A three-level inverter fault diagnosis method and system

Abstract

The application relates to a three-level inverter fault diagnosis method and system. The method comprises the following steps: constructing a current state observer model according to a three-level inverter state space model; obtaining a current residual signal through the current state observer model, and establishing a current residual information table; constructing a sample set according to the current residual signal; the sample set comprises a training sample set and a test sample set; the training sample set is pretreated to generate current residual data; the current residual data is subjected to feature extraction according to the current residual information table to generate a current residual signal fault feature data set; a three-level inverter fault diagnosis model is constructed according to the current residual signal fault feature data set and a random forest model; and fault recognition and positioning are performed according to the three-level inverter fault diagnosis model. The application can complete fault diagnosis recognition and positioning of single tubes, homophase multiple tubes and heterophase multiple tubes, and has faster diagnosis speed and higher accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of inverter fault diagnosis, in particular to a three-level inverter fault diagnosis method and system. BACKGROUND

[0002] With the large-scale development of new energy, the grid connection of new energy represented by wind power and photovoltaic power generation needs to be realized through inverters. Due to the important role played by inverters in the field of new energy, the demand for high-power inverters is increasing. Compared with two-level inverters, three-level inverters have the advantages of flexible control, high voltage resistance, small harmonics, and output current waveform tending to be sinusoidal. Especially the neutral point clamped (NPC) type inverter is widely used due to its simple control and stable performance. However, due to the variable and harsh working conditions of three-level inverters, inverter system failures occur frequently. Among all the faults of inverters, the probability of open circuit fault of insulated gate bipolar transistor (IGBT) is the highest. If the fault source cannot be found and accurately processed in time, not only economic losses will be caused, but also serious casualties may occur. Therefore, the research on three-level inverter fault diagnosis method plays an important role in production safety.

[0003] Nowadays, many experts and scholars have studied the inverter fault diagnosis technology. The inverter fault diagnosis methods are mainly divided into three categories: model-based method, signal-based method and data-driven method. The model-based method makes good use of system information, and the fault diagnosis result is reliable, but it has a high dependence on system model. The more accurate the system model is, the higher the fault diagnosis accuracy is. However, in actual situations, for systems that cannot be accurately modeled and are difficult to model, the fault accuracy cannot be guaranteed. The signal-based method uses prior knowledge of signals for fault diagnosis, which does not require an accurate system model, but it is highly dependent on the prior knowledge of the system, and the diagnosis result is easily affected by noise. The data-driven method can perform fault diagnosis without prior knowledge and accurate system model, but it needs a large amount of high-quality sample data for training. The more sample data, the higher the accuracy. It is difficult to establish a sample set covering all fault types. For fault types not involved in the fault sample, the diagnosis method is difficult to give accurate results, and there is also the problem of uneven distribution of fault types in sample data.

[0004] Therefore, the application provides a three-level inverter fault diagnosis method based on model and data hybrid driving. The model-based and data-based fault diagnosis methods are combined to fully utilize the advantages of both. By designing a current observer-based NPC three-level inverter model, a current residual sample data set containing different fault types is obtained, and the current residual sample data set is processed, a random forest classifier is used, a fault diagnosis model is constructed, and three-level inverter fault diagnosis is completed. Compared with the model-based and data-driven single fault diagnosis method, the model and data hybrid driving-based fault diagnosis method has good robustness, faster training speed, lower computational burden, and fewer data samples used, and further improves the accuracy of fault diagnosis. SUMMARY

[0005] To this end, the technical problem to be solved by the application is to overcome the high dependence on system models in the prior art. The more accurate the system model is, the higher the fault diagnosis accuracy is. However, in actual situations, for systems that cannot be accurately modeled and are difficult to model, the fault accuracy is difficult to guarantee.

[0006] To solve the above technical problems, the first aspect of the application provides a three-level inverter fault diagnosis method, which comprises:

[0007] a current state observer model is constructed according to a three-level inverter state space model;

[0008] a current residual signal is obtained through the current state observer model, and a current residual information table is established;

[0009] a sample set is constructed according to the current residual signal; the sample set comprises a training sample set and a test sample set;

[0010] the training sample set is preprocessed to generate current residual data;

[0011] the current residual data is feature extracted according to the current residual information table to generate a current residual signal fault feature data set;

[0012] a three-level inverter fault diagnosis model is constructed according to the current residual signal fault feature data set and a random forest model;

[0013] fault recognition and positioning are performed according to the three-level inverter fault diagnosis model.

