CN113204280B - Method, system, equipment and medium for diagnosing power failure - Google Patents
Method, system, equipment and medium for diagnosing power failure Download PDFInfo
- Publication number
- CN113204280B CN113204280B CN202110501950.2A CN202110501950A CN113204280B CN 113204280 B CN113204280 B CN 113204280B CN 202110501950 A CN202110501950 A CN 202110501950A CN 113204280 B CN113204280 B CN 113204280B
- Authority
- CN
- China
- Prior art keywords
- data
- source
- fault
- domain data
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
- G06F1/26—Power supply means, e.g. regulation thereof
- G06F1/30—Means for acting in the event of power-supply failure or interruption, e.g. power-supply fluctuations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Supply And Distribution Of Alternating Current (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a method, a system, equipment and a storage medium for diagnosing power failure, wherein the method comprises the following steps: collecting original multi-source data of a power supply in a normal state and a fault state, and preprocessing the original multi-source data to obtain multi-source data; performing depth feature extraction on the multi-source data in an offline stage to obtain source domain data, and obtaining a pre-training model according to the source domain data; performing depth feature extraction on the multi-source data in an online stage to obtain source domain data and target domain data respectively, and reducing distribution difference of the source domain data and the target domain data to update the pre-training model into a training model; and predicting the fault type of the power supply according to the training model. The application adopts domain self-adaptive transfer learning to intelligently identify the UPS state of the data center, can automatically diagnose the type of UPS fault, has high robustness and saves manpower.
Description
Technical Field
The present application relates to the field of power supplies, and more particularly, to a method, system, computer device, and readable medium for diagnosing power failure.
Background
With the continuous acceleration of the digitizing process and the continuous progress of artificial intelligence technology, the demands of data centers are increasing. The Modularized Data Center (MDC) is a new concept adopted for coping with the changes of servers such as cloud computing, virtualization, centralization, high densification and the like and improving the operation efficiency of the data center, and has the advantages of rapid deployment, green energy conservation and flexible expansion compared with the traditional data center. The state monitoring and abnormality early warning of the data center are important components of intelligent operation and maintenance management. With the increasing size and complexity of data centers, the difficulty of operation and maintenance management is increasing. This is mainly due to the enhanced association and coupling between the systems, and the problem of one link not only affects the subsystem, but also may cause abnormality of the associated subsystem. In a data center, a UPS (Uninterruptible Power Supply ) is an intermediate device that connects an ac power source to a powered device, and can output stable electrical energy as a backup power system for protecting data security in a data center server. Uncertainty factors in the system, such as fluctuation of grid voltage, output overload, change of ambient temperature and the like, can cause problems of overcurrent of a UPS inverter, abnormal rectifier and the like, and the abnormal conditions can cause abnormal power supply of the UPS to influence the safe operation of a data center, so that the condition monitoring and fault diagnosis are necessary.
The traditional UPS fault diagnosis method mainly identifies the fault type through a modeling method based on expert knowledge and machine learning. The modeling method based on expert knowledge needs to establish a large number of knowledge rules, and parameters in the judgment rules are set according to the expert knowledge to have certain randomness, so that the method is difficult to establish an accurate fault diagnosis model. The UPS fault diagnosis based on machine learning obtains key indexes representing the UPS state through feature extraction and feature selection, and on the basis, the UPS state is identified through a machine learning intelligent identification model. The method mainly comprises the steps of manually extracting the characteristics, has certain subjectivity, and the quality of the selected characteristics has direct influence on fault diagnosis results.
In recent years, with the progress of deep learning technology, it has been a remarkable result in the fields of image processing and recognition. However, there are few studies in the UPS fault diagnosis field, and there are the following problems: 1) The UPS fault data of the data center are less in serious unbalance compared with normal data; 2) The UPS data of the data center is affected by external environment and other factors and is in dynamic change, so that the training data and the actual field data are distributed in an inconsistent manner. 3) Training the data model is difficult to adapt to the distribution of field data and thus produces poor recognition results.
