CN113807461A - Transformer fault diagnosis method based on Bayesian network - Google Patents
Transformer fault diagnosis method based on Bayesian network Download PDFInfo
- Publication number
- CN113807461A CN113807461A CN202111139051.9A CN202111139051A CN113807461A CN 113807461 A CN113807461 A CN 113807461A CN 202111139051 A CN202111139051 A CN 202111139051A CN 113807461 A CN113807461 A CN 113807461A
- Authority
- CN
- China
- Prior art keywords
- probability
- transformer
- fault
- protection action
- action
- 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 26
- 238000003745 diagnosis Methods 0.000 title claims abstract description 15
- 238000011208 chromatographic data Methods 0.000 claims abstract description 7
- 239000007789 gas Substances 0.000 claims description 48
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 claims description 24
- 239000004215 Carbon black (E152) Substances 0.000 claims description 17
- 229930195733 hydrocarbon Natural products 0.000 claims description 17
- 150000002430 hydrocarbons Chemical class 0.000 claims description 16
- VGGSQFUCUMXWEO-UHFFFAOYSA-N Ethene Chemical compound C=C VGGSQFUCUMXWEO-UHFFFAOYSA-N 0.000 claims description 15
- 239000005977 Ethylene Substances 0.000 claims description 15
- HSFWRNGVRCDJHI-UHFFFAOYSA-N alpha-acetylene Natural products C#C HSFWRNGVRCDJHI-UHFFFAOYSA-N 0.000 claims description 9
- 125000002534 ethynyl group Chemical group [H]C#C* 0.000 claims description 9
- OTMSDBZUPAUEDD-UHFFFAOYSA-N Ethane Chemical compound CC OTMSDBZUPAUEDD-UHFFFAOYSA-N 0.000 claims description 6
- 239000001257 hydrogen Substances 0.000 claims description 5
- 229910052739 hydrogen Inorganic materials 0.000 claims description 5
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 description 10
- 239000012528 membrane Substances 0.000 description 7
- 238000012544 monitoring process Methods 0.000 description 7
- 230000001537 neural effect Effects 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 238000012423 maintenance Methods 0.000 description 6
- 238000007476 Maximum Likelihood Methods 0.000 description 5
- 238000007689 inspection Methods 0.000 description 5
- 230000007257 malfunction Effects 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000004587 chromatography analysis Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000013021 overheating Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000035882 stress Effects 0.000 description 2
- 238000004804 winding Methods 0.000 description 2
- 238000000342 Monte Carlo simulation Methods 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- -1 hydrogen hydrocarbon Chemical class 0.000 description 1
- 238000009413 insulation Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000000541 pulsatile effect Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000012421 spiking Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000008646 thermal stress Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- 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
-
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Power Engineering (AREA)
- Protection Of Transformers (AREA)
Abstract
The transformer fault diagnosis method based on the Bayesian network comprises the following steps: acquiring oil chromatographic data of a transformer to be tested, and classifying by using a learning pulse neurolemma system to obtain transformer state categories; calculating the classification probability of each transformer state category; calculating the posterior probability by taking the classification probability of each transformer state category as the prior probability of the Bayesian network model; and taking the transformer state class with the maximum posterior probability as the transformer fault diagnosis result. The method has the advantages that whether the transformer fails or not and the severity of the failure can be judged more comprehensively under the conditions of considering main protection, backup protection and failure and misoperation of the breaker.
Description
Technical Field
The invention relates to the technical field of fault diagnosis of power systems and transformers, in particular to a transformer fault diagnosis method based on a Bayesian network.
Background
With the development of the on-line monitoring technology of the transformer, diagnosing the fault type of the transformer according to the oil chromatographic information becomes a normalized state monitoring means of the transformer. When some sudden faults occur to the transformer, the electrical quantity or the non-electrical quantity inside the transformer is changed, and the relay protection device sends alarm information, however, the alarm and the fault type are difficult to be rapidly reported and identified only by means of the current transformer oil color spectrum online monitoring method. The learning pulse neurolemma system can realize transformer fault classification and early warning by taking oil chromatographic data as input, but the classification of the learning pulse neurolemma system does not consider main protection, backup protection and classification under the condition of circuit breaker failure and misoperation, and the accuracy is still to be improved.
Disclosure of Invention
The invention aims to provide a transformer fault diagnosis method based on a Bayesian network.
