CN110852509A - Fault prediction method and device of IGBT module and storage medium - Google Patents

Fault prediction method and device of IGBT module and storage medium Download PDF

Info

Publication number
CN110852509A
CN110852509A CN201911094431.8A CN201911094431A CN110852509A CN 110852509 A CN110852509 A CN 110852509A CN 201911094431 A CN201911094431 A CN 201911094431A CN 110852509 A CN110852509 A CN 110852509A
Authority
CN
China
Prior art keywords
igbt module
fault
time
historical
value
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
Application number
CN201911094431.8A
Other languages
Chinese (zh)
Inventor
李鲲鹏
陈飞
李祎璞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Haopeng Intelligent Technology Co Ltd
Original Assignee
Beijing Haopeng Intelligent Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Haopeng Intelligent Technology Co Ltd filed Critical Beijing Haopeng Intelligent Technology Co Ltd
Priority to CN201911094431.8A priority Critical patent/CN110852509A/en
Publication of CN110852509A publication Critical patent/CN110852509A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/26—Testing of individual semiconductor devices
    • G01R31/2601—Apparatus or methods therefor
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Abstract

The embodiment of the application discloses a method, a device and a storage medium for predicting the faults of an IGBT module, wherein the method comprises the following steps: acquiring a collector-emitter electrical signal of the IGBT module when the IGBT module works at a first time; determining a fault characteristic value of the IGBT module at a first time based on a collector-emitter electric signal of the IGBT module when the IGBT module works at the first time, wherein the fault characteristic value comprises a turn-off instantaneous peak voltage value of the IGBT module and a collector-emitter saturation voltage drop value of the IGBT module; and inputting the fault characteristic value of the IGBT module at the first time into a pre-trained fault prediction model, predicting the fault result of the IGBT module when the IGBT module works at the second time, wherein the time difference between the first time and the second time is equal to a preset time value. According to the method and the device, the health state of the IGBT module can be judged, and the time when the IGBT module breaks down can also be judged, so that the accuracy and the reliability of the IGBT module fault prediction are improved.

Description

Fault prediction method and device of IGBT module and storage medium
Technical Field
The present disclosure relates to the field of fault analysis technologies, and in particular, to a method and an apparatus for predicting a fault of an IGBT module, and a storage medium.
Background
An Insulated Gate Bipolar Transistor (IGBT) is a composite fully-controlled voltage-driven power Semiconductor device, integrates the advantages of Metal-Oxide-Semiconductor Field Effect transistors (MOSFETs) and Giant Transistors (GTRs), has the characteristics of high input impedance, easiness in driving, high switching speed, low conduction voltage drop, high voltage and large current resistance and the like, is an ideal power element of a high-power electronic device, and is widely applied to electric energy conversion occasions of repeated acceleration, deceleration, starting/stopping and the like of rail transit, aerospace, new energy automobiles, wind power generation and the like.
In the application of the IGBT module in the high-power electronic device, on one hand, the IGBT module has large working voltage and current, and high power consumption and chip working junction temperature; on the other hand, due to repeated heating and cooling, and the inherent influence of the IGBT module on overheating, overvoltage and overshoot interference does not have good bearing capacity, the IGBT module is easy to lose effectiveness under extremely complex and harsh external stress generated by different application environments, and the service life is shortened. In order to ensure that the IGBT module can operate safely and reliably for a long period of time, the state of the IGBT module needs to be monitored and predicted online.
At present, IGBT module fault prediction research is based on stable and linear univariate or single information simpler prediction. In practical application, the fault development of the IGBT module is non-smooth and non-linear non-single variable or non-single information. Therefore, the current failure prediction technology of the IGBT module has poor effect in practical application, the prediction result is inaccurate, and the IGBT module cannot be guided to be maintained according to the situation.
Disclosure of Invention
The embodiment of the application provides a method, a device and a system for predicting the fault of an IGBT module and a storage medium, so as to improve the accuracy of the fault prediction of the IGBT module.
In a first aspect, an embodiment of the present application provides a method for predicting a fault of an IGBT module, including:
acquiring a collector-emitter electrical signal of the IGBT module when the IGBT module works at a first time;
determining a fault characteristic value of the IGBT module at a first time based on a collector-emitter electrical signal of the IGBT module when the IGBT module works at the first time, wherein the fault characteristic value comprises a turn-off instantaneous spike voltage value of the IGBT module and a saturation voltage drop value of a collector-emitter of the IGBT module;
inputting the fault characteristic value of the IGBT module at the first time into a pre-trained fault prediction model, and predicting a fault result of the IGBT module when the IGBT module works at a second time, wherein the second time is the future time of the first time, and the time difference between the first time and the second time is equal to a preset time value.
In a possible implementation manner of the first aspect, the method further includes:
acquiring historical collector-emitter electrical signals of the IGBT module when the IGBT module works in historical time;
determining a historical fault characteristic value of the IGBT module based on a historical collector-emitter electrical signal of the IGBT module;
and training the fault prediction model by using the historical fault characteristic value of the IGBT module.
In a possible implementation manner of the first aspect, the training the fault prediction model by using the historical fault feature value of the IGBT module includes:
sampling historical fault characteristic values of the IGBT module by taking the preset time value as a time sampling interval to obtain a plurality of sampled historical fault characteristic values;
and training the fault prediction model by using the plurality of sampled historical fault characteristic values.
In a possible implementation manner of the first aspect, the sampling the historical fault characteristic value of the IGBT module with the preset time value as a time sampling interval to obtain a plurality of sampled historical fault characteristic values includes:
dividing the historical fault characteristic value of the IGBT module into a plurality of sampling intervals by taking the preset time value as a time sampling interval;
and aiming at each sampling interval, taking the median value or the mean value of the historical fault characteristic values of the IGBT module in the sampling interval as the historical fault characteristic value of the sampling interval, and obtaining a plurality of sampled historical fault characteristic values.
In a possible implementation manner of the first aspect, before sampling the historical fault feature value of the IGBT module by using the preset time value as a time sampling interval, the method further includes:
and sequencing the historical fault characteristic values of the IGBT modules according to a time sequence.
In one possible implementation manner of the first aspect, the fault prediction model includes: the IGBT module state evaluation system comprises a characteristic prediction model and an IGBT module state evaluation model, wherein the output of the characteristic prediction model is the input of the IGBT module state evaluation model.
In a possible implementation manner of the first aspect, the training the fault prediction model by using the plurality of sampled historical fault feature values includes:
extracting a plurality of first fault characteristic values and a plurality of second fault characteristic values in the plurality of sampled historical fault characteristic values, wherein the first fault characteristic values correspond to the second faults one by one, and the time difference between the corresponding first fault characteristic values and the corresponding second fault characteristic values is the preset time value;
aiming at each first fault characteristic value in a plurality of first fault characteristic values, taking the first fault characteristic value as input, taking a second fault characteristic value corresponding to the first fault characteristic value as output, and training the characteristic prediction model;
extracting a plurality of third fault characteristic values in the plurality of sampled historical fault characteristic values and fault state information corresponding to each piece of third fault characteristic value information, wherein the fault characteristic values further comprise the fault state information of the IGBT module;
and aiming at each third fault characteristic value in the plurality of third fault characteristic values, taking the third fault characteristic value as input, taking fault state information corresponding to the third fault characteristic value as output, and training the IGBT module state evaluation model.
Optionally, the second fault characteristic value is the same as the third fault characteristic value.
In a possible implementation manner of the first aspect, before determining the historical fault characteristic value of the IGBT module based on the historical collector-emitter electrical signal of the IGBT module, the method includes:
removing bad points in historical collector-emitter electrical signals of the IGBT module;
determining a historical fault signature value for the IGBT module based on historical collector-emitter electrical signals for the IGBT module, comprising:
and determining a historical fault characteristic value of the IGBT module based on the historical collector-emitter electrical signal of the IGBT module after the dead pixel is removed.
In one possible implementation manner of the first aspect, the removing the dead pixel in the historical collector-emitter electrical signal of the IGBT module includes:
acquiring an arithmetic mean value of historical collector-emitter electrical signals of the IGBT module;
and removing the electric signals of which the deviation from the arithmetic mean value is larger than a preset value in the historical collector-emitter electric signals of the IGBT module.
In a possible implementation manner of the first aspect, the determining a historical fault characteristic value of the IGBT module based on the historical collector-emitter electrical signal of the IGBT module after dead pixel removal includes:
and determining all historical turn-off instantaneous peak voltage values corresponding to the IGBT module entering a turn-off region and all historical saturation voltage drop values corresponding to the IGBT module entering a saturation region in each working cycle of the IGBT module from the historical collector-emitter electric signals of the IGBT module after dead spots are removed.
In a possible implementation manner of the first aspect, before the training of the fault prediction model by using the historical fault feature values of the IGBT module, the method further includes:
carrying out normalization processing on the historical fault characteristic value of the IGBT module;
training the fault prediction model using historical fault feature values of the IGBT module, including:
and training the fault prediction model by using the normalized historical fault characteristic value of the IGBT module.
In a possible implementation manner of the first aspect, the inputting the fault characteristic value of the IGBT module at the first time into a fault prediction model trained in advance to predict the fault result of the IGBT module when the IGBT module operates at the second time includes:
inputting the fault characteristic value of the IGBT module at the first time into the pre-trained characteristic prediction model, and predicting the fault characteristic value of the IGBT module when the IGBT module works at the second time;
and inputting the fault characteristic value of the IGBT module when the IGBT module works at the second time into the IGBT module state evaluation model, and predicting the fault state information of the IGBT module when the IGBT module works at the second time.
Optionally, the fault state information of the IGBT module includes: normal, alert and dangerous.
In a possible implementation manner of the first aspect, the method further includes:
and predicting whether the state parameters of the IGBT module are suddenly changed when the IGBT module works at the second time based on the fault characteristic value of the IGBT module when the IGBT module works at the second time.
In a second aspect, an embodiment of the present application provides a failure prediction apparatus for an IGBT module, including:
the acquisition module is used for acquiring collector-emitter electrical signals when the IGBT module works at the first time;
the determining module is used for determining a fault characteristic value of the IGBT module at a first time based on a collector-emitter electric signal when the IGBT module works at the first time, wherein the fault characteristic value comprises a turn-off instantaneous spike voltage value of the IGBT module and a collector-emitter saturation voltage drop value of the IGBT module;
and the prediction module is used for inputting the fault characteristic value of the IGBT module at the first time into a pre-trained fault prediction model, and predicting a fault result of the IGBT module when the IGBT module works at a second time, wherein the second time is the future time of the first time, and the time difference between the first time and the second time is equal to a preset time value.
In a possible implementation manner of the second aspect, the method further includes a training module;
the obtaining module is further used for obtaining historical collector-emitter electrical signals when the IGBT module works in historical time;
the determination module is further used for determining a historical fault characteristic value of the IGBT module based on a historical collector-emitter electrical signal of the IGBT module;
and the training module is used for training the fault prediction model by using the historical fault characteristic value of the IGBT module.
In one possible implementation manner of the second aspect, the training module includes an obtaining unit and a training unit;
the obtaining unit is used for sampling the historical fault characteristic values of the IGBT module by taking the preset time value as a time sampling interval to obtain a plurality of sampled historical fault characteristic values;
and the training unit is used for training the fault prediction model by using the plurality of sampled historical fault characteristic values.
In a possible implementation manner of the second aspect, the obtaining unit is specifically configured to divide the historical fault feature value of the IGBT module into a plurality of sampling intervals by taking the preset time value as a time sampling interval; and aiming at each sampling interval, taking the median value or the mean value of the historical fault characteristic values of the IGBT module in the sampling interval as the historical fault characteristic value of the sampling interval, and obtaining a plurality of sampled historical fault characteristic values.
In a possible implementation manner of the second aspect, the obtaining unit is further configured to sort the historical fault feature values of the IGBT modules in a time sequence.
In one possible implementation manner of the second aspect, the fault prediction model includes: the IGBT module state evaluation system comprises a characteristic prediction model and an IGBT module state evaluation model, wherein the output of the characteristic prediction model is the input of the IGBT module state evaluation model.
In a possible implementation manner of the second aspect, the training unit is specifically configured to extract a plurality of first fault feature values and a plurality of second fault feature values in the plurality of sampled historical fault feature values, where the first fault feature values correspond to the second faults one by one, and a time difference between the corresponding first fault feature values and the corresponding second fault feature values is the preset time value; aiming at each first fault characteristic value in a plurality of first fault characteristic values, taking the first fault characteristic value as input, taking a second fault characteristic value corresponding to the first fault characteristic value as output, and training the characteristic prediction model; extracting a plurality of third fault characteristic values in the plurality of sampled historical fault characteristic values and fault state information corresponding to each piece of third fault characteristic value information, wherein the fault characteristic values further comprise the fault state information of the IGBT module; and aiming at each third fault characteristic value in the plurality of third fault characteristic values, taking the third fault characteristic value as input, taking fault state information corresponding to the third fault characteristic value as output, and training the IGBT module state evaluation model.
Optionally, the second fault characteristic value is the same as the third fault characteristic value.
In one possible implementation of the second aspect, the apparatus includes a removal module;
the removing module is used for removing dead pixels in historical collector-emitter electrical signals of the IGBT module;
the determination module is used for determining the historical fault characteristic value of the IGBT module based on the historical collector-emitter electrical signal of the IGBT module after dead pixels are removed.
In a possible implementation manner of the second aspect, the removing module is specifically configured to obtain an arithmetic average of historical collector-emitter electrical signals of the IGBT module; and removing the electric signals of which the deviation from the arithmetic mean value is larger than a preset value in the historical collector-emitter electric signals of the IGBT module.
In a possible implementation manner of the second aspect, the determining module is specifically configured to determine, from the historical collector-emitter electrical signals of the IGBT module after the dead pixel is removed, all historical turn-off instantaneous spike voltage values corresponding to when the IGBT module enters a turn-off region and all historical collector-emitter saturation voltage drop values corresponding to when the IGBT module enters a saturation region in each operating cycle of the IGBT module.
In a possible implementation manner of the second aspect, the apparatus further includes a normalization module:
the normalization module is used for normalizing the historical fault characteristic value of the IGBT module;
and the training module is used for training the fault prediction model by using the historical fault characteristic value of the IGBT module after normalization processing.
In a possible implementation manner of the second aspect, the prediction module is specifically configured to input a fault characteristic value of the IGBT module at the first time into the feature prediction model trained in advance, and predict a fault characteristic value of the IGBT module when the IGBT module operates at the second time; and inputting the fault characteristic value of the IGBT module when the IGBT module works at the second time into the IGBT module state evaluation model, and predicting the fault state information of the IGBT module when the IGBT module works at the second time.
Optionally, the fault state information of the IGBT module includes: normal, alert and dangerous.
In a possible implementation manner of the second aspect, the predicting module is further configured to predict whether the state parameter changes suddenly when the IGBT module operates at the second time based on the fault characteristic value of the IGBT module operating at the second time.
In a third aspect, the present embodiment provides a failure prediction apparatus for an IGBT module, where the apparatus exists in the form of a chip product, and the apparatus includes a processor and a memory, where the memory is configured to be coupled to the processor and store necessary program instructions and data of the apparatus, and the processor is configured to execute the program instructions stored in the memory, so that the apparatus executes the method according to the first aspect.
In a fourth aspect, an embodiment of the present application provides an electronic device, including a processor and a memory;
the memory for storing a computer program;
the processor is configured to execute the computer program to implement the method for predicting a fault of the IGBT module according to any one of the first aspect.
In a fifth aspect, the present application provides a computer storage medium, where the storage medium includes computer instructions, and when the instructions are executed by a computer, the computer is enabled to implement the method for predicting the failure of the IGBT module according to any one of the first aspect.
In a sixth aspect, the present application provides a computer program product, where the computer program product includes a computer program, the computer program is stored in a readable storage medium, the computer program can be read by at least one processor of a computer from the readable storage medium, and the at least one processor executes the computer program to make the computer implement the method for predicting a failure of an IGBT module according to any one of the first aspect.
According to the method, the device and the storage medium for predicting the fault of the IGBT module, the collector-emitter electrical signal of the IGBT module is obtained when the IGBT module works at the first time; determining a fault characteristic value of the IGBT module at a first time based on a collector-emitter electric signal of the IGBT module when the IGBT module works at the first time, wherein the fault characteristic value comprises a turn-off instantaneous peak voltage value of the IGBT module and a collector-emitter saturation voltage drop value of the IGBT module; and then, inputting the fault characteristic value of the IGBT module at the first time into a pre-trained fault prediction model, and predicting the fault result of the IGBT module when the IGBT module works at the second time, wherein the second time is the future time of the first time, and the time difference between the first time and the second time is equal to a preset time value. According to the method and the device, the fault condition of the IGBT module at the second time in the future is predicted by using two parameters, namely the turn-off instantaneous peak voltage value of the IGBT module at the first time and the saturation voltage drop value of the collector-emitter, so that the health state of the IGBT module can be judged, the time when the IGBT module fails can also be judged, and the accuracy and the reliability of the fault prediction of the IGBT module are improved.
Drawings
Fig. 1 is a schematic view of an application scenario according to an embodiment of the present application;
fig. 2 is a flowchart of a fault prediction method for an IGBT module according to an embodiment of the present application;
FIG. 3 is a flowchart of a method for training a fault prediction model according to an embodiment of the present disclosure;
FIG. 4 is a flow chart of another training method of a fault prediction model according to an embodiment of the present disclosure;
fig. 5 is a schematic structural diagram of a failure prediction module according to an embodiment of the present application;
FIG. 6 is a flow chart of another training method of a fault prediction model according to an embodiment of the present disclosure;
fig. 7 is a flowchart of a fault prediction method for an IGBT module according to an embodiment of the present application;
fig. 8 is a schematic diagram of a failure prediction apparatus of an IGBT module according to an embodiment of the present application;
fig. 9 is a schematic diagram of a failure prediction apparatus of an IGBT module according to another embodiment of the present application;
fig. 10 is a schematic diagram of a failure prediction apparatus of an IGBT module according to another embodiment of the present application;
fig. 11 is a schematic diagram of a failure prediction apparatus of an IGBT module according to another embodiment of the present application;
fig. 12 is a schematic structural diagram of a failure prediction apparatus of an IGBT module according to an embodiment of the present application;
fig. 13 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
The basic concepts and prior art related to the present application will first be briefly described as follows.
Optional repair refers to repair or replacement of a component that has degraded performance or will fail before the failure of the electronic system occurs, and taking preventive or protective measures before the failure does not occur. The on-demand maintenance is established on the basis of deeply researching the failure principle of an electronic system, monitoring the effective state of the system and performing trend analysis by using a correct algorithm. The condition-based maintenance not only can avoid the occurrence of serious catastrophic production accidents in advance, but also brings economic benefits of reduced maintenance cost, and the method has the advantages of automation, high efficiency, small required logistics scale and particular importance in the military field with high reliability requirement.
As a key component of the power system, the reliability of the IGBT affects the operational stability of the entire system equipment, so if the IGBT module fault prediction and health management can be realized, the purpose of on-demand maintenance is achieved, and the IGBT module fault prediction and health management method has great significance for improving the reliability of the power system, reducing maintenance and repair costs, avoiding safety accidents, and the like.
The existing IGBT module fault prediction method comprises a qualitative analysis method and a quantitative analysis method, wherein the quantitative analysis method mainly comprises the following steps: model-based methods, knowledge-based methods, and data-based methods.
The data-based method comprises an autoregressive prediction method, a gray prediction method, a multi-layer hierarchical method, a chaotic time sequence prediction method, a hidden Markov model, a machine learning (neural network and support vector machine) and a statistical process monitoring method. The IGBT mechanism model is difficult to establish, expert knowledge is difficult to obtain, and the IGBT mechanism model and the expert knowledge are not beneficial to predicting the IGBT module faults. The data-based method is completely based on industrial field data, and the implicit information in the data is mined, so that the method has wide engineering application value. The method has the advantages of widest application range and minimum cost. Therefore, the data-based method is the most practical, and has become a research hotspot and development trend in the field of IGBT module fault prediction. However, the existing prediction method based on data is mainly based on a stable and linear single variable, the fusion of data and information is not considered, the actual failure of the IGBT module is a non-stable and non-linear random process, and each characteristic parameter has different specificities for reflecting the failure reason, so that the method cannot accurately predict the failure of the IGBT module.
In order to solve the above technical problem, an embodiment of the present application provides a method for predicting a fault of an IGBT module, where the fault characteristics of the IGBT module during operation of the IGBT module at a first time are obtained, the fault characteristics of the IGBT module at the first time are input into a pre-trained fault characteristic prediction model, the pre-trained fault prediction model is used to predict a development trend of the fault characteristics of the IGBT module, and a fault of the IGBT module at a future time (for example, at a second time) is evaluated, so that not only a health state of the IGBT module can be predicted, but also when the IGBT module fails can be predicted, and accuracy and reliability of fault prediction of the IGBT module are improved.
In the present embodiment, the phrase "B corresponding to a" means that B is associated with a. In one implementation, B may be determined from a. It should also be understood that determining B from a does not mean determining B from a alone, but may be determined from a and/or other information.
In the description of the present application, "plurality" means two or more than two unless otherwise specified.
In addition, in order to facilitate clear description of technical solutions of the embodiments of the present application, in the embodiments of the present application, terms such as "first" and "second" are used to distinguish the same items or similar items having substantially the same functions and actions. Those skilled in the art will appreciate that the terms "first," "second," etc. do not denote any order or quantity, nor do the terms "first," "second," etc. denote any order or importance.
The method for predicting the fault of the IGBT module according to the embodiment of the present application is described in detail below with specific examples.
Fig. 1 is a schematic view of an application scenario related to an embodiment of the present application, including: IGBT module, sensor and electronic equipment.
The sensor is used for collecting collector-emitter electrical signals when the IGBT module works. Alternatively, the sensor may be a current sensor, and the collector-emitter electrical signal acquired by the current sensor when the IGBT module operates is collector-emitter current. Alternatively, the sensor may be a voltage sensor, for example, a broadband voltage sensor, and the collected collector-emitter electrical signal when the IGBT module operates is the collector-emitter voltage. Optionally, the output range of the voltage sensor is ± 10V, and the sampling rate is 10 Mbps.
The electronic device is used for acquiring collector-emitter electrical signals acquired by the sensor when the IGBT module works, executing the method of the embodiment of the application, predicting the health state of the IGBT module, and predicting when and what kind of faults the IGBT module will occur.
Fig. 2 is a flowchart of a fault prediction method for an IGBT module according to an embodiment of the present application, and as shown in fig. 2, the method according to the embodiment of the present application includes:
s101, acquiring a collector-emitter electric signal when the IGBT module works at the first time.
The main execution body of the embodiment of the application is a device with a function of predicting the fault of the IGBT module, for example, a fault prediction device of the IGBT module. The failure prediction device of the IGBT module may be a single electronic device shown in fig. 1, or may be a component in the electronic device shown in fig. 1, for example, a processor in the electronic device shown in fig. 1.
The embodiment of the present application takes an execution subject as an electronic device as an example for description.
The first time may be any time when the IGBT module operates, and optionally, the first time may be a current operating time of the IGBT module. For example, the embodiment of the present application expects to use the collector-emitter electrical signal when the IGBT module is operated for the previous week to predict whether the IGBT module will fail after one week, and the first time is the current week.
In the embodiment of the application, as shown in fig. 1, the sensor acquires the collector-emitter electrical signal of the IGBT module in real time when the IGBT module operates at the first time, and sends the acquired collector-emitter electrical signal of the IGBT module when the IGBT module operates at the first time to the electronic device.
S102, determining a fault characteristic value of the IGBT module at a first time based on a collector-emitter electric signal of the IGBT module when the IGBT module works at the first time.
Wherein the fault characteristic values comprise turn-off instantaneous spike voltage values of the IGBT module and collector-emitter saturation voltage drop values of the IGBT module.
Specifically, a turn-off instantaneous peak voltage value when the IGBT module enters a turn-off region in each working cycle of the IGBT module and a saturation voltage drop value of the collector-emitter when the IGBT module performs saturation region are obtained from a collector-emitter electrical signal when the IGBT module operates at the first time. One working cycle of the IGBT module is the time length from one turn-on to turn-off of the IGBT module.
According to the embodiment of the application, at least two characteristic parameters, namely the turn-off instantaneous peak voltage value of the IGBT module and the collector-emitter saturation voltage drop value of the IGBT module, are considered when the fault of the IGBT module is analyzed, and each characteristic parameter has different specificities for reflecting the fault reason, so that compared with the existing fault prediction method only considering a single characteristic parameter, the accuracy of the fault prediction of the IGBT module is greatly improved.
S103, inputting the fault characteristic value of the IGBT module at the first time into a fault prediction model trained in advance, and predicting a fault result of the IGBT module when the IGBT module works at the second time.
The second time is a future time of the first time, and the time difference between the first time and the second time is equal to a preset time value.
The electronic device shown in the embodiment of the application may call a pre-trained fault prediction model, for example, the electronic device includes a processor and a memory, the pre-trained fault prediction model is stored in the memory, and the processor may call the pre-trained fault prediction model from the memory.
Optionally, the failure prediction module may be a common Recurrent Neural Network (RNN).
Optionally, the fault prediction model may be a Long Short Term memory network (LSTM) model.
Optionally, the fault prediction model may be a recent fault prediction model, for example, a daily fault prediction module, and may predict a case where the IGBT module fails on a certain future day.
Optionally, the fault prediction model may be a medium-term fault prediction model, for example, a week fault prediction module, which may predict a case where the IGBT module fails in a certain week in the future.
Optionally, the fault prediction model may be a long-term fault prediction model, for example, a monthly fault prediction module, which may predict a fault condition of the IGBT module in a future month.
Namely, the fault prediction model of the embodiment of the application is used for predicting the situation that the IGBT module has a fault when operating at the second time, and the time difference between the second time and the first time is a preset time difference. For example, when the predicted time difference is 1 week, the fault prediction model may predict that the IGBT module transmits a fault when operating at a second time after one week, according to the fault characteristic value of the IGBT module at the current first time. Or, when the predicted time difference is one month, the fault prediction model may predict a fault condition of the IGBT module when the IGBT module operates at a second time after one month according to the fault characteristic value of the IGBT module at the current first time.
The predicted fault result of the IGBT module operating at the second time may include: whether the change trend of the fault characteristic value of the IGBT module during the operation at the second time exceeds the preset trend or not and whether the IGBT module fails during the operation at the second time, which failure occurs, such as warning or danger, etc.
Optionally, the health state of the IGBT module may be predicted according to a variation trend of the fault characteristic value of the IGBT module, for example, when the variation trend of the fault characteristic value of the IGBT module during operation at the second time meets a preset trend, it is indicated that the health state of the IGBT module during operation at the second time is good, and if the variation trend of the fault characteristic value of the IGBT module during operation at the second time does not meet the preset trend, it is indicated that the health state of the IGBT module during operation at the second time is not good, and a technician may perform operations such as maintenance or replacement on the IGBT module.
According to the fault prediction method of the IGBT module, the collector-emitter electrical signal of the IGBT module is obtained when the IGBT module works at the first time; determining a fault characteristic value of the IGBT module at a first time based on a collector-emitter electric signal of the IGBT module when the IGBT module works at the first time, wherein the fault characteristic value comprises a turn-off instantaneous peak voltage value of the IGBT module and a collector-emitter saturation voltage drop value of the IGBT module; and then, inputting the fault characteristic value of the IGBT module at the first time into a pre-trained fault prediction model, and predicting the fault result of the IGBT module when the IGBT module works at the second time, wherein the second time is the future time of the first time, and the time difference between the first time and the second time is equal to a preset time value. According to the method and the device, the fault condition of the IGBT module at the second time in the future is predicted by using two parameters, namely the turn-off instantaneous peak voltage value of the IGBT module at the first time and the saturation voltage drop value of the collector-emitter, so that the health state of the IGBT module can be judged, the time when the IGBT module fails can also be judged, and the accuracy and the reliability of the fault prediction of the IGBT module are improved.
The following describes the training process of the fault prediction model in detail with reference to specific embodiments.
Fig. 3 is a flowchart of a training method of a fault prediction model according to an embodiment of the present application, and based on the embodiment, as shown in fig. 3, a training process of the fault prediction model according to the present application includes:
s201, obtaining a historical collector-emitter electric signal of the IGBT module when the IGBT module works in historical time.
In practical application, in order to monitor the operating state of the IGBT module, collector-emitter electrical signals are collected and stored in real time when the IGBT module is operating.
Alternatively, the historical collector-emitter electrical signals of the IGBT module operating at the historical time may be stored in a database or other storage device, and the electronic device shown in fig. 1 may obtain the historical collector-emitter electrical signals of the IGBT module operating at the historical time from the database or other storage device.
The historical time is any time before the current time.
Optionally, in order to facilitate subsequent model training, the historical collector-emitter electrical signals of the IGBT module during operation in the historical time are sorted according to a time sequence to obtain a time sequence of the historical collector-emitter electrical signals of the IGBT module during operation in the historical time, for example, the time sequence of the historical collector-emitter electrical signals of the IGBT module during operation in the historical time is: u. of1,u2,…uN。
S202, determining a historical fault characteristic value of the IGBT module based on the historical collector-emitter electric signal of the IGBT module.
Specifically, after the historical collector-emitter electrical signal of the IGBT module during the operation in the historical time is obtained according to the above steps, the turn-off instantaneous peak voltage value of the IGBT module and the saturation voltage drop value of the collector-emitter and other fault characteristics are obtained from the historical collector-emitter electrical signal of the IGBT module. The specific implementation process of this step is the same as the implementation process of S102 described above, and reference may be made to the description of S102 described above.
In order to distinguish from S102, in this step, the fault characteristic value of the IGBT module determined based on the historical collector-emitter electrical signal of the IGBT module is recorded as the historical fault characteristic value of the IGBT module, for example, the turn-off instantaneous spike voltage value and the collector-emitter saturation voltage drop value of the IGBT module determined based on the historical collector-emitter electrical signal of the IGBT module are recorded as the historical turn-off instantaneous spike voltage value and the historical collector-emitter saturation voltage drop value of the IGBT module.
In some embodiments, in training the fault prediction model, in order to improve the accuracy of model training, the historical collector-emitter electrical signals of the IGBT module may be preprocessed, specifically, before S202, the method further includes step B:
and B, removing dead spots in the historical collector-emitter electrical signals of the IGBT module.
The above-mentioned dead spots in the historical collector-emitter electrical signals of the IGBT module may be understood as points deviating from the overall variation trend of the historical collector-emitter electrical signals, for example, when a sensor collecting the collector-emitter electrical signals fails, the collected collector-emitter electrical signals of the IGBT module are inaccurate, and the inaccurate collector-emitter electrical signals may form dead spots.
In an example, the above-mentioned manner of removing the dead pixel in the historical collector-emitter electrical signal of the IGBT module may be to draw a variation curve of the historical collector-emitter electrical signal of the IGBT module, and remove a point far away from the variation curve as the dead pixel.
In another example, the above-mentioned manner of removing the dead pixel in the historical collector-emitter electrical signal of the IGBT module may include the following steps B1 and B2:
and B1, acquiring the arithmetic mean value of the historical collector-emitter electric signals of the IGBT module.
Specifically, the obtained historical collector-emitter electrical signal of the IGBT module is assumed to be: u. of1,u2,…uNAnd N is the number of the collected historical collector-emitter electrical signals of the IGBT module, and is a positive integer greater than or equal to 1. Determining an arithmetic mean of historical collector-emitter electrical signals of an IGBT module according to the following equation (1)
Figure BDA0002267863180000142
Figure BDA0002267863180000141
And step B2, removing the electric signals of which the deviation from the arithmetic mean value is larger than a preset value in the historical collector-emitter electric signals of the IGBT module.
Specifically, each of the historical collector-emitter electrical signals u of the IGBT module is determined according to the following equation (2)iArithmetic mean of historical collector-emitter electrical signals with IGBT module
Figure BDA0002267863180000151
Deviation therebetween:
Figure BDA0002267863180000152
the deviation e of the above formula (2)iComparing with a preset value, and further removing the deviation e between the historical collector-emitter electrical signals of the IGBT module and the arithmetic mean valueiAn electrical signal greater than a preset value. The preset value is not particularly limited, and is determined according to actual needs.
Optionally, the preset value may be 3 times of the root mean square deviation of the historical collector-emitter electrical signal of the IGBT module, and at this time, the root mean square deviation of the historical collector-emitter electrical signal of the IGBT module is determined according to the following formula (3):
Figure BDA0002267863180000153
each historical collector-emitter electrical signal uiArithmetic mean of historical collector-emitter electrical signals with IGBT module
Figure BDA0002267863180000154
Deviation e betweeniComparing with the root mean square deviation sigma of the historical collector-emitter electrical signal of the IGBT module by 3 times, specifically, if
Figure BDA0002267863180000155
Then u isiGross errors should be discarded if
Figure BDA0002267863180000156
Then u isiIt should be kept as normal data.
In this implementation, after the dead pixel in the historical collector-emitter electrical signal of the IGBT module is removed according to step B, step S202 may be replaced by step C.
And step C, determining a historical fault characteristic value of the IGBT module based on the historical collector-emitter electrical signal of the IGBT module after the dead pixel is removed.
In a specific example, the step C includes: and determining all historical turn-off instantaneous peak voltage values corresponding to the IGBT module entering a turn-off region and all historical saturation voltage drop values corresponding to the IGBT module entering a saturation region in each working cycle of the IGBT module from the historical collector-emitter electric signals of the IGBT module after the dead pixel is removed.
Optionally, the above-mentioned manner for determining all historical turn-off instantaneous peak voltage values corresponding to when the IGBT enters the turn-off region in each operating cycle of the IGBT module is specifically: extracting historical turn-off instantaneous peak voltage when the IGBT enters a turn-off region in each working cycle of the IGBT module according to the following formula (4) from the historical collector-emitter electrical signal of the IGBT module after dead points are removed:
where k is the kth duty cycle of the IGBT module, NkThe IGBT module in the kth working cycle is acquired from the historical collector-emitter electrical signals of the IGBT module after dead spots are removedNumber of collector-emitter voltage signal data, P, of blocks in turn-off regionkCorresponding u when the IGBT module starts to enter the turn-off region in the kth working cyclemSequence number of time series, where umThe time sequence is the historical collector-emitter electrical signal u of the IGBT modulenAnd removing the time sequence with the dead pixel in the time sequence.
Optionally, the above-mentioned manner for determining the historical saturation voltage drop values of all collector-emitter electrodes corresponding to the IGBT module entering the saturation region in each operating cycle of the IGBT module is specifically: extracting the historical collector-emitter saturation voltage drop of the IGBT module when the IGBT module enters a saturation region in each working cycle of the IGBT module according to the following formula (5) from the historical collector-emitter electrical signal of the IGBT module after dead spots are removed
Figure BDA0002267863180000161
Wherein M iskIs umThe number of collector-emitter voltage signal data, L, of the IGBT module in the saturation region in the kth working cycle acquired in the time sequencekCorresponding u when the IGBT module starts to enter the saturation region in the kth working cyclemSequence number of time series.
S203, training the fault prediction model by using the historical fault characteristic value of the IGBT module.
In the embodiment of the application, the fault result corresponding to each historical fault characteristic value is known.
Specifically, after the historical fault characteristic value of the IGBT module is obtained according to the steps, the historical fault characteristic value of the IGBT module is used for training the fault prediction model. For example, each historical fault characteristic value of the IGBT module is used as an input of the fault prediction model, and a fault result corresponding to the historical fault characteristic value is used as an output of the fault prediction module, so that each parameter of the fault prediction model is adjusted, and the fault prediction model is trained.
In some embodiments, as can be seen from the above description, the fault prediction model of the embodiment of the present application may be a near-term, medium-term or long-term fault prediction model, so that in order to enable the trained fault prediction model to accurately predict the fault condition of the IGBT module at a time in the future expected by the user, the fault training model is trained using the corresponding historical fault feature value. For example, when the fault prediction model is used for predicting the fault condition of the IGBT module after one week, the fault model is trained by using historical fault characteristic values at intervals of one week, if the fault prediction model is used for predicting the fault condition of the IGBT module after one month, the fault prediction model is trained by using historical fault characteristic values at intervals of one month, and if the fault prediction model is used for predicting the fault condition of the IGBT module after one year, the fault model is trained by using historical fault characteristic values at intervals of one year.
Based on the above description, when the historical fault feature value is used for the fault prediction model, the historical fault feature value needs to be resampled according to the characteristics of the fault prediction model in the embodiment of the application. Specifically, as shown in fig. 4, at this time, S203 may include S2031 and S2032:
s2031, sampling historical fault characteristic values of the IGBT module by taking the preset time value as a time sampling interval, and obtaining a plurality of sampled historical fault characteristic values.
For example, if the preset time value is one week, that is, the fault prediction model is used for predicting the fault condition of the IGBT after one week, the historical fault feature values of the IGBT module are sampled by using one week as a time sampling interval to obtain a plurality of sampled historical fault feature values, the plurality of sampled historical fault feature values are smoothly arranged according to time, and the time difference corresponding to two adjacent sampled historical fault feature values is one week.
Optionally, before sampling, the historical fault characteristic values of the IGBT modules may be sorted according to a time sequence.
In one example, the sampling process may be to collect a historical fault feature value every preset time value.
In another example, the sampling process may be that a preset time value is used as a time sampling interval, and the historical fault characteristic value of the IGBT module is divided into a plurality of sampling intervals. And aiming at each sampling interval, taking the median value or the mean value of the historical fault characteristic values of the IGBT module in the sampling interval as the historical fault characteristic value of the sampling interval, and obtaining a plurality of sampled historical fault characteristic values.
Optionally, the sampled historical fault feature value is an input sample of the fault prediction model, and may be represented in a matrix form. For example, a multivariate and multidimensional fault feature time series matrix is constructed by using the sampled historical fault feature values as column vectors (or row vectors) and using the time series numbers corresponding to the sampled historical fault feature values as row time series vectors (or column time series vectors).
S2032, training the fault prediction model by using the plurality of sampled historical fault characteristic values.
Specifically, according to the above steps, historical fault feature values of the IGBT module are sampled to obtain a plurality of sampled historical fault feature values, and the fault prediction model is trained using the plurality of sampled historical fault feature values, where a specific process is similar to that in S203, and reference may be made to the description in S203, and details are not repeated here.
Alternatively, the above S2031 may be performed after the above step C.
In some embodiments, before the step S203 trains the fault prediction model by using the historical fault feature value of the IGBT module, the method according to the embodiment of the present application may further include:
and D, normalizing the historical fault characteristic value of the IGBT module.
Specifically, the historical fault characteristic value of the IGBT module is converted into the range of [ a, b ] according to the following formula (6):
Figure BDA0002267863180000181
wherein x (k) can represent historical turn-off transient spike voltage fault characteristics V of the IGBT moduleP(k) And IGBT moduleHistorical collector-emitter saturation voltage drop VCE(sat)(k)，xminSample representing the smallest of the historical fault characteristic values, x, of the IGBT modulemaxRepresenting the largest sample among the historical fault characteristic values of the IGBT module.
In this case, in S203, the failure prediction model may be trained by using the normalized historical failure feature value of the IGBT module.
Optionally, step D may be executed before S2031 or after S2031, which is not limited in this embodiment of the application.
According to the embodiment of the application, historical collector-emitter electrical signals of an IGBT module working in historical time are obtained; determining a historical fault characteristic value of the IGBT module based on a historical collector-emitter electrical signal of the IGBT module; and training a fault prediction model by using the historical fault characteristic value of the IGBT module, so that the trained fault prediction model can predict the condition that the IGBT module sends faults in the future time.
In some embodiments of the present application, as shown in fig. 5, the fault prediction model includes a characteristic prediction model and an IGBT module state evaluation model, where an output of the characteristic prediction model is an input of the IGBT module state evaluation model, the characteristic prediction model is used to predict a fault characteristic development trend of the IGBT module at a future time, and the IGBT module state evaluation model is used to predict whether the IGBT module will be in fault at the future time and what kind of fault is sent.
At this time, in step S203, the training fault prediction model includes two parts, the first part is the training fault prediction model using the historical fault feature value of the IGBT module, and the second part is the training IGBT module state evaluation model using the historical fault feature value of the IGBT module. Specifically, as shown in fig. 6, S203 includes S301 to S304, where S301 and S302 are processes for training a characteristic prediction model using the historical fault characteristic values of the IGBT modules, and S303 and S304 are processes for training a state evaluation model of the IGBT modules using the historical fault characteristic values of the IGBT modules.
S301, extracting a plurality of first fault characteristic values and a plurality of second fault characteristic values in the plurality of sampled historical fault characteristic values.
The first fault characteristic value and the second fault characteristic value correspond to each other one by one, and the time difference between the corresponding first fault characteristic value and the corresponding second fault characteristic value is a preset time value.
For example, assuming that the preset time value is one week, the time-sequenced historical fault characteristic values after a plurality of samples are shown in table 1,
TABLE 1
Working time Post-adoption historical fault characteristic value
T1 b1
T2 b2
…… ……
Tm bm
The difference between the two adjacent operating times is equal to a preset time difference, for example, equal to one week, where T2 is the future time of T1, and assuming that the first fault characteristic value is b1, the second fault characteristic value corresponding to the first fault characteristic value b1 is b2, the time difference corresponding to b1 and b2 is the preset time difference, that is, the time difference between T2 and T1 is the preset time difference, and m is a positive integer greater than or equal to 2.
In this way, a plurality of first fault characteristic values and a plurality of second fault characteristic values can be obtained from a plurality of sampled historical fault characteristic values, and the plurality of first fault characteristic values and the second fault characteristic value corresponding to each first fault characteristic value form a training sample of the feature prediction model.
S302, aiming at each first fault characteristic value in the plurality of first fault characteristic values, taking the first fault characteristic value as input, taking a second fault characteristic value corresponding to the first fault characteristic value as output, and training the characteristic prediction model.
Specifically, according to the above method, a plurality of pairs of first fault characteristic values and second fault characteristic values can be obtained, for each first fault characteristic value, the first fault characteristic value is used as an input of a characteristic prediction model, the second fault characteristic value corresponding to the first fault characteristic value is used as an output of the characteristic prediction model, and parameters of the characteristic prediction model are adjusted to realize training of the characteristic prediction model.
Optionally, the feature prediction model may include: a network structure of one input layer, two hidden layers and one output layer. Optionally, the input layer adopts 2 nerve units, the first hidden layer adopts 50 neurons, the second hidden layer adopts 100 neurons, and the output layer adopts one nerve unit.
In one example, refining the training process of the feature prediction model, the training samples for training the fault feature model may be divided into: a training sample set and a testing sample set.
The training sample set may be a training sample set which takes two fault characteristic values of the sampled historical turn-off instantaneous peak voltage and the historical collector-emitter saturation voltage drop of the IGBT module as column vectors, takes a time sequence obtained by resampling the same fault characteristic as a row time sequence vector, and forms a two-variable and multi-dimensional fault characteristic time sequence matrix as the characteristic prediction model.
The test sample set may be a multivariate and multidimensional test sample set for testing the feature prediction model obtained by delaying a row time series vector (or a column time series vector) of a training sample set of the prediction model.
For example, when the preset time difference is one week, the historical collector-emitter electrical signals obtained when the IGBT modules arranged in time sequence operate at the historical time are recorded as: u1, u2 and u3 … un, firstly, with 7 days as a time window (or sampling interval), collecting historical collector-emitter electrical signals of u1 to u7, u8 to u14 …, and the like by adopting a method of S2031 as a training sample set of the feature prediction model. And taking 7 days as a time window (or sampling interval), and collecting historical collector-emitter electric signals of u 2-u 8, u 9-u 15 … and the like as a test sample set of the characteristic prediction model by adopting the method of S2031. According to the training data obtained by the method, the training data used for training the feature prediction model each time can be ensured to be different, and the training speed and the reliability of the feature prediction model are improved.
Therefore, the training sample set can be input into the feature prediction module, parameters of the feature prediction model are adjusted, and the accuracy of the trained feature prediction model is tested by using the test sample set.
S303, extracting a plurality of third fault characteristic values in the plurality of sampled historical fault characteristic values and fault state information corresponding to each piece of third fault characteristic value information.
The fault characteristic value further comprises fault state information of the IGBT module.
Optionally, the fault state information of the IGBT module includes: normal, alert and dangerous.
And S304, aiming at each third fault characteristic value in the plurality of third fault characteristic values, taking the third fault characteristic value as input, taking fault state information corresponding to the third fault characteristic value as output, and training the IGBT module state evaluation model.
In the embodiment of the present application, the third fault characteristic value includes a historical turn-off instantaneous spike voltage value of the IGBT module and a historical collector-emitter saturation voltage drop value of the IGBT module.
Therefore, the IGBT module state evaluation model comprises two inputs and one output, the historical turn-off instantaneous peak voltage value of the IGBT module and the saturation voltage drop value of the historical collector-emitter are used as the two inputs of the IGBT module state evaluation model, the fault state information corresponding to the historical turn-off instantaneous peak voltage value of the IGBT module and the saturation voltage drop value of the historical collector-emitter is used as the output, and the parameters of the IGBT module state evaluation model are adjusted, so that the trained IGBT module state evaluation model can accurately predict the fault state information of the IGBT module at the future moment.
Alternatively, as shown in fig. 5, the output of the characteristic prediction model is the input of the IGBT module state evaluation model, and therefore, in order to reduce the selection workload of the training data of the IGBT module state evaluation model, the second fault characteristic value may be used as the third fault characteristic value.
Optionally, the training sample for training the IGBT module state evaluation model includes a training sample set and a testing sample set, where the training sample set is used to adjust parameters in the IGBT module state evaluation model, and the testing sample set is used to test whether the trained IGBT module state evaluation model is accurate.
In one example, the embodiment of the application may train the feature prediction model and the IGBT module state evaluation model by using a back propagation algorithm.
Specifically, for a training sample of the feature prediction model, a Mean Absolute Error (MAE) loss function is used for testing the feature prediction model, a hidden layer performs back propagation on a gradient obtained by calculating the loss function, all weights in a formula are adjusted, an Adam algorithm is used for generating an optimization parameter for each iteration learning, the parameters of the feature prediction model are adjusted until the loss function is converged to obtain appropriate network parameters, and the feature prediction model is stored.
Training the IGBT module state evaluation model by using the training sample test sample of the IGBT module state evaluation model, adjusting parameters of the IGBT module state evaluation model to obtain appropriate network parameters, and storing the IGBT module state evaluation model.
In the embodiment, the feature prediction model and the IGBT module state evaluation model are trained by using the training sample set according to a back propagation algorithm, the prediction effects of the feature prediction model and the IGBT module state evaluation model are tested by using the test sample set, if the test effects meet requirements, the next step is carried out, otherwise, the steps are repeated until a satisfactory prediction effect is obtained, and the finally obtained feature prediction model and the IGBT module state evaluation model are stored. After the model training is finished, the output layer carries out processing such as inverse normalization on the result, and the predicted value is restored to be in a time sequence data format consistent with the input.
According to the embodiment of the application, when a fault prediction model comprises a characteristic prediction model and an IGBT module state evaluation model, a plurality of first fault characteristic values and a plurality of second fault characteristic values in a plurality of sampled historical fault characteristic values are extracted, the first fault characteristic values are used as input aiming at each first fault characteristic value in the plurality of first fault characteristic values, the second fault characteristic values corresponding to the first fault characteristic values are used as output, the characteristic prediction model is trained, wherein the first fault characteristic values correspond to second faults one by one, and the time difference between the corresponding first fault characteristic values and the corresponding second fault characteristic values is a preset time value; meanwhile, a plurality of third fault characteristic values in the plurality of sampled historical fault characteristic values and fault state information corresponding to each piece of third fault characteristic value information are extracted, the third fault characteristic values are used as input aiming at each third fault characteristic value in the plurality of third fault characteristic values, the fault state information corresponding to the third fault characteristic values is used as output, an IGBT module state evaluation model is trained, the fault characteristic values further comprise the fault state information of the IGBT module, and therefore accurate training of a characteristic prediction model and the IGBT module state evaluation model is achieved.
Based on the fault prediction model shown in fig. 5, in this case, in step S103, inputting the fault characteristic value of the IGBT module at the first time into a fault prediction model trained in advance, and predicting a fault result when the IGBT module operates at the second time, as shown in fig. 7, the method may include:
s401, inputting the fault characteristic value of the IGBT module at the first time into the pre-trained characteristic prediction model, and predicting the fault characteristic value of the IGBT module when the IGBT module works at the second time.
S402, inputting the fault characteristic value of the IGBT module when the IGBT module works at the second time into the IGBT module state evaluation model, and predicting the fault state information of the IGBT module when the IGBT module works at the second time.
In practical application, when the fault condition of the IGBT module during operation at the second time in the future is predicted, the collector-emitter electrical signal of the IGBT module during operation at the first time is obtained, and the fault characteristic value of the IGBT module at the first time is extracted from the collector-emitter electrical signal of the IGBT module during operation at the first time by using the method in S102. And then, inputting the fault characteristic value of the IGBT module at the first time into the trained characteristic prediction model, wherein the fault characteristic value of the IGBT module when the IGBT module works at the second time in the future can be predicted by the characteristic prediction model. Then, the characteristic prediction model inputs the predicted fault characteristic value of the IGBT module when the IGBT module works at the second time into the IGBT module state evaluation model, and the IGBT module state evaluation model predicts the fault state information of the IGBT module when the IGBT module works at the second time.
Optionally, the characteristic prediction model outputs the predicted fault characteristic value of the IGBT module when the IGBT module operates at the second time in the future to the electronic device, and the electronic device or the user may determine, according to the predicted fault characteristic value of the IGBT module when the IGBT module operates at the second time in the future, whether the state parameter of the IGBT module changes suddenly when the IGBT module operates at the second time, and determine whether the development trend of the fault characteristic value is abnormal.
Similarly, the electronic device or the user can predict the fault state information of the IGBT module when the IGBT module operates at the second time according to the state evaluation model of the IGBT module, and determine whether the IGBT module will have a fault when the IGBT module operates at the second time, what kind of fault occurs, and the like.
That is, according to the embodiment of the application, the change situation of the fault characteristic value of the IGBT module in the second time in the future can be predicted based on the fault characteristic value of the IGBT module in the first time, whether the IGBT module will have a fault in the second time, and what kind of fault is sent, so that the fault of the IGBT module can be accurately predicted.
Fig. 8 is a schematic diagram of a failure prediction apparatus of an IGBT module according to an embodiment of the present application, where the failure prediction apparatus of the IGBT module is applied to an electronic device, and the failure prediction apparatus of the IGBT module may be the electronic device, or may be a component (e.g., an integrated circuit, a chip, or the like) of the electronic device, as shown in fig. 8, the failure prediction apparatus 100 of the IGBT module includes:
the obtaining module 110 is configured to obtain a collector-emitter electrical signal when the IGBT module operates at a first time;
a determining module 120, configured to determine a fault characteristic value of the IGBT module at a first time based on a collector-emitter electrical signal of the IGBT module when the IGBT module operates at the first time, where the fault characteristic value includes a turn-off instantaneous spike voltage value of the IGBT module and a collector-emitter saturation voltage drop value of the IGBT module;
the prediction module 130 is configured to input the fault feature value of the IGBT module at the first time into a fault prediction model trained in advance, and predict a fault result of the IGBT module when the IGBT module operates at a second time, where the second time is a future time of the first time, and a time difference between the first time and the second time is equal to a preset time value.
The fault prediction apparatus of the IGBT module according to the embodiment of the present application may be configured to implement the technical solution of the above method embodiment, and its implementation principle and technical effect are similar, which are not described herein again.
Fig. 9 is a schematic diagram of a failure prediction apparatus for an IGBT module according to another embodiment of the present application, on the basis of the above embodiment, the apparatus further includes a training module 140;
the obtaining module 110 is further configured to obtain a historical collector-emitter electrical signal of the IGBT module when the IGBT module operates in a historical time;
the determining module 120 is further configured to determine a historical fault characteristic value of the IGBT module based on a historical collector-emitter electrical signal of the IGBT module;
the training module 140 is configured to train the fault prediction model by using the historical fault feature value of the IGBT module.
The fault prediction apparatus of the IGBT module according to the embodiment of the present application may be configured to implement the technical solution of the above method embodiment, and its implementation principle and technical effect are similar, which are not described herein again.
Fig. 10 is a schematic diagram of a failure prediction apparatus of an IGBT module according to another embodiment of the present application, on the basis of the above embodiment, the training module 140 includes an obtaining unit 141 and a training unit 142;
the obtaining unit 141 is configured to sample the historical fault feature value of the IGBT module by using the preset time value as a time sampling interval, and obtain a plurality of sampled historical fault feature values;
the training unit 142 is configured to train the fault prediction model using the plurality of sampled historical fault feature values.
In a possible implementation manner, the obtaining unit 141 is specifically configured to divide the historical fault feature value of the IGBT module into a plurality of sampling intervals by taking the preset time value as a time sampling interval; and aiming at each sampling interval, taking the median value or the mean value of the historical fault characteristic values of the IGBT module in the sampling interval as the historical fault characteristic value of the sampling interval, and obtaining a plurality of sampled historical fault characteristic values.
In a possible implementation manner, the obtaining unit 141 is further configured to sort the historical fault feature values of the IGBT modules in a time sequence.
In one possible implementation, the fault prediction model includes: the IGBT module state evaluation system comprises a characteristic prediction model and an IGBT module state evaluation model, wherein the output of the characteristic prediction model is the input of the IGBT module state evaluation model.
In a possible implementation manner, the training unit 142 is specifically configured to extract a plurality of first fault feature values and a plurality of second fault feature values in the plurality of sampled historical fault feature values, where the first fault feature values correspond to the second faults one by one, and a time difference between the corresponding first fault feature values and the corresponding second fault feature values is the preset time value; aiming at each first fault characteristic value in a plurality of first fault characteristic values, taking the first fault characteristic value as input, taking a second fault characteristic value corresponding to the first fault characteristic value as output, and training the characteristic prediction model; extracting a plurality of third fault characteristic values in the plurality of sampled historical fault characteristic values and fault state information corresponding to each piece of third fault characteristic value information, wherein the fault characteristic values further comprise the fault state information of the IGBT module; and aiming at each third fault characteristic value in the plurality of third fault characteristic values, taking the third fault characteristic value as input, taking fault state information corresponding to the third fault characteristic value as output, and training the IGBT module state evaluation model.
Optionally, the second fault characteristic value is the same as the third fault characteristic value.
The fault prediction apparatus of the IGBT module according to the embodiment of the present application may be configured to implement the technical solution of the above method embodiment, and its implementation principle and technical effect are similar, which are not described herein again.
Fig. 11 is a schematic diagram of a failure prediction apparatus of an IGBT module according to another embodiment of the present application, and on the basis of the above embodiment, the apparatus further includes a removal module 150;
the removing module 150 is configured to remove a dead pixel in the historical collector-emitter electrical signal of the IGBT module;
the determining module 120 is configured to determine a historical fault characteristic value of the IGBT module based on the historical collector-emitter electrical signal of the IGBT module after the dead pixel is removed.
In a possible implementation manner, the removing module 150 is specifically configured to obtain an arithmetic average of historical collector-emitter electrical signals of the IGBT module; and removing the electric signals of which the deviation from the arithmetic mean value is larger than a preset value in the historical collector-emitter electric signals of the IGBT module.
In a possible implementation manner, the determining module 120 is specifically configured to determine, from the historical collector-emitter electrical signals of the IGBT module after the dead pixel is removed, all historical turn-off instantaneous spike voltage values corresponding to when the IGBT module enters a turn-off region and all historical collector-emitter saturation voltage drop values corresponding to when the IGBT module enters a saturation region in each operating cycle of the IGBT module.
In some embodiments, the apparatus further comprises a normalization module 160:
the normalization module 160 is configured to perform normalization processing on the historical fault characteristic value of the IGBT module;
the training module 140 is configured to train the fault prediction model by using the normalized historical fault feature value of the IGBT module.
In a possible implementation manner, the prediction module 130 is specifically configured to input the fault characteristic value of the IGBT module at the first time into the pre-trained characteristic prediction model, and predict the fault characteristic value of the IGBT module when the IGBT module operates at the second time; and inputting the fault characteristic value of the IGBT module when the IGBT module works at the second time into the IGBT module state evaluation model, and predicting the fault state information of the IGBT module when the IGBT module works at the second time.
Optionally, the fault state information of the IGBT module includes: normal, alert and dangerous.
In a possible implementation manner, the predicting module 130 is further configured to predict whether the state parameter changes suddenly when the IGBT module operates at the second time based on the fault characteristic value when the IGBT module operates at the second time.
The fault prediction apparatus of the IGBT module according to the embodiment of the present application may be configured to implement the technical solution of the above method embodiment, and its implementation principle and technical effect are similar, which are not described herein again.
Fig. 12 is a schematic structural diagram of a failure prediction apparatus of an IGBT module according to an embodiment of the present application. The failure prediction device 700 of the IGBT module exists in the product form of a chip, the failure prediction device of the IGBT module structurally includes a processor 701 and a memory 702, the memory 702 is configured to be coupled with the processor 701, the memory 702 stores necessary program instructions and data of the device, and the processor 701 is configured to execute the program instructions stored in the memory 702, so that the device performs the functions of the electronic device in the above-mentioned method embodiments.
The failure prediction device of the IGBT module according to the embodiment of the present application may be configured to implement the technical solutions of the electronic devices in the above method embodiments, and the implementation principles and technical effects thereof are similar and will not be described herein again.
Fig. 13 is a schematic structural diagram of an electronic device according to an embodiment of the present application. The electronic device 600 may implement the functions executed by the electronic device in the above method embodiments, and the functions may be implemented by hardware, or may be implemented by hardware executing corresponding software. The hardware or software includes one or more modules or units corresponding to the above functions.
In one possible design, the electronic device 600 includes a processor 601, a transceiver 602, and a memory 603 in its structure, and the processor 601 is configured to support the electronic device 600 to perform the corresponding functions of the above-described method. The transceiver 602 is used to support communication between the electronic device 600 and other electronic devices or servers. The electronic device 600 may further comprise a memory 603, the memory 603 being adapted to be coupled to the processor 601 and to store program instructions and data necessary for the electronic device 600.
When the electronic device 600 is powered on, the processor 601 may read the program instructions and data in the memory 603, interpret and execute the program instructions, and process the data of the program instructions. When data needs to be transmitted, the processor 601 outputs data to be transmitted to the transceiver 602, and the transceiver 602 transmits the data to be transmitted to the outside. When data is transmitted to the electronic device, the transceiver 602 outputs the received data to the processor 601, and the processor 601 processes the data.
Those skilled in the art will appreciate that fig. 13 shows only one memory 603 and one processor 601 for ease of illustration. In an actual electronic device 600, there may be multiple processors 601 and multiple memories 603. The memory 603 may also be referred to as a storage medium or a storage device, etc., which is not limited in this application.
The electronic device of the embodiment of the application can be used for executing the technical scheme of the electronic device in the above method embodiments, and the implementation principle and the technical effect are similar, which are not described herein again.
Based on such understanding, the technical solution of the present application may be substantially implemented or contributed by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) or a processor (processor) to execute all or part of the steps of the apparatus according to the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, cause the processes or functions described in accordance with the embodiments of the application to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing embodiments of the apparatuses, and are not described herein again. In addition, the device embodiments and the device embodiments may also refer to each other, and the same or corresponding contents in different embodiments may be referred to each other, which is not described in detail.

Claims (18)

1. A fault prediction method for an IGBT module is characterized by comprising the following steps:
acquiring a collector-emitter electrical signal of the IGBT module when the IGBT module works at a first time;
determining a fault characteristic value of the IGBT module at a first time based on a collector-emitter electrical signal of the IGBT module when the IGBT module works at the first time, wherein the fault characteristic value comprises a turn-off instantaneous spike voltage value of the IGBT module and a saturation voltage drop value of a collector-emitter of the IGBT module;
inputting the fault characteristic value of the IGBT module at the first time into a pre-trained fault prediction model, and predicting a fault result of the IGBT module when the IGBT module works at a second time, wherein the second time is the future time of the first time, and the time difference between the first time and the second time is equal to a preset time value.
2. The method of claim 1, further comprising:
acquiring historical collector-emitter electrical signals of the IGBT module when the IGBT module works in historical time;
determining a historical fault characteristic value of the IGBT module based on a historical collector-emitter electrical signal of the IGBT module;
and training the fault prediction model by using the historical fault characteristic value of the IGBT module.
3. The method of claim 2, wherein the training the fault prediction model using historical fault signature values for the IGBT modules comprises:
sampling historical fault characteristic values of the IGBT module by taking the preset time value as a time sampling interval to obtain a plurality of sampled historical fault characteristic values;
and training the fault prediction model by using the plurality of sampled historical fault characteristic values.
4. The method according to claim 3, wherein the sampling the historical fault characteristic value of the IGBT module by taking the preset time value as a time sampling interval to obtain a plurality of sampled historical fault characteristic values comprises:
dividing the historical fault characteristic value of the IGBT module into a plurality of sampling intervals by taking the preset time value as a time sampling interval;
and aiming at each sampling interval, taking the median value or the mean value of the historical fault characteristic values of the IGBT module in the sampling interval as the historical fault characteristic value of the sampling interval, and obtaining a plurality of sampled historical fault characteristic values.
5. The method according to claim 4, wherein before sampling the historical fault characteristic value of the IGBT module by taking the preset time value as a time sampling interval, the method further comprises the following steps:
and sequencing the historical fault characteristic values of the IGBT modules according to a time sequence.
6. The method of claim 5, wherein the fault prediction model comprises: the IGBT module state evaluation system comprises a characteristic prediction model and an IGBT module state evaluation model, wherein the output of the characteristic prediction model is the input of the IGBT module state evaluation model.
7. The method of claim 6, wherein the training the fault prediction model using the plurality of sampled historical fault signature values comprises:
extracting a plurality of first fault characteristic values and a plurality of second fault characteristic values in the plurality of sampled historical fault characteristic values, wherein the first fault characteristic values correspond to the second faults one by one, and the time difference between the corresponding first fault characteristic values and the corresponding second fault characteristic values is the preset time value;
aiming at each first fault characteristic value in a plurality of first fault characteristic values, taking the first fault characteristic value as input, taking a second fault characteristic value corresponding to the first fault characteristic value as output, and training the characteristic prediction model;
extracting a plurality of third fault characteristic values in the plurality of sampled historical fault characteristic values and fault state information corresponding to each piece of third fault characteristic value information, wherein the fault characteristic values further comprise the fault state information of the IGBT module;
and aiming at each third fault characteristic value in the plurality of third fault characteristic values, taking the third fault characteristic value as input, taking fault state information corresponding to the third fault characteristic value as output, and training the IGBT module state evaluation model.
8. The method of claim 7, wherein the second fault signature value is the same as the third fault signature value.
9. The method according to any one of claims 2-8, wherein before determining the historical fault signature of the IGBT module based on the historical collector-emitter electrical signals of the IGBT module, the method comprises:
removing bad points in historical collector-emitter electrical signals of the IGBT module;
determining a historical fault signature value for the IGBT module based on historical collector-emitter electrical signals for the IGBT module, comprising:
and determining a historical fault characteristic value of the IGBT module based on the historical collector-emitter electrical signal of the IGBT module after the dead pixel is removed.
10. The method of claim 9, wherein the removing the dead spots in the historical collector-emitter electrical signals of the IGBT module comprises:
acquiring an arithmetic mean value of historical collector-emitter electrical signals of the IGBT module;
and removing the electric signals of which the deviation from the arithmetic mean value is larger than a preset value in the historical collector-emitter electric signals of the IGBT module.
11. The method of claim 9, wherein determining the historical fault signature of the IGBT module based on the historical collector-emitter electrical signal of the IGBT module after dead pixel removal comprises:
and determining all historical turn-off instantaneous peak voltage values corresponding to the IGBT module entering a turn-off region and all historical saturation voltage drop values corresponding to the IGBT module entering a saturation region in each working cycle of the IGBT module from the historical collector-emitter electric signals of the IGBT module after dead spots are removed.
12. The method of claim 2, wherein before training the fault prediction model using historical fault signature values for the IGBT modules, the method further comprises:
carrying out normalization processing on the historical fault characteristic value of the IGBT module;
training the fault prediction model using historical fault feature values of the IGBT module, including:
and training the fault prediction model by using the normalized historical fault characteristic value of the IGBT module.
13. The method according to any one of claims 3-9, wherein the inputting the fault characteristic value of the IGBT module at the first time into a fault prediction model trained in advance to predict the fault result of the IGBT module when operating at the second time comprises:
inputting the fault characteristic value of the IGBT module at the first time into the pre-trained characteristic prediction model, and predicting the fault characteristic value of the IGBT module when the IGBT module works at the second time;
and inputting the fault characteristic value of the IGBT module when the IGBT module works at the second time into the IGBT module state evaluation model, and predicting the fault state information of the IGBT module when the IGBT module works at the second time.
14. The method of claim 13, wherein the fault status information of the IGBT module comprises: normal, alert and dangerous.
15. The method of claim 13, further comprising:
and predicting whether the state parameters of the IGBT module are suddenly changed when the IGBT module works at the second time based on the fault characteristic value of the IGBT module when the IGBT module works at the second time.
16. A failure prediction apparatus of an IGBT module, characterized by comprising:
the acquisition module is used for acquiring collector-emitter electrical signals when the IGBT module works at the first time;
the determining module is used for determining a fault characteristic value of the IGBT module at a first time based on a collector-emitter electric signal when the IGBT module works at the first time, wherein the fault characteristic value comprises a turn-off instantaneous spike voltage value of the IGBT module and a collector-emitter saturation voltage drop value of the IGBT module;
and the prediction module is used for inputting the fault characteristic value of the IGBT module at the first time into a pre-trained fault prediction model, and predicting a fault result of the IGBT module when the IGBT module works at a second time, wherein the second time is the future time of the first time, and the time difference between the first time and the second time is equal to a preset time value.
17. An electronic device comprising a processor and a memory;
the memory for storing a computer program;
the processor for executing the computer program to implement the method for predicting a failure of an IGBT module according to any one of claims 1 to 15.
18. A computer storage medium, characterized in that the storage medium comprises computer instructions which, when executed by a computer, cause the computer to implement the method of fault prediction of an IGBT module according to any one of claims 1-15.
CN201911094431.8A 2019-11-11 2019-11-11 Fault prediction method and device of IGBT module and storage medium Pending CN110852509A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911094431.8A CN110852509A (en) 2019-11-11 2019-11-11 Fault prediction method and device of IGBT module and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911094431.8A CN110852509A (en) 2019-11-11 2019-11-11 Fault prediction method and device of IGBT module and storage medium

Publications (1)

Publication Number Publication Date
CN110852509A true CN110852509A (en) 2020-02-28

Family

ID=69601191

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911094431.8A Pending CN110852509A (en) 2019-11-11 2019-11-11 Fault prediction method and device of IGBT module and storage medium

Country Status (1)

Country Link
CN (1) CN110852509A (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112214951A (en) * 2020-10-19 2021-01-12 武汉大学 Sectional type LSTM microgrid inverter IGBT reliability evaluation method and system
CN112213660A (en) * 2020-08-25 2021-01-12 广西电网有限责任公司南宁供电局 Method for predicting residual life of power electronic device in UPS system
CN113625144A (en) * 2021-08-11 2021-11-09 北京信息科技大学 IGBT fault prediction method and system
CN113848445A (en) * 2021-10-11 2021-12-28 中国华能集团清洁能源技术研究院有限公司 Method for monitoring short-circuit and open-circuit faults of IGBT (insulated Gate Bipolar transistor) in SST (Voltage tester) based on Vce
CN117031235A (en) * 2023-07-31 2023-11-10 中南大学 IGBT fault diagnosis method and device based on time domain reflection signals
CN117590190A (en) * 2024-01-18 2024-02-23 国网山东省电力公司营销服务中心(计量中心) Photovoltaic inverter IGBT aging fault prediction method and system

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH07128395A (en) * 1992-10-15 1995-05-19 Nippon Inter Electronics Corp Inspection method for semiconductor chip
WO2001095354A1 (en) * 2000-06-07 2001-12-13 Abb Ab A method and a device for prediction of a zero-crossing of an alternating current
US20130177041A1 (en) * 2012-01-11 2013-07-11 Abb Research Ltd System and method for monitoring in real time the operating state of an igbt device
CN104237761A (en) * 2013-06-13 2014-12-24 通用电气公司 Failure mode detection and protection system and method of insulated gate bipolar transistor
CN106597245A (en) * 2016-11-04 2017-04-26 山东科技大学 IGBT fault monitoring device and method
CN107621782A (en) * 2017-05-08 2018-01-23 安徽理工大学 A kind of method for diagnosing faults of grid bipolar transistor (IGBT) module
CN108957271A (en) * 2017-05-26 2018-12-07 许继集团有限公司 A kind of IGBT driving over current fault monitoring method and device
CN110266296A (en) * 2019-06-24 2019-09-20 北京机械设备研究所 The driving circuit of IGBT a kind of and IGBT guard method
DE102018110291A1 (en) * 2018-04-27 2019-10-31 Gottfried Wilhelm Leibniz Universität Hannover Method for determining the aging and for switching off or function restriction of a bipolar transistor and device with a bipolar transistor

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH07128395A (en) * 1992-10-15 1995-05-19 Nippon Inter Electronics Corp Inspection method for semiconductor chip
WO2001095354A1 (en) * 2000-06-07 2001-12-13 Abb Ab A method and a device for prediction of a zero-crossing of an alternating current
US20130177041A1 (en) * 2012-01-11 2013-07-11 Abb Research Ltd System and method for monitoring in real time the operating state of an igbt device
CN104237761A (en) * 2013-06-13 2014-12-24 通用电气公司 Failure mode detection and protection system and method of insulated gate bipolar transistor
CN106597245A (en) * 2016-11-04 2017-04-26 山东科技大学 IGBT fault monitoring device and method
CN107621782A (en) * 2017-05-08 2018-01-23 安徽理工大学 A kind of method for diagnosing faults of grid bipolar transistor (IGBT) module
CN108957271A (en) * 2017-05-26 2018-12-07 许继集团有限公司 A kind of IGBT driving over current fault monitoring method and device
DE102018110291A1 (en) * 2018-04-27 2019-10-31 Gottfried Wilhelm Leibniz Universität Hannover Method for determining the aging and for switching off or function restriction of a bipolar transistor and device with a bipolar transistor
CN110266296A (en) * 2019-06-24 2019-09-20 北京机械设备研究所 The driving circuit of IGBT a kind of and IGBT guard method

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
李为柱: "《2000版ISO9000族标准统计技术应用教程》", 31 March 2001 *
李宁: "基于机器学习算法的IGBT模块故障预测技术研究", 《中国优秀博硕士学位论文全文数据库(硕士) 信息科技辑》 *
林强: "《机器学习、深度学习与强化学习》", 31 May 2019 *
高志强: "《深度学习 从入门到实战》", 30 June 2018, 中国铁道出版社 *

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112213660A (en) * 2020-08-25 2021-01-12 广西电网有限责任公司南宁供电局 Method for predicting residual life of power electronic device in UPS system
CN112214951A (en) * 2020-10-19 2021-01-12 武汉大学 Sectional type LSTM microgrid inverter IGBT reliability evaluation method and system
CN112214951B (en) * 2020-10-19 2022-05-24 武汉大学 Sectional type LSTM microgrid inverter IGBT reliability evaluation method and system
CN113625144A (en) * 2021-08-11 2021-11-09 北京信息科技大学 IGBT fault prediction method and system
CN113848445A (en) * 2021-10-11 2021-12-28 中国华能集团清洁能源技术研究院有限公司 Method for monitoring short-circuit and open-circuit faults of IGBT (insulated Gate Bipolar transistor) in SST (Voltage tester) based on Vce
CN117031235A (en) * 2023-07-31 2023-11-10 中南大学 IGBT fault diagnosis method and device based on time domain reflection signals
CN117590190A (en) * 2024-01-18 2024-02-23 国网山东省电力公司营销服务中心(计量中心) Photovoltaic inverter IGBT aging fault prediction method and system

Similar Documents

Publication Publication Date Title
CN110852509A (en) Fault prediction method and device of IGBT module and storage medium
US10685159B2 (en) Analog functional safety with anomaly detection
CN111046564B (en) Residual life prediction method for two-stage degraded product
US11243524B2 (en) System and method for unsupervised root cause analysis of machine failures
Lindemann et al. Anomaly detection and prediction in discrete manufacturing based on cooperative LSTM networks
CN105071983A (en) Abnormal load detection method for cloud calculation on-line business
JP5827426B1 (en) Predictive diagnosis system and predictive diagnosis method
JP5827425B1 (en) Predictive diagnosis system and predictive diagnosis method
CN111045894B (en) Database abnormality detection method, database abnormality detection device, computer device and storage medium
CN104390657A (en) Generator set operating parameter measuring sensor fault diagnosis method and system
CN102042848A (en) Prediction method of multi-functional parameter accelerated degradation testing product life based on multivariate hybrid time sequence analysis
JP5771317B1 (en) Abnormality diagnosis apparatus and abnormality diagnosis method
Xu et al. An adaptive gamma process based model for residual useful life prediction
US11657148B2 (en) Event analysis in an electric power system
De Simone et al. LSTM-based failure prediction for railway rolling stock equipment
CN105488335A (en) Lyapunov exponent based power system load prediction method and apparatus
CN114970665A (en) Model training method, electrolytic capacitor residual life prediction method and system
CN105718733A (en) Fault predicting method based on fuzzy nearness and particle filter
CN115600695B (en) Fault diagnosis method for metering equipment
Alonso-González et al. A big data architecture for fault prognostics of electronic devices: application to power MOSFETs
JP5771318B1 (en) Abnormality diagnosis apparatus and abnormality diagnosis method
CN111382891B (en) Short-term load prediction method and short-term load prediction device
CN113553232B (en) Technology for carrying out unsupervised anomaly detection on operation and maintenance data through online matrix image
CN116720983A (en) Power supply equipment abnormality detection method and system based on big data analysis
Ji et al. Reliability Study on Control Unit of Metro Train Auxiliary Inverter Based on Improved Monte Carlo Algorithm

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20200228