key network marker under the key network day part (i)i, as (Ii-Ii-1) >
0.05 and (Ii-Ii+1) > 0.05 when, it is meant that complex network model corresponding to transformer is faced in period (i) state
Boundary's variation, by stable state Ii-1Through critical state IiIt is converted into defect state Ii+1, issue warning signal at this time.
2. a kind of pre- police of transformer early defect based on oil dissolved gas on-line monitoring according to claim 1
Method, which is characterized in that the step I specifically:
Every kind of representative gases that Gases Dissolved in Transformer Oil on-Line Monitor Device is monitored are mapped as in complex network model
A node, take full mutual contact mode between node.
3. a kind of pre- police of transformer early defect based on oil dissolved gas on-line monitoring according to claim 1
Method, which is characterized in that the step II specifically:
Step II-1. establishes every kind based on historical data sequence acquired in Gases Dissolved in Transformer Oil on-Line Monitor Device
The prediction model of types of gases；
The prediction data that step II-2. obtains currently practical monitoring data and the prediction model established based on historical data carries out
Significance difference analysis；
If certain gas measured value of step II-3. deviates from expected dynamic trajectory and has significant difference with desired value,
Think that the variation of the gas may influence the state of transformer complex network model, which is key node.
4. a kind of pre- police of transformer early defect based on oil dissolved gas on-line monitoring according to claim 3
Method, which is characterized in that the step II-2 specifically:
Step II-2-1. sets the period (i) and shares the measured data of 7 groups of representative gases and the prediction data of 7 groups of representative gases, including
The gas data of each sampled point under the period；Variance analysis is carried out to each gas respectively using t method of inspection, if level of significance α
=0.05, it is (1) p respectively that the corresponding p value of each gas arranges from small to large, p (2), p (3), p (4), p (5), p (6), p (7),
Middle p value reflection measured data and the prediction data probability that there was no significant difference, if p > α two groups of data differences without significant meaning,
Otherwise two groups of data differences have significant meaning；
Step II-2-2. combination false discovery rate carries out FDR correction to p value, judges whether the p value of gas meets:
M=1,2 ..., 7 indicates 7 kinds of gases, tentatively assert that the gas is key node if meeting, wherein p (m) represents gas
Corresponding p value；
Step II-2-3. combines two multiple method of changing further to screen the tentatively selected gas variable being changed significantly, if full
Foot:
Then think that the gas measured value deviates from expected dynamic trajectory and has significant difference with desired value, corresponding node is
Key node；Wherein A1i, A2iGas A is represented in the measured value and predicted value of period (i), SD (A1i) it is the gas in the period (i)
The standard deviation of measured value, SD (A2i) it is standard deviation of the gas in period (i) predicted value.
5. a kind of pre- police of transformer early defect based on oil dissolved gas on-line monitoring according to claim 1
Method, which is characterized in that the step III specifically:
The average that step III-1. calculates the key network of present period (i) is poor
M is key node number in network in formula, and N is sampling point number, XijIndicate network node within the period (i) (j)
Normalization data in a sampling time point,It is averaged for the gas data period (i) in key network
Value；
Step III-2. calculates the average Pearson correlation coefficient in the key network of present period (i) between each node
R in formulamIndicate the related coefficient in key network between m-th of node and other interior nodes of network；
Step III-3. calculates the average Pearson came between node and other non-key nodes in the key network of present period (i)
Related coefficient
R in formulamIndicate the related coefficient in key network between m-th of node and other outer nodes of network；
Step III-4. is according to calculated result, if meeting following critical characteristic:
1) average of key network is poorWithCompared to dramatically increasing；
2) the average Pearson correlation coefficient in key network between each nodeWithCompared to increase；
3) the average Pearson correlation coefficient between the node in key network and other non-key nodesWithCompared to reduction；
Then complex network model corresponding to transformer may reach the Near The Critical Point of state transformation, and be obtained according to calculated result
The quantized value I of key network marker under day part (i)i, the dynamic change by monitoring the key network marker helps
In the pre-warning signal of detection of complex network model critical transitions.
