WO2018120077A1

WO2018120077A1 - Three-level inverter fault diagnosis method based on empirical mode decomposition and decision tree rvm

Info

Publication number: WO2018120077A1
Application number: PCT/CN2016/113643
Authority: WO
Inventors: 陶洪峰; 周超超; 童亚军; 刘艳; 沈建强
Original assignee: 江南大学
Priority date: 2016-12-26
Filing date: 2016-12-30
Publication date: 2018-07-05
Also published as: NL2020015B1; NL2020015A; CN106682303B; CN106682303A

Abstract

A diode neutral point clamped three-level inverter fault diagnosis method based on empirical mode decomposition and a decision tree RVM comprises: for a diode neutral point clamped three-level inverter fault diagnosis problem in a photovoltaic power generation system, firstly analyzing the operation condition of an inverter main circuit and performing fault classification; then extracting each signal component by means of an empirical mode decomposition method by using upper, middle and lower bridge arm voltages as measurement signals, and then calculating the corresponding energy and energy entropy, thus generating a decision tree RVM classification model using a particle swarm clustering algorithm, and finally realizing the fault diagnosis of the photovoltaic diode neutral point clamped three-level inverter. In the diode neutral point clamped three-level inverter fault diagnosis method based on empirical mode decomposition and a decision tree RVM, there is no need to set parameters, the number of classification models is relatively small, the operation efficiency is high and the diagnosis precision is high, and the robustness is strong.

Description

A Three-level Inverter Fault Diagnosis Method Based on Empirical Mode Decomposition and Decision Tree RVM

Technical field

The invention relates to the field of power electronic device fault diagnosis, in particular to a diode midpoint clamp type three-level inverter fault diagnosis method based on empirical mode decomposition and decision tree RVM.

Background technique

With the advancement of photovoltaic power generation technology and the scale of photovoltaic power generation grid-connected operation, the optimization, improvement and operation cost of photovoltaic power generation system have seriously restricted the development of photovoltaic power generation industry. Among them, although the cost of the photovoltaic inverter is not high, the inverter is always due to the fragility of the power electronic device used in the inverter circuit, the complicated control of the inverter circuit, the frequent on-off control, and the harsh external environment. The weak link that is prone to failure in the whole system is prone to faults such as overvoltage, overcurrent, power tube short circuit and open circuit, and these situations are seriously related to the safe operation of the entire photovoltaic power generation system. In order to prevent more serious accidents caused by faults, timely detection of faulty equipment and determination of the cause and location of equipment failures are not only conducive to reducing economic losses, but also conducive to the development of maintenance personnel. At the same time, it is possible to realize stable, efficient and safe operation of the photovoltaic power generation system, which is of great significance for promoting the large-scale development of photovoltaic power generation in China.

With the gradual application of different types and structures of inverters in photovoltaic power generation systems, the reliability, stability and maintainability of their work become more and more important. According to the data, 38% of all grid-connected inverter faults are caused by power tube damage in the inverter main circuit. Common inverter faults mainly include short-circuit faults and open-circuit faults. Short-circuit faults are usually protected by hardware circuits in microseconds. However, most short-circuit faults do not immediately cause system shutdown, but cause other devices. The second failure eventually caused the system to fail. When the inverter fails, the physical quantities such as voltage and current in the circuit change with respect to the normal state. Therefore, the inverter power tube open circuit fault diagnosis method can be divided into two types according to different detection signals: current and voltage. Fault diagnosis method. The current-based fault diagnosis method does not require an additional sensor, but in many cases, the current is related to the load. When it is no-load or light-loaded, the current method has a low diagnostic accuracy. The voltage method investigates the fault of the inverter phase voltage, line voltage or bridge arm voltage from the normal state, and needs to increase the sensor. However, this also has many advantages: stronger robustness to noise and load, and false positive rate. Lower and less diagnostic time.