Therefore, the rule base is difficult to establish by adopting the rule method fault diagnosis of expert knowledge, and the parameter setting has strong subjectivity; the fault diagnosis adopting the machine learning modeling method has the defects of complicated manual feature extraction, strong subjectivity and fault diagnosis accuracy depending on feature selection quality; the traditional deep learning UPS fault diagnosis method has the defects that the adaptability of the abnormal distribution data model is poor, and the fault prediction accuracy is affected by the data distribution difference.
Disclosure of Invention
In view of the above, an object of the embodiments of the present application is to provide a method, a system, a computer device and a computer readable storage medium for diagnosing a power failure, wherein the method and the system amplify small sample failure data to increase the number of failure samples by one-dimensional Wasserstein generation of an countermeasure network 1D-WGAN; deep feature extraction is carried out on original UPS multi-source data by combining a variable scale sliding window method with a DBN network; and the distribution difference between the UPS source domain data and the target domain data is reduced by adopting an MK-MMD characteristic measurement criterion, the model parameters are updated in an online self-adaptive mode, and the updated network model predicts UPS faults online, so that the defects in the prior art can be effectively avoided.
Based on the above objects, an aspect of the embodiments of the present application provides a method for diagnosing a power failure, including the steps of: collecting original multi-source data of a power supply in a normal state and a fault state, and preprocessing the original multi-source data to obtain multi-source data; performing depth feature extraction on the multi-source data in an offline stage to obtain source domain data, and obtaining a pre-training model according to the source domain data; performing depth feature extraction on the multi-source data in an online stage to obtain source domain data and target domain data respectively, and reducing distribution difference of the source domain data and the target domain data to update the pre-training model into a training model; and predicting the fault type of the power supply according to the training model.
In some embodiments, the preprocessing the raw multi-source data to obtain multi-source data includes: and amplifying the original multi-source data in the fault state to obtain first multi-source data, and combining the first multi-source data with the original multi-source data in the normal state to obtain the multi-source data.
In some embodiments, the augmenting the original multi-source data in the fault state to obtain first multi-source data includes: constructing a generating model and a judging model; inputting noise data in the original multi-source data under the fault state into the generation model to obtain a generation sample, and inputting the generation sample and the fault data in the original multi-source data under the fault state into the discrimination model to distinguish; responding to the judging model to distinguish fault data from a generated sample, adjusting the generated model according to the difference between the generated sample and the fault data and returning to the previous step; and responsive to the discrimination model failing to distinguish between fault data and a generated sample, augmenting the fault data with the generated model.
In some embodiments, the online stage performing depth feature extraction on the multi-source data to obtain source domain data and target domain data, respectively, includes: extracting local features of the multi-source data by adopting a variable-scale sliding window method and an RBM network; and fusing the local features by adopting a DBN network model to extract deep features.
In some implementations, the reducing the distribution variance of the source domain data and the target domain data to update the pre-training model to a training model includes: and reducing the distribution difference of the source domain data and the target domain data by adopting MK-MMD characteristic measurement criteria.
In some embodiments, the reducing the distribution variance of the source domain data and the target domain data using MK-MMD feature metric criteria includes: the source domain features and the target domain features are differentially measured using a linear combination of the plurality of base kernels.
In some embodiments, the reducing the distribution variance of the source domain data and the target domain data using MK-MMD feature metric criteria includes: network parameters are updated by back propagation to reduce the difference between the source domain and the target domain.
In another aspect of an embodiment of the present application, there is provided a system for diagnosing a power failure, including: the preprocessing module is configured to collect original multi-source data of the power supply in a normal state and a fault state, and preprocess the original multi-source data to obtain multi-source data; the creation module is configured to perform depth feature extraction on the multi-source data in an offline stage to obtain source domain data, and obtain a pre-training model according to the source domain data; the updating module is configured to perform depth feature extraction on the multi-source data in an online stage to respectively obtain source domain data and target domain data, and reduce distribution difference of the source domain data and the target domain data so as to update the pre-training model into a training model; and the execution module is configured to predict the fault type of the power supply according to the training model.