The technical scheme for realizing the purpose of the invention is as follows:
the transformer fault diagnosis method based on the Bayesian network comprises the following steps,
step 1: acquiring oil chromatographic data of a transformer to be tested, and classifying the oil chromatographic data by using a learning pulse neurolemma system to obtain transformer state categories, namely normal, medium and low temperature faults, partial discharge faults, high temperature faults, low energy discharge faults and high energy discharge faults;
step 2: calculating the classification probability of each transformer state category;
and step 3: calculating the posterior probability by taking the classification probability of each transformer state category as the prior probability of the Bayesian network model;
3.1 Transformer State class Normal, posterior probability
Wherein, P (h) is the classification probability that the transformer state class is normal; e.g. of the type1、e2And e3Respectively representing a main protection action, a backup protection action and a breaker action; p (e)1|e2,e3,h)、P(e2|e1,e3H) and P (e)3|e1,e2H) are conditional probabilities, respectively; p (e)1,e2,e3) To full probability, P (e)1,e2,e3)＝P(e1|h)P(e2|e1)P(e3|e1,e2) (ii) a Wherein, P (e)1H) is the probability of primary protection action in normal state, P (e)2|e1) Probability of backup protection action under primary protection action, P (e)3|e1,e2) The action probability of the circuit breaker under the main protection action and the backup protection action;
3.2 Transformer State class Medium-Low temperature Fault or partial discharge Fault, posterior probability
Wherein, P (h) is the classification probability of the transformer state classification as medium-low temperature fault or partial discharge fault; e.g. of the type1、e2、e3And e4Respectively representing differential protection action, light gas protection action, backup protection action and breaker action; p (e)1|e2,e3,e4,h)，P(e2|e1,e3,e4,h)，P(e3|e1,e2,e4H) and P (e)4|e1,e2,e3H) are conditional probabilities, respectively; p (e)1,e2,e3,e4) To full probability, P (e)1,e2,e3,e4)＝P(e1|h)P(e2|h)P(e3|e1)P(e4|e1,e2,e3) (ii) a Wherein, P (e)1| h) as medium-low temperature faultOr the probability of differential protection action under partial discharge fault, P (e)2H) is the probability of light gas protection action under medium-low temperature fault or partial discharge fault, P (e)3|e1) For the probability of a backup protection action under differential protection action, P (e)4|e1,e2,e3) The circuit breaker action probability under the differential protection action, the light gas protection action and the backup protection action is obtained;
3.3 Transformer State class high temperature Fault, Low energy discharge Fault or high energy discharge Fault, posterior probability
Wherein, p (h) is the classification probability that the transformer state category is a high-temperature fault, a low-energy discharge fault or a high-energy discharge fault; e.g. of the type1、e2、e3And e4Respectively representing differential protection action, heavy gas protection action, backup protection action and breaker action; p (e)1|e2,e3,e4,h)，P(e2|e1,e3,e4,h)，P(e3|e1,e2,e4H) and P (e)4|e1,e2,e3H) are conditional probabilities, respectively; p (e)1,e2,e3,e4) To full probability, P (e)1,e2,e3,e4)＝P(e1|h)P(e2|h)P(e3|e1)P(e4|e1,e2,e3) (ii) a Wherein, P (e)1I h) is the probability of differential protection action under high-temperature fault, low-energy discharge fault or high-energy discharge fault, P (e)2H) is the probability of heavy gas protection action under high-temperature fault, low-energy discharge fault or high-energy discharge fault, P (e)3|e1) For the probability of a backup protection action under differential protection action, P (e)4|e1,e2,e3) The breaker action probability under the differential protection action, the heavy gas protection action and the backup protection action;
and 4, step 4: and taking the transformer state class with the maximum posterior probability as the transformer fault diagnosis result.
According to a further technical scheme, oil chromatographic data of the transformer is seven non-coding ratios of the content of dissolved gas in transformer oil, namely methane/hydrogen, ethylene/ethane, acetylene/ethylene, acetylene/total hydrocarbon, ethylene/total hydrocarbon, methane/total hydrocarbon, (ethylene + methane)/total hydrocarbon, wherein the total hydrocarbon is the sum of the contents of methane, ethane, ethylene and acetylene.
Compared with the prior art, the method has the beneficial effects that whether the transformer fails or not and the severity of the failure can be judged more comprehensively under the conditions of considering main protection, backup protection and circuit breaker failure and misoperation.
Drawings
Fig. 1 is a bayesian network structure in a failure-free mode.
Fig. 2 is a bayesian network structure in a light failure mode.
Fig. 3 is a bayesian network structure in a critical failure mode.
Fig. 4 is a schematic diagram of a learning pulsatile neural membrane system classification process.