In the fault diagnosis of power electronic devices, the selection and extraction of fault feature vectors has always been diagnosed. The key is that it directly affects the accuracy of the fault diagnosis results. Photovoltaic three-level inverters have many switching devices, and the types of faults are complicated. The large number of signals measured are unsteady signals. Therefore, it is necessary to adopt a feature extraction method suitable for processing non-stationary signals in the process of fault diagnosis. The empirical mode decomposition method is just such a method.

On the other hand, designing a well-structured classifier for state recognition is another key step in fault diagnosis. At present, the pattern recognition methods for fault detection and diagnosis are mainly statistical pattern recognition and neural network recognition. At the same time, intelligent diagnosis algorithms such as extreme learning machines and support vector machines also show great application potential. However, traditional statistical pattern recognition methods have their own limitations. There are many important problems in neural network technology that have not been solved theoretically. Extreme learning machines require a large number of samples for training. Support vector machines are suitable for solving small samples and non- Linear and high-dimensional pattern recognition, but there are still many parameters that need to be selected by experience. Parameters such as penalty coefficient and kernel function radius have a great influence on the diagnostic accuracy. The related vector machine (RVM) is a learning machine based on the Bayesian framework. It does not need to set the penalty factor, and there is no case that the support vector machine has caused learning due to improper setting parameters. And the algorithm can also solve the pattern recognition problem of high dimensional, nonlinear and small samples, and has a good application prospect.

Summary of the invention

It is an object of the present invention to provide a diode midpoint clamped three-level inverter fault diagnosis method based on empirical mode decomposition and decision tree RVM.

A fault diagnosis method for diode midpoint clamped three-level inverter based on empirical mode decomposition and decision tree RVM, including: constructing diode midpoint clamped three-level inverter circuit model and classifying faults Extracting the open-circuit fault eigenvector of the three-level inverter circuit; constructing a three-level inverter fault diagnosis decision tree; constructing a correlation vector machine fault classification decision tree model, and finally realizing a photovoltaic diode midpoint clamped three-level inverter Troubleshooting.

The first step is to establish a model of the photovoltaic diode midpoint clamped three-level inverter circuit and classify the fault. The main circuit of the three-level inverter consists of three-phase bridge arms with two clamp capacitors, twelve main switch tubes, twelve freewheeling diodes and six midpoint clamp diodes. The three-level inverter circuit has two remarkable features: the output voltage waveform synthesized by multiple levels, the harmonic content is greatly reduced compared with the conventional two levels, and the output voltage output waveform is improved; the voltage rating of the switching tube The value is half of the voltage on the DC bus, allowing the low voltage switch to be used in high voltage converters.

Since the three phases of the photovoltaic diode midpoint clamped three-level inverter circuit are symmetrical, the phase A is taken as an example, and the others are similar. Mainly discuss open circuit faults of three-level inverter circuit faults, including IGBT turn-on Road, series fuse blown and trigger pulse loss fault, while also considering the midpoint clamp diode open circuit, the fault classification is as follows, a total of four categories of thirteen categories.

1) The system is fault-free and has a small class.

2) Single clamp diodes are open circuited in two sub-categories.

3) A single power device is open, that is, open in any of the four power tubes, for a total of four sub-categories.

4) Two devices are open, there are two cases: First, the two power tubes that are open are not in the same bridge arm. This situation can be attributed to a single device fault on different bridge arms. Refer to the third single power device for open circuit. Fault classification; Second, the two switching tubes of the fault are in the same bridge arm, that is, any two of the four power tubes are open, and there are six sub-categories.

The second step: extracting the open circuit fault feature vector of the three-level inverter circuit. During the analysis of the signal, the energy distributed over time scales and time scales is the two most important parameters of the signal. When the power circuit of the inverter circuit is open, its voltage signal has a large difference in energy of signals in the same frequency band compared with the voltage signal of the normal system. The energy of each frequency component of the signal contains rich fault information, and the change of energy of one or several frequency components represents a fault, so fault analysis can be performed according to the change of energy of each frequency band.