In yet another aspect of the embodiment of the present application, there is also provided a computer apparatus, including: at least one processor; and a memory storing computer instructions executable on the processor, which when executed by the processor, perform the steps of the method as above.
In yet another aspect of the embodiments of the present application, there is also provided a computer-readable storage medium storing a computer program which, when executed by a processor, implements the method steps as described above.
The application has the following beneficial technical effects:
(1) The domain self-adaptive transfer learning is adopted to intelligently identify the UPS state of the data center, so that the type of UPS fault can be automatically diagnosed, the robustness is high, and the labor is saved;
(2) The UPS state of the data center can be predicted in real time, and the early warning result is fed back to an administrator through a mail or a mobile terminal to realize intelligent early warning.
Drawings
In order to more clearly illustrate the embodiments of the application or the technical solutions in the prior art, the drawings that are necessary for the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the application and that other embodiments may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of an embodiment of a method for diagnosing power failure provided by the present application;
fig. 2 is a schematic diagram of a network structure for generating 1D-WGAN fault data according to an embodiment of the present application;
FIG. 3 is a schematic diagram of a hardware configuration of an embodiment of a computer device for diagnosing power failure according to the present application;
FIG. 4 is a schematic diagram of an embodiment of a computer storage medium for diagnosing power failure according to the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the following embodiments of the present application will be described in further detail with reference to the accompanying drawings.
It should be noted that, in the embodiments of the present application, all the expressions "first" and "second" are used to distinguish two entities with the same name but different entities or different parameters, and it is noted that the "first" and "second" are only used for convenience of expression, and should not be construed as limiting the embodiments of the present application, and the following embodiments are not described one by one.
In view of the above object, according to a first aspect of the embodiments of the present application, an embodiment of a method for diagnosing a power failure is provided. Fig. 1 is a schematic diagram of an embodiment of a method for diagnosing power failure provided by the present application. As shown in fig. 1, the embodiment of the present application includes the following steps:
s1, collecting original multi-source data of a power supply in a normal state and a fault state, and preprocessing the original multi-source data to obtain multi-source data;
s2, performing depth feature extraction on the multi-source data in an offline stage to obtain source domain data, and obtaining a pre-training model according to the source domain data;
s3, performing depth feature extraction on the multi-source data in an online stage to obtain source domain data and target domain data respectively, and reducing distribution difference of the source domain data and the target domain data to update the pre-training model into a training model; and
s4, predicting the fault type of the power supply according to the training model.
Aiming at the defects of the prior art, the application provides a data center UPS fault diagnosis method based on domain self-adaptive transfer learning and small sample data. The method comprises the steps of generating a countermeasure network 1D-WGAN through one-dimensional Wasserstein to amplify small sample fault data and improve the number of fault samples; deep feature extraction is carried out on original UPS multi-source data by combining a variable scale sliding window method with a DBN network; and reducing the distribution difference between the UPS source domain data and the target domain data by adopting an MK-MMD characteristic measurement criterion, and adaptively updating model parameters on line, and predicting UPS faults on line by the updated network model.
And collecting original multi-source data of the power supply in a normal state and a fault state, and preprocessing the original multi-source data to obtain multi-source data.
In some embodiments, the preprocessing the raw multi-source data to obtain multi-source data includes: and amplifying the original multi-source data in the fault state to obtain first multi-source data, and combining the first multi-source data with the original multi-source data in the normal state to obtain the multi-source data.
In order to select data representing UPS faults, the application fuses UPS multi-source data to carry out fault feature extraction, and the selected data are shown in the following table:
| sequence number | Data | Name of the name |
| 1 | Input current A | Input_Current_A |
| 2 | Input current B | Input_Current_B |
| 3 | Input current C | Input_Current_C |
| 4 | Output current A | Output_Current_A |
| 5 | Output voltage A | Output_Voltage_A |
| 6 | Output load | Output_Workload |
| 7 | UPS temperature | UPS_Temperature |
| 8 | Ambient temperature | ENV_Temperatutre |
| 9 | Ambient humidity | ENV_Humidity |
In practical application, the application provides a 1D-WGAN network model for amplifying fault data in order to increase the number of fault samples, wherein the UPS fault data generated on site are less and have larger difference with normal sample data.
In some embodiments, the augmenting the original multi-source data in the fault state to obtain first multi-source data includes: constructing a generating model and a judging model; inputting noise data in the original multi-source data under the fault state into the generation model to obtain a generation sample, and inputting the generation sample and the fault data in the original multi-source data under the fault state into the discrimination model to distinguish; responding to the judging model to distinguish fault data from a generated sample, adjusting the generated model according to the difference between the generated sample and the fault data and returning to the previous step; and responsive to the discrimination model failing to distinguish between fault data and a generated sample, augmenting the fault data with the generated model. The generative model comprises an input layer, a one-dimensional convolution layer, a full connection layer and an output layer. Wherein the input layer UPS noise data and the output layer generated UPS fault data are marked as X i '＝(x i1 ',x i2 ',x i3 ',...x i9 '), i=1, 2,..n. Where i is the number of samples. The discriminant model comprises an input layer, a one-dimensional convolution layer, a Maxpooling1D layer, a flame layer, a Dropout layer, a full connection layer and an output layer. Wherein the input layer is a real UPS fault X i ＝(x i1 ,x i2 ,x i3 ,...x i9 ) I=1, 2, 3..n and UPS fault data X generated thereby i . The output is true fault data or generates predictions of fault data.
The Wasserstein distance metric function adopted by WGAN is used for trueThe measure between real data and the generated data distribution, the loss function is as follows: generator loss function:the arbiter loss function is:the function of the loss function is to increase the distribution distance between the real fault sample and the generated fault sample as much as possible when the generator is fixed c Approximately the wasperstein distance between the two. When the discriminator is fixed, the distribution distance between the generated fault sample and the real fault sample is reduced as much as possible g Approximately the wasperstein distance between the two.
The overall network structure for generating fault data is shown in fig. 2. As shown in fig. 2, the multi-source data is divided into UPS fault data and UPS noise data, the UPS fault data is directly input as a real sample to a discrimination model, the UPS noise data is input as a generated sample to the discrimination model through a generation model, if the discrimination model can identify the real sample and the generated sample, the generation model is not perfect enough, and further adjustment is needed. The generated model may be further adjusted based on the output of the discriminant model. Until the discrimination model fails to identify the true sample and the generated sample, the generated model may be used to augment the UPS fault data.
And in an offline stage, extracting depth features of the multi-source data to obtain source domain data, and obtaining a pre-training model according to the source domain data.
And in an online stage, extracting depth features of the multi-source data to obtain source domain data and target domain data respectively, and reducing distribution difference of the source domain data and the target domain data to update the pre-training model into a training model.
In some embodiments, the online stage performing depth feature extraction on the multi-source data to obtain source domain data and target domain data, respectively, includes: extracting local features of the multi-source data by adopting a variable-scale sliding window method and an RBM network; and fusing the local features by adopting a DBN network model to extract deep features.
The traditional DBN network fault identification method is characterized in that original data are preprocessed and features are manually extracted and input into a network model to judge fault types, but the method cannot fully utilize the original data information, and therefore, the method is used for extracting local features by combining a sliding window method with an RBM network. Recording the length of single sample data as L, the length of a sliding window unit as L, and the sliding step length as T, wherein the number of visible layer nodes in the RBM network is as follows:
the method comprises the steps of extracting features of original data through a sliding window method and an RBM network to obtain:
will F RBM The features are input into a DBN network for deep feature extraction to obtain:
because the test data and the training data have different working conditions to cause different data distribution, MK-MMD is adopted for F DBN The feature performs domain adaptive learning to reduce the difference between the two data distributions. And network parameters are updated online to identify UPS faults.
In some implementations, the reducing the distribution variance of the source domain data and the target domain data to update the pre-training model to a training model includes: and reducing the distribution difference of the source domain data and the target domain data by adopting MK-MMD characteristic measurement criteria.
In some embodiments, the reducing the distribution variance of the source domain data and the target domain data using MK-MMD feature metric criteria includes: the source domain features and the target domain features are differentially measured using a linear combination of the plurality of base kernels.
In some embodiments, the reducing the distribution variance of the source domain data and the target domain data using MK-MMD feature metric criteria includes: network parameters are updated by back propagation to reduce the difference between the source domain and the target domain.
In order to reduce the difference between test data and training data in data distribution, the application carries out self-adaptive migration learning on source domain data and target domain data from a feature layer. The specific method comprises the following steps:
(1) Adopting a variable-scale sliding window method and an RBM network to extract local characteristics of UPS multi-source data;
(2) Fusing the local features by adopting a DBN network model to extract deep features;
(3) Reducing the feature domain difference and updating model parameters by adopting MK-MMD feature measurement criteria;
(4) The updated network parameters are used for online reasoning and fault identification of the test data.
Wherein the MK-MMD feature metric function is as follows:
wherein φ (·) represents a value that maps k (x) to the kernel s ,x t )＝<φ(x s ),φ(x t )>And (5) relevant feature mapping. k (x) s ,x t ) Expressed as a linear combination of l basis kernels:
and carrying out difference measurement on the source domain characteristics and the target domain characteristics by using linear combination of the l basic kernels, and taking the distance measurement parameters into the loss function, and reducing the difference between the source domain and the target domain by updating the network parameters through back propagation.
And predicting the fault type of the power supply according to the training model.
For a clearer description of the technical solution of the present application, the steps of the present application will now be described by way of example:
(1) Collecting multi-source data of the UPS in three states of normal, bypass faults and hardware faults;
(2) 1D-WGAN is adopted to amplify the small sample fault data so as to improve the number of UPS fault samples;
(3) Carrying out local feature extraction on UPS multi-source data by using offline data and adopting a variable-scale sliding window method and an RBM network;
(4) Fusing local features by using offline data and adopting a DBN network model to extract deep features;
(5) Training the network model by using the offline data to obtain a pre-training model;
(6) Inputting the online data and the offline data into a pre-training model, executing a variable-scale sliding window method and carrying out local feature extraction on UPS multi-source data by an RBM network;
(7) Inputting local features of online data and offline data into a DBN network to output deep features;
(8) Performing distance measurement on deep features of online data and offline data by MK-MMD, calculating the overall loss, updating through counter-propagating network parameters, and outputting a UPS state by the updated network model;
(9) And finishing the deployment of the web end and the checking of the real-time prediction result of the UPS state by adopting a flash.
The following table shows the accuracy rates of the UPS fault diagnosis under different methods in the training state and the testing state:
| Method | training accuracy | Accuracy of test |
| DBN | 93％ | 62％ |
| LRBM+DBN | 94％ | 74％ |
| LRBM+DBN+1D-WGAN | 92％ | 78％ |
| LRBM+DBN+1D-WGAN+MK-MMD (application) | 98％ | 91％ |
Wherein, DBN is the traditional method for extracting artificial characteristics and then adopts DBN identification method; the LRBM adopts RBM local feature extraction and is matched with a DBN identification method; the LRBM+DBN+1D-WGAN is a RBM local feature extraction method after fault sample amplification and is matched with a DBN identification method; the LRBM+DBN+1D-WGAN+MK-MMD is the UPS fault diagnosis result provided by the application, and the best performance can be achieved on training data and on-site verification data according to the application.
