CN102570392A - Method for identifying exciting inrush current of transformer based on improved probability neural network - Google Patents
Method for identifying exciting inrush current of transformer based on improved probability neural network Download PDFInfo
- Publication number
- CN102570392A CN102570392A CN2012100131545A CN201210013154A CN102570392A CN 102570392 A CN102570392 A CN 102570392A CN 2012100131545 A CN2012100131545 A CN 2012100131545A CN 201210013154 A CN201210013154 A CN 201210013154A CN 102570392 A CN102570392 A CN 102570392A
- Authority
- CN
- China
- Prior art keywords
- current
- neural network
- inrush current
- transformer
- smoothing factor
- 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
Images
Landscapes
- Protection Of Transformers (AREA)
Abstract
The invention relates to a method for identifying an exciting inrush current of a transformer based on an improved probability neural network. The method comprises the following steps of: optimizing smooth factors of the probability neural network by using a genetic algorithm; simulating a model which is constructed in Matlab/Simulink to obtain current waveform; and by taking the wavelet transformation energy of the exciting inrush current and an internal failure current as network input, identifying a failure mode. The method has the advantage that: by applying an intelligent technology to determination of the exciting inrush current, the failure identification capacity of the exciting inrush current is greatly improved.
Description
Technical field
The present invention relates to a kind of transformer production detection technique, particularly a kind of based on the transformer excitation flow discrimination method that improves probabilistic neural network.
Background technology
Power transformer is being undertaken the important function of electric energy transmitting and voltage transformation as one of the important component part in the electric power system and key equipment, and its running status directly has influence on the fail safe and the stability of whole electric power system.Just towards the electric pressure development of bigger transmission line capability and Geng Gao, power transformer also develops towards big capacity, high-tension direction for the power industry develop rapidly of China at present, electric power system.Yet capacity is big more, and grade is high more, makes the failure rate of power transformer also high more, because the power transformer internal structure is complicated, and running environment is special, in the middle of power transformer operation over a long time, it is unavoidable breaking down simultaneously.If being in the large-scale power transformer of electric power system central status breaks down; Can have a strong impact on the production of society and the safety of people's lives and properties; Can restrict development and national economy greatly, cause serious economy loss, therefore the relaying protection that requires transformer had higher reliability.
For a long time, longitudinal differential protection is considered to the most perfect main protection of power transformer always, and it is based on Kirchhoff's current law (KCL) (KCL), and the unsymmetrical current that produces when utilizing fault moves, have good, the highly sensitive characteristics of selectivity.But the performance factor of transformer differential protection is not high always, does not reach the requirement of modern electric relaying protection far away.According to the statistics of pertinent literature, 1996~2005 years are during the decade, and 220kv and the total action frequency of above transformer differential protection are 2247 times, and wherein the correct operation number of times is 1673 times, incorrect operation 574 times, and average performance factor is 74.45%.
The protection of transformer development is also relatively backward, and it protects precision maneuver rate still quite low, and other main equipments in the electric power system are like generator protection, line protection etc.Causing present this result, staff's operational issue (design, make, the debugging of adjusting, the error of operation maintenance aspects) is arranged, the deficiency of management supervision aspect is also arranged, but the reason of searching to the bottom, is deficiency on the principle of power transformer differential protection scheme.The application of differential protection on pure circuit arrangement such as generator, transmission line is fairly simple; Effect is more satisfactory; But the main protection performance as power transformer is not fine, and how to prevent becomes the most critical issue of having to face by the differential protection misoperation that unsymmetrical current causes.The theoretical foundation of differential protection is a kirchhoff electric current theorem, but the inner employing of transformer is iron core-closed, come down to a non-linear element, so kirchhoff electric current theorem is no longer suitable in essence.
The exciting current of transformer is a kind of main source of differential protection unsymmetrical current.When transformer normally moved, its exciting current was generally 1%~3% of rated current, can ignore, and through the suitable threshold value of adjusting, differential protection is the accurately internal fault of differentiating transformer and external fault just.But when no-load transformer closes a floodgate; When perhaps external fault excision back terminal voltage was recovered suddenly, the numerical value of magnetizing inrush current was very big, possibly compare mutually with short circuit current sometimes; If there is not other safeguard measure, so big unsymmetrical current must cause differential protection misoperation.Therefore; How fast and accurately the principal contradiction of current transformer differential protection concentrates on differentiating transformer exciting surge and internal fault current; Have only and excise internal fault current timely and accurately, just can avoid bigger power loss, guarantee the reliability of supply of electric power.For this reason, Chinese scholars has been carried out a large amount of explorations, has proposed many new principles, new method is used for differential protection, mainly comprises secondary harmonic brake principle, interval angle principle, waveform symmetry principle and magnetic flux characteristic braking principle etc.At present, the tranformer protection that in system, disposes mainly is to adopt secondary harmonic brake principle and electric current interval angle principle to discern the existence of magnetizing inrush current.Because the interval angle principle is high to hardware requirement, in the microcomputer transformer differential protection, adopt the secondary harmonic brake principle more.This principle determines whether Blocking Differential Protection through the size of the ratio of second harmonic in the poor streaming current and first-harmonic, but since in the magnetizing inrush current content of second harmonic by a plurality of parameter determining and alter a great deal, so be difficult to suitably select restraint coefficient.Particularly along with a large amount of uses of modern high-power transformer; The excitation property of transformer changes a lot; Cause that the content of second harmonic reduces in the magnetizing inrush current, and the development of high-power transformer and long distance power transmission, the second harmonic when making internal fault again in the transient current increases; Thereby cause harmonic braking principle identification difficulty, misoperation often occurs.
Therefore, it is urgent to use new principle and method to realize that the differentiation of transformer excitation flow has reality, to improve the accuracy and the performance of transformer differential protection.Begin from the nineties, the application that the Protection Technology of China has got into the epoch microcomputer of Microcomputer Protection has caused the deep reform in relaying protection field, has also brought the new opportunity of Protection Technology development.At identification transformer excitation flow technical elements, many new principles and solution have been proposed.There is the newest fruits that much combines artificial intelligence field to propose in these methods.Such as wavelet theory, fuzzy mathematics, neural net or the like, the appearance and the breakthrough of the intelligent protection that these technological develop rapidlys are transformer provide solid foundation.
Summary of the invention
The present invention be directed to the problem of the exciting current of transformer to the importance of raising transformer quality; Proposed a kind of based on the transformer excitation flow discrimination method that improves probabilistic neural network; (Probability Neural Network PNN) as the core classification device, differentiates the magnetizing inrush current and the internal fault current of power transformer to adopt probabilistic neural network; Apply to intellectual technology in the differentiation of magnetizing inrush current, and obtained good achievement.
Technical scheme of the present invention is: a kind of based on the transformer excitation flow discrimination method that improves probabilistic neural network, specifically comprise the steps:
1) waveform of modeling and simulating transformer excitation flow and short circuit current: adopt the SimPowerSystems module library among the Matlab/Simulink, parameter is a system default, obtains magnetizing inrush current and two kinds of current waveforms of short circuit;
2) respectively magnetizing inrush current and short circuit current are carried out wavelet analysis; Extract energy feature: set sample frequency; Current signal to sampling carries out small echo four times; Decompose the energy of getting each high band and form characteristic vector:
; Characteristic vector is carried out normalization:
; The characteristic vector of identification magnetizing inrush current and short circuit current has formed the sample space of being made up of the characteristic vector after the normalization;
3) smoothing factor of genetic algorithm optimization probabilistic neural network
: in limited sample space, extract the smoothing factor that can reflect whole sample space, utilize the optimal value that obtains smoothing factor behind the genetic algorithm optimization;
4) based on the transformer differential protection scheme of artificial neural net: magnetizing inrush current and the energy of internal fault current after wavelet decomposition are imported as characteristic vector, the neural net after optimizing is carried out training and testing, to carry out pattern recognition.
Said genetic algorithm optimization smoothing factor mainly contains following step:
A: the span of setting smoothing factor; Produce initial N bar chromosome immediately and form initial population;
, and establish current algebraically t=1;
B: according to the smoothing factor that obtains by all chromosomes; Make up the PNN network; Calculate classification correct number t and error function
; M is the quantity of training sample, promptly calculates chromosomal fitness function;
C: select winning individuality, intersect, mutation operation, obtain population of following generation;
D: establish current algebraically t=t+1;
Beneficial effect of the present invention is: the present invention is based on the transformer excitation flow discrimination method that improves probabilistic neural network, apply to intellectual technology in the differentiation of magnetizing inrush current, the method has improved magnetizing inrush current Fault Identification ability greatly.
Description of drawings
Fig. 1 is probabilistic neural network (PNN) structure chart;
Fig. 2 is single supply electric power system figure;
Magnetizing inrush current oscillogram when Fig. 3 is the three-phase transformer idle-loaded switching-on;
Fig. 4 is three-phase transformer short circuit current waveform figure;
Fig. 5 is the transformer differential protection system schematic that the present invention is based on the transformer excitation flow discrimination method design that improves probabilistic neural network.
Embodiment
Adopt genetic algorithm that the smoothing factor of probabilistic neural network is optimized; Model emulation through in Matlab/Simulink, building obtains current waveform; The wavelet transformation energy of magnetizing inrush current and internal fault current is imported as network, carried out Fault Pattern Recognition.
Probabilistic neural network is a kind of feedforward neural network.The probabilistic neural network structure is a kind of according to Parzen parameter probability density function (probability density function, pdf) neural net of estimation and Bayes classification rule.The training program of probabilistic neural network standard should comprise the training set of all patterns.These characteristics make probabilistic neural network compare the feedforward reverse transmittance nerve network training speed faster.The unique shortcoming of probabilistic neural network is essential a large amount of storage training mode.Along with calculator memory has become very cheaply effective again, nowadays the cost of big capacity storage and space size have not all become problem.Probabilistic neural network is widely used in fields such as pattern recognition, Nonlinear Mapping, fault detect and classification.
The structure of probabilistic neural network is as shown in Figure 1.It is four layers of feed-forward neural net, and the I layer is an input layer, and the II layer is for seeking knowledge layer, and the III layer is a mode layer, and the IV layer is an output layer, can be used as the grader of an optimum.In probabilistic neural network, Gaussian function is because it is functional and be easy to calculate, usually as activation primitive.The activation primitive of probabilistic neural network is to obtain according to the probability density function that training sample obtains, and concrete steps are following:
If
is d dimension training sample vector; Belong to the i class; Wherein
, k is the number of all categories.According to the Bayes criterion, posterior probability
can calculate like this:
Wherein,
; For being under the jurisdiction of certain type conditional probability density function;
is prior probability, and
is the probability density function of input vector.
A certain type posterior probability is maximum, and then sample belongs to this classification.
to all has
Parzen in 1962 proposed a kind of from known random sample the method for estimated probability density function, as long as number of samples is abundant, the function that this method obtained can approach former probability density function continuously smooth.The PDF estimator that is obtained by the Parzen method is:
Wherein,
is j training sample in the i class; M is the quantity of training sample;
is smoothing factor.
Different with the BP network, the learning algorithm of probabilistic neural network is not adjusted the connection weights between the neuron in training process, and data sample is depended in the study of network fully, only needs to confirm smoothing parameter.Under the situation that the input sample is confirmed, correctly select spreading coefficient or smoothing factor, can improve the classification performance of probabilistic neural network effectively.Usually, people can carry out the estimation of empirical formula statistics to smoothing factor, but have increased amount of calculation like this, and choose the process complicacy, are difficult to obtain optimum smoothing factor parameter.Therefore, this paper chooses the very competent genetic algorithm of optimization and confirms smoothing factor, in the hope of reaching good classifying quality.
Genetic algorithm is (Genetic Algorithm, GA) a kind of optimization searching method based on natural selection and gene genetics principle.The theory of biologic evolution that it will " be selected the superior and eliminated the inferior; the survival of the fittest " is introduced in the coded strings colony of parameter formation to be optimized; According to certain adaptation value function and a series of genetic manipulation individuality is screened, thereby make the high individuality of adaptation value be retained, form new colony; New colony comprises the bulk information of previous generation, and has introduced the new individuality that is superior to previous generation.Go round and begin again like this, the fitness of individual in population improves constantly, till meeting some requirements.
When utilizing genetic algorithm to be optimized, at first to smoothing factor
formation of encoding chromosome.This paper adopts real coding, because real coding has shorter code length than binary coding, and has avoided decode procedure, and precision is higher.After each genetic manipulation is accomplished, make up the PNN network, calculate the error function of neural net, confirm the fitness function that each is individual according to Different Individual.Error function can be set at:
Wherein,
is desired output corresponding and the training set sample;
is the current output after being trained by the probabilistic neural network that chromosome obtains; Q is the number of training sample set, and s is the fault type number.Calculate each chromosomal fitness value, in order to weighing the quality of each chromosome performance, thereby select well behaved chromosome,, finally obtain the optimal value of smoothing factor through after the iteration repeatedly.
Among the present invention, the author utilizes the electric power system tool box (SPS) among the Matlab/Simulink to set up the analogue system model, and shown in Figure 2 is unidirectional power supply electric power system.It is by three-phase voltage source (rated capacity is 250MVA), the three-phase two winding transformer (40MVA of a 110kv;
, 110/11 kV) and the three-phase load composition.It is to be noted that the connection type of current transformer of primary side and secondary side is a star, the phase problem of both sides electric current can pass through relay in house software calculation correction.
In the fault type of all transformers, the shared ratio of inner shorted-turn fault is quite big, has also directly determined the performance of differential relay.In the simulation process, secondary side winding equivalence one-tenth three windings with the two winding transformer that turn-to-turn short circuit takes place come the emulation shorted-turn fault.This paper has mainly carried out emulation to following several kinds of situation:
(1) changes combined floodgate initial phase angle and remanent magnetism and produce magnetizing inrush current.The excursion of switching angle is 0~360, and step-length is 30 degree.When iron core remanent magnetism is rated voltage 0%~80% of Peak Value of Magnetic Leakage Flux.
(2) different turn-to-turn short circuits than the time (5%~50%) the internal fault short circuit, change short-circuit-type: single-line to ground fault, two phase ground short circuit, line to line fault and three-phase ground short circuit.
Fig. 3 is depicted as typical magnetizing inrush current waveform, and Fig. 4 is a three-phase transformer fault current waveform.
From figure, can find out significantly that fault current and exciting current have tangible difference, fault current is the continuous interval angle that do not exist, and waveform is symmetrical sine wave.After fft analysis, can find magnetizing inrush current except that containing a large amount of first-harmonic and aperiodic component, contain a large amount of high order harmonic components, be main wherein, and internal fault current is main with first-harmonic mainly with second harmonic, higher harmonic content is low.The content of the each harmonic that is comprised just because of magnetizing inrush current and internal fault current is different, and after wavelet decomposition, the energy that is distributed in each high band is different with current signal.
According to Ba Shiwa (Parseval) theorem, in signal processing, energy and the energy in frequency domain of signal in the time frequency domain is conservation.Therefore, signal is after the small echo multilayer is decomposed, and the energy of primary signal can be expressed as:
Wherein,
,
are respectively approximate signal coefficient and the detail signal coefficient under each yardstick after the wavelet decomposition.After being about to signal f (t) wavelet decomposition, the quadratic sum of its approximate signal coefficient and detail signal coefficient equals the energy of primary signal on time domain.
Calculate for ease, this paper chooses the energy of detail section as scaling function.To discrete signal, obtain the wavelet coefficient
under each yardstick through dyadic wavelet transform.Definition details energy function:
Tectonic energy measure feature vector:
, m is a yardstick.Characteristic vector is carried out normalization:
; Then T is the feature space that is generated by characteristic vector after the normalization, can be used as the characteristic vector of identification magnetizing inrush current and short circuit current.
In the magnetizing inrush current, except that first-harmonic and non-same period component, containing a large amount of higher harmonic currents, is main with second harmonic wherein, also have the part triple-frequency harmonics sometimes, and the internal short-circuit electric current is main with first-harmonic mainly, and content of high order harmonic is lower.Different just because of magnetizing inrush current with the each harmonic that the internal short-circuit electric current is comprised, current signal carried out wavelet decomposition after, the energy that is distributed in each high band is also different.
Step 1: the waveform of modeling and simulating transformer excitation flow and short circuit current:
Each module of simulation model directly adopts the SimPowerSystems module library among the Matlab/Simulink, and parameter is system default basically, obtains two kinds of current waveforms like Fig. 3 and 4.
Short-circuit current among the waveform of 3 kinds of transformer excitation flows of comparison diagram and Fig. 4 can be found significantly, and magnetizing inrush current has following characteristics:
(1) magnetizing inrush current is a peaked wave, wherein contains the aperiodic component and the higher harmonic components of suitable composition.Be main with secondary and triple-frequency harmonics in the high order harmonic component, the ratio of second harmonic component very significantly, and As time goes on, its proportion increases on the contrary to some extent, and has at least a phase second harmonic component very big, possibly surpass 60% of fundametal compoment.And first few possibly is partial to a side of time shaft fully in the cycle;
Voltage initial phase angle when (2) amplitude of magnetizing inrush current drops into no-load transformer is relevant;
(3) magnetizing inrush current first few in the cycle waveform be interrupted,
interval angle was arranged in each cycle;
(4) magnetizing inrush current is relevant with the amount of capacity of transformer for the multiple of rated current amplitude.Transformer capacity is big more, and magnetizing inrush current is more little to rated current amplitude rhythm multiple;
(5) factors such as size, transformer capacity and core material of impedance are relevant between the time constant of magnetizing inrush current decay and transformer to power supply.Generally speaking, the capacity of transformer is big more, or the closer to power supply, its die-away time is long more.Iron core is saturated more, and reactance value is more little, and it is fast more to decay.
Step 2: respectively magnetizing inrush current and short circuit current are carried out wavelet analysis, extract energy feature:
The sample frequency that the present invention selects is 1kHz, can be distributed to individually on each high band in order to make each harmonic, needs signal is carried out wavelet decomposition four times.Choosing of wavelet basis also has certain influence to test result, and through repeatedly emulation, the author chooses the better sym4 wavelet function of performance.
The energy of getting each high band forms characteristic vector:
.Characteristic vector is carried out normalization:
then formed the sample space of being made up of the characteristic vector after the normalization.
Step 3: the smoothing factor of genetic algorithm optimization probabilistic neural network:
In limited pattern sample, extract the smoothing factor that can reflect whole sample space, and present smoothing factor is estimated all the method based on experience estimation or very limited sample cluster, and can not the probability nature in space very intactly be expressed.And genetic algorithm can have no the potential knowledge of discovery system under the situation of priori, and has obtained success widely at aspects such as function optimization, control automatically, image recognition and machine learning.
Mainly contain following step through the genetic algorithm optimization smoothing factor:
(1) span of setting smoothing factor; Produce initial N bar chromosome immediately and form initial population;
, and establish current algebraically t=1;
(2) according to the smoothing factor that obtains by all chromosomes, make up the PNN network, calculate classification correct number and error, promptly calculate chromosomal fitness function;
(3) select winning individuality, intersect, mutation operation, obtain population of following generation;
(4) establish current algebraically t=t+1;
Step 4: based on the transformer differential protection scheme of artificial neural net:
A kind of new transformer differential protection scheme has been proposed; At first utilize genetic algorithm that the smoothing factor of probabilistic neural network is optimized; Magnetizing inrush current and the energy of internal fault current after wavelet decomposition are imported as characteristic vector; Neural net is carried out training and testing, to carry out pattern recognition.
Shown in Figure 5 is transformer differential protection scheme based on artificial neural net, is used for discerning magnetizing inrush current and internal fault current.The current value 1 that Circuit Fault on Secondary Transformer collects by wavelet decomposition 2 after, handle through normalization, form sample space.By its characteristic quantity of each network extraction, according to corresponding topological structure and learning rules, train, then test sample book is made corresponding fault mode and judge.In order to compare with the classifying quality that passes through the probabilistic neural network 5 after optimizing, this paper has also chosen probabilistic neural network 4 and BP neural net 3.Can know by above analysis; The input number of nodes of three kinds of networks is identical; Be the energy feature vector number of each high band; The output node number is 1, i.e. network diagnosis result 6, comprises exciting current 7, turn-to-turn short circuit 8, single-phase earthing 9, two phase ground 10, line to line fault 11 and 12 6 results of three-phase ground connection.Through after a large amount of debugging, the classification error number of the classifying quality of three kinds of neural nets three kinds of neural nets as shown in table 1, the total number of the digitized representation test sample book of back.
The classification accuracy of contrast diverse network can be found, though the BP neural net also has good performance, could confirm network configuration through experiment constantly, and will pass through ability trained behind a large amount of iterationses; The training process of probabilistic neural network is once accomplished, and need not carry out the weights adjustment, has better generalization ability than traditional BP neural net, and convergence rate is faster, selects suitable smoothing factor can improve the performance of probabilistic neural network; Probabilistic neural network through after optimizing has certain adaptive learning ability, can select optimum smoothing parameter according to different situations, has improved the Fault Identification ability of network greatly.
Claims (2)
1. the transformer excitation flow discrimination method based on the improvement probabilistic neural network is characterized in that, specifically comprises the steps:
1) waveform of modeling and simulating transformer excitation flow and short circuit current: adopt the SimPowerSystems module library among the Matlab/Simulink, parameter is a system default, obtains magnetizing inrush current and two kinds of current waveforms of short circuit;
2) respectively magnetizing inrush current and short circuit current are carried out wavelet analysis; Extract energy feature: set sample frequency; Current signal to sampling carries out small echo four times; Decompose the energy of getting each high band and form characteristic vector:
; Characteristic vector is carried out normalization:
; The characteristic vector of identification magnetizing inrush current and short circuit current has formed the sample space of being made up of the characteristic vector after the normalization;
3) smoothing factor of genetic algorithm optimization probabilistic neural network
: in limited sample space, extract the smoothing factor that can reflect whole sample space, utilize the optimal value that obtains smoothing factor behind the genetic algorithm optimization;
4) based on the transformer differential protection scheme of artificial neural net: magnetizing inrush current and the energy of internal fault current after wavelet decomposition are imported as characteristic vector, the neural net after optimizing is carried out training and testing, to carry out pattern recognition.
2. said based on the transformer excitation flow discrimination method that improves probabilistic neural network according to claim 1, it is characterized in that said genetic algorithm optimization smoothing factor mainly contains following step:
A: the span of setting smoothing factor; Produce initial N bar chromosome immediately and form initial population;
, and establish current algebraically t=1;
B: according to the smoothing factor that obtains by all chromosomes; Make up the PNN network; Calculate classification correct number t and error function
; M is the quantity of training sample, promptly calculates chromosomal fitness function;
C: select winning individuality, intersect, mutation operation, obtain population of following generation;
D: establish current algebraically t=t+1;
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2012100131545A CN102570392A (en) | 2012-01-17 | 2012-01-17 | Method for identifying exciting inrush current of transformer based on improved probability neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2012100131545A CN102570392A (en) | 2012-01-17 | 2012-01-17 | Method for identifying exciting inrush current of transformer based on improved probability neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102570392A true CN102570392A (en) | 2012-07-11 |
Family
ID=46415133
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2012100131545A Pending CN102570392A (en) | 2012-01-17 | 2012-01-17 | Method for identifying exciting inrush current of transformer based on improved probability neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102570392A (en) |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103398769A (en) * | 2013-08-05 | 2013-11-20 | 国家电网公司 | Transformer on-line fault detecting method based on sampling integrated SVM (support vector machine) under wavelet GGD (general Gaussian distribution) feather and unbalanced K-mean value |
CN103412221A (en) * | 2013-08-13 | 2013-11-27 | 昆明理工大学 | Transformer excitation surge current identification method based on time-frequency characteristic quantities |
CN103884966A (en) * | 2014-04-15 | 2014-06-25 | 河海大学常州校区 | Power distribution network low-current single-phase earth fault positioning method based on neural network |
WO2016019593A1 (en) * | 2014-08-06 | 2016-02-11 | 浙江群力电气有限公司 | Method and apparatus for identifying causes for cable overcurrent |
CN106327357A (en) * | 2016-08-17 | 2017-01-11 | 深圳先进技术研究院 | Load identification method based on improved probabilistic neural network |
CN107301271A (en) * | 2017-05-23 | 2017-10-27 | 哈尔滨工业大学深圳研究生院 | A kind of steel wire rope outer layer damages Magnetic Flux Leakage Inspecting Quantitative algorithm |
CN108267684A (en) * | 2018-01-15 | 2018-07-10 | 哈尔滨理工大学 | A kind of Converter Fault Diagnosis method |
CN108599104A (en) * | 2018-05-11 | 2018-09-28 | 国网山东省电力公司东营供电公司 | Tranformer protection New Algorithm for Inrush Current Identification of Transformer |
CN109795713A (en) * | 2019-02-14 | 2019-05-24 | 四川大学 | Aileron actuator method for diagnosing faults based on Simulink model |
CN110084106A (en) * | 2019-03-19 | 2019-08-02 | 中国地质大学(武汉) | Microgrid inverter method for diagnosing faults based on wavelet transformation and probabilistic neural network |
CN110705031A (en) * | 2019-09-06 | 2020-01-17 | 西南交通大学 | Excitation inrush current identification method based on second-order Taylor coefficient |
CN112307918A (en) * | 2020-10-21 | 2021-02-02 | 华北电力大学 | Diagnosis method for transformer direct-current magnetic biasing based on fuzzy neural network |
CN112580700A (en) * | 2020-12-04 | 2021-03-30 | 杭州佳速度产业互联网有限公司 | Data correction method, system and storage medium of electric power Internet of things meter |
CN112989734A (en) * | 2021-02-25 | 2021-06-18 | 中国人民解放军海军航空大学 | Equipment analog circuit fault diagnosis method based on probabilistic neural network |
CN113030824A (en) * | 2021-03-03 | 2021-06-25 | 云南海力特电气自动化有限公司 | Integrated debugging device of online monitoring system of capacitive equipment of intelligent substation |
CN114076855A (en) * | 2020-08-10 | 2022-02-22 | 施耐德电器工业公司 | Inrush current detection method and apparatus for transformer, and computer-readable storage medium |
CN116451544A (en) * | 2023-05-31 | 2023-07-18 | 南通大学 | Intelligent optimization design method for high-power high-frequency transformer |
RU2803942C2 (en) * | 2018-11-12 | 2023-09-22 | Альцхем Тростберг Гмбх | Method of reducing gases emission from on-farm fertilizers |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101599634A (en) * | 2009-06-12 | 2009-12-09 | 西安理工大学 | Based on the transformer excitation flow of S-conversion and the discrimination method of fault current |
-
2012
- 2012-01-17 CN CN2012100131545A patent/CN102570392A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101599634A (en) * | 2009-06-12 | 2009-12-09 | 西安理工大学 | Based on the transformer excitation flow of S-conversion and the discrimination method of fault current |
Non-Patent Citations (4)
Title |
---|
吴浩: "基于人工神经网络的变压器差动保护研究", 《中国优秀硕士学位论文全文数据库》 * |
田有文等: "基于小波变换的变压器励磁涌流识别方法", 《沈阳农业大学学报》 * |
蔡曲林: "基于概率神经网络的模式识别", 《中国优秀硕士学位论文全文数据库》 * |
雷正伟等: "基于遗传优化的概率神经网络预测方法", 《自动化与仪表》 * |
Cited By (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103398769A (en) * | 2013-08-05 | 2013-11-20 | 国家电网公司 | Transformer on-line fault detecting method based on sampling integrated SVM (support vector machine) under wavelet GGD (general Gaussian distribution) feather and unbalanced K-mean value |
CN103398769B (en) * | 2013-08-05 | 2014-12-31 | 国家电网公司 | Transformer on-line fault detecting method based on sampling integrated SVM (support vector machine) under wavelet GGD (general Gaussian distribution) feather and unbalanced K-mean value |
CN103412221A (en) * | 2013-08-13 | 2013-11-27 | 昆明理工大学 | Transformer excitation surge current identification method based on time-frequency characteristic quantities |
CN103884966A (en) * | 2014-04-15 | 2014-06-25 | 河海大学常州校区 | Power distribution network low-current single-phase earth fault positioning method based on neural network |
WO2016019593A1 (en) * | 2014-08-06 | 2016-02-11 | 浙江群力电气有限公司 | Method and apparatus for identifying causes for cable overcurrent |
CN106327357A (en) * | 2016-08-17 | 2017-01-11 | 深圳先进技术研究院 | Load identification method based on improved probabilistic neural network |
CN107301271A (en) * | 2017-05-23 | 2017-10-27 | 哈尔滨工业大学深圳研究生院 | A kind of steel wire rope outer layer damages Magnetic Flux Leakage Inspecting Quantitative algorithm |
CN107301271B (en) * | 2017-05-23 | 2020-05-29 | 哈尔滨工业大学深圳研究生院 | Steel wire rope outer layer damage magnetic flux leakage detection quantification method |
CN108267684A (en) * | 2018-01-15 | 2018-07-10 | 哈尔滨理工大学 | A kind of Converter Fault Diagnosis method |
CN108267684B (en) * | 2018-01-15 | 2020-03-10 | 哈尔滨理工大学 | Frequency converter fault diagnosis method |
CN108599104A (en) * | 2018-05-11 | 2018-09-28 | 国网山东省电力公司东营供电公司 | Tranformer protection New Algorithm for Inrush Current Identification of Transformer |
RU2803942C2 (en) * | 2018-11-12 | 2023-09-22 | Альцхем Тростберг Гмбх | Method of reducing gases emission from on-farm fertilizers |
CN109795713A (en) * | 2019-02-14 | 2019-05-24 | 四川大学 | Aileron actuator method for diagnosing faults based on Simulink model |
CN110084106A (en) * | 2019-03-19 | 2019-08-02 | 中国地质大学(武汉) | Microgrid inverter method for diagnosing faults based on wavelet transformation and probabilistic neural network |
CN110705031A (en) * | 2019-09-06 | 2020-01-17 | 西南交通大学 | Excitation inrush current identification method based on second-order Taylor coefficient |
CN114076855A (en) * | 2020-08-10 | 2022-02-22 | 施耐德电器工业公司 | Inrush current detection method and apparatus for transformer, and computer-readable storage medium |
CN112307918A (en) * | 2020-10-21 | 2021-02-02 | 华北电力大学 | Diagnosis method for transformer direct-current magnetic biasing based on fuzzy neural network |
CN112580700A (en) * | 2020-12-04 | 2021-03-30 | 杭州佳速度产业互联网有限公司 | Data correction method, system and storage medium of electric power Internet of things meter |
CN112580700B (en) * | 2020-12-04 | 2021-07-30 | 杭州佳速度产业互联网有限公司 | Data correction method, system and storage medium of electric power Internet of things meter |
CN112989734A (en) * | 2021-02-25 | 2021-06-18 | 中国人民解放军海军航空大学 | Equipment analog circuit fault diagnosis method based on probabilistic neural network |
CN113030824A (en) * | 2021-03-03 | 2021-06-25 | 云南海力特电气自动化有限公司 | Integrated debugging device of online monitoring system of capacitive equipment of intelligent substation |
CN116451544A (en) * | 2023-05-31 | 2023-07-18 | 南通大学 | Intelligent optimization design method for high-power high-frequency transformer |
CN116451544B (en) * | 2023-05-31 | 2023-11-07 | 南通大学 | Intelligent optimization design method for high-power high-frequency transformer |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102570392A (en) | Method for identifying exciting inrush current of transformer based on improved probability neural network | |
CN110311376B (en) | Dynamic safety assessment comprehensive model and space-time visualization method for power system | |
Zhang et al. | A fault diagnosis model of power transformers based on dissolved gas analysis features selection and improved krill herd algorithm optimized support vector machine | |
CN103136585B (en) | Weighting Voronoi diagram substation planning method based on chaotic and genetic strategy | |
CN105653764B (en) | Power grid safety accident risk class online evaluation and method for early warning | |
CN102510044A (en) | Excitation inrush current identification method based on wavelet transformation and probabilistic neural network (PNN) | |
CN108152651A (en) | Transmission line malfunction integrated recognition method based on GMAPM and SOM-LVQ-ANN | |
Ozgonenel et al. | Power transformer protection based on decision tree approach | |
Luo et al. | Stacked auto-encoder-based fault location in distribution network | |
CN111652479B (en) | Data driving method for dynamic security assessment of power system | |
CN104616061B (en) | Island detection method based on wavelet packet logarithmic energy entropy and genetic algorithm optimization | |
CN103972883B (en) | Closed loop network point selection method and system | |
CN104614638A (en) | Grounding line selection method for small current system | |
CN104538222A (en) | High-voltage switch phase selection controller based on artificial neural network and method | |
Tripathy | Power Transformer Differential Protection Based on Neural Network Principal Component Analysis, Harmonic Restraint and Park′ s Plots | |
CN109064060B (en) | Method for evaluating relay protection misoperation risk caused by magnetizing inrush current | |
CN104578050A (en) | Transient stability strongly-correlated power transmission section identification method for power grid | |
Abohagar et al. | Back propagation neural network aided wavelet transform for high impedance fault detection and faulty phase selection | |
Balaga et al. | Numerical differential protection of power transformer using ANN as a pattern classifier | |
Quan et al. | The application of Bayesian network theory in transformer condition assessment | |
Shu et al. | A detection method of high impedance arcing fault for distribution network with distributed generation based on CEEMDAN and TEO algorithm | |
Halim et al. | Hysteretic iron-core modelling for inrush current transients using Jiles-Atherton model | |
Ji et al. | Diagnosis of Power Transformer Internal Fault Based on Denoising Autoencoder and Optimized Convolutional Neural Network | |
Wu et al. | The magnetic inrush current and internal fault types recognition in transformer based on probabilistic neural network | |
CN106548287A (en) | A kind of comprehensive estimation method of intelligent substation design |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20120711 |