[0014] In an embodiment of the application, the step of obtaining a current residual signal through the current state observer model and establishing a current residual information table comprises:

[0015] a current observation value is obtained through the current state observer model;

[0016] generating a current residual value according to the difference between the current observation value and the current actual value;

[0017] generating a current residual information table according to the current residual value.

[0018] In an embodiment of the present application, the expression of the three-level inverter state space model is:

[0019]

[0020] wherein A, B, C are known 3-order real matrices, the main diagonal elements of A are -R / L, and the rest elements are 0, the main diagonal elements of B are 1 / L, and the rest elements are 0, C is a unit matrix; x(t) is a state vector, and u(t) is a control input vector.

[0021] In an embodiment of the present application, the expression of x(t), u(t) is:

[0022]

[0023] wherein i a (t), i b (t) and i c (t) are A-phase, B-phase and C-phase load output phase currents, which constitute the state vector x(t). u an (t), u bn (t) and u cn (t) are A-phase, B-phase and C-phase load output phase voltages, which constitute the input vector u(t). t is time.

[0024] In an embodiment of the present application, the expression of the current state observer model is:

[0025]

[0026]

[0027] wherein x(t) is a state vector, is an observation value of x(t), H is an error compensation matrix, A, C are matrices, the main diagonal elements of A are -R / L, and the rest elements are 0, C is a unit matrix, y(t) is an output vector, is an observation value of y(t). is a derivative with respect to time, is an observation value of y(t).

[0028] In an embodiment of the present application, the step of pre-processing the training sample set comprises:

[0029] The training sample set is preprocessed according to a wavelet transform and a Park vector module normalization algorithm.

[0030] The second aspect of the present application provides a three-level inverter fault diagnosis system, comprising a data acquisition module, a first calculation module and a second calculation module.

[0031] The data acquisition module is configured to: construct a current state observer model according to a three-level inverter state space model; acquire a current residual signal through the current state observer model and establish a current residual information table; and construct a sample set according to the current residual signal; the sample set comprises a training sample set and a test sample set.

[0032] The first calculation module is configured to: preprocess the training sample set to generate current residual data; and extract features from the current residual data according to the current residual information table to generate a current residual signal fault feature data set.

[0033] The second calculation module is configured to: construct a three-level inverter fault diagnosis model according to the current residual signal fault feature data set and a random forest model; and perform fault identification and positioning according to the three-level inverter fault diagnosis model.

[0034] The third aspect of the present application provides an electronic device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method described in the first aspect or any possible implementation manner of the first aspect when executing the program.

[0035] The fourth aspect of the present application provides a non-transitory computer readable storage medium having a computer program stored thereon, wherein the computer program is executable by a processor to implement the method described in the first aspect or any possible implementation manner of the first aspect.

[0036] The above technical solutions of the present application have the following advantages compared with the prior art:

[0037] This invention discloses a fault diagnosis method and system for a three-level inverter. It employs a hybrid model- and data-driven approach to diagnose three-level inverter faults. The current residual signal generated by the current state observer model serves as the original dataset, and its feature values ​​are used to train the model. This combines data-driven and model-driven methods, avoiding the need for manually determining thresholds compared to model-based methods, thus improving the universality of the diagnostic method. Compared to data-driven methods, it uses fewer data samples and has higher diagnostic efficiency. Wavelet transform and Park vector modulus normalization algorithms are used to preprocess the training sample set, and feature extraction is performed using a residual signal fault feature table. This results in a more distinctive data feature distribution, improving the accuracy of fault diagnosis. The residual data generated by the current state observer model is used to train a random forest model, which, compared to traditional data-driven methods, offers advantages such as higher model accuracy, less data usage, and shorter training time, significantly improving fault diagnosis accuracy. It also exhibits anti-interference and robustness. This application can complete fault diagnosis, identification, and location for single-transistor, in-phase multi-transistor, and out-of-phase multi-transistor systems, with faster diagnosis speed and higher accuracy. Attached Figure Description

[0038] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0039] Figure 1 This is a flowchart of a three-level inverter fault diagnosis method provided by the present invention;

[0040] Figure 2 This invention provides a method for diagnosing faults in a three-level inverter and a topology diagram of a three-level inverter in the system.

