CN103995528A - Intelligent self-repairing technology for main circuit of power converter - Google Patents
Intelligent self-repairing technology for main circuit of power converter Download PDFInfo
- Publication number
- CN103995528A CN103995528A CN201410197879.3A CN201410197879A CN103995528A CN 103995528 A CN103995528 A CN 103995528A CN 201410197879 A CN201410197879 A CN 201410197879A CN 103995528 A CN103995528 A CN 103995528A
- Authority
- CN
- China
- Prior art keywords
- fault
- main circuit
- circuit
- power
- self
- 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
Landscapes
- Inverter Devices (AREA)
- Analogue/Digital Conversion (AREA)
Abstract
The invention relates to an intelligent self-repairing technology for a main circuit of a power converter, and belongs to the field of circuit testing. A method includes the following steps that (1), the main circuit is simulated and analyzed to determine the fault type, the number, the testing nodes (namely voltage or current information) and the like of power tubes of the main circuit; (2), fault information is collected through the testable nodes, off-line training of a classifier is carried out by implementing feature extraction, and a fault dictionary is set up and stored in a storage of a CPU; (3), the CPU collects related information of the power tubes of the main circuit and carries out feature extraction (consistent with that in the step (2) ), calculation and analysis are carried out by means of the fault dictionary to judge whether the power tubes fail or not, if not, the power tubes are monitored again, if yes, the positions of the power tubes are judged, the failed power tube is turned off, and the standby power tube is switched to, so that functions of the main circuit and self-repairing of the structure are achieved. The intelligent self-repairing technology is high in flexibility and reliability and suitable for occasions with high requirements for reliability of the power tubes.
Description
Technical field
The present invention relates to a kind of fault diagnosis and self-repair technology of power converter main circuit, belong to the field of fault prognostics and health management (Prognostics and health management is called for short PHM) aspect.
Background technology
Power converter mainly consists of power tube and additional device, the reliability of power tube substantially can the whole changer system in left and right reliability, thereby its fault diagnosis and location technology has almost all obtained broad research and application in every field, although relevant Research on Fault Diagnosis Technology is more, to the self-repair technology research of main circuit but seldom.Self-repair technology belongs to the PHM field of power converter, is circuit fault diagnosis and the Some Key Technologies of prediction in decision-making, is also one of the focus of research and trend.
At present, the intellectual technology based on data-driven is emphasis and the trend in diagnosing main circuit fault of power converter research.Wherein, support vector machine classifier (Support Vector Machines Classifier, be called for short SVC) is strong with its extrapolability, small-sample learning ability is strong, to advantages such as data dimension are insensitive, becomes the research emphasis of fault diagnosis field.Conventional binary SVC (Binary SVC is called for short BSVC) can not be directly used in many merotypes identification, must carry out certain redesign and structure just can complete this type of work.At present, main mentality of designing is directly to combine BSVC, to form a SVC that can carry out multi-mode classification, and " one against rest " SVC (be called for short OVRSVC) for example.
Summary of the invention
In order to address the above problem, the present invention proposes a kind of Smart self-repairing method of the power converter main circuit based on OVRSVC, the method utilizes DSC to implement signals collecting and processing, and by means of OVRSVC, carry out the intelligent trouble Fast Classification of power tube, according to positioning result, implement the selfreparing of power tube.This technology can significantly improve the reliability of power converter main circuit, in a lot of important events, has very important using value.
The present invention adopts following technical scheme for solving its technical matters:
A Smart Self-repairing for power converter main circuit, comprises the steps (1)～(4):
(1) power converter main circuit to be measured is carried out to Testability Analysis, determine type, the number of power tube fault, determine the information types such as voltage, electric current can survey node and collection;
(2) carry out off-line operation, comprise the structure that data acquisition, fault signature extract, sorter is trained (adopting PC to realize) and fault dictionary, described fault signature extracts, and adopts fractional order Fourier analysis and the Feature Selection based on genetic algorithm;
(3) when carrying out the state on_line monitoring of side circuit, DSC utilizes and can gather failure message by measuring point, and utilize step 2) in same fault signature extracting method implement failure message compression and extract, and utilize OVRSVC to implement the working state evaluation of power converter main circuit, determine position and the number of fault power pipe.