Modeling the main circuit of a diode midpoint clamped three-level inverter using space vector pulse width modulation (SVPWM) and neutral point potential control, and modeling the EMD of the bridge arm voltage when various faults occur. Decompose, select the first n IMF components and residual amount, and then calculate the energy of each IMF component and residual amount. Let the energy E _{i of} each component

Where c _i,k (i=1,2,...,n+1;k=1,2,...,J) are the amplitudes of the first n IMF components and the J discrete points of the residual amount. After obtaining the energy of each bridge arm voltage, a feature vector can be constructed, wherein the feature vector T ₁ is:

T ₁ =[E ₀ E ₁ ... E _n+1 ] (2) Considering that the value of energy tends to be large, the normalization process is improved for the convenience of subsequent classification.

At the same time, based on the IMF energy, calculate the corresponding IMF energy entropy

Where p _i =E _i /E _z is the energy of the i-th component as a percentage of the total signal energy

Based on the above parameters, the fault feature vector is defined as:

T ₁ '=[E ₀ /E E ₁ /E ... E _n+1 /E H ₁ ] (6)

Using the same method to process the upper and lower arms, the eigenvectors T ₂ ′ and T ₃ ′ can be obtained respectively, and the fault eigenvectors are defined as:

T=[T ₁ ' T ₂ ' T ₃ '] (7)

The bridge arm voltage under each fault condition is extracted according to the above process, and finally the data sample is constructed.

The third step: construct a particle swarm clustering fault diagnosis decision tree. As mentioned earlier, there are 13 types of faults in a three-level inverter. To build a decision tree, a clustering algorithm is needed to continuously divide the fault into two categories until the subclass contains only one sample type. Specifically:

The initial class is processed first, and all training samples are taken as the initial class. The clustering algorithm is used to divide it into two subclasses; then the subclass is judged. If the subclass contains only one sample type, the algorithm ends, otherwise the clustering is continued. The algorithm performs clustering until all subclasses contain only one sample type.

The key to constructing a decision tree lies in the choice of clustering algorithm. Here, the particle swarm clustering algorithm is adopted. The particle swarm clustering algorithm needs to be initialized first, randomly initialize the particle swarm, set relevant parameters, and then perform random classification. Randomly classify each sample, calculate parameters such as fitness and cluster center, and set the initial velocity of the particle to zero. In this way, the individual particle optimal position p _id and the global optimal position p _gd can be obtained from the initial particle group. According to the clustering center coding of particles, according to the nearest neighbor rule, the clustering of each sample is determined, and according to the new clustering, the new clustering center is calculated and the fitness is updated. Again fitness comparison, if it is better than the optimal location of p _id, update p _id; if it is better than the global optimum position p _gd, update p _gd. If the maximum number of iterations is reached, the algorithm ends, otherwise iteration continues.

By summarizing the results of the clustering, the structure of the fault diagnosis decision tree can be constructed to provide a basis for the training objects of the RVM.

The fourth step: construct a correlation vector machine fault classification decision tree model. The data samples are divided into training sets and test sets according to a ratio of 3:7, and the training set is trained according to the decision tree structure obtained in the previous step. After the training is completed, the test set is used to test, and the diagnostic accuracy, average training time and average test time are obtained, and finally the fault diagnosis of the photovoltaic diode midpoint clamp type three-level inverter is realized.

The beneficial effects of the invention are:

1) The three-level inverter fault diagnosis method based on empirical mode decomposition and decision tree RVM proposed by the present invention is based on the idea of data driving, combining empirical mode decomposition, particle swarm clustering and correlation vector machine algorithm. To achieve photovoltaic inverters, especially photovoltaic diode midpoint clamped three-level inverter Fault diagnosis.

2) The present invention performs feature extraction by an empirical mode decomposition algorithm, which is an adaptive algorithm, which is very suitable for analyzing non-stationary and nonlinear signals. At the same time, it is not necessary to select the parameter values according to experience like wavelet analysis, and the fault information can be characterized by extracting the energy of each IMF component and the energy entropy of the signal as the fault feature vector.