[0041] Figure 3 This is a functional block diagram of a three-level inverter fault diagnosis system provided by the present invention. Detailed Implementation

[0042] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0043] Reference Figure 1 and Figure 2 As shown, in a first aspect, the present invention provides a method for diagnosing faults in a three-level inverter, the method comprising:

[0044] S100: Construct a current state observer model based on the state-space model of a three-level inverter;

[0045] In step S100, the expression of the three-level inverter state space model is:

[0046]

[0047] Wherein, A, B, C are known 3-order real matrices, the main diagonal elements of A are -R / L, and the rest elements are 0, the main diagonal elements of B are 1 / L, and the rest elements are 0, C is a unit matrix; x(t) is a state vector, and u(t) is a control input vector. The expressions of x(t) and u(t) are:

[0048]

[0049] Wherein, i a (t), i b (t) and i c (t) are A-phase, B-phase and C-phase load output phase currents, which constitute the state vector x(t). u an (t), u bn (t) and u cn (t) are A-phase, B-phase and C-phase load output phase voltages, which constitute the input vector u(t). t is time.

[0050] The expression of the current state observer model is:

[0051]

[0052] Wherein, x(t) is a state vector, is the observation value of x(t), H is an error compensation matrix, A and C are matrices, the main diagonal elements of A are -R / L, and the rest elements are 0, C is a unit matrix, y(t) is an output vector, is the observation value of y(t). The derivative of t is is the observation value of y(t).

[0053] In actual application scenarios, based on the requirements of fault diagnosis, a current state observer model is designed according to the NPC three-level inverter state space model, wherein the expression of the NPC three-level inverter state space model is shown in formula (1), the expressions of x(t) and u(t) in formula (1) can be calculated by formula (2), and the expression of the designed current state observer model is shown in formula (3).

[0054] The NPC three-level inverter is composed of A-phase, B-phase and C-phase bridge arms, as shown in Figure 2The topology of a three-level inverter is shown. Ud is a DC side voltage source, C1 and C2 are two equivalent capacitors on the DC side, which have the functions of buffering energy and voltage division. The inverter has three-phase bridge arms, each bridge arm has four IGBT power switches, a total of 12 power switches, i.e. A phase is Sa1, Sa2, Sa3, Sa4, B phase is Sb1, Sb2, Sb3, Sb4, C phase is Sc1, Sc2, Sc3, Sc4. Each bridge arm has four freewheeling diodes, a total of 12 freewheeling diodes, i.e. A phase is Da1, Da2, Da3, Da4, B phase is Db1, Db2, Db3, Db4, C phase is Dc1, Dc2, Dc3, Dc4. Each bridge arm has two clamping diodes, a total of six clamping diodes, i.e. A phase is Da5, Da6, B phase is Db5, Db6, C phase is Dc5, Dc6, which clamps the DC side voltage to make the power tube only withstand Ud / 2 voltage, prolong the service life of the switch device, and at the same time provide a path for the current in the O state of the inverter circuit. Use IA, IB, IC to represent three-phase load current, and the load is a resistive and inductive load.

[0055] S200: Obtain a current residual error signal through the current state observer model, and establish a current residual error information table;

[0056] In step S200, the step of obtaining a current residual error signal through the current state observer model and establishing a current residual error information table includes: obtaining a current observation value through the current state observer model; generating a current residual error value according to the difference between the current observation value and the actual current value; and generating a current residual error information table according to the current residual error value.

[0057] In actual application scenarios, the current observation value is obtained according to the current state observer model, the current observation value is subtracted from the actual current value to generate a current residual error value, and a current residual error information table is established according to all current residual error values. When the NPC three-level inverter fails, its operating state will deviate compared to the normal situation, and different power switch open circuit faults will have different deviations. Therefore, the difference between the current observation value of the current state observer model and the actual current value, i.e. the current residual error value, can be compared to obtain information about IGBT faults at different positions. The expression of the current residual error value is shown in formula (4).

[0058]

[0059] Where Δx represents the difference of three-phase current, and Δu represents the difference of three-phase voltage.

[0060] Table 1: Current residual error information table.

[0061] State Fault switch tube A, B, C three-phase current residual relationship Normal operation No Δi b = Δi a = Δi c = 0]]> Single tube fault Sb1 Δi b = -2Δi a = -2Δi c ≤ 0]]> Single tube fault Sb2 Δi b = -2Δi a = -2Δi c ≤ 0 Single tube fault Sb3 Δi b = -2Δi a = -2Δi c ≥ 0 Single tube fault Sb4 Δi b = -2Δi a = -2Δi c ≥ 0]]>

[0062] According to the expression of the current residual value, assuming that the B-phase single tube fails, the B-phase single tube fault current residual information table is shown in Table 1. According to the current residual information table, when the NPC three-level inverter works normally, the three-phase current residual values are all zero. However, when the three-phase residual values are not zero, it is known that the NPC three-level inverter power switch tube open circuit fault occurs. According to the symbol of the current residual value and the three-phase current residual relationship, the fault phase and the fault power switch tube can be accurately located. Thus, the single tube open circuit fault diagnosis of the inverter is completed.