(4) DSC, according to the positional information of fault power pipe, sends control signal, implements the backup of power tube and replaces, thereby realize selfreparing by self-repair circuit.
Beneficial effect of the present invention is as follows:
(1) by DSC, implement signal and process and control, can significantly improve speed and the efficiency of information processing, and can realize flexibly upgrading and the transformation of systemic-function.
(2) by the advantage of GA algorithm, can select to be applicable to the required nuclear parameter of follow-up OVRSVC and corresponding fault signature set, and significantly packed data is tieed up, and reduces the requirement to DSC storage space, thereby can reduce the computation complexity of DSC, improve the counting yield of on-line monitoring.
(3) by by DSC and artificial intelligence technology, the status monitoring of power converter main circuit and the automaticity of selfreparing be can greatly improve, human cost and stop time reduced.This technology can be applied in the plate level and system-level level of power converter main circuit.
Accompanying drawing explanation
The Smart self-repairing flow process of Fig. 1 power converter main circuit.
The reparation schematic diagram of x power tube of Fig. 2.
Embodiment
Below in conjunction with accompanying drawing, the invention is described in further detail.
The present invention has designed a kind of Smart self-repairing method of power converter main circuit, and the method mainly comprises two flow process parts (flow process I and flow process II), respectively as shown in the two-way dotted line in Fig. 1.
One, the enforcement of flow process I
(1) testability analysis (measuring point and information type are selected).
By circuit is carried out to Testability Analysis, for determining the type (open fault or short trouble) of power tube fault and number (having how many power tube devices is potential sources of trouble, whether will consider the combination of two or above power tube etc.).Suppose to need the total N kind (normal state that comprises circuit) of fault mode one herein.Above-mentioned these work can adopt simulation software (for example Saber) to carry out, while carrying out emulation, need to be at software inhouse drawing principle figure, and arrange according to the fault mode that may occur.Software analysis can provide the Voltage-output information of all node surveyed, and also comprises the information such as electric current of each branch road.Therefore, gather these information and carry out qualitative or quantitative test, tell which node and information type and be conducive to carry out fault information acquisition, and can differentiate under the prerequisite of all fault modes, guarantee that the utilized nodes surveyed and failure message number of types are minimum.
(2) data acquisition.
Because the OVRSVC in the present invention adopts supervised learning method to train, therefore need the fault sample of some.Both obtaining of sample can adopt physical data sample acquisition, also can adopt emulation to be obtained.If adopt the mode of emulation, need to consider the impact (simulating by Monte Carlo technology) of component tolerance in practical application.
When carrying out actual hardware data collection, also need special sensor type and the precision of considering that diagnosis is required: if collection is current signal, needs, by I/V circuit, current signal is converted to voltage and gather again.The sampling rate of ADC, precision and data point number etc. can determine according to actual needs, such as, the AD converter precision of data acquisition is 12, sample rate f
s=500KHz.
(3) feature extraction and selection.
The data sample collecting is carried out to feature extraction, and with packed data the interference of removing high frequency noise, feature extracting method herein adopts fractional order Fourier (Fractional Fourier Transformation is called for short FrFT) analytical technology.For continuous signal f (t), the continuous transformation formula of FrFT is:
Wherein:
At κ
αin, n is integer, and j is the imaginary symbols (j2=-1) in plural number, and cot, csc are cotangent and cosecant function, and a is twiddle factor, and
p is the exponent number of fractional order, and t is time variable, and u is the parameter of above formula kernel function, and δ is impulse function, and n is integer (0,1,2 etc.), and π is circular constant constant.
The span of p is generally: 0～1, when p gets different values, obtain different coefficients.State when these coefficients have reflected the frequency domain space transition of this node signal from from original signal space to Fourier, when circuit produces dissimilar fault, the output signal that can survey node all can produce different variations, correspondingly these fractional order coefficient of dissociation all can produce different values, for fault diagnosis, are therefore all valuable information.Therefore, in the present invention, these coefficient of dissociation are for the feature as fault diagnosis, and the number of features of supposing all coefficient of dissociation is N (the original decomposition intrinsic dimensionality that is each fault sample is N).
Be not that all features all have contribution to follow-up fault diagnosis and pattern classification, therefore, need to select to N feature (selection work wants off-line to carry out, and generally on PC, realizes), being also advantageous in that of doing like this reduced number of features, reduced the complexity that DSC calculates.Adopt the feature selection approach based on GA herein.Shown in the key step of selecting:
<1> sets individuality (being the feature locations of sample and the scope of the nuclear parameter) coded system of GA algorithm, all adopts binary coding mode herein.Fitness function is herein Euclidean distance tolerance, is set as:
di is the Euclidean distance between i fault mode class sample average and all the other all fault mode class sample averages under current feature, and min (.) represents to get minimum value function, and Φ (.) is Nonlinear Mapping kernel function.
In this step, all sample standard deviations need to (adopt radial basis kernel function herein by a Nonlinear Mapping kernel function Φ (a), a is the width parameter of radial basis kernel function) be mapped into higher dimensional space, Euclidean distance is also the distance metric in this higher dimensional space.Therefore, in the search procedure of GA algorithm, the size (affecting the nuclear parameter size of follow-up SVC) of the corresponding eigenwert of the larger f value position that can produce (also recording the corresponding eigenwert size in this position simultaneously) and nuclear parameter a.
<2> operation GA algorithm, finally obtains one group of optimum solution (may be also suboptimum), size and the sample characteristics of the nuclear parameter a that record now obtains.
(4) sorter training and fault dictionary storage.
Build OVRSVC, adopt standard support vector machine for example, to the sample training of each fault mode (process of training is carried out by means of application software such as PCs, Matlab software), the BSVC number of such SVC is consistent with fault mode number.By training, can obtain the training parameter (comprising support vector, deviation and nuclear parameter etc.) of each BSVC, and be stored in the external non-volatile memory of DSC, form fault dictionary.
Two, the enforcement of flow process II
The data acquisition of this flow process, two parts of feature extraction and flow process I are substantially similar.Difference is, the feature extraction in flow process II is determined fault signature in direct computing technique flow process I, and no longer carries out the feature selecting work based on GA algorithm.
<1> localization of fault.
DSC, by after data acquisition and feature extraction, has obtained the status information of current power inverter main circuit, and DSC by by means of fault dictionary, can calculate the output of OVRSVC, thereby whether the state that can adjudicate current power pipe breaks down.The state if current system does not break down, system index belongs to normal range of operation, does not need to move, and directly returns and continues monitoring system state.If malfunction has appearred in system, show that power tube goes wrong, by means of the output of the OVRSVC positional information of failure judgement power tube easily.
<2> controls reparation.
Control the schematic diagram of reparation as shown in Figure 2.Suppose that current x power tube (is numbered the pipe of Mx0 in figure, total n of the backup of this pipe, number consecutively is Mx1～Mxn) there is fault, DSC sends control signal (being generally digital signal), by circuit, controlled multi-way switch is moved, switch is switched to Sx1 from Sx0, and Mx1 is switched to system from backup, operation topological structure and the function of system are all remained unchanged, thereby completed self-repair procedure.
The process of selfreparing is not unlimited, and the number of times that can complete selfreparing is definite by the number n of backup pipe, and the size of n is often wanted the factors such as considering cost size and reliability requirement.
Claims (3)
1. the self-repair technology based on power converter main circuit, is characterized in that, comprises following steps;
(1) utilize digital signal controller (Digital signal controller, be called for short DSC) powerful digital signal processing function, utilize analog to digital converter (Analog to digital converter, abbreviation ADC) Acquisition Circuit can be surveyed the status informations such as voltage and current (need to be transformed to information of voltage through I/V, I/V represents current/voltage translation circuit) of Nodes herein;
(2) DSC carries out pre-service to collecting the signal of memory inside, for extracting fault characteristic information;
(3) fault characteristic information is inputed to pattern classifier, to the duty of whole power converter main circuit is assessed, if circuit working belongs to normal condition, return to step (1); If circuit working is undesired, provide the particular location of fault power pipe, enter next step;
(4) DSC is according to the number of fault power pipe and position, send control signal, control circuit internal actions, makes to back up power tube and replaces fault power pipe, can keep like this topological structure of whole main circuit constant, realize the selfreparing of power converter main circuit structure and function.
2. a kind of self-repair technology based on power converter main circuit according to claim 1, it is characterized in that the Signal Pretreatment technology (being that fault signature extracts) in described step (2), calculated off-line is to determine fault signature in advance for needs, and step is as follows:
(2.1) all status informations (comprising normal operating conditions) sample is carried out to fractional order Fourier analysis, obtain many stack features vector, suppose that number of features is N.All samples of corresponding certain fault mode of each group vector;
(2.2) adopt genetic algorithm (Genetic algorithm is called for short GA algorithm) to find suitable Feature Combination.Wherein, the individuality of GA algorithm adopts binary coding mode (, in the crossover and mutation computing of GA, meaning adding or deleting of certain feature); Fitness function is set as:
di is the Euclidean distance between i fault mode class sample average and all the other all fault mode class sample averages under current feature, and min (.) represents to get minimum value function.In this step, all samples need to be mapped into higher dimensional space by a Nonlinear Mapping kernel function Φ (a) (parameter that a is kernel function), and Euclidean distance is also the distance metric in this higher dimensional space.
(2.3) under some selected nuclear parameters (being the nuclear parameter of follow-up support vector machine classifier), will obtain the feature set of an optimum (or approach optimum), and selected feature is exactly the failure message of the required extraction of subsequent treatment.
3. a kind of self-repair technology based on power converter main circuit according to claim 1, it is characterized in that power tube in described step (4) drives and self-repair circuit is controlled by DSC, wherein, self-repair circuit mainly consists of multi-way switch, backup power tube and DSC control circuit.The number of backup power tube is definite by the reliability index requirement of inverter main circuit, and whole system can complete the number of times of selfreparing and also by the number of backup power tube, be determined.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410197879.3A CN103995528B (en) | 2014-05-09 | 2014-05-09 | Intelligent self-repairing technology for main circuit of power converter |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410197879.3A CN103995528B (en) | 2014-05-09 | 2014-05-09 | Intelligent self-repairing technology for main circuit of power converter |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103995528A true CN103995528A (en) | 2014-08-20 |
| CN103995528B CN103995528B (en) | 2017-04-19 |
Family
ID=51309720
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410197879.3A Expired - Fee Related CN103995528B (en) | 2014-05-09 | 2014-05-09 | Intelligent self-repairing technology for main circuit of power converter |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103995528B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104578416A (en) * | 2014-12-30 | 2015-04-29 | 深圳市科陆电子科技股份有限公司 | Energy storage monitoring system |
| CN105182219A (en) * | 2015-09-06 | 2015-12-23 | 南京航空航天大学 | Power converter fault classification method based on Hamming error correcting code support vector machine |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101614786A (en) * | 2009-07-07 | 2009-12-30 | 南京航空航天大学 | Power electronic circuit on-line intelligence method for diagnosing faults based on FRFT and IFSVC |
| CN101907681A (en) * | 2010-07-15 | 2010-12-08 | 南京航空航天大学 | Analog circuit dynamic online failure diagnosing method based on GSD-SVDD |
| CN102435910A (en) * | 2011-09-14 | 2012-05-02 | 南京航空航天大学 | Power electronic circuit health monitoring method based on support vector classification |
| CN103091603A (en) * | 2013-01-14 | 2013-05-08 | 华北电力大学 | Breakdown intelligent classification and positioning method of electric transmission line |
-
2014
- 2014-05-09 CN CN201410197879.3A patent/CN103995528B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101614786A (en) * | 2009-07-07 | 2009-12-30 | 南京航空航天大学 | Power electronic circuit on-line intelligence method for diagnosing faults based on FRFT and IFSVC |
| CN101907681A (en) * | 2010-07-15 | 2010-12-08 | 南京航空航天大学 | Analog circuit dynamic online failure diagnosing method based on GSD-SVDD |
| CN102435910A (en) * | 2011-09-14 | 2012-05-02 | 南京航空航天大学 | Power electronic circuit health monitoring method based on support vector classification |
| CN103091603A (en) * | 2013-01-14 | 2013-05-08 | 华北电力大学 | Breakdown intelligent classification and positioning method of electric transmission line |
Non-Patent Citations (1)
| Title |
|---|
| 姚睿，等: "空间飞行器集成化电子系统芯片级在线进化修复", 《中国宇航学会深空探测技术专业委员会第三届学术会议论文集》, 1 November 2006 (2006-11-01) * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104578416A (en) * | 2014-12-30 | 2015-04-29 | 深圳市科陆电子科技股份有限公司 | Energy storage monitoring system |
| CN105182219A (en) * | 2015-09-06 | 2015-12-23 | 南京航空航天大学 | Power converter fault classification method based on Hamming error correcting code support vector machine |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103995528B (en) | 2017-04-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Borghesi et al. | A semisupervised autoencoder-based approach for anomaly detection in high performance computing systems | |
| CN101614786B (en) | Online intelligent fault diagnosis method of power electronic circuit based on FRFT and IFSVC | |
| Nguyen et al. | Smart grid line event classification using supervised learning over PMU data streams | |
| CN111459700A (en) | Method and apparatus for diagnosing device failure, diagnostic device, and storage medium | |
| da Silva et al. | A new methodology for multiple incipient fault diagnosis in transmission lines using QTA and Naïve Bayes classifier | |
| CN103745119A (en) | Oil-immersed transformer fault diagnosis method based on fault probability distribution model | |
| CN112330165B (en) | Power grid transient stability evaluation method and system based on feature separation type neural network | |
| CN105095756A (en) | Method and device for detecting portable document format document | |
| Mustafa et al. | Fault identification for photovoltaic systems using a multi-output deep learning approach | |
| Silva et al. | Assets predictive maintenance using convolutional neural networks | |
| CN110968703B (en) | Method and system for constructing abnormal metering point knowledge base based on LSTM end-to-end extraction algorithm | |
| CN115374842A (en) | Mechanical bearing fault classification method and system based on generation countermeasure network | |
| Harish et al. | Fault detection and classification for wide area backup protection of power transmission lines using weighted extreme learning machine | |
| Barreto et al. | Artificial neural network approach for fault detection and identification in power systems with wide area measurement systems | |
| CN103995528A (en) | Intelligent self-repairing technology for main circuit of power converter | |
| Kumar et al. | Deep Learning based Fault Detection in Power Transmission Lines | |
| CN109889258B (en) | Optical network fault checking method and equipment | |
| CN116307269A (en) | Photovoltaic power generation power prediction method and device based on artificial intelligence | |
| Elkanishy et al. | An FPGA decision tree classifier to supervise a communication SoC | |
| Bariya et al. | Data-driven topology estimation with limited sensors in radial distribution feeders | |
| CN116298670A (en) | Intelligent fault positioning method and system suitable for multi-branch distribution line | |
| Su et al. | PSR-LSTM model for weak pulse signal detection | |
| Othman et al. | Online fault detection for power system using wavelet and PNN | |
| Zhou et al. | Convolutional Neural Network and Data Augmentation Method for Electricity Theft Detection | |
| Liu et al. | State of health estimation of lithium-ion batteries based on multi-feature extraction and temporal convolutional network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170419 Termination date: 20210509 |