3) The invention adopts the fault diagnosis model structure of the decision tree RVM. The decision tree structure only needs to construct fewer classification models to complete the fault diagnosis task, and the RVM algorithm has fewer vectors and tests than the SVM algorithm. The time is shorter, the sparsity is stronger, and the training samples and the features with less features are more robust and do not need to set parameters.

DRAWINGS

Figure 1 shows the fault diagnosis process of the diode midpoint clamped three-level inverter.

Figure 2 shows the main circuit topology of the diode midpoint clamped three-level inverter

Figure 3 shows the A-phase topology of the inverter main circuit

Figure 4 shows the bridge voltage for a single device failure.

Figure 5 shows the bridge voltage when two devices are simultaneously open.

Figure 6 shows the bridge voltage when a single clamp diode is open.

Figure 7 shows the EMD decomposition result when the inverter is normal.

Figure 8 is a fault eigenvector histogram of the inverter when it is normal.

Figure 9 is a decision tree structure diagram after clustering

detailed description

The invention will be further described below in conjunction with the accompanying drawings.

The three-level inverter fault diagnosis flowchart based on the empirical mode decomposition and decision tree RVM of the present invention is shown in FIG. 1. The specific implementation of the method of the present invention includes the following steps:

Figure 2 shows the topology diagram of the main circuit of the diode midpoint clamped three-level inverter. To simplify the analysis, only the operating state of the A phase in the inverter inverter state is studied. The circuit topology is shown in Figure 3. Show. In the figure, the solid line is the positive direction of the current, and the broken line is the negative direction of the current. After ignoring the power-on voltage drop of the power device, the potential of point A of P state is always equal to the potential of point P, and the potential of point A of point O is always equal to the potential of point O, N state A The point potential is always equal to the N point potential.

According to the topology structure, the faults are classified into four categories and thirteen subclasses, that is, the fault classification of the diode midpoint clamp type three-level inverter.

1) The inverter circuit has no faults, and the power devices work normally, a total of a small class.

2) Open one of the single clamp diodes VD _a5 and VD _a6 in two sub-categories.

2) Single device open circuit, that is, power tubes S _a1 , S _a2 , S _a3 , S _a4 , a total of four sub-categories.

3) The two devices are open. There are two subclasses of this type. One is that the two power tubes that are open are not in the same bridge arm. You can refer to the third type of open circuit and do not count the fault classification. The second is the two powers of the open circuit. The tubes are on the same bridge arm, ie the power tubes (S _a1 , S _a2 ), (S _a1 , S _a3 ), (S _a1 , S _a4 ), (S _a2 , S _a3 ), (S _a2 , S _a4 ) or S _a3 , S _a4 ) Any group of open circuits, a total of six sub-categories. In summary, the fault classification and corresponding labels are shown in Table 1.

Table 1 Fault classification

A diode midpoint clamped three-phase three-level inverter model is established. The SVPWM control cooperative midpoint potential control technology is used to control the three-phase operating state of the inverter, and the three-level inverter is driven to complete the inverter operation. Selecting the bridge arm voltage as the research object, the bridge arm voltage under various fault conditions can be obtained as shown in Fig. 4 and Fig. 5. Compared with Fig. 4(c) and Fig. 5(a), it can be found that S _a2 and (S _a1 , S _a2 ) The level logic of the two is the same, which is caused by the structure of the circuit itself, so it is necessary to introduce a new measuring point, that is, the voltage of the upper arm, as shown in Fig. 6. EMD decomposition is performed on each bridge arm voltage, and each bridge arm voltage is decomposed into 4 IMF components and 1 residual amount. The EMD decomposition result of the bridge arm voltage under normal conditions is shown in FIG. 7 . After decomposing, the energy of the signal is calculated. After the dimension is unified, the energy entropy is calculated, and finally the fault eigenvector of the single bridge arm voltage is constructed. The single fault feature vector is integrated, and the overall fault feature vector is constructed in the order of middle, upper and lower, and the data samples are constructed according to different fault types. The histogram of the fault eigenvector when the inverter is working normally is shown in Fig. 8.