6. a kind of pre- police of transformer early defect based on oil dissolved gas on-line monitoring according to claim 5
Method, which is characterized in that the correlation coefficient r between the key network interior nodes between exterior node in key networkmAnd RmCalculating it is public
Formula specifically:
Pearson correlation coefficient between key network interior nodes are as follows:
Wherein m ≠ k, XimIndicate the normalization data of (m) a key node in key network in present period (i), YikIt indicates
In period (i) in key network (k) a key node normalization data,WithIndicate corresponding node in present period
(i) mean value on all sampled points；
Pearson correlation coefficient calculation formula in key network between other non-key nodes are as follows:
Wherein XimIndicate the normalization data of (m) a key node in key network in present period (i), ZikIndicate the period
(i) normalization data of (k) a non-key node in,WithIndicate corresponding node all samplings in present period (i)
Mean value on point.
7. a kind of pre- police of transformer early defect based on oil dissolved gas on-line monitoring according to claim 5
Method, which is characterized in that the calculating of first critical characteristic specifically:
Judge that average is poor using multiple method of changingWhether dramatically increase:
In formulaWithThe respectively key network standard deviation mean value of period (i) and period (i-1), usually with 2
~3 times of differences are threshold value, work as FcThink that incrementss are smaller when < 2, as 2 < FcThink to increase when < 4, works as FcThink to increase when > 4
It measures larger.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811275492.XA CN109490685B (en) | 2018-10-30 | 2018-10-30 | Early defect early warning method of transformer based on-line monitoring of dissolved gas in oil |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811275492.XA CN109490685B (en) | 2018-10-30 | 2018-10-30 | Early defect early warning method of transformer based on-line monitoring of dissolved gas in oil |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109490685A true CN109490685A (en) | 2019-03-19 |
| CN109490685B CN109490685B (en) | 2020-10-09 |
Family
ID=65691748
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811275492.XA Active CN109490685B (en) | 2018-10-30 | 2018-10-30 | Early defect early warning method of transformer based on-line monitoring of dissolved gas in oil |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109490685B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110188309A (en) * | 2019-05-31 | 2019-08-30 | 华侨大学 | Oil-immersed power transformer defect method for early warning based on Hidden Markov Model |
| CN111579731A (en) * | 2020-06-28 | 2020-08-25 | 华侨大学 | Early transformer defect early warning method based on combined model |
| CN112861350A (en) * | 2021-02-08 | 2021-05-28 | 华侨大学 | Temperature overheating defect early warning method for stator winding of water-cooled steam turbine generator |
| CN116631527A (en) * | 2023-06-20 | 2023-08-22 | 齐丰科技股份有限公司 | Method for predicting increment and trend of chromatographic gas component of transformer oil |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102123164A (en) * | 2010-12-27 | 2011-07-13 | 安徽继远电网技术有限责任公司 | On-line monitoring management system for dissolved gas in transformer oil |
| CN104502488A (en) * | 2015-01-06 | 2015-04-08 | 国家电网公司 | Method for evaluating insulation aging degree of oil-immersed type power transformer |
| CN105891629A (en) * | 2016-03-31 | 2016-08-24 | 广西电网有限责任公司电力科学研究院 | Transformer equipment fault identification method |
| CN106649204A (en) * | 2016-09-18 | 2017-05-10 | 广东电网有限责任公司电力科学研究院 | Method for identifying exception of dissolved gas in transformer oil |
| WO2017116091A1 (en) * | 2015-12-31 | 2017-07-06 | 주식회사 효성 | Method for diagnosing oil-immersed transformer |
-
2018
- 2018-10-30 CN CN201811275492.XA patent/CN109490685B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102123164A (en) * | 2010-12-27 | 2011-07-13 | 安徽继远电网技术有限责任公司 | On-line monitoring management system for dissolved gas in transformer oil |
| CN104502488A (en) * | 2015-01-06 | 2015-04-08 | 国家电网公司 | Method for evaluating insulation aging degree of oil-immersed type power transformer |