Fig. 5 is a flow chart of bayesian network parameter estimation.
Detailed Description
The method comprises the steps of establishing a Bayesian network model according to an action relation between a transformer fault type and a relay protection device, obtaining a transformer fault prior probability based on a learning pulse neural membrane system, and estimating a transformer relay protection action condition probability based on maximum likelihood estimation and diagnosing the transformer fault.
A transformer fault diagnosis method based on a learning pulse neurolemma system and a Bayesian network comprises the following steps:
s10, establishing a Bayesian network model according to the relation between the transformer fault type and the relay protection device;
a node of the bayesian network represents a random variable E ═ (E)1,e2,…,en) And the like variable H ═ H1,h2,…,hm) The expression and reasoning of the knowledge are realized through the Bayes theory, and the formula of the reasoning process is as follows:
in the formula, P (e)1,e2,…,en) For joint probability of input data, P (h)j|e1,e2,…,en) Is a class hjA posteriori probability of P (e)i|e1,e2,…,ei-1,hj) For inputting data eiConditional probability of (c), P (h)j) Is a class variable hjPrior probability of.
According to the posterior probabilities of different categories, the output category is judged by using the maximized posterior probability, and the calculation method comprises the following steps:
in the invention, the transformer has two protection types of main protection and backup protection according to the function of the transformer protection device. The main protection includes differential protection and light (heavy) gas protection. The backup protection is that when the fault happens and the main protection does not act, the backup protection acts to open the breaker. Because the relay protection device of the transformer can quickly and accurately make corresponding measures according to different running states of the transformer, a Bayesian network can be established according to the fault information of the transformer and the action information of the relay protection device, and then the running state of the relay protection device can be judged. According to the severity of the fault development of the transformer, a Bayesian network for diagnosing the fault of the transformer is established from three aspects of no fault, slight fault and serious fault.
(1) Transformer fault reasoning network in fault-free mode
According to different voltage grades of the transformer, when the content of the dissolved gas in the oil does not exceed the attention value, the transformer is considered to be in a normal state. With the extension of the service time of the transformer, when the gas content in the transformer exceeds the attention value, the fault of the transformer is not considered to occur. If the gas production rate is stable, the transformer is considered to be aged normally and belongs to a normal state. Since no fault occurs inside the transformer, theoretically, the relay protection device inside the transformer does not operate, and if a certain device operates, it is considered that the device performs a malfunction, and the bayesian network structure is shown in fig. 1.
(2) Transformer fault reasoning network in slight fault mode
The slight faults of the transformer mainly comprise medium-low temperature overheating faults and partial discharge faults, and are externally represented by poor contact or overheating and oil flow static electricity caused by circulation current. Since the energy density at the fault occurrence point is not high and the transformer oil is not decomposed violently, an alarm signal is sent out by the light gas protection, and the heavy gas protection does not act. When a circular current occurs inside the transformer, a differential current occurs at the outlet of the transformer, which causes a differential protection action of the transformer, thereby causing a trip accident. Therefore, when the transformer has a slight fault, the light gas protection or differential protection action is mainly adopted, when the main protection refuses to act, the connection between the transformer and the outside is cut off by the backup protection action, and the Bayesian network structure is shown as figure 2.
(3) Transformer fault reasoning network under severe fault mode
When faults such as low-energy (high-energy) discharge, multipoint grounding and the like occur in the transformer, the transformer oil is decomposed more violently due to high internal temperature, so that the faults are considered to be serious. Under the combined action of electrical stress and thermal stress, transformer oil rapidly deteriorates and releases a large amount of hydrogen hydrocarbon gas. The large amount of generated gas causes heavy gas protection action of the transformer, thereby avoiding continuous spread of faults. When a short-circuit fault occurs inside the transformer, the current at the outlet of the transformer is unbalanced, so that the differential protection action of the transformer is caused. Fig. 3 shows a bayesian network structure when a critical fault occurs inside the transformer.
And S11, acquiring the prior probability of the Bayesian network under the transformer fault condition by utilizing the learning optimization pulse neural membrane system.