As mentioned above, the particle swarm clustering algorithm is used to divide the fault samples. For example, the result of the first partition is that the data samples with

labels

0, 1, 4, 5, 6, and 14 are classified into one category; Data samples for 2, 3, 12, 13, 23, 24, and 34 are classified as another category. The structure of the first layer of the decision tree and the training samples of the corresponding classification model RVM1 are also determined, and so on. After the division is completed, the decision tree is constructed. The final result is shown in Figure 9. It can be seen from the figure that for the 13 fault classification problems, only 12 classification models need to be constructed using the decision tree structure, and if a one-to-one structure is adopted, 78 classification models need to be constructed. At the same time, in the test model, the decision tree structure only needs to perform 2 to 6 classification operations, and the one-to-one structure still needs 78 classification operations. In summary, the decision tree structure will undoubtedly be greatly reduced. Reduce the number of models to build, reduce computation time, and improve computational efficiency.

The data samples are divided into training and test sets at a ratio of 3:7. According to the constructed decision tree structure, RVM1~RVM12 are trained respectively, and a total of 12 correlation vector machine classification models are constructed. In order to verify the anti-interference ability of the algorithm, the original data is compared with the white noise of the signal amplitude of 10% and 15%, and the BP neural network (BPNN) and the extreme learning machine are also compared horizontally. , ELM), one-to-one structure (1vs.1) correlation vector machine and decision tree support vector machine (DT-SVM) training, test time and diagnostic accuracy, the final fault diagnosis results are summarized as Table 2 and Table 3 are shown.

Table 2 Troubleshooting results (10% white noise)

Table 3 Troubleshooting results (15% white noise)

The above embodiments are merely illustrative of the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms may be made based on the above description. Change or change.

Claims

The object of the present invention is to provide a diode midpoint clamp type three-level inverter fault diagnosis method based on empirical mode decomposition and decision tree RVM;

A fault diagnosis method for diode midpoint clamped three-level inverter based on empirical mode decomposition and decision tree RVM, including: constructing diode midpoint clamped three-level inverter circuit model and classifying faults Extracting the open-circuit fault eigenvector of the three-level inverter circuit; constructing a three-level inverter fault diagnosis decision tree; constructing a correlation vector machine fault classification decision tree model, and finally realizing a photovoltaic diode midpoint clamped three-level inverter Fault diagnosis;

The first step is to establish a model of the photovoltaic diode midpoint clamped three-level inverter circuit and classify the fault; the three-level inverter main circuit is composed of three-phase bridge arms, with two clamp capacitors and twelve One main switch tube, twelve freewheeling diodes and six midpoint clamp diodes; the three-level inverter circuit has two distinguishing features: an output voltage waveform synthesized by multiple levels, and a conventional two-level phase Ratio, the harmonic content is greatly reduced, and the output voltage output waveform is improved; the voltage rating of the switching tube is half of the voltage on the DC bus, so that the low voltage switch tube can be applied to the high voltage converter;

Since the three-phase of the photovoltaic diode midpoint clamped three-level inverter circuit is symmetrical, the phase A is taken as an example, and the other phases are similar; the open circuit fault of the three-level inverter circuit fault is mainly discussed, including the IGBT open circuit, The series fuse blows and triggers the pulse loss fault, and also considers the case of the midpoint clamp diode open circuit. The faults are classified as follows, and there are four categories and thirteen subclasses;

1) The system is fault-free and has a small class;

2) A single clamp diode is open, in two sub-categories;

3) A single power device is open, that is, open in any of the four power tubes, a total of four sub-categories;

4) Two devices are open, there are two cases: First, the two power tubes that are open are not in the same bridge arm. This situation can be attributed to a single device fault on different bridge arms. Refer to the third single power device for open circuit. Fault classification; second, the two switching tubes of the fault are in the same bridge arm, that is, any two of the four power tubes are open, and there are six sub-categories;

The second step: extracting the open-circuit fault feature vector of the three-level inverter circuit; in the signal analysis process, the time scale and the energy distributed with the time scale are the two most important parameters of the signal; when the inverter circuit power tube is open, Compared with the voltage signal of the normal system, the voltage signal of the same frequency band has a large difference; the energy of each frequency component of the signal contains rich fault information, and the energy of one or several frequency components is changed. Represents a fault, so it can be based on the change of energy in each frequency band. Line failure analysis;

Modeling the main circuit of a diode midpoint clamped three-level inverter using space vector pulse width modulation (SVPWM) and neutral point potential control, and modeling the EMD of the bridge arm voltage when various faults occur. Decompose, select the first n IMF components and residual amount, and then calculate the energy of each IMF component and residual amount; set the energy E i of each component

Where c i,k (i=1,2,...,n+1;k=1,2,...,J) are the amplitudes of the first n IMF components and the J discrete points of the residual amount; After the energy of the bridge arm voltage, a feature vector can be constructed, wherein the feature vector T 1 is:

T 1 =[E 0 E 1 ... E n+1 ] (2)

Considering that the value of energy tends to be large, the normalization process is improved for the convenience of subsequent classification.

At the same time, based on the IMF energy, calculate the corresponding IMF energy entropy

Where p i =E i /E z is the energy of the i-th component as a percentage of the total signal energy

Based on the above parameters, the fault feature vector is defined as:

T 1 '=[E 0 /E E 1 /E ... E n+1 /E H 1 ] (6)

Using the same method to process the upper and lower arms, the eigenvectors T 2 ′ and T 3 ′ can be obtained respectively, and the fault eigenvectors are defined as:

T=[T 1 ' T 2 ' T 3 '] (7)

The bridge arm voltage in each fault condition is extracted according to the above process, and finally the data sample is constructed;

The third step: constructing a particle swarm clustering fault diagnosis decision tree; as mentioned above, there are 13 fault types in the three-level inverter. To build a decision tree, a clustering algorithm is needed to continuously divide the fault into two. Class, until the subclass contains only one sample type, which is specifically:

The initial class is processed first, and all training samples are taken as the initial class. The clustering algorithm is used to divide it into two subclasses; then the subclass is judged. If the subclass contains only one sample type, the algorithm ends, otherwise the clustering is continued. The algorithm performs clustering until all subclasses contain only one sample type;

The key to constructing the decision tree lies in the choice of clustering algorithm. Particle clustering algorithm is used here. The particle swarm clustering algorithm needs to be initialized first, randomly initialize the particle swarm, set relevant parameters, and then perform random classification to randomly sample each sample. Classification, calculation of fitness, clustering center and other parameters, the initial velocity of the particle is set to zero; thus, the individual particle optimal position p id and the global optimal position p gd can be obtained according to the initial particle group; Central coding, according to the nearest neighbor rule, determine the clustering of each sample, and calculate the new cluster center according to the new clustering, update the fitness; once again compare the fitness, if it is better than the individual optimal position p id , update p id ; if it is better than the global optimal position p gd , update p gd ; if the maximum number of iterations is reached, the algorithm ends, otherwise iteratively continues;

By summarizing the results of the clustering, the structure of the fault diagnosis decision tree can be constructed to provide a basis for the training objects of the RVM.

The fourth step: construct the correlation vector machine fault classification decision tree model; divide the data sample into training set and test set according to the ratio of 3:7, and train the training set according to the decision tree structure obtained in the previous step; after the training is completed, use the test The set is tested and the indicators such as diagnostic accuracy, average training time and average test time are obtained, and finally the fault diagnosis of the photovoltaic diode midpoint clamped three-level inverter is realized.