| WO2017116091A1 (en) * | 2015-12-31 | 2017-07-06 | 주식회사 효성 | Method for diagnosing oil-immersed transformer |
| CN105891629A (en) * | 2016-03-31 | 2016-08-24 | 广西电网有限责任公司电力科学研究院 | Transformer equipment fault identification method |
| CN106649204A (en) * | 2016-09-18 | 2017-05-10 | 广东电网有限责任公司电力科学研究院 | Method for identifying exception of dissolved gas in transformer oil |
Non-Patent Citations (1)
| Title |
|---|
| RUI LIU等: "Identifying early-warning signals of critical transitions with strong noise by dynamical network markers", 《SCIENTIFIC REPORTS》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110188309A (en) * | 2019-05-31 | 2019-08-30 | 华侨大学 | Oil-immersed power transformer defect method for early warning based on Hidden Markov Model |
| CN110188309B (en) * | 2019-05-31 | 2022-08-30 | 华侨大学 | Oil-immersed power transformer defect early warning method based on hidden Markov model |
| CN111579731A (en) * | 2020-06-28 | 2020-08-25 | 华侨大学 | Early transformer defect early warning method based on combined model |
| CN112861350A (en) * | 2021-02-08 | 2021-05-28 | 华侨大学 | Temperature overheating defect early warning method for stator winding of water-cooled steam turbine generator |
| CN112861350B (en) * | 2021-02-08 | 2023-03-07 | 华侨大学 | Temperature overheating defect early warning method for stator winding of water-cooled steam turbine generator |
| CN116631527A (en) * | 2023-06-20 | 2023-08-22 | 齐丰科技股份有限公司 | Method for predicting increment and trend of chromatographic gas component of transformer oil |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109490685B (en) | 2020-10-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109490685A (en) | A kind of transformer early defect method for early warning based on oil dissolved gas on-line monitoring | |
| CN105891629B (en) | A kind of discrimination method of transformer equipment failure | |
| CN105512474B (en) | A kind of method for detecting abnormality of Transformer's Condition Monitoring data | |
| CN104007336B (en) | A kind of transformer online monitoring information aggregation method based on Internet of Things | |
| CN103245861B (en) | A kind of transformer fault diagnosis method based on Bayesian network | |
| CN102221651B (en) | Fault on-line diagnosis and early warning method of flameproof dry-type transformer for mine | |
| CN107907783B (en) | Transformer fault integrated diagnostic system and diagnostic method based on fuzzy association rules | |
| CN106405475A (en) | Electric energy meter abnormity diagnosis method | |
| CN108051709A (en) | Transformer state online evaluation analysis method based on artificial intelligence technology | |
| CN106443556A (en) | Method for intelligently diagnosing electric energy meter | |
| CN109490726A (en) | Electric power transformer insulated state evaluating method based on Clouds theory | |
| CN103513139A (en) | Power transformer fault intelligent diagnosis technology, method and device | |
| CN110188309A (en) | Oil-immersed power transformer defect method for early warning based on Hidden Markov Model | |
| CN103218662A (en) | Transformer fault diagnosis method based on back propagation (BP) neural network | |
| CN103576061A (en) | Method for discharge fault diagnosis of transformer | |
| CN108508320A (en) | Arc grounding fault identification method based on harmonic energy and wave distortion feature | |
| CN105512962A (en) | Method for comprehensively evaluating insulation status of gas insulated switchgear (GIS) | |
| CN102944811B (en) | Detection method of transformer winding deformation | |
| CN109188082A (en) | A kind of Transformer condition evaluation based on BP neural network | |
| CN107886171A (en) | A kind of circuit-breaker status inline diagnosis method and system based on PMU data | |
| CN106950450B (en) | The method for early warning and equipment of protective relaying device secondary circuit failure | |
| CN103616616B (en) | Fault detection method for large-scale grounding grid | |
| Bo et al. | Research on an online monitoring system for efficient and accurate monitoring of mine water | |
| CN110988562A (en) | Method for predicting transformer fault through vibration | |
| CN106933097A (en) | A kind of Fault Diagnosis for Chemical Process method based on multi-level optimization PCC SDG |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |