Fault risk prediction method and device

Abstract

The invention provides a fault risk prediction method and device. The method comprises the following steps: determining a target prediction variable and future prediction time for performing fault risk prediction on to-be-tested equipment; obtaining at least one fault prediction model corresponding to the target prediction variable and a first data set formed by data which is obtained before future prediction time and has influence on the fault of the to-be-tested equipment in the aspect of the target prediction variable; obtaining a prediction value of each fault prediction model for performing fault risk prediction on the to-be-tested equipment based on the first data set; and performing fault risk prediction on the to-be-tested equipment based on all the prediction values. According to the method, the fault risk in the future prediction time can be predicted by using the fault prediction model based on the data obtained before the future prediction time, so that the possible fault of the to-be-detected equipment can be found in advance, and the possible fault of the to-be-detected equipment can be maintained in advance in time; and loss caused by real occurrence of a fault is avoided, and the applicability is better.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a fault risk prediction method and device. Background Art

[0002] In the metallurgy, steel, power generation, petroleum, chemical industry, water supply and other industries, high-voltage inverters are widely used to drive fans, pumps, compressors, and various large rotating machinery, becoming one of the main means of energy saving and automation transformation and upgrading. With the development of artificial intelligence and Internet of Things technology, modern industry has higher and higher requirements for high-voltage inverters. In addition to requiring high-frequency switching technology and intelligent control technology of high-voltage inverters, the requirements for the reliability and maintenance technology of high-voltage inverters are also getting higher and higher.

[0003] However, due to the various use environments of high-voltage inverters and different levels of human operation, various faults may occur during the use of high-voltage inverters. Once a fault occurs, the speed regulation function of the high-voltage inverter cannot be used normally, or even equipment damage and line stoppage may occur, resulting in huge losses. Therefore, it is very important to discover possible faults of high-voltage inverters as early as possible and to troubleshoot or repair them in advance.

[0004] Based on this, how to discover the possible failure risks of the high-voltage inverter as early as possible so that the possible failures can be checked or repaired in advance has become a technical problem that needs to be solved urgently by technical personnel in this field. Summary of the invention

[0005] In order to solve the problem of how to discover the possible fault risks of a high-voltage inverter as early as possible so that the possible faults can be checked or repaired in advance, an embodiment of the present application provides a fault risk prediction method and device.

[0006] In a first aspect, an embodiment of the present application provides a fault risk prediction method, the method comprising: determining a target prediction variable and a future prediction time for fault risk prediction of a device under test; obtaining at least one fault prediction model corresponding to the target prediction variable; obtaining a first data set; the first data set comprising data obtained before the future prediction time, which has an impact on the fault of the device under test in terms of the target prediction variable; obtaining a prediction value of each fault prediction model for fault risk prediction of the device under test based on the first data set; and performing fault risk prediction on the device under test based on all the prediction values.

[0007] In one possible implementation, the device to be tested is a frequency converter in a target system; the target system includes a power generation system, an energy extraction system or a building system; the target prediction variable includes at least one of the following variables: input voltage, input current, output voltage, output current, operating frequency, control cabinet temperature, power cabinet temperature, transformer cabinet temperature, and humidity inside the frequency converter cabinet.

[0008] In one possible implementation, the future prediction time is the next moment after the prediction start moment; the obtaining of the first data set includes: obtaining a second data set; the second data set includes data obtained at the prediction start moment that has an impact on the failure of the device under test in terms of the target prediction variable; and determining that the second data set is the first data set.

[0009] In a possible implementation, the future prediction time is a first time period starting from the prediction start time; the duration of the first time period is a preset duration; the obtaining of the first data set includes: obtaining a third data set; the third data set includes data obtained in a second time period that has an impact on the failure of the device under test in terms of the target prediction variable; the end time of the second time period is the prediction start time, and the duration of the second time period is the preset duration; and the third data set is determined to be the first data set.

[0010] In one possible implementation, obtaining at least one fault prediction model corresponding to the target prediction variable includes: obtaining a fourth data set; the fourth data set includes historical data that is correlated with the failure of the device under test in terms of the target prediction variable; based on a preset feature group division rule, dividing the data included in the fourth data set into at least one feature group; based on the at least one feature group, constructing the at least one fault prediction model.

[0011] In a possible implementation, the obtaining of the fourth data set includes: obtaining a fifth data set; the fifth data set includes first historical business data, first historical time series data, and / or first historical fault alarm data that are correlated with the failure of the device under test in terms of the target prediction variable, and second historical business data, second historical time series data, and / or second historical fault alarm data that are correlated with the failure of an associated device of the device under test in terms of the target prediction variable; preprocessing the data included in the fifth data set to obtain a sixth data set; performing feature expansion based on the sixth data set to obtain a seventh data set; and determining that the union of the sixth data set and the seventh data set is the fourth data set.

[0012] In one possible implementation, the preprocessing of the data included in the fifth data set includes: based on a preset business data format, format normalization processing of the first historical business data and the second historical business data; data cleaning processing of the first historical time series data and the second historical time series data; based on a preset fault alarm data format, format conversion processing of the first historical fault alarm data and the second historical fault alarm data; normalization processing of the numerical data in the fifth data set.

[0013] In one possible implementation, constructing the at least one fault prediction model based on the at least one feature group includes: using a principal component analysis method or a kernel principal component analysis method to perform dimensionality reduction processing on the at least one feature group to obtain a reduced dimensionality group of each feature group; each reduced dimensionality group includes data corresponding to a principal component that has a fault risk impact on the target prediction variable of the device under test; and constructing the at least one fault prediction model based on each reduced dimensionality group.

[0014] In a possible implementation manner, the data included in the first data set corresponds to the principal component.

[0015] In the second aspect, an embodiment of the present application also provides a fault risk prediction device, which includes: a determination module, used to determine a target prediction variable and a future prediction time for fault risk prediction of a device under test; a first acquisition module, used to acquire at least one fault prediction model corresponding to the target prediction variable; a second acquisition module, used to acquire a first data set; the first data set includes data obtained before the future prediction time, which has an impact on the fault of the device under test in terms of the target prediction variable; a third acquisition module, used to acquire the predicted value of each fault prediction model for fault risk prediction of the device under test based on the first data set; a prediction module, used to predict the fault risk of the device under test based on all the predicted values.

[0016] In one possible implementation, the device to be tested is a frequency converter in a target system; the target system includes a power generation system, an energy extraction system or a building system; the target prediction variable includes at least one of the following variables: input voltage, input current, output voltage, output current, operating frequency, control cabinet temperature, power cabinet temperature, transformer cabinet temperature, and humidity inside the frequency converter cabinet.

[0017] In one possible implementation, the future prediction time is the next moment after the prediction start moment; the second acquisition module is used to acquire the first data set, specifically: the second acquisition module is used to: acquire the second data set; the second data set includes data obtained at the prediction start moment that has an impact on the failure of the device under test in terms of the target prediction variable; determine that the second data set is the first data set.

[0018] In a possible implementation, the future prediction time is a first time period starting from the prediction start time; the duration of the first time period is a preset duration; the second acquisition module is used to acquire a first data set, specifically: the second acquisition module is used to: acquire a third data set; the third data set includes data obtained within the second time period that has an impact on the failure of the device under test in terms of the target prediction variable; the end time of the second time period is the prediction start time, and the duration of the second time period is the preset duration; the third data set is determined to be the first data set.

[0019] In one possible implementation, the first acquisition module is used to acquire at least one fault prediction model corresponding to the target prediction variable, specifically: the first acquisition module is used to: acquire a fourth data set; the fourth data set includes historical data that is correlated with the failure of the device under test in terms of the target prediction variable; based on a preset feature group division rule, divide the data included in the fourth data set into at least one feature group; based on the at least one feature group, construct the at least one fault prediction model.

[0020] In a possible implementation, the first acquisition module is used to acquire a fourth data set, specifically: the first acquisition module is used to: acquire a fifth data set; the fifth data set includes first historical business data, first historical time series data, and / or first historical fault alarm data that are correlated with the failure of the device under test in terms of the target prediction variable, and second historical business data, second historical time series data, and / or second historical fault alarm data that are correlated with the failure of an associated device of the device under test in terms of the target prediction variable; preprocess the data included in the fifth data set to obtain a sixth data set; perform feature expansion based on the sixth data set to obtain a seventh data set; determine that the union of the sixth data set and the seventh data set is the fourth data set.

[0021] In one possible implementation, the first acquisition module is used to preprocess the data included in the fifth data set, specifically: the first acquisition module is used to: based on a preset business data format, perform format normalization processing on the first historical business data and the second historical business data; perform data cleaning processing on the first historical time series data and the second historical time series data; based on a preset fault alarm data format, perform format conversion processing on the first historical fault alarm data and the second historical fault alarm data; and perform normalization processing on the numerical data in the fifth data set.

[0022] In one possible implementation, the first acquisition module is used to construct the at least one fault prediction model based on the at least one feature group, specifically: the first acquisition module is used to: use the principal component analysis method or the kernel principal component analysis method to perform dimensionality reduction processing on the at least one feature group to obtain a reduced dimensionality group of each feature group; each reduced dimensionality group includes data corresponding to the principal component that has a fault risk impact on the target prediction variable of the device under test; based on each reduced dimensionality group, construct the at least one fault prediction model.

[0023] In a possible implementation manner, the data included in the first data set corresponds to the principal component.

[0024] In a third aspect, an embodiment of the present application further provides an electronic device, comprising a memory and a processor; the memory is used to store a computer program product; the processor is used to execute the computer program product stored in the memory, and when the computer program product is executed, the method of the above-mentioned first aspect is implemented.

[0025] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed, the method of the first aspect described above is implemented.

[0026] In summary, in order to solve the problem of how to discover the possible fault risks of high-voltage inverters as early as possible so that possible faults can be checked or repaired in advance, the embodiment of the present application provides a fault risk prediction method and device. In this method, the target prediction variable and future prediction time for fault risk prediction of the device under test can be selected; then, at least one fault prediction model corresponding to the target prediction variable and a first data set consisting of data obtained before the future prediction time that has an impact on the fault of the device under test in terms of the target prediction variable can be obtained; then, at least one fault prediction model obtained can be used to predict the fault risk of the device under test based on the first data set, and the predicted value can be output; finally, all the predicted values ​​can be combined to predict the fault risk of the device under test.

[0027] In this way, a fault prediction model can be used to predict the fault risk at the future prediction time based on the data obtained before the future prediction time, so as to discover possible faults in the equipment under test in advance, and then timely repair the possible faults in the equipment under test in advance, making the equipment run safer and avoiding the losses caused by the actual occurrence of faults. The equipment also has a longer service life and better applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0029] Figure 1 A schematic diagram of a flow chart of a fault risk prediction method provided in an embodiment of the present application;

[0030] Figure 2 A flowchart of a method for obtaining at least one fault prediction model corresponding to a target prediction variable provided in an embodiment of the present application;

[0031] Figure 3 A flowchart of a method for obtaining a fourth data set provided in an embodiment of the present application;

[0032] Figure 4 A flowchart of a method for predicting the failure risk of a device under test based on all predicted values ​​provided in an embodiment of the present application;

[0033] Figure 5 A structural block diagram of a fault risk prediction device provided in an embodiment of the present application;

[0034] Figure 6 A structural block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0035] The present disclosure is further described in detail below through the accompanying drawings and embodiments. Through these descriptions, the characteristics and advantages of the present disclosure will become more clear and distinct.

[0036] The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise noted.

[0037] In addition, the technical features involved in different embodiments of the present disclosure described below can be combined with each other as long as they do not conflict with each other.

[0038] To facilitate understanding, the application scenarios of the technical solutions provided in this application are first described below.

[0039] In the metallurgy, steel, power generation, petroleum, chemical industry, water supply and other industries, high-voltage inverters are widely used to drive fans, pumps, compressors, and various large rotating machinery, becoming one of the main means of energy saving and automation transformation and upgrading. Once a high-voltage inverter fails, it may cause huge losses. Therefore, it is very important to find possible faults of high-voltage inverters as early as possible and repair them in advance.

[0040] However, the current maintenance methods for high-voltage inverters are usually post-maintenance or preventive regular maintenance. Among them, post-maintenance refers to maintenance after a fault occurs, which often causes large losses. The interval between regular maintenance is difficult to grasp, and it will also cause "over-maintenance" or "under-maintenance" due to large differences in different fields and scenarios, which will bring great safety problems and losses to production operations. It can be seen that there is currently no more accurate method to detect possible faults of high-voltage inverters as early as possible, so as to promptly repair possible faults of high-voltage inverters in advance, thereby avoiding losses caused by faults.

[0041] The inventors have found in actual applications and research that the application scenarios of high-voltage inverters can include the following three:

[0042] In the first application scenario, the high-voltage inverter is a device that has just been connected to an actual application scenario. For example, the actual application scenario may be a power generation system, an energy mining system, or a building system.

[0043] In this application scenario, the high-voltage inverter usually does not have time-series operation data (referred to as time-series data) nor historical operation and maintenance data such as fault alarm data. Therefore, it is impossible to detect the fault of the high-voltage inverter in advance and repair the fault in time based on the operation status of the high-voltage inverter or the history of fault alarms.

[0044] However, the inventors have found that when the device attributes are the same or similar, they usually have similar fault conditions. For example, two high-voltage inverters of the same model have the same or similar information such as the time of fault occurrence, the probability of fault occurrence, and the fault content. For another example, two high-voltage inverters in the same environment have similar information such as the time of fault occurrence, the probability of fault occurrence, and the fault content.

[0045] In this way, fault prediction can be performed in advance for a high-voltage inverter that has just been connected to an actual application scenario based on the fault conditions of high-voltage inverters with the same or similar properties.

[0046] In the second application scenario, the high-voltage inverter is a running device. In this application scenario, the high-voltage inverter usually has timing data. For example, the timing data of the high-voltage inverter may include input voltage, output voltage, input current, output current, etc. In this way, fault prediction can be performed in advance for the running high-voltage inverter based on the timing data of the high-voltage inverter.

[0047] In the third application scenario, the high-voltage inverter is a device that has been running for a period of time. In this application scenario, the high-voltage inverter usually has time series data and fault alarm data. Among them, the fault alarm data usually has a label indicating the time when the fault occurred, the fault content and other information. In this way, based on the time series data, fault alarm data and other information of the high-voltage inverter, fault prediction can be performed in advance for the high-voltage inverter that has been running for a period of time.

[0048] Based on this, in order to be able to discover the possible fault risks of high-voltage inverters as early as possible, so that the possible faults of high-voltage inverters can be checked and repaired in advance, the embodiment of the present application provides a fault risk prediction method and device. In this method, the target prediction variable and future prediction time for fault risk prediction of the device under test can be selected; then at least one fault prediction model corresponding to the target prediction variable and a first data set consisting of data obtained before the future prediction time that has an impact on the fault of the device under test in terms of the target prediction variable can be obtained; then, at least one fault prediction model obtained can be used to predict the fault risk of the device under test based on the first data set, and the predicted value can be output; finally, all the predicted values ​​can be combined to predict the fault risk of the device under test.

[0049] In this way, a fault prediction model can be used to predict the fault risk at the future prediction time based on the data obtained before the future prediction time, so as to discover possible faults in the equipment under test in advance, and then timely repair the possible faults in the equipment under test in advance, making the equipment run safer and avoiding the losses caused by the actual occurrence of faults. The equipment also has a longer service life and better applicability.

[0050] The following describes an embodiment of the fault risk prediction method provided by the present application in conjunction with the accompanying drawings.

[0051] See also Figure 1 , Figure 1A flowchart of a fault risk prediction method provided in an embodiment of the present application. The method can be applied to terminal devices, servers, or other devices with data processing functions (such as control devices installed in buildings, control devices installed in power grid management platforms, control devices installed in energy extraction systems, etc.), and the present application does not limit this. The following uses terminal devices as an example to introduce the embodiments of the present application. Figure 1 As shown, the method may include the following steps:

[0052] Step S101: Determine target prediction variables and future prediction time for fault risk prediction of the device under test.

[0053] Among them, the device to be tested refers to the device for which the fault risk prediction is to be performed. Optionally, the device to be tested may be a frequency converter in the target system. Optionally, the frequency converter may be a high-voltage frequency converter. Optionally, the target system may include a power generation system, an energy extraction system or a building system, etc. It is understandable that the device to be tested may also be other devices for which the fault risk prediction is to be performed. The target system may also be other systems in which the device to be tested is installed or arranged. This application does not limit this.

[0054] In actual application, the terminal device can predict the fault risk of the device under test from a certain variable of the device under test. Alternatively, the terminal device can also predict the fault risk of the device under test from multiple variables of the device under test. For example, the terminal device can predict the fault risk of the device under test from the variable of input voltage, and predict whether the device under test will have an input voltage fault. Based on this, in an embodiment of the present application, the variable of the device under test for which the fault risk is to be predicted is recorded as a target prediction variable. The number of target prediction variables can be one or more.

[0055] Optionally, the target prediction variable may include at least one of the following variables: input voltage, input current, output voltage, output current, operating frequency, control cabinet temperature, power cabinet temperature, transformer cabinet temperature, humidity inside the inverter cabinet, etc.

[0056] It is understandable that the above target prediction variables are only examples. The target prediction variables may also include other variables. In addition, the target prediction variables corresponding to different devices under test may be different.

[0057] In addition, in actual application, the terminal device can predict the failure risk of the device under test at the next moment after the prediction start moment. That is, the terminal device can predict the failure risk of the device under test at a certain moment in the future. Alternatively, the terminal device can also predict the failure risk of the device under test within a period of time after the prediction start moment. That is, the terminal device can predict the failure risk of the device under test within a period of time in the future.

[0058] Based on this, optionally, the future prediction time can be the next moment of the prediction start moment. Optionally, the future prediction time can also be the first time period from the prediction start moment. Optionally, the duration of the first time period is a preset duration. The preset duration can be set according to the needs of the actual application scenario. For example, the preset duration can be one day, one week, one month, three months, or six months.

[0059] Step S102: Obtain at least one fault prediction model corresponding to the target prediction variable.

[0060] It should be noted that the number of target prediction variables can be one or more, but the implementation steps of obtaining at least one fault prediction model corresponding to each target prediction variable are the same. Figure 2 , Figure 2 In the example of any target prediction variable, the implementation method of obtaining at least one fault prediction model corresponding to the target prediction variable is described. The methods of obtaining at least one fault prediction model corresponding to other target prediction variables can refer to Figure 2 The embodiment shown.

[0061] like Figure 2 As shown, obtaining at least one fault prediction model corresponding to the target prediction variable can be implemented in the following manner:

[0062] Step S201: Acquire a fourth data set.

[0063] The fourth data set includes historical data that is correlated with failures of the device under test in terms of the target prediction variable.

[0064] In one possible implementation, Figure 3 As shown, obtaining the fourth data set can be achieved in the following manner:

[0065] Step S301: Acquire a fifth data set.

[0066] The fifth data set includes original historical data related to the failure of the device under test in terms of the target prediction variable, and original historical data related to the failure of the associated device of the device under test in terms of the target prediction variable. The original historical data may include historical business data of the corresponding device, historical time series data detected by sensors or other detection elements, and / or historical fault alarm data that has been recorded. Since the business data of the device is usually fixed data or information and does not change over time, for the device under test and the associated device, the historical business data is the same as the current business data or future business data, and can all be called business data.

[0067] In combination with the above content, it can be known that, optionally, the device under test can be a device that has just been connected to the actual application scenario. For example, the device under test can be a frequency converter that has just been installed in the power generation system. In this application scenario, the device under test does not have the historical time series data that has been detected and the historical fault alarm data that has been recorded, but only has the historical business data. Therefore, when the device under test is a device that has just been connected to the actual application scenario, the original historical data that is relevant to the fault of the device under test in terms of the target prediction variable only includes the historical business data of the device under test.

[0068] Optionally, the device under test may also be a running device. For example, the device under test may be a running frequency converter in an energy extraction system. In this application scenario, the device under test has historical time series data that has been detected. Therefore, when the device under test is a running device, the original historical data that is relevant to the failure of the device under test in terms of the target prediction variable may include historical business data and historical time series data of the device under test.

[0069] Optionally, the device under test may also be a device that has been running for a period of time. In this application scenario, the device under test may have historical time series data that has been detected and historical fault alarm data that has been recorded. Therefore, when the device under test is a device that has been running for a period of time, the original historical data that is relevant to the fault of the device under test in terms of the target prediction variable may include historical business data, historical time series data, and historical fault alarm data of the device under test.

[0070] Optionally, the associated devices of the device under test may include devices with the same or similar properties as the device under test. Exemplarily, the associated devices of the device under test may include devices with the same device type as the device type of the device under test. For example, when the device under test is a frequency converter, the associated devices of the device under test may also be frequency converters. For another example, when the device under test is a frequency converter, the associated devices may be frequency converters with the same model type as the model type of the device under test. For another example, when the device under test is a frequency converter, the associated devices may be frequency converters with the same manufacturer information as the manufacturer information of the device under test, etc.

[0071] Optionally, the associated devices of the device under test may also include devices whose environmental information is the same as that of the device under test. For example, when the device under test is a frequency converter, the associated devices of the device under test may be other frequency converters in the same frequency converter module as the device under test or other devices in the same control unit (e.g., a power distribution control room of a building) as the device under test.

[0072] It should be understood that the associated devices of the device under test may also include other devices having the same or similar properties as the device under test, which are not listed one by one here.

[0073] Optionally, the associated devices of the device to be tested may include devices that have just been connected to an actual application scenario, devices that are in operation, and / or devices that have been in operation for a period of time.

[0074] Based on this, the original historical data that is relevant to the failure of the associated device of the device under test in terms of the target prediction variable may include historical business data, historical time series data, and / or historical fault alarm data of the associated device.

[0075] In summary, the fifth data set may include historical business data (recorded as first historical business data), historical time series data (recorded as first historical time series data), and / or historical fault alarm data (first historical fault alarm data) that are relevant to the failure of the device under test in terms of the target prediction variable, as well as historical business data (recorded as second historical business data), historical time series data (recorded as second historical time series data), and / or historical fault alarm data (recorded as second historical fault alarm data) that are relevant to the failure of the associated devices of the device under test in terms of the target prediction variable.

[0076] Exemplarily, the first historical business data may include data for indicating the device information and rated information of the device under test. Optionally, the device information may include basic information and basic parameters. Optionally, the basic information may include the device type, voltage level, inverter level, load type, model of the module (or module) where the inverter is located, maintenance method, heat dissipation method, bypass cabinet method, variable voltage and variable frequency (VVVF) system start-up method, etc. Optionally, the basic parameters may include usage information of the device under test, deployment time, deployment location, customer information of customers using the device under test, etc. Optionally, the rated information may include the rated power, rated input voltage, rated input phase number, rated output current, etc. of the device under test.

[0077] Exemplarily, the first historical time series data may include data obtained before the prediction start time, such as environmental factors (or environmental information), operating parameters (or operation and maintenance parameters), and state parameters, which are used to indicate the operation of the device under test at each detection time (or acquisition time). Optionally, the environmental factors may include the ambient temperature, ambient humidity, ventilation conditions, dust conditions, etc. of the device under test at each detection time before the prediction start time. The operating parameters may include the inverter state, input voltage, output voltage, input current, output current, given frequency, output frequency, input power factor, input active power, input reactive power, output active power, motor speed, output torque, power cabinet temperature, transformer cabinet temperature, transformer temperature, etc. of the device under test at each detection time before the prediction start time. Optionally, the state parameters may include the operating state of the device under test at each detection time before the prediction start time, the state of the unit to which the device under test belongs, the given mode, etc.

[0078] Exemplarily, the first historical fault alarm data may include data obtained before the prediction start time, which is used to indicate that the device under test has a fault in terms of the target prediction variable. Optionally, the first historical fault alarm data may include information such as fault point information, fault alarm level, fault alarm type, and fault record information. Among them, the fault point information may include information such as the target prediction variable (or parameter name) of the device under test, the unit to which the device under test belongs, and the device level to which the device under test belongs. The fault alarm level may include three levels: high, medium, and low. It should be understood that the fault alarm level can also be set to other level modes according to the needs of the application scenario. For example, it is set to 1, 2, 3, 4, 5, five levels, etc., and this application does not limit this. The fault alarm type may include device status, numerical threshold, etc. For example, the device status may be an overload state, etc. The numerical threshold may be used to indicate the threshold of the occurrence of a fault. That is, when the target prediction variable of the device under test is greater than the numerical threshold, it indicates that the device under test has a fault in terms of the target prediction variable. The fault record information may include the time when the fault occurs, etc.

[0079] Exemplarily, the second historical service data may include data for indicating device information and rated information of an associated device of the device under test. The specific contents of the device information and rated information of the associated device may refer to the contents of the aforementioned embodiment, and will not be described in detail here.

[0080] Exemplarily, the second historical time series data may include data obtained before the prediction start time, which is used to indicate the environmental factors, operating parameters, and state parameters of the associated device of the device under test at each detection time. The specific contents of the environmental factors, operating parameters, and state parameters of the associated device at each detection time can refer to the contents of the aforementioned embodiment, and will not be repeated here.

[0081] Exemplarily, the second historical fault alarm data may include data obtained before the prediction start time and used to indicate that the associated device of the device under test has a fault in the target prediction variable. The specific content of the second historical fault alarm data can refer to the content of the above embodiment, which will not be repeated here.

[0082] It should be noted that the fifth data set may also include other data or information. For example, the fifth data set may also include rule setting information related to target prediction variables of the device under test and / or associated devices. The rule setting information refers to rules related to failures of the device under test and / or associated devices in terms of target prediction variables, determined based on prior knowledge or expert experience.

[0083] For example, assuming that the device to be tested is a frequency converter, and the target prediction variable includes the temperature of the transformer cabinet, it is known from prior knowledge or expert experience that when the temperature of the transformer cabinet corresponding to the frequency converter is within the range of 40 to 45 degrees, the probability of the frequency converter failure risk is 80%; when the temperature of the transformer cabinet is greater than 45 degrees, the probability of the frequency converter failure risk is 98%. Then the following rule setting information can be determined: when the temperature of the transformer cabinet is within the range of 40 to 45 degrees, the probability of the frequency converter failure risk is 80%; when the temperature of the transformer cabinet is greater than 45 degrees, the probability of the frequency converter failure risk is 98%, etc.

[0084] Optionally, the terminal device can detect and obtain the first historical time series data through a sensor or other detection element, and record and obtain the first historical fault alarm data. Then the terminal device can pre-store the first historical business data, the first historical time series data, the first historical fault alarm data, the rule setting information related to the target prediction variable of the device to be tested, etc., together with the device to be tested and the target prediction variable, in the terminal device or other storage device that is communicatively connected to the terminal device. Similarly, the terminal device can also pre-store the second historical business data, the second historical time series data, the second historical fault alarm data, the rule setting information related to the target prediction variable of the associated device, etc., together with the associated device and the target prediction variable, in the terminal device or other storage device that is communicatively connected to the terminal device. Then, when the terminal device executes the step of obtaining the fifth data set, it can read the data included in the fifth data set from the terminal device or other storage device that is communicatively connected to the terminal device, thereby obtaining the fifth data set.

[0085] Step S302: pre-process the data included in the fifth data set to obtain a sixth data set.

[0086] In one possible implementation, preprocessing of the data included in the fifth data set can be achieved in the following manner: based on a preset business data format, format normalization processing is performed on the first historical business data and the second historical business data; data cleaning processing is performed on the first historical time series data and the second historical time series data; based on a preset fault alarm data format, format conversion processing is performed on the first historical fault alarm data and the second historical fault alarm data; and normalization processing is performed on the numerical data in the fifth data set.

[0087] Optionally, the preset business data format can be set according to the business data format required when building a fault prediction model in an actual application scenario. After obtaining the fifth data set, the format of the first historical business data and the second historical business data included in the fifth data set needs to be normalized to the preset business data format. The first historical business data and the second historical business data after format normalization can be used to build a fault prediction model.

[0088] Optionally, data cleaning processing is performed on the first historical time series data and the second historical time series data, which can be implemented in the following manner:

[0089] The non-numeric data in the first historical time series data and the second historical time series data are cleaned using a pre-set cleaning rule (referred to as a first cleaning rule), wherein the first cleaning rule is usually pre-set based on the experience of experts.

[0090] For the numerical data in the first historical time series data and the second historical time series data, data cleaning can be performed in the following two ways:

[0091] For numerical data with corresponding cleaning rules (referred to as second cleaning rules) preset, data cleaning processing is performed according to the second cleaning rules.

[0092] For numerical data without corresponding second cleaning rules, data cleaning processing can be performed according to existing abnormal data detection and correction methods, which will not be described in detail here.

[0093] Optionally, the preset fault alarm data format can be set according to the fault alarm data format required when building a fault prediction model in an actual application scenario. After acquiring the fifth data set, the format of the first historical fault alarm data and the second historical fault alarm data included in the fifth data set needs to be converted to the preset fault alarm data format. The first historical fault alarm data and the second historical fault alarm data after the format conversion processing can be used to build a fault prediction model.

[0094] Optionally, the normalization processing of the numerical data included in the fifth data set can be implemented in the following manner: determine each numerical data as the first data respectively; then normalize each first data in the following manner: determine the variable corresponding to the first data, recorded as the first variable; determine the median μ of all the data corresponding to the first variable in the numerical data included in the fifth data set; after filtering outliers for all the data corresponding to the first variable in the numerical data included in the fifth data set, calculate the standard deviation σ of all the filtered data corresponding to the first variable; use the following formula Normalize the first data, where X is the first data, X * is the data obtained after normalizing the first data.

[0095] After the data included in the fifth data set is preprocessed in the above manner, the obtained data set is recorded as the sixth data set.

[0096] Step S303: perform feature expansion based on the sixth data set to obtain a seventh data set.

[0097] After obtaining the sixth data set, the terminal device can perform feature expansion based on the data included in the sixth data set to obtain more feature data that can be used to build a fault prediction model. In addition, the terminal device can record the data set consisting of the feature data obtained by feature expansion as the seventh data set. Further, the terminal device can construct new feature data based on the data included in the sixth data set, thereby achieving feature expansion.

[0098] Exemplarily, the terminal device can perform feature expansion from at least one of the following dimensions: classification labels, operation and maintenance parameters, health indicators, time-domain derived features, statistical analysis indicators, etc., to obtain more feature data for building a fault prediction model.

[0099] Among them, feature expansion from the dimension of classification labels refers to constructing labels of multiple preset label categories of the device to be tested and associated devices based on the data included in the sixth data set. Optionally, multiple preset label categories may include label categories such as point attribute type, model type, device type, system type, industry category, fault category, and fault level. In other words, the point attribute type label of the device to be tested can be constructed based on the data indicating the point attribute type of the device to be tested included in the sixth data set. Similarly, the model type label, device type label, system type label, industry category label, fault category label, and fault level label of the device to be tested can also be constructed based on the data indicating the model type, device type, system type, industry category, fault category, and fault level of the device to be tested included in the sixth data set. Similarly, the point attribute type label, model type label, device type label, system type label, industry category label, fault category label, and fault level label of the associated device can also be constructed.

[0100] Feature expansion from the dimension of operation and maintenance parameters refers to constructing new feature data of the device to be tested and associated devices based on the following operation and maintenance parameters (or operating parameters) of the device to be tested and associated devices included in the sixth data set: inverter state, input voltage, output voltage, input current, output current, given frequency, output frequency, input power factor, input active power, input reactive power, output active power, motor speed, output torque, power cabinet temperature, transformer cabinet temperature, transformer temperature, etc. Exemplarily, the constructed new feature data may include: overvoltage, ultra-high voltage, undervoltage, overcurrent, speed peak factor, speed pulse factor, speed margin factor, overtemperature, ultra-high temperature and other feature data. It should be understood that the new feature data constructed from the dimension of operation and maintenance parameters may also include other data, which are not listed here one by one.

[0101] Feature expansion from the dimension of health indicators refers to determining the remaining life of the device under test and the unit aging index and the like of the associated devices based on the data included in the sixth data set, thereby constructing feature data such as the remaining life of the device under test and the unit aging index and the like of the associated devices.

[0102] Feature expansion from the dimension of time-domain derived features refers to constructing dimensional feature data and dimensionless feature data of the device under test and associated devices based on the historical time series data included in the sixth data set. Optionally, the dimensional feature data may include maximum value, minimum value, mean, variance, standard deviation (or called mean square error), mean square error, root mean square error, etc. Among them, the maximum value, minimum value and mean value can characterize the energy situation of the corresponding historical time series data. Variance and standard deviation can reflect the degree of discreteness of the corresponding historical time series data. Mean square error and root mean square error can reflect the degree of discreteness when the corresponding historical time series data changes.

[0103] For example, the maximum value, minimum value, mean value, variance, standard deviation, mean square error, root mean square error and other characteristic data of the target prediction variables of the device to be tested and the associated devices can be constructed. It should be understood that the mean value, variance, standard deviation, mean square error, root mean square error and other characteristic data of other variables included in the sixth data set can also be constructed, and this application does not limit this.

[0104] Optionally, the dimensionless characteristic data may include peak factor, pulse factor, margin factor, kurtosis factor, waveform factor, and skewness factor. Among them, the peak factor and pulse factor can reflect whether there is vibration signal impact. The margin factor can reflect whether the device under test or the associated equipment is worn. The skewness factor and kurtosis factor can both characterize the response to the vibration signal impact characteristics.

[0105] Feature expansion from the dimension of statistical analysis indicators refers to constructing feature data with statistical characteristics of the device under test and associated devices (which can be referred to as statistical data) based on the data included in the sixth data set. For example, the following statistical data of the device under test and associated devices can be constructed based on the data included in the sixth data set: cumulative power consumption, cumulative minor faults, cumulative major faults, cumulative operating time, average fault maintenance time, time from the last fault state, etc. It should be understood that other statistical data can also be constructed, which will not be listed here one by one.

[0106] Step S304: Determine that the union of the sixth data set and the seventh data set is the fourth data set.

[0107] After the terminal device obtains the sixth data set and the seventh data set, it may determine the union of the sixth data set and the seventh data set as the fourth data set, thereby obtaining the fourth data set.

[0108] Step S202: Divide the data included in the fourth data set into at least one feature group based on a preset feature group division rule.

[0109] In actual applications, since different features have different influences on the model and fault risk prediction results, in order to improve the quality of features and models, the terminal device can set preset feature group division rules in advance according to the modeling requirements in the actual application scenario. In this way, feature data with high correlation with the model can be screened more accurately in the future, making the constructed fault prediction model more accurate.

[0110] Based on this, after the terminal device obtains the fourth data set, it can divide the data included in the fourth data set (or also referred to as feature data) into at least one feature group based on a preset feature group division rule.

[0111] Exemplarily, the terminal device may divide the data included in the fourth data set into at least one of the following feature groups based on preset feature group division rules: a mechanism rule group, a statistical analysis group, a category label group, a time series prediction group, and a classification prediction group.

[0112] Among them, the mechanism rule group includes the equipment information, rated information, environmental information and rule setting information of the equipment to be tested and the associated equipment in the fourth data set. The statistical analysis group includes the fault statistical characteristics of the equipment to be tested and the associated equipment in the fourth data set. Optionally, the fault statistical characteristics may include characteristic data such as cumulative minor faults, cumulative major faults, average fault maintenance time, and time from the last fault state. The category label group includes labels of multiple preset label categories of the equipment to be tested and the associated equipment in the fourth data set. The timing prediction group includes the operation and maintenance parameters of the equipment to be tested and the associated equipment in the fourth data set. The classification prediction group includes the feature data obtained by feature expansion from the dimensions of operation and maintenance parameters, time domain derived features, statistical analysis indicators, and classification labels in the fourth data set.

[0113] Step S203: construct at least one fault prediction model corresponding to the target prediction variable based on the at least one feature group.

[0114] In combination with the foregoing content, it can be seen that different features have different impacts on the model and fault risk prediction results. In order to improve the feature quality and model performance, after the data included in the fourth data set is divided into at least one feature group, feature selection and dimensionality reduction can be performed on each feature group.

[0115] Based on this, in a possible implementation, at least one fault prediction model corresponding to the target prediction variable is constructed based on at least one feature group, which can be implemented in the following manner: using the principal component analysis method or the kernelized principal component analysis (KPCA) method, at least one feature group is subjected to dimensionality reduction processing to obtain dimensionality reduction groups of each feature group, each dimensionality reduction group includes data corresponding to the principal component that has a fault risk impact on the target prediction variable of the device to be tested (can be referred to as principal component data for short); based on each of the dimensionality reduction groups, at least one fault prediction model is constructed. After dimensionality reduction processing is performed on at least one feature group to obtain dimensionality reduction groups corresponding to each feature group, the dimensionality reduction group only includes principal component data that has a strong correlation impact on the fault of the device to be tested in terms of the target prediction variable, and the fault risk prediction result of the constructed fault prediction model is more accurate.

[0116] Optionally, the at least one fault prediction model corresponding to the constructed target prediction variable may include a mechanism rule model, a statistical analysis model, and / or a prediction model.

[0117] Optionally, the mechanism rule model, statistical analysis model and prediction model can all be constructed based on the data and / or information included in the dimensionality reduction group corresponding to one or more feature groups in the mechanism rule group, statistical analysis group, category label group, time series prediction group and classification prediction group.

[0118] Optionally, the mechanism rule model may include one or more. Exemplarily, assuming that the device to be tested is a frequency converter, the target prediction variables include transformer cabinet temperature and load type, etc. Then the mechanism rule model may include a transformer cabinet temperature model and a load type model, etc. Further, the transformer cabinet temperature model may also include one or more. For example, the transformer cabinet temperature model may include a model constructed based on relevant data and / or information of the device to be tested, a model constructed based on data and / or information of associated devices with the same attributes as the device to be tested, a model constructed based on data and / or information of associated devices with the same environment as the device to be tested, etc. Similarly, the load type model may also include one or more.

[0119] Exemplarily, one or more temperature thresholds may be set in the constructed transformer cabinet temperature model. For example, the temperature threshold may include two thresholds of 40 and 45 degrees Celsius (referred to as degrees). When the input value in the transformer cabinet temperature model is between 40 and 45 degrees, the predicted value of the fault risk output by the transformer cabinet temperature model is 80%; or, when the input value in the transformer cabinet temperature model is greater than 45 degrees, the predicted value of the fault risk output by the transformer cabinet temperature model is 98%, etc.

[0120] Exemplarily, one or more load types may be set in the constructed load type model. For example, the load type may include an overload type. When the input value in the load type model is an overload type, the predicted value of the failure risk output by the load type model is 75%, etc.

[0121] It should be noted that the mechanism rule model can be constructed according to the requirements of the device to be tested and the target prediction variable in the actual application scenario. It should be understood that the mechanism rule model can also include other models, which are not listed here one by one.

[0122] Optionally, the statistical analysis model may include a historical statistical model and a same-label analogy model. Optionally, the historical statistical model and the same-label analogy model may each include one or more.

[0123] Optionally, the historical statistical model may be constructed based on historical fault alarm data of the device under test in terms of the target prediction variable and / or historical fault alarm data of an associated device of the device under test in terms of the target prediction variable.

[0124] For example, assuming that the device to be tested is a frequency converter, and the target prediction variable includes output voltage, the terminal device can statistically analyze the probability of the frequency converter failing in the future unit time based on the historical fault alarm data of the frequency converter in terms of output voltage, and / or the historical fault alarm data of the frequency converter's associated devices in terms of output voltage, and then build a model for determining the probability of the frequency converter failing in the future unit time based on the statistical analysis results.

[0125] Optionally, when the future prediction time is the next moment of the prediction start moment, the unit time may be seconds or minutes, etc. Optionally, when the future preset time is the first time period from the prediction start moment, the unit time may be the length of the first time period, for example, the unit time may be days, weeks, months, years, etc.

[0126] Optionally, the same-label analog model may include one or more models of the same-label variable category. Exemplarily, the same-label variable category may include categories such as the same-attribute variable, the same-model variable, the same-device variable, and the same-system variable. In other words, the same-label analog model may include the same-attribute variable model, the same-model variable model, the same-device variable model, and the same-system variable model. It should be understood that the same-label analog model may also include other models of the same-label variable category according to the needs of the actual application scenario, and the present application does not limit this.

[0127] Among them, when constructing a same-attribute variable model, based on the labels of multiple preset label categories of the device to be tested and the associated devices, first, from the data and / or information included in each dimensionality reduction group, filter out the data and / or information related to the fault in the target prediction variable of the first associated device, wherein the point attribute type label of the first associated device is the same as the point attribute type label of the device to be tested; then, based on the data and / or information related to the fault in the target prediction variable of the first associated device, construct a same-attribute variable model.

[0128] Similarly, based on the labels of multiple preset label categories of the device to be tested and the associated devices, first, the data and / or information related to the failure of the target prediction variables of the corresponding associated devices can be screened out from the data and / or information included in the dimensionality reduction group, wherein the corresponding associated devices are associated devices whose model type labels, device type labels, and system type labels are the same as those of the device to be tested; then, based on the data and / or information related to the failure of the target prediction variables of the corresponding associated devices, the same model variable model, the same device variable model, and the same system variable model are constructed.

[0129] Optionally, the prediction model may include an autoregressive prediction model and a classification prediction model. Optionally, the autoregressive prediction model and the classification prediction model may each include one or more.

[0130] Optionally, the autoregressive prediction model can be a Neural Prophet algorithm model. The Neural Prophet algorithm model is an upgraded algorithm model based on neural network and Prophet (prophet algorithm). Optionally, the autoregressive prediction model can also be other similar algorithm models of the Neural Prophet algorithm model. Optionally, the autoregressive prediction model can also be an algorithm model obtained by fusing the Neural Prophet algorithm model with other similar algorithm models.

[0131] Optionally, the autoregressive prediction model can be used to directly predict the failure risk of the device under test in terms of the target prediction variable at the next moment after the prediction starts or within the first time period from the prediction starts. Optionally, the autoregressive prediction model can also be used to predict the value of the target prediction variable of the device under test at the next moment after the prediction starts, or to predict the value of the target prediction variable of the device under test in the first time period from the prediction starts, and then use the mechanism rule model and / or statistical analysis model to predict the failure risk of the device under test in terms of the target prediction variable at the next moment after the prediction starts or within the first time period from the prediction starts.

[0132] Optionally, the classification prediction model may include at least one of the following classification models: RF (random forest) algorithm model, GBDT (gradient boosting iterative decision tree) algorithm model, XGBoost (extreme gradient boosting) algorithm model, SVM (support vector machine) algorithm model, ridge regression algorithm model, LSTM (long short-term memory neural network) algorithm model, etc.

[0133] Optionally, the classification prediction model can be used to directly predict the failure risk of the device under test in terms of the target prediction variable at the next moment after the prediction starts or within the first time period from the prediction starts. Optionally, the classification prediction model can also be used to predict the value of the target prediction variable of the device under test at the next moment after the prediction starts, or to predict the value of the target prediction variable of the device under test in the first time period from the prediction starts, and then use the mechanism rule model and / or statistical analysis model to predict the failure risk of the device under test in terms of the target prediction variable at the next moment after the prediction starts or within the first time period from the prediction starts.

[0134] It should be noted that more or fewer fault prediction models than the above can be constructed according to the needs of actual application scenarios, and this application does not impose any restrictions on this.

[0135] according to Figure 2 The method shown in FIG. 1 can construct at least one fault prediction model corresponding to each target prediction variable. Figure 2According to the method shown, at least one fault prediction model corresponding to each target prediction variable is obtained to better meet the needs of actual application scenarios, and the subsequent use of the fault prediction model to predict fault risks is more accurate.

[0136] In a possible implementation, at least one corresponding fault prediction model may be pre-built for different variables of different devices. Then, the constructed fault prediction model is stored in the terminal device or other storage devices connected to the terminal device in communication with the terminal device together with the device and the variable. In this way, when the terminal device executes step S102, it can directly read at least one fault prediction model corresponding to the target prediction variable of the device to be tested from the terminal device or other storage devices connected to the terminal device in communication with the terminal device, so that the required fault prediction model can be obtained more quickly, which can speed up the efficiency of subsequent fault risk prediction.

[0137] In one possible implementation, the terminal device can also update at least one pre-built fault prediction model corresponding to each target prediction variable of the terminal device, so as to quickly obtain a more accurate fault prediction model, and the efficiency and accuracy of fault risk prediction can be improved, and the applicability is better.

[0138] Step S103: Acquire a first data set.

[0139] The first data set may include data obtained before the future prediction time that has an impact on the failure of the device under test in terms of the target prediction variable. Optionally, the data included in the first data set may correspond to the principal components determined by the principal component analysis method or the kernel principal component analysis method when constructing at least one failure prediction model corresponding to the target prediction variable. In other words, the data included in the first data set are all strongly correlated with the failure of the device under test in terms of the target prediction variable.

[0140] In a possible implementation, the future prediction time is the next moment after the prediction start moment. That is, in this application scenario, it is necessary to predict the failure risk of the device under test in terms of the target prediction variable at the next moment. Then, obtaining the first data set can be implemented in the following manner: obtaining a second data set, the second data set including data obtained at the prediction start moment that has an impact on the failure of the device under test in terms of the target prediction variable; determining the second data set to be the first data set.

[0141] Optionally, the data of the device under test corresponding to the principal component obtained at the start time of the prediction can be obtained, and the set consisting of these data can be determined as the second data set. The data of the device under test corresponding to the principal component can include all or part of the business data of the device under test, all or part of the time series data, and / or all or part of the fault alarm data. The contents of the business data, time series data and fault alarm data can all refer to the contents of the aforementioned first historical business data, first historical time series data and first historical fault alarm data, which will not be described in detail here.

[0142] In a possible implementation, the future prediction time is the first time period from the prediction start time. That is to say, in this application scenario, it is necessary to predict the failure risk of the device under test in terms of the target prediction variable in the next first time period. Then, obtaining the first data set can be implemented in the following manner: obtaining a third data set, the third data set includes data obtained in the second time period that has an impact on the failure of the device under test in terms of the target prediction variable, the end time of the second time period is the prediction start time, and the duration of the second time period is the preset duration; determining the third data set to be the first data set.

[0143] Optionally, data of the device under test corresponding to the main component obtained in the second time period may be acquired, and a set consisting of these data may be determined as a third data set.

[0144] Step S104: obtaining a prediction value of each fault prediction model for predicting the fault risk of the device under test based on the first data set.

[0145] In a possible implementation, after the terminal device obtains at least one fault prediction model and the first data set corresponding to the target prediction variable, the predicted value of each fault prediction model for predicting the fault risk of the device under test based on the first data set can be obtained in the following manner: extract the input value corresponding to each fault prediction model from the first data set, and then obtain the predicted value output by each fault prediction model. Furthermore, when extracting the input value corresponding to each fault prediction model from the first data set, based on the principal component corresponding to the fault prediction model, the data corresponding to the principal component can be extracted from the first data set as the input value of the fault prediction model.

[0146] Exemplarily, assuming that the device to be tested is a frequency converter, and the target prediction variable is the input voltage, the obtained fault prediction model includes: a mechanism rule model of input voltage, a historical statistical model of input voltage, a variable model of the same attribute of input voltage, a variable model of the same model of input voltage, an autoregressive prediction model of input voltage, and a classification prediction model of input voltage. In addition, the future prediction time is the next moment of the prediction start time. Then the terminal device needs to extract the input voltage of the frequency converter obtained at the prediction start time from the first data set, the input voltage of the associated device with the same point attribute type as the frequency converter, and the input voltage of the associated device with the same model type as the frequency converter. Then, the terminal device can input the input voltage of the frequency converter into the mechanism rule model of input voltage, the historical statistical model of input voltage, the autoregressive prediction model of input voltage, and the classification prediction model of input voltage, respectively, input the input voltage of the associated device with the same point attribute type as the frequency converter into the variable model of the same attribute of input voltage, and input the input voltage of the associated device with the same model type as the frequency converter into the variable model of the same model of input voltage, and obtain the predicted value output by each model.

[0147] In one possible implementation, obtaining the predicted value of each fault prediction model for predicting the fault risk of the device under test based on the first data set can also be implemented in the following manner: preprocessing the data included in the first data set to obtain a preprocessed data set corresponding to the first data set; extracting the input value corresponding to each fault prediction model from the preprocessed data set, and then obtaining the predicted value output by each fault prediction model.

[0148] Among them, the implementation method of preprocessing the data included in the first data set and the implementation method of extracting the input values ​​corresponding to each fault prediction model from the preprocessed data set can all refer to the contents of the aforementioned embodiments and will not be repeated here.

[0149] According to this implementation method, the predicted value of each fault prediction model for fault risk prediction of the device under test based on the first data set is obtained. Since the input value is more accurate, the obtained output value is also more accurate, which can improve the accuracy of the fault risk prediction of the device under test in terms of the target prediction variable and has better applicability.

[0150] Step S105: Based on all the predicted values, predict the failure risk of the device under test.

[0151] In one possible implementation, when the target predictor variable is one, such as Figure 4 As shown, based on all the predicted values, the failure risk prediction of the device under test can be achieved in the following way:

[0152] Step S401: determine the first model dimension.

[0153] The first model dimension refers to a model dimension corresponding to at least one fault prediction model corresponding to a target prediction variable of the device under test.

[0154] In actual application, according to the actual application scenario, it is possible to pre-analyze and determine from which model dimensions the fault prediction model needs to be constructed for different variables of devices with different attributes. In addition, it is also possible to pre-determine which types (or types) of fault prediction models can be constructed under each model dimension. Then, the devices, variables, model dimensions and types of fault prediction models of each attribute are stored together in a corresponding manner in a terminal device or a storage device connected to the terminal device.

[0155] Exemplarily, the model dimensions may include mechanism rule dimensions, historical statistical dimensions, same-label analogy dimensions, autoregressive prediction dimensions, and classification prediction dimensions. The mechanism rule model, historical statistical model, same-label analogy model, autoregressive prediction model, and classification prediction model correspond to the mechanism rule dimension, historical statistical dimension, same-label analogy dimension, autoregressive prediction dimension, and classification prediction dimension, respectively. It should be understood that, according to the needs of actual application scenarios, the model dimensions may also include more or fewer dimensions than the aforementioned, and this application does not limit this.

[0156] After the terminal device obtains at least one fault prediction model corresponding to the target prediction variable of the device under test, it can obtain the model dimension corresponding to at least one fault prediction model corresponding to the target prediction variable of the device under test from the pre-stored model dimensions, and then determine the obtained model dimension as the first model dimension. It should be noted that the number of the first model dimensions can be one or more.

[0157] Step S402: Based on all the predicted values, determine the mean of the predicted values ​​under each first model dimension.

[0158] The mean of the predicted values ​​under each first model dimension refers to the mean of the predicted values ​​output by the target fault prediction model corresponding to the first model dimension. The target fault prediction model corresponding to each first model dimension refers to the fault prediction model corresponding to the first model dimension in at least one fault prediction model corresponding to the target prediction variable of the device under test.

[0159] In a possible implementation, the mean of the prediction values ​​under each first model dimension may be the arithmetic mean of the prediction values ​​output by the target fault prediction model corresponding to the first model dimension.

[0160] Based on this, based on all the predicted values, determining the mean predicted value under each first model dimension can be achieved in the following manner: extracting the predicted values ​​output by all target fault prediction models corresponding to each first model dimension from all the predicted values; calculating the arithmetic mean of all the predicted values ​​corresponding to each first model dimension; and determining the arithmetic mean of all the predicted values ​​corresponding to each first model dimension as the mean predicted value under the first model dimension.

[0161] Exemplarily, it is assumed that the first model dimension corresponding to at least one fault prediction model corresponding to the target prediction variable of the device under test includes a mechanism rule dimension, a historical statistical dimension, a same-label analog dimension, an autoregressive prediction dimension, and a classification prediction dimension. Then, when the terminal device executes step S402, it can calculate the arithmetic mean of the predicted values ​​output by all mechanism rule models in at least one fault prediction model corresponding to the target prediction variable of the device under test, and determine the arithmetic mean as the predicted value mean under the mechanism rule dimension. Similarly, the terminal device can also calculate the arithmetic mean of the predicted values ​​output by all historical statistical models in at least one fault prediction model corresponding to the target prediction variable of the device under test, and determine the arithmetic mean as the predicted value mean under the historical statistical dimension. The terminal device can also calculate the arithmetic mean of the predicted values ​​output by all same-label analog models in at least one fault prediction model corresponding to the target prediction variable of the device under test, and determine the arithmetic mean as the predicted value mean under the same-label analog dimension. The terminal device can also calculate the arithmetic mean of the predicted values ​​output by all autoregressive prediction models in at least one fault prediction model corresponding to the target prediction variable of the device under test, and determine the arithmetic mean as the predicted value mean under the autoregressive prediction dimension. The terminal device may also calculate the arithmetic mean of the prediction values ​​output by all classification prediction models in at least one fault prediction model corresponding to the target prediction variable of the device under test, and determine the arithmetic mean as the prediction value mean under the classification prediction dimension.

[0162] In a possible implementation, the mean of the prediction values ​​under each first model dimension may also be a weighted average of the prediction values ​​output by the target fault prediction model corresponding to the first model dimension.

[0163] In this application scenario, for different model dimensions of different variables of different devices, it is possible to pre-analyze and determine which fault prediction models can be constructed under each model dimension, and configure a weighting coefficient or a weighting base value for each fault prediction model. Then, the weighting coefficients or weighting base values ​​of the devices, variables, model dimensions, and each fault prediction model under each model dimension can be stored in a terminal device or a storage device that is communicatively connected to the terminal device. Thereafter, when the terminal device executes step S402, it can retrieve the weighting coefficients or weighting base values ​​corresponding to each target fault prediction model under each first model dimension from the terminal device or the storage device that is communicatively connected to the terminal device, thereby further determining the mean of the predicted values ​​under each first model dimension.

[0164] Exemplarily, as shown in Table 1 below, the first model dimension may include a same-label analogy dimension. The target fault prediction model corresponding to the same-label analogy dimension may include a same-attribute variable model, a same-model variable model, a same-device variable model, and a same-system variable model. The weighting coefficients corresponding to the same-attribute variable model, the same-model variable model, the same-device variable model, and the same-system variable model are shown in Table 1.

[0165] Table 1

[0166] Models with the same label analogy dimension Weighting coefficient Same attribute variable model 0.5 Same model variable model 0.3 Same device variable model 0.1 Same system variable model 0.1

[0167] Combined with Table 1, it can be seen that the mean predicted value under the same label analogy dimension is equal to (0.5A+0.3B+0.1C+0.1D) / 4, where A represents the predicted value output by the same attribute variable model, B represents the predicted value output by the same model variable model, C represents the predicted value output by the same equipment variable model, and D represents the predicted value output by the same system variable model.

[0168] Based on this, when determining the mean of the predicted values ​​under each first model dimension based on all the predicted values, determining the mean of the predicted values ​​under any first model dimension can also be implemented in the following manner: extracting the predicted values ​​output by each target fault prediction model corresponding to the first model dimension from all the predicted values; reading the weighting coefficients corresponding to each target fault prediction model under the first model dimension from the pre-stored weighting coefficients; and calculating the weighting coefficients according to the following formula: Calculate the mean of the predicted values ​​under the first model dimension, where S′ represents the mean of the predicted values ​​under the first model dimension, N represents the number of target fault prediction models corresponding to the first model dimension, and w i represents the weight coefficient corresponding to the i-th target fault prediction model, w i ≥0, s i Represents the predicted value output by the i-th target fault prediction model.

[0169] Optionally, the following formula may be used according to the pre-stored weighted base value: Calculate the weight coefficient corresponding to the i-th target fault prediction model under the first model dimension. i represents the weight coefficient corresponding to the i-th target fault prediction model, k i represents the weighted base value corresponding to the i-th target fault prediction model, and N represents the number of target fault prediction models corresponding to the first model dimension.

[0170] Step S403: Based on the mean of all prediction values, generate a fusion fault risk value of the device under test in terms of the target prediction variable.

[0171] In one possible implementation, based on the mean of all predicted values, a fusion is generated to generate a failure risk value of the device under test in terms of the target predicted variable. This can be achieved in the following way: the arithmetic mean of the mean of all predicted values ​​is calculated, and the arithmetic mean is determined to be the failure risk value of the device under test in terms of the target predicted variable.

[0172] In one possible implementation, weighting coefficients or weighted base values ​​may be set in advance for each model dimension corresponding to different variables of different devices, and then the devices, variables, model dimensions and the weighting coefficients or weighted base values ​​of the model dimensions may be stored together in a terminal device or a storage device that is communicatively connected to the terminal device.

[0173] Based on this, based on the mean of all predicted values, the failure risk value of the device under test in terms of the target prediction variable is generated by fusion, which can also be implemented in the following way: read the weighting coefficients corresponding to each first model dimension corresponding to the target prediction variable from the pre-stored weighting coefficients; according to the following formula Calculate the failure risk value of the device under test in terms of the target prediction variable, where V represents the failure risk value of the device under test in terms of the target prediction variable, M represents the number of the first model dimensions corresponding to the target prediction variable, and o h represents the weight coefficient corresponding to the hth first model dimension, o h ≥0, p h Represents the mean of the predicted values ​​corresponding to the h-th first model dimension.

[0174] Optionally, the weighting coefficients corresponding to the first model dimensions corresponding to the target prediction variable may be calculated according to the pre-stored weighted base values. The specific method may refer to the contents of the aforementioned embodiments and will not be described again here.

[0175] In one possible implementation, when there are multiple target predictor variables, Figure 4The method shown performs risk prediction on the failure of the device under test in terms of any target prediction variable, and determines the failure risk value of the device under test in terms of the target prediction variable.

[0176] In one possible implementation, when there are multiple target predictor variables, Figure 4 The method shown determines the failure risk value of the device under test in any target prediction variable, then calculates the arithmetic mean or weighted average of all the failure risk values, and determines the arithmetic mean or weighted average as the final failure risk value of the device under test. The specific implementation method can refer to the content of the aforementioned embodiment, which will not be described in detail here.

[0177] In the fault risk prediction method provided by the present application, a target prediction variable and a future prediction time for fault risk prediction of the device under test can be selected; then, at least one fault prediction model corresponding to the target prediction variable and a first data set consisting of data obtained before the future prediction time and having an impact on the fault of the device under test in terms of the target prediction variable can be obtained; thereafter, the at least one fault prediction model obtained can be used to perform fault risk prediction on the device under test respectively based on the first data set, and a prediction value can be output; finally, all the prediction values ​​can be integrated to perform fault risk prediction on the device under test.

[0178] In this way, a fault prediction model can be used to predict the fault risk at the future prediction time based on the data obtained before the future prediction time, so as to discover possible faults in the equipment under test in advance, and then timely repair the possible faults in the equipment under test in advance, making the equipment run safer and avoiding the losses caused by the actual occurrence of faults. The equipment also has a longer service life and better applicability.

[0179] It can be understood that the above embodiments are only examples and can be modified in actual implementation. Those skilled in the art can understand that the modification methods of the above embodiments without creative labor all fall within the protection scope of the present application and will not be described in detail in the embodiments.

[0180] Based on the same inventive concept, an embodiment of the present application also provides a fault risk prediction device. Since the principle of the problem solved by the fault risk prediction device is similar to that of the aforementioned fault risk prediction method, the implementation of the fault risk prediction device can refer to the implementation of the aforementioned fault risk prediction method, and the repeated parts will not be repeated.

[0181] See also Figure 5 , Figure 5 This is a structural block diagram of a fault risk prediction device provided in an embodiment of the present application. Figure 5As shown, the fault risk prediction device 500 may include: a determination module 501, a first acquisition module 502, a second acquisition module 503, a third acquisition module 504 and a prediction module 505.

[0182] The determination module 501 may be used to determine target prediction variables and future prediction time for fault risk prediction of the device under test.

[0183] The first acquisition module 502 may be configured to acquire at least one fault prediction model corresponding to the target prediction variable.

[0184] The second acquisition module 503 may be used to acquire a first data set; the first data set includes data obtained before the future prediction time and having an impact on the failure of the device under test in terms of the target prediction variable.

[0185] The third acquisition module 504 may be used to acquire a prediction value of each fault prediction model for predicting the fault risk of the device under test based on the first data set.

[0186] The prediction module 505 may be used to predict the failure risk of the device under test based on all the prediction values.

[0187] In one possible implementation, the device to be tested is a frequency converter in a target system; the target system includes a power generation system, an energy extraction system or a building system; the target prediction variable includes at least one of the following variables: input voltage, input current, output voltage, output current, operating frequency, control cabinet temperature, power cabinet temperature, transformer cabinet temperature, and humidity inside the frequency converter cabinet.

[0188] In one possible implementation, the future prediction time is the next moment after the prediction start moment; the second acquisition module 503 is used to acquire the first data set, specifically: the second acquisition module 503 is used to: acquire the second data set; the second data set includes data obtained at the prediction start moment that has an impact on the failure of the device under test in terms of the target prediction variable; determine that the second data set is the first data set.

[0189] In one possible implementation, the future prediction time is a first time period starting from the prediction start time; the duration of the first time period is a preset duration; the second acquisition module 503 is used to acquire a first data set, specifically: the second acquisition module 503 is used to: acquire a third data set; the third data set includes data obtained within the second time period that has an impact on the failure of the device under test in terms of the target prediction variable; the end time of the second time period is the prediction start time, and the duration of the second time period is the preset duration; the third data set is determined to be the first data set.

[0190] In one possible implementation, the first acquisition module 502 is used to acquire at least one fault prediction model corresponding to the target prediction variable, specifically: the first acquisition module 502 is used to: acquire a fourth data set; the fourth data set includes historical data that is correlated with the failure of the device under test in terms of the target prediction variable; based on a preset feature group division rule, divide the data included in the fourth data set into at least one feature group; based on the at least one feature group, construct the at least one fault prediction model.

[0191] In a possible implementation, the first acquisition module 502 is used to acquire a fourth data set, specifically: the first acquisition module 502 is used to: acquire a fifth data set; the fifth data set includes first historical business data, first historical time series data, and / or first historical fault alarm data that are correlated with the failure of the device under test in terms of the target prediction variable, and second historical business data, second historical time series data, and / or second historical fault alarm data that are correlated with the failure of an associated device of the device under test in terms of the target prediction variable; preprocess the data included in the fifth data set to obtain a sixth data set; perform feature expansion based on the sixth data set to obtain a seventh data set; determine that the union of the sixth data set and the seventh data set is the fourth data set.

[0192] In one possible implementation, the first acquisition module 502 is used to preprocess the data included in the fifth data set, specifically: the first acquisition module 502 is used to: based on a preset business data format, perform format normalization processing on the first historical business data and the second historical business data; perform data cleaning processing on the first historical time series data and the second historical time series data; based on a preset fault alarm data format, perform format conversion processing on the first historical fault alarm data and the second historical fault alarm data; and perform normalization processing on the numerical data in the fifth data set.

[0193] In one possible implementation, the first acquisition module 502 is used to construct the at least one fault prediction model based on the at least one feature group, specifically: the first acquisition module 502 is used to: use the principal component analysis method or the kernel principal component analysis method to perform dimensionality reduction processing on the at least one feature group to obtain a reduced dimensionality group of each feature group; each reduced dimensionality group includes data corresponding to the principal component that has a fault risk impact on the target prediction variable of the device under test; based on each reduced dimensionality group, construct the at least one fault prediction model.

[0194] In a possible implementation manner, the data included in the first data set corresponds to the principal component.

[0195] See also Figure 6 , Figure 6 This is a structural block diagram of an electronic device provided in an embodiment of the present application. Figure 6 As shown, the electronic device 600 may include a processor 601 and a memory 602; the memory 602 may be coupled to the processor 601. It is worth noting that Figure 6 is exemplary; other types of structures may also be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0196] In a possible implementation, the function of the fault risk prediction device 500 may be integrated into the processor 601. The processor 601 may be configured to perform the following operations:

[0197] Determine the target prediction variables and future prediction time for failure risk prediction of the equipment under test;

[0198] Acquire at least one fault prediction model corresponding to the target prediction variable;

[0199] Acquire a first data set; the first data set includes data obtained before the future prediction time and having an impact on the failure of the device under test in terms of the target prediction variable;

[0200] Obtaining a prediction value of each of the fault prediction models for predicting the fault risk of the device under test based on the first data set;

[0201] Based on all the predicted values, a failure risk prediction is performed on the device under test.

[0202] In another possible implementation, the fault risk prediction device 500 may be configured separately from the processor 601 . For example, the fault risk prediction device 500 may be configured as a chip connected to the processor 601 , and the fault risk prediction is implemented through the control of the processor 601 .

[0203] In addition, in some optional implementations, the electronic device 600 may also include: a communication module, an input unit, an audio processor, a display, a power supply, etc. It is worth noting that the electronic device 600 does not necessarily include Figure 6 In addition, the electronic device 600 may also include Figure 6 For components not shown, reference may be made to the prior art.

[0204] In some optional implementations, the processor 601 is sometimes also referred to as a controller or an operation control, and may include a microprocessor or other processor device and / or logic device, which receives input and controls the operation of various components of the electronic device 600.

[0205] The memory 602 may be, for example, a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory or other suitable devices. The memory 602 may store the above-mentioned information related to the fault risk prediction device 500, and may also store a program for executing the related information. The processor 601 may execute the program stored in the memory 602 to implement information storage or processing.

[0206] The input unit can provide input to the processor 601. The input unit is, for example, a key or a touch input device. The power supply can be used to provide power to the electronic device 600. The display can be used to display display objects such as images and text. The display can be, for example, an LCD display, but is not limited thereto.

[0207] The memory 602 may be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It may also be a memory that saves information even when the power is off, can be selectively erased, and is provided with more data, examples of which are sometimes referred to as EPROMs, etc. The memory 602 may also be some other type of device. The memory 602 includes a buffer memory (sometimes referred to as a buffer). The memory 602 may include an application / function storage unit for storing application programs and function programs or processes for executing the operation of the electronic device 600 through the processor 601.

[0208] The memory 602 may also include a data storage unit for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit of the memory 602 may include various drivers for the communication function of the computer device and / or for executing other functions of the computer device (such as a messaging application, a contact book application, etc.).

[0209] The communication module is a transmitter / receiver that sends and receives signals via an antenna. The communication module (transmitter / receiver) is coupled to the processor 601 to provide input signals and receive output signals, which may be the same as in a conventional mobile communication terminal.

[0210] Based on different communication technologies, multiple communication modules may be provided in the same computer device, such as a cellular network module, a Bluetooth module and / or a wireless local area network module, etc. The communication module (transmitter / receiver) is also coupled to a speaker and a microphone via an audio processor to provide an audio output via the speaker and receive an audio input from the microphone, thereby realizing a common telecommunication function. The audio processor may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor is also coupled to the processor 601, so that recording can be performed on the machine through the microphone, and the sound stored on the machine can be played through the speaker.

[0211] The embodiments of the present application also provide a computer-readable storage medium capable of implementing all the steps in the fault risk prediction method in the above embodiments. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, all the steps in the fault risk prediction method in the above embodiments are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0212] Determine the target prediction variables and future prediction time for failure risk prediction of the equipment under test;

[0213] Acquire at least one fault prediction model corresponding to the target prediction variable;

[0214] Acquire a first data set; the first data set includes data obtained before the future prediction time and having an impact on the failure of the device under test in terms of the target prediction variable;

[0215] Obtaining a prediction value of each of the fault prediction models for predicting the fault risk of the device under test based on the first data set;

[0216] Based on all the predicted values, a failure risk prediction is performed on the device under test.

[0217] Although the present application provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-creative labor. The order of steps listed in the embodiments is only one way of executing the order of many steps and does not represent the only execution order. When the actual device or client product is executed, it can be executed in the order of the method shown in the embodiments or the drawings or in parallel (for example, in a parallel processor or multi-threaded processing environment).

[0218] It should be understood by those skilled in the art that the embodiments of this specification may be provided as methods, devices (systems) or computer program products. Therefore, the embodiments of this specification may take the form of complete hardware embodiments, complete software embodiments or embodiments combining software and hardware. Moreover, the application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0219] The present application is described with reference to the flowcharts and / or block diagrams of the methods, apparatus (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0220] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0221] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0222] Each embodiment in this specification is described in a progressive manner, and the same and similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. In this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or equipment. The orientation or position relationship indicated by the terms "upper", "lower", etc. is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application. Unless otherwise expressly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two components. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to the specific circumstances. It should be noted that the embodiments and features in the embodiments in this application can be combined with each other unless there is a conflict. The present application is not limited to any single aspect, nor to any single embodiment, nor to any combination and / or permutation of these aspects and / or embodiments. Moreover, each aspect and / or embodiment of the present application can be used alone or in combination with one or more other aspects and / or embodiments thereof.

[0223] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application, and they should all be included in the scope of the claims and specification of the present application.

[0224] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "install", "connect" and "connect" should be understood in a broad sense. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0225] The present application has been described above in conjunction with preferred embodiments, but these embodiments are only exemplary and serve only as an illustration. On this basis, various replacements and improvements may be made to the present application, all of which fall within the scope of protection of the present application.

Claims

1. A method for predicting failure risk, characterized in that: The method comprises: Determine the target prediction variables and future prediction time for failure risk prediction of the equipment under test; Acquire at least one fault prediction model corresponding to the target prediction variable; Acquire a first data set; the first data set includes data obtained before the future prediction time and having an impact on the failure of the device under test in terms of the target prediction variable; Obtaining a prediction value of each of the fault prediction models for predicting the fault risk of the device under test based on the first data set; Based on all the predicted values, a failure risk prediction is performed on the device under test.

2. The method according to claim 1, characterized in that The device to be tested is a frequency converter in a target system; the target system includes a power generation system, an energy extraction system or a building system; the target prediction variable includes at least one of the following variables: input voltage, input current, output voltage, output current, operating frequency, control cabinet temperature, power cabinet temperature, transformer cabinet temperature, and humidity inside the frequency converter cabinet.

3. The method according to claim 1 or 2, characterized in that The future prediction time is the next moment after the prediction start moment; The obtaining of the first data set includes: Acquire a second data set; the second data set includes data obtained at the start time of the prediction that has an impact on the failure of the device under test in terms of the target prediction variable; Determine that the second data set is the first data set.

4. The method according to claim 1 or 2, characterized in that: The future prediction time is a first time period from the prediction start time; the duration of the first time period is a preset duration; The obtaining of the first data set includes: Acquire a third data set; the third data set includes data obtained in a second time period that has an impact on the failure of the device under test in terms of the target prediction variable; the end time of the second time period is the prediction start time, and the duration of the second time period is the preset duration; The third data set is determined to be the first data set.

5. The method according to claim 1 or 2, characterized in that: The obtaining of at least one fault prediction model corresponding to the target prediction variable includes: Acquire a fourth data set; the fourth data set includes historical data that is correlated with the failure of the device under test in terms of the target prediction variable; Based on a preset feature group division rule, dividing the data included in the fourth data set into at least one feature group; Based on the at least one feature group, the at least one fault prediction model is constructed.

6. The method according to claim 5, characterized in that The acquiring of the fourth data set includes: Acquire a fifth data set; the fifth data set includes first historical business data, first historical time series data, and / or first historical fault alarm data that are correlated with the failure of the device under test in terms of the target prediction variable, and second historical business data, second historical time series data, and / or second historical fault alarm data that are correlated with the failure of an associated device of the device under test in terms of the target prediction variable; Preprocessing the data included in the fifth data set to obtain a sixth data set; Perform feature expansion based on the sixth data set to obtain a seventh data set; A union of the sixth data set and the seventh data set is determined to be the fourth data set.

7. The method according to claim 6, characterized in that The preprocessing of the data included in the fifth data set includes: Based on a preset business data format, performing format normalization processing on the first historical business data and the second historical business data; Performing data cleaning processing on the first historical time series data and the second historical time series data; Based on a preset fault alarm data format, performing format conversion processing on the first historical fault alarm data and the second historical fault alarm data; The numerical data in the fifth data set is normalized.

8. The method according to claim 5, characterized in that The constructing the at least one fault prediction model based on the at least one feature group includes: Using a principal component analysis method or a kernel principal component analysis method, performing dimensionality reduction processing on the at least one feature group to obtain a dimensionality reduction group of each feature group; each dimensionality reduction group includes data corresponding to a principal component that has a failure risk impact on the target prediction variable of the device under test; Based on each of the dimensionality reduction groups, at least one fault prediction model is constructed.

9. The method according to claim 8, characterized in that The first data set includes data corresponding to the principal components.

10. A fault risk prediction device, characterized in that: The device comprises: A determination module, used to determine target prediction variables and future prediction time for failure risk prediction of the equipment under test; A first acquisition module, used to acquire at least one fault prediction model corresponding to the target prediction variable; A second acquisition module is used to acquire a first data set; the first data set includes data obtained before the future prediction time and having an impact on the failure of the device under test in terms of the target prediction variable; A third acquisition module is used to obtain a prediction value of each fault prediction model for predicting the fault risk of the device under test based on the first data set; A prediction module is used to predict the failure risk of the device under test based on all the prediction values.

11. The device according to claim 10, characterized in that The device to be tested is a frequency converter in a target system; the target system includes a power generation system, an energy extraction system or a building system; the target prediction variable includes at least one of the following variables: input voltage, input current, output voltage, output current, operating frequency, control cabinet temperature, power cabinet temperature, transformer cabinet temperature, and humidity inside the frequency converter cabinet.

12. The device according to claim 10 or 11, characterized in that The future prediction time is the next moment after the prediction start moment; The second acquisition module is used to acquire the first data set, specifically: The second acquisition module is used for: Acquire a second data set; the second data set includes data obtained at the start time of the prediction that has an impact on the failure of the device under test in terms of the target prediction variable; Determine that the second data set is the first data set.

13. The device according to claim 10 or 11, characterized in that The future prediction time is a first time period from the prediction start time; the duration of the first time period is a preset duration; The second acquisition module is used to acquire the first data set, specifically: The second acquisition module is used for: Acquire a third data set; the third data set includes data obtained in a second time period that has an impact on the failure of the device under test in terms of the target prediction variable; the end time of the second time period is the prediction start time, and the duration of the second time period is the preset duration; The third data set is determined to be the first data set.

14. The device according to claim 10 or 11, characterized in that The first acquisition module is used to acquire at least one fault prediction model corresponding to the target prediction variable, specifically: The first acquisition module is used for: Acquire a fourth data set; the fourth data set includes historical data that is correlated with the failure of the device under test in terms of the target prediction variable; Based on a preset feature group division rule, dividing the data included in the fourth data set into at least one feature group; Based on the at least one feature group, the at least one fault prediction model is constructed.

15. The device according to claim 14, characterized in that The first acquisition module is used to acquire the fourth data set, specifically: The first acquisition module is used for: Acquire a fifth data set; the fifth data set includes first historical business data, first historical time series data, and / or first historical fault alarm data that are correlated with the failure of the device under test in terms of the target prediction variable, and second historical business data, second historical time series data, and / or second historical fault alarm data that are correlated with the failure of an associated device of the device under test in terms of the target prediction variable; Preprocessing the data included in the fifth data set to obtain a sixth data set; Perform feature expansion based on the sixth data set to obtain a seventh data set; A union of the sixth data set and the seventh data set is determined to be the fourth data set.

16. The device according to claim 15, characterized in that The first acquisition module is used to preprocess the data included in the fifth data set, specifically: The first acquisition module is used for: Based on a preset business data format, performing format normalization processing on the first historical business data and the second historical business data; Performing data cleaning processing on the first historical time series data and the second historical time series data; Based on a preset fault alarm data format, performing format conversion processing on the first historical fault alarm data and the second historical fault alarm data; The numerical data in the fifth data set is normalized.

17. The device according to claim 14, characterized in that The first acquisition module is used to construct the at least one fault prediction model based on the at least one feature group, specifically: The first acquisition module is used for: Using a principal component analysis method or a kernel principal component analysis method, performing dimensionality reduction processing on the at least one feature group to obtain a dimensionality reduction group of each feature group; each dimensionality reduction group includes data corresponding to a principal component that has a failure risk impact on the target prediction variable of the device under test; Based on each of the dimensionality reduction groups, at least one fault prediction model is constructed.

18. The device according to claim 17, characterized in that The first data set includes data corresponding to the principal components.

19. An electronic device, characterized in that: The electronic device comprises: A memory for storing a computer program product; A processor is used to execute a computer program product stored in the memory, and when the computer program product is executed, the method described in any one of claims 1 to 9 is implemented.

20. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, which, when executed, implement the method described in any one of claims 1 to 9.

Images