The presence of electrical, thermal, mechanical, etc. stresses during transformer operation gradually degrades the transformer insulation and releases large quantities of hydrogen and hydrocarbon gases. When the content of dissolved gas in the oil does not exceed the noted value, it is considered as a normal state. Defining the significance level to be 0.5 per thousand according to historical statistical data, and when the content of dissolved gas in oil is greater than an attention value, classifying the transformer states by using a data set defined by K to 6 transformer states (N, T1T2, T3, PD, D1 and D2) and using a learning pulse neurolemma system (LSNP) algorithm, wherein the classified states are normal (N), slight faults (medium-low temperature fault (T1T2) and partial discharge fault (PD)) and severe faults (high-temperature fault (T3), low-energy discharge fault (D1) or high-energy discharge fault (D2)) respectively. The classification probability is calculated as
Wherein d iskRepresents the kth state data point xiTo another state data point xjK is 1,2, …, K.
The classification process is shown in fig. 4.
And determining the input of the learning pulse neurolemma system according to the relation between the dissolved gas in the transformer oil and the transformer fault. The model parameters selected by the invention are methane/hydrogen, ethylene/ethane, acetylene/ethylene, acetylene/total hydrocarbon, ethylene/total hydrocarbon, methane/total hydrocarbon, (ethylene + methane)/total hydrocarbon, wherein the total hydrocarbon is the sum of the contents of methane, ethane, ethylene and acetylene, and the output is the probability of the transformer state. The model parameters for learning the spiking neural membrane system are as follows: the number of ascending orders is selected to be 4, the learning rate is set to be 0.1, and the number of iterations is set to be 1000. And calculating according to the parameters of the learning pulse neural membrane system to obtain the state of the transformer and the classification probability corresponding to the state of the transformer, and then taking the classification probability as the prior probability of the Bayesian network model.
S12: and (4) estimating the conditional probability (the movement rejection rate and the movement error rate of a transformer relay protection device (circuit breaker)) of the Bayesian network by using a Monte Carlo algorithm and a maximum likelihood estimation theory according to the estimated value obtained after statistics.
The conditional probability of the intermediate node of the Bayesian network model described in the invention represents the states of the relay protection device and the circuit breaker, and a small amount of action rejection or misoperation is inevitable because the relay protection device cannot be always in an ideal state. However, the phenomenon of the operation rejection or the false operation of the relay protection device has strong randomness, so the state and the operation probability of the relay protection device are simulated by using the monte carlo algorithm and the maximum likelihood estimation, and the algorithm flow chart is shown in fig. 5.
The malfunction of the relay protection device of the transformer refers to a situation that the relay protection device does not act according to a command for some reason when the transformer normally operates. The calculation method of the misoperation rate of the relay protection device is shown in formula (3):
in the formula, N1Indicating the number of times that the relay protection device is not operating according to the setting requirement, NfoIndicating the number of times that the relay protection device is not operating according to the setting requirement, PfoAnd the error rate of the relay protection device is shown.
The failure of the transformer relay protection device refers to the situation that the relay protection device cannot execute setting action according to a command when the transformer fails. The calculation method of the rejection rate of the transformer relay protection device is shown as the formula (4):
in the formula, N2Indicating the number of times that the relay protection device acts according to the setting requirement, NroIndicating the number of times that the relay protection device is not operating according to the setting requirement, ProAnd the rejection rate of the relay protection device is shown.
According to the analysis of the statistical data, the failure rate of the transformer relay protection device is shown in table 1. Under the condition that the transformer protection device and the fault probability are mutually independent, firstly, a Monte Carlo simulation algorithm is used for simulating the action condition of the relay protection device of the transformer in different states, and then, the maximum likelihood estimation is used for calculating the fault correction value of the relay protection device and the breaker of the transformer according to the simulation result, and the result is shown in a table 2:
TABLE 1 Relay protection device and Circuit breaker failure rates
TABLE 2 corrected failure rates of the relay protection devices and circuit breakers
The operation failure and the malfunction of the relay protection device and the breaker mean that the relay protection device and the breaker should operate but not operate and that the relay protection device and the breaker should not operate but operate. Suppose e1，e2，e3And h denotes main protection, backup protection, breaker and fault, respectively, then P (e)1)，P(e2)，P(e3) And p (h) respectively represent corresponding probabilities. The relationship between the conditional probability and the rejection rate and the malfunction rate is shown in table 3.
TABLE 3 relationship between conditional probability and rejection and false action rates
And S13, acquiring the oil chromatogram on-line monitoring information of the transformer and the action information of the relay protection device, and calculating the posterior probability of the fault information through the Bayesian network.