[0063] S300: constructing a sample set according to the current residual signal; the sample set includes a training sample set and a test sample set;

[0064] In step S300, the IGBT pulse signal is given, and the current residual signal is collected from the current state observer model. The sample set is constructed according to the collected current residual signal, wherein the sample set includes a training sample set and a test sample set.

[0065] S400: preprocessing the training sample set to generate current residual data;

[0066] In step S400, the training sample set is preprocessed according to the wavelet transform and Park vector module normalization algorithm.

[0067] In actual application scenarios, the training sample set in the constructed sample set is preprocessed by the wavelet transform and Park vector module normalization algorithm to generate current residual data. When preprocessing the training sample set, the three-phase current residual values (Δi a , Δi b , Δi c ) need to be converted into two-phase current residual values (Δi d , Δi q ), and the expression is shown in equation (5).

[0068]

[0069] The normalized three-phase current residual i kpark (k is a, b, c), and the expression is shown in equation (6).

[0070]

[0071] wherein, is the Park vector module, and the expression is shown in equation (7).

[0072]

[0073] S500: feature extraction is performed on the current residual data according to the current residual information table, and a current residual signal fault feature dataset is generated;

[0074] In step S500, the current residual data generated after preprocessing is feature extracted according to the current residual information table, and a current residual signal fault feature dataset is generated according to the extracted features. When the current residual data is feature extracted, the extracted features can be summarized into a current residual signal fault feature table. The current residual signal fault feature table is shown in Table 2.

[0075] Table 2: Current residual signal fault feature table.

[0076]

[0077] where x(i) is the normalized three-phase current residual, n is the number of current residual samples, and i is the sample number of the current residual.

[0078] S600: a three-level inverter fault diagnosis model is constructed according to the current residual signal fault feature dataset and the random forest model;

[0079] In step S600, since the training sample set X and the test sample set X' are constructed, after feature extraction, each sample in the training sample set X has 36 current residual signal feature values (12 feature values for each phase current residual, and a total of 3 phases). Then, a three-level inverter fault diagnosis model is constructed according to the current residual signal fault feature dataset and the random forest model. The random forest model is a random sampling of multiple decision trees with replay, and the final classification result is determined by the classification results of the multiple decision trees. Since the random forest is a basic overall classifier with multiple decision trees, a feature subset is randomly extracted in the random forest. Specifically, m samples are randomly extracted from the original current residual signal, and then n feature values are randomly extracted from the signal features, and the best feature is selected for node partitioning. In the new sample set (with m samples and n feature values), the probability step Gini of node m is shown in formula (8).

[0080]

[0081] where n is the number of feature values, p n is the probability of drawing n classes. Select the best feature and the partition point. Based on the Gini maximization principle, select a feature and partition the feature node into left and right child nodes. Recursion is performed on the right and left child nodes, and the process is repeated until the training sample set is preprocessed to generate current residual data, thereby constructing a decision tree with n samples and k features.

[0082] S700: fault identification and positioning according to the three-level inverter fault diagnosis model.

[0083] In step S700, the test sample set containing 36 current residual characteristic values is put into the trained three-level inverter fault diagnosis model for testing, the diagnosis result is obtained, that is, the fault positioning is completed, and the output result state of the random forest model is compared with the actual state of the test set to obtain the accuracy of fault diagnosis.

[0084] Referring to Figure 3 According to the second aspect, the present application provides a three-level inverter fault diagnosis system, comprising: a data acquisition module 100, a first calculation module 200 and a second calculation module 300;

[0085] The data acquisition module 100 is configured to: construct a current state observer model according to a three-level inverter state space model; obtain a current residual signal through the current state observer model, and establish a current residual information table; construct a sample set according to the current residual signal; the sample set includes a training sample set and a test sample set;

[0086] The first calculation module 200 is configured to: preprocess the training sample set to generate current residual data; according to the current residual information table, the current residual data is extracted to generate a current residual signal fault feature data set;

[0087] The second calculation module 300 is configured to: construct a three-level inverter fault diagnosis model according to the current residual signal fault feature data set and a random forest model; according to the three-level inverter fault diagnosis model, fault identification and positioning are performed.

[0088] The effects of the above system in applying the foregoing method can be referred to the description in the foregoing method embodiments, which will not be repeated here.

[0089] Based on the same application concept, in a third aspect, the present application provides an electronic device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to realize the method described in the first aspect or any one of the possible implementation manners in the first aspect.

[0090] Based on the same application concept, in a fourth aspect, the present application provides a non-transitory computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to realize the method described in the first aspect or any one of the possible implementation manners in the first aspect.

[0091] Those skilled in the art will appreciate that embodiments of the application can be devised for a method, a system, or a computer program product. Accordingly, the present application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0092] The present application is described in reference to the flowchart illustrations and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0093] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0094] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0095] Obviously, the above-described embodiments are only examples and are not intended to limit the present application. Other variations and modifications can be made to the embodiments described above based on the principles of the present application. The present application is not limited to the embodiments described above. The scope of the present application is defined by the appended claims.

Claims

1. A three-level inverter fault diagnosis method, characterized by, The method comprises: a current state observer model is constructed according to a three-level inverter state space model; a current residual signal is obtained through the current state observer model, and a current residual information table is established; a sample set is constructed according to the current residual signal; the sample set comprises a training sample set and a test sample set; the training sample set is preprocessed to generate current residual data; feature extraction is performed on the current residual data according to the current residual information table to generate a current residual signal fault feature data set; a three-level inverter fault diagnosis model is constructed according to the current residual signal fault feature data set and a random forest model; fault recognition and positioning are performed according to the three-level inverter fault diagnosis model; the current residual data is generated, including randomly extracting m samples in the original current residual signal, then randomly extracting n feature values from the signal features, and selecting the best features for node segmentation; in the new sample set, the probability distribution of node m is as follows: ； wherein n is the number of characteristic values, is the probability of drawing n classes; selecting the best feature and the split point; based on the Gini maximization principle, a feature is selected, and the node of the feature is divided into left and right child nodes; recursion is performed on the right and left child nodes, and the process is repeated until the training sample set is preprocessed to generate the current residual error data.

2. The fault diagnostic method for a three-level inverter according to claim 1, characterized by, the step of obtaining a current residual signal through the current state observer model and establishing a current residual information table comprises: a current observation value is obtained through the current state observer model; a current residual value is generated according to the difference between the current observation value and the actual current value; a current residual information table is generated according to the current residual value.

3. The fault diagnostic method for a three-level inverter according to claim 1, characterized by, The expression of the three-level inverter state space model is: where A, B, C are known 3x3 real matrices, the main diagonal elements of A are , the remaining elements are 0, the main diagonal elements of B are , the remaining elements are 0, C is the identity matrix; is the state vector, is the control input vector.

4. The fault diagnostic method for a three-level inverter according to claim 3, characterized by, , The expression is: wherein, , and are the A-phase, B-phase, C-phase load output phase currents, which form the state vector ; , and are the A-phase, B-phase, C-phase load output phase voltages, which form the input vector ; t is time.

5. The fault diagnostic method for a three-level inverter according to claim 1, characterized by, The expression of the current state observer model is: wherein, is a state vector, is an observation of , H is an error compensation matrix, the main diagonal elements of A are , the remaining elements are 0, C is an identity matrix, is an output vector, is a derivative with respect to time, is an observation of 6. The fault diagnostic method of a three-level inverter according to claim 1, wherein The step of preprocessing the training sample set comprises: The training sample set is preprocessed according to a wavelet transform and a Park vector module normalization algorithm.

7. A three-level inverter fault diagnosis system for performing a three-level inverter fault diagnosis method according to any one of claims 1 to 6, characterized by It comprises: a data acquisition module, a first calculation module and a second calculation module; the data acquisition module is configured to construct a current state observer model according to a three-level inverter state space model, and obtain a current residual signal through the current state observer model and establish a current residual information table; a sample set is constructed according to the current residual signal; the sample set comprises a training sample set and a test sample set; the first calculation module is configured to preprocess the training sample set to generate current residual data; feature extraction is performed on the current residual data according to the current residual information table to generate a current residual signal fault feature data set; the second calculation module is configured to construct a three-level inverter fault diagnosis model according to the current residual signal fault feature data set and a random forest model, and perform fault recognition and positioning according to the three-level inverter fault diagnosis model.

8. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to realize the steps of the three-level inverter fault diagnosis method according to any one of claims 1 to 6. 9.A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the three-level inverter fault diagnosis method according to any one of claims 1 to 6.

Images