The posterior probability of each case is calculated separately from the established bayesian network structure and the bayesian network calculation formula in S10. (1) Posterior probability calculation without failure
The prior probability is respectively defined as P (h), and the main protection action, the backup protection action and the breaker action are respectively e1，e2And e3(ii) a The total probability distribution of which is P (e)1,e2,e3)＝P(e1|h)P(e2|e1)P(e3|e1,e2). With a conditional probability of P (e)1|e2,e3,h)，P(e2|e1,e3H) and P (e)3|e1,e2H). The posterior probability is calculated as
Wherein, P (e)1|h)，P(e2|e1) And P (e)3|e1,e2) The main protection operation probability (1-0.26% ═ 0.9974) in the case of a fault, the backup protection operation probability (0.87%) in the case of the main protection operation, and the breaker operation probability (1-1.43% ═ 0.9857) in the case of the main protection operation and the backup protection operation, respectively.
(2) Posterior probability calculation in the case of minor faults
The prior probability is respectively defined as P (h), the differential protection action, the light gas protection action (the differential protection and the light gas protection belong to the main protection), the backup protection action and the breaker action are respectively e1,e2,e3And e4(ii) a The total probability distribution of which is P (e)1,e2,e3,e4)＝P(e1|h)P(e2|h)P(e3|e1)P(e4|e1,e2,e3). With a conditional probability of P (e)1|e2,e3,e4,h)，P(e2|e1,e3,e4,h)，P(e3|e1,e2,e4H) and P (e)4|e1,e2,e3H). The posterior probability is calculated as
Wherein, P (e)1|h)，P(e2|h)，P(e3|e1) And P (e)4|e1,e2,e3) The probability of the differential protection operation in the case of a fault (1-0.26%: 0.9974), the probability of the light gas protection operation in the case of a fault (1-0.26%: 0.9974), the probability of the backup protection operation in the case of the main protection operation (0.87%), and the probability of the circuit breaker operation in the case of the main protection operation and the backup protection operation (1-1.43%: 0.9857), respectively.
(3) Posterior probability calculation under severe fault conditions
The prior probability is respectively defined as P (h), the differential protection action, the heavy gas protection action (the differential protection and the light gas protection belong to the main protection), the backup protection action and the breaker action are respectively e1,e2,e3And e4(ii) a The total probability distribution of which is P (e)1,e2,e3,e4)＝P(e1|h)P(e2|h)P(e3|e1)P(e4|e1,e2,e3). With a conditional probability of P (e)1|e2,e3,e4,h)，P(e2|e1,e3,e4,h)，P(e3|e1,e2,e4H) and P (e)4|e1,e2,e3H). The posterior probability is calculated as
Wherein, P (e)1|h)，P(e2|h)，P(e3|e1) And P (e)4|e1,e2,e3) The probability of the differential protection operation in the case of a fault (1-0.26%: 0.9974), the probability of the heavy gas protection operation in the case of a fault (1-0.26%: 0.9974), the probability of the backup protection operation in the case of the main protection operation (0.87%), and the probability of the circuit breaker operation in the case of the main protection operation and the backup protection operation (1-1.43%: 0.9857), respectively.
And S14, judging the operation state according to the maximum posterior probability and giving a maintenance strategy.
The invention provides a transformer fault diagnosis method based on a learning pulse neurolemma system and a Bayesian network, which comprises the steps of establishing a Bayesian network model according to the relation between the transformer fault type and a relay protection device; acquiring the prior probability of the Bayesian network under the condition of transformer fault by utilizing a learning optimization pulse neural membrane system; according to an estimated value given in a document, a Monte Carlo algorithm and a maximum likelihood estimation theory are used for estimating conditional probability (the rejection rate and the misoperation rate of a transformer relay protection device (circuit breaker)) of the Bayesian network; acquiring oil chromatography online monitoring information of a transformer and action information of a relay protection device, and calculating posterior probability of fault information through a Bayesian network; and judging the operation state according to the maximum posterior probability and giving out a maintenance strategy. Therefore, the method provided by the invention has the following advantages:
(1) the method for fusing the relay protection information and the oil-gas information of the transformer establishes that the transformer fault diagnosis based on the learning pulse neurolemma system and the Bayesian network can more comprehensively judge whether the transformer has faults and the severity of the faults.
(2) The Bayesian network is easy to process incomplete data sets, reflects a probability relation model among data in the whole database, and particularly has a good effect on data with causal relation.
In the following, the procedure of diagnosing the fault of the transformer based on the learning pulse neurolemma system and the bayesian network is further explained by using three embodiments.
Example 1: the light gas protection is found to frequently act and alarm information is sent out in the daily inspection of a certain transformer, and the breaker does not act. The oil and gas data obtained after the transformer oil sample is taken for inspection are shown in table 4:
TABLE 4 oil chromatogram content values μ L/L
The cause of the transformer failure was calculated using a learning pulse neurolemma system since the content of dissolved gases in the oil exceeded the attention value specified in the guidelines. After calculation, the probability of the transformer having a fault is 99.98%, and the probability of the transformer having a fault is 99.95% of the serious fault. And inputting the prior probabilities of the three faults into the Bayesian network to calculate the posterior probability of the transformer fault. The probability of no fault of the transformer is almost 0 through calculation; the probability of slight fault of the transformer is 11.09%; due to the light gas protection alarm, the light gas protection alarm is temporarily set to be in the initial stage of serious fault, and the posterior probability of the light gas protection alarm is approximately equal to 1 by reasoning according to the network form of the figure 2. And judging that the fault belongs to the initial stage of the serious fault and the action of the relay protection device is correct according to the maximum posterior hypothesis method. And after subsequent overhaul, the high-temperature fault caused by poor contact of the phase splitting joint is found, and the action of the relay protection device is correct.
Example 2: abnormal sound appears in the voltage regulation process of one 220kV transformer in a certain place, a relay protection device of the transformer and a circuit breaker do not act, and the content value obtained after oil sampling analysis is shown in a table 5:
TABLE 5 oil chromatogram content values μ L/L
The oil and gas data in table 5 far exceed the attention values thereof, so the fault probability of the transformer is calculated by using a learning pulse neural membrane system, and the prior probability of the transformer fault is 94.14% and the prior probability of the serious fault is 80.02% after calculation. The prior probability of the transformer fault is input into the Bayesian network to obtain the posterior probability of the transformer fault under the fault-free model, the posterior probability of the transformer fault is only 4.02%, the posterior probability of other two conditions is close to 0, and the transformer is immediately determined to possibly have a certain fault, but the probability of fault propagation is smaller. While the oil chromatography monitoring is enhanced and the stability of the regional power grid is guaranteed, a professional is dispatched to enter the transformer body to check the fault reason. The DC resistance of the transformer winding is firstly tested in the maintenance process, and no abnormal state is found. After the transformer is drained, a person enters the body to check, the trace of arc burning of a certain phase change-over switch is found, and the winding is free of abnormality. After expert discussion, the transformer can be considered to be operated in a charged mode. After an accident happens for one month, operation and maintenance personnel replace an on-load switch of the transformer and fill oil in vacuum, and the monitoring value of the transformer oil chromatogram is recovered to be normal.
In this example, if the fault is only according to the regulation in the oil immersed transformer (reactor) condition inspection guide, the class A or B power failure inspection should be carried out immediately. However, the transformer does not have to be immediately overhauled in power failure in actual operation, but has been operated in a live state for one month.
Example 3: when a certain transformer suddenly generates a heavy gas protection tripping accident in the operation process, the content value of dissolved gas in oil obtained after oil sampling and inspection is as follows:
TABLE 6 oil chromatogram content values μ L/L
Because the oil gas content does not exceed the attention value, the transformer is considered to be in a normal state, and the significance level is 0.5 per thousand. The probability that the transformer is in a normal state is found to be 88.91% according to Bayesian model reasoning, and the transformer body is considered to be free of faults. And in the subsequent maintenance link, the tripping accident caused by the gas protection defect is found, and the transformer normally operates after the gas protection is replaced.
The embodiment fully illustrates that the method provided by the invention can effectively judge the development trend of the transformer fault and provide scientific basis for the formulation of the transformer maintenance plan by depending on oil and gas data information and considering the conditions of main protection, backup protection and circuit breaker failure and misoperation.
Claims (2)
1. The transformer fault diagnosis method based on the Bayesian network is characterized by comprising the following steps of,
step 1: acquiring oil chromatographic data of a transformer to be tested, and classifying the oil chromatographic data by using a learning pulse neurolemma system to obtain transformer state categories, namely normal, medium and low temperature faults, partial discharge faults, high temperature faults, low energy discharge faults and high energy discharge faults;
step 2: calculating the classification probability of each transformer state category;
and step 3: calculating the posterior probability by taking the classification probability of each transformer state category as the prior probability of the Bayesian network model;
3.1 Transformer State class Normal, posterior probability
Wherein, P (h) is the classification probability that the transformer state class is normal; e.g. of the type1、e2And e3Respectively representing a main protection action, a backup protection action and a breaker action; p (e)1|e2，e3，h)、P(e2|e1，e3H) and P (e)3|e1，e2H) are conditional probabilities, respectively; p (e)1，e2，e3) To full probability, P (e)1，e2，e3)＝P(e1|h)P(e2|e1)P(e3|e1，e2) (ii) a Wherein, P (e)1H) is the probability of primary protection action in normal state, P (e)2|e1) Probability of backup protection action under primary protection action, P (e)3|e1，e2) The action probability of the circuit breaker under the main protection action and the backup protection action;
3.2 Transformer State class Medium-Low temperature Fault or partial discharge Fault, posterior probability
Wherein, P (h) is the classification probability of the transformer state classification as medium-low temperature fault or partial discharge fault; e.g. of the type1、e2、e3And e4Respectively showing differential protection action, light gas protection action and backup protectionActions and breaker actions; p (e)1|e2，e3，e4，h)，P(e2|e1，e3，e4，h)，P(e3|e1，e2，e4H) and P (e)4|e1，e2，e3H) are conditional probabilities, respectively; p (e)1，e2，e3，e4) To full probability, P (e)1，e2，e3，e4)＝P(e1|h)P(e2|h)P(e3|e1)P(e4|e1，e2，e3) (ii) a Wherein, P (e)1I h) is the differential protection action probability under medium-low temperature fault or partial discharge fault, P (e)2H) is the probability of light gas protection action under medium-low temperature fault or partial discharge fault, P (e)3|e1) For the probability of a backup protection action under differential protection action, P (e)4|e1，e2，e3) The circuit breaker action probability under the differential protection action, the light gas protection action and the backup protection action is obtained;
3.3 Transformer State class high temperature Fault, Low energy discharge Fault or high energy discharge Fault, posterior probability
Wherein, p (h) is the classification probability that the transformer state category is a high-temperature fault, a low-energy discharge fault or a high-energy discharge fault; e.g. of the type1、e2、e3And e4Respectively representing differential protection action, heavy gas protection action, backup protection action and breaker action; p (e)1|e2，e3，e4，h)，P(e2|e1，e3，e4，h)，P(e3|e1，e2，e4H) and P (e)4|e1，e2，e3H) are conditional probabilities, respectively; p (e)1，e2，e3，e4) To full probability, P (e)1，e2，e3，e4)＝P(e1|h)P(e2|h)P(e3|e1)P(e4|e1，e2，e3) (ii) a Wherein, P (e)1I h) is the probability of differential protection action under high-temperature fault, low-energy discharge fault or high-energy discharge fault, P (e)2H) is the probability of heavy gas protection action under high-temperature fault, low-energy discharge fault or high-energy discharge fault, P (e)3|e1) For the probability of a backup protection action under differential protection action, P (e)4|e1，e2，e3) The breaker action probability under the differential protection action, the heavy gas protection action and the backup protection action;
and 4, step 4: and taking the transformer state class with the maximum posterior probability as the transformer fault diagnosis result.
2. The bayesian network based transformer fault diagnosis method according to claim 1, wherein the oil chromatogram data of the transformer are seven uncoded ratios of the content of dissolved gas in the transformer oil, namely, methane/hydrogen, ethylene/ethane, acetylene/ethylene, acetylene/total hydrocarbon, ethylene/total hydrocarbon, methane/total hydrocarbon, (ethylene + methane)/total hydrocarbon, wherein the total hydrocarbon is the sum of the contents of methane, ethane, ethylene and acetylene.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111139051.9A CN113807461A (en) | 2021-09-27 | 2021-09-27 | Transformer fault diagnosis method based on Bayesian network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111139051.9A CN113807461A (en) | 2021-09-27 | 2021-09-27 | Transformer fault diagnosis method based on Bayesian network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113807461A true CN113807461A (en) | 2021-12-17 |
Family
ID=78896880
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111139051.9A Pending CN113807461A (en) | 2021-09-27 | 2021-09-27 | Transformer fault diagnosis method based on Bayesian network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113807461A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113625088A (en) * | 2021-08-17 | 2021-11-09 | 国网四川省电力公司电力科学研究院 | Transformer fault diagnosis method based on learning pulse neurolemma system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103197177A (en) * | 2013-03-20 | 2013-07-10 | 山东电力集团公司济宁供电公司 | Transformer fault diagnosis analysis method based on bayesian network |
| CN103293421A (en) * | 2013-06-07 | 2013-09-11 | 云南电力调度控制中心 | Power grid fault diagnostic model and diagnostic method thereof |
| CN104297589A (en) * | 2014-09-29 | 2015-01-21 | 国家电网公司 | Method for diagnosing fault of oil-immersed transformer on basis of rough set and bayesian network |
| CN105678337A (en) * | 2016-01-12 | 2016-06-15 | 国网技术学院 | Information fusion method in intelligent transformer station fault diagnosis |
| CN109142969A (en) * | 2018-07-20 | 2019-01-04 | 西南交通大学 | A kind of power transmission line fault phase selection based on Continuous Hidden Markov Model |
| CN111598150A (en) * | 2020-05-12 | 2020-08-28 | 国网四川省电力公司电力科学研究院 | Transformer fault diagnosis method considering operation state grade |
-
2021
- 2021-09-27 CN CN202111139051.9A patent/CN113807461A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103197177A (en) * | 2013-03-20 | 2013-07-10 | 山东电力集团公司济宁供电公司 | Transformer fault diagnosis analysis method based on bayesian network |
| CN103293421A (en) * | 2013-06-07 | 2013-09-11 | 云南电力调度控制中心 | Power grid fault diagnostic model and diagnostic method thereof |
| CN104297589A (en) * | 2014-09-29 | 2015-01-21 | 国家电网公司 | Method for diagnosing fault of oil-immersed transformer on basis of rough set and bayesian network |
| CN105678337A (en) * | 2016-01-12 | 2016-06-15 | 国网技术学院 | Information fusion method in intelligent transformer station fault diagnosis |
| CN109142969A (en) * | 2018-07-20 | 2019-01-04 | 西南交通大学 | A kind of power transmission line fault phase selection based on Continuous Hidden Markov Model |
| CN111598150A (en) * | 2020-05-12 | 2020-08-28 | 国网四川省电力公司电力科学研究院 | Transformer fault diagnosis method considering operation state grade |
Non-Patent Citations (1)
| Title |
|---|
| 王健等: ""计及继电保护与油气信息的变压器故障推理"" * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113625088A (en) * | 2021-08-17 | 2021-11-09 | 国网四川省电力公司电力科学研究院 | Transformer fault diagnosis method based on learning pulse neurolemma system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106908674B (en) | Transformer state evaluation method based on multi-state quantity prediction | |
| CN107609725B (en) | Power transmission network risk assessment method considering transformer substation influence | |
| CN101425686B (en) | Electrical power system on-line safety and stability evaluation forecast failure collection adaptive selection method | |
| CN110210095B (en) | Power distribution network reliability index calculation method based on mixed integer linear programming | |
| Ahmed et al. | Power transformer fault diagnosis using fuzzy logic technique based on dissolved gas analysis | |
| CN109490726A (en) | Electric power transformer insulated state evaluating method based on Clouds theory | |
| CN110443037B (en) | Power monitoring network security situation perception method based on improved AHP method | |
| CN110163477A (en) | A kind of method and system for the operating status determining power equipment based on Condition Monitoring Data | |
| CN112116276B (en) | Substation operation risk assessment method considering time-varying state of electric main equipment | |
| CN106548265A (en) | A kind of Transmission Network Reliability appraisal procedure based on cascading failure accident chaining search | |
| CN114370380A (en) | Wind turbine generator fault diagnosis method considering meteorological factors | |
| CN108011358A (en) | A kind of power distribution network multinode intelligent fault diagnosis method based on Situation Awareness | |
| CN115453267A (en) | Fault diagnosis system for electric power information system | |
| CN109064071A (en) | A kind of secondary system of intelligent substation methods of risk assessment based on shielding failure | |
| CN116050817A (en) | Power system risk assessment method based on secondary coupling multiple fault model | |
| CN113807461A (en) | Transformer fault diagnosis method based on Bayesian network | |
| CN106597845B (en) | A kind of power transmission network method for diagnosing faults based on multiple-objection optimization | |
| CN115902510A (en) | Power failure fault early warning method and system for transformer area | |
| CN110443481A (en) | Distribution power automation terminal state evaluation system and method based on mixing K- nearest neighbor algorithm | |
| Gracheva et al. | Study of Probability Characteristics of the Reliability of Electrical Equipment in Internal Power Supply Systems | |
| Altaher et al. | Reliability investigation of digital substation networks design using fmea technique | |
| Zeng et al. | Reliability assessment on switchgear and its controller based on fault tree analysis | |
| CN113033985A (en) | Single-element fault risk assessment and rectification optimization method | |
| Zapata et al. | Reliability assessment of substations using stochastic point processes and Monte Carlo simulation | |
| Shang et al. | Finite State Machine model of fault diagnosis for distribution system under time sequence constraints |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20211217 |