The application provides a data center UPS fault diagnosis method based on domain self-adaptive transfer learning and small sample data. The method comprises the steps of amplifying and improving the number of fault samples by generating a countermeasure network 1D-WGAN (peer-to-peer network) to small sample fault data; deep feature extraction is carried out on original UPS multi-source data by combining a variable scale sliding window method with a DBN network; and reducing the distribution difference between the UPS source domain data and the target domain data by adopting an MK-MMD characteristic measurement criterion, and adaptively updating model parameters on line, and predicting UPS faults on line by the updated network model.
It should be noted that, in the foregoing embodiments of the method for diagnosing power failure, the steps may be intersected, replaced, added and subtracted, so that the method for diagnosing power failure by using these reasonable permutation and combination changes should also fall within the scope of the present application, and the scope of the present application should not be limited to the embodiments.
In view of the above object, a second aspect of the embodiments of the present application provides a system for diagnosing a power failure, including: the preprocessing module is configured to collect original multi-source data of the power supply in a normal state and a fault state, and preprocess the original multi-source data to obtain multi-source data; the creation module is configured to perform depth feature extraction on the multi-source data in an offline stage to obtain source domain data, and obtain a pre-training model according to the source domain data; the updating module is configured to perform depth feature extraction on the multi-source data in an online stage to respectively obtain source domain data and target domain data, and reduce distribution difference of the source domain data and the target domain data so as to update the pre-training model into a training model; and the execution module is configured to predict the fault type of the power supply according to the training model.
In some embodiments, the preprocessing module is configured to: and amplifying the original multi-source data in the fault state to obtain first multi-source data, and combining the first multi-source data with the original multi-source data in the normal state to obtain the multi-source data.
In some embodiments, the preprocessing module is configured to: constructing a generating model and a judging model; inputting noise data in the original multi-source data under the fault state into the generation model to obtain a generation sample, and inputting the generation sample and the fault data in the original multi-source data under the fault state into the discrimination model to distinguish; responding to the judging model to distinguish fault data from a generated sample, adjusting the generated model according to the difference between the generated sample and the fault data and returning to the previous step; and responsive to the discrimination model failing to distinguish between fault data and a generated sample, augmenting the fault data with the generated model.
In some embodiments, the update module is configured to: extracting local features of the multi-source data by adopting a variable-scale sliding window method and an RBM network; and fusing the local features by adopting a DBN network model to extract deep features.
In some embodiments, the update module is configured to: and reducing the distribution difference of the source domain data and the target domain data by adopting MK-MMD characteristic measurement criteria.
In some embodiments, the update module is configured to: the source domain features and the target domain features are differentially measured using a linear combination of the plurality of base kernels.
In some embodiments, the update module is configured to: network parameters are updated by back propagation to reduce the difference between the source domain and the target domain.
In view of the above object, a third aspect of the embodiments of the present application provides a computer device, including: at least one processor; and a memory storing computer instructions executable on the processor, the instructions being executable by the processor to perform the steps of: s1, collecting original multi-source data of a power supply in a normal state and a fault state, and preprocessing the original multi-source data to obtain multi-source data; s2, performing depth feature extraction on the multi-source data in an offline stage to obtain source domain data, and obtaining a pre-training model according to the source domain data; s3, performing depth feature extraction on the multi-source data in an online stage to obtain source domain data and target domain data respectively, and reducing distribution difference of the source domain data and the target domain data to update the pre-training model into a training model; s4, predicting the fault type of the power supply according to the training model.
In some embodiments, the preprocessing the raw multi-source data to obtain multi-source data includes: and amplifying the original multi-source data in the fault state to obtain first multi-source data, and combining the first multi-source data with the original multi-source data in the normal state to obtain the multi-source data.
In some embodiments, the augmenting the original multi-source data in the fault state to obtain first multi-source data includes: constructing a generating model and a judging model; inputting noise data in the original multi-source data under the fault state into the generation model to obtain a generation sample, and inputting the generation sample and the fault data in the original multi-source data under the fault state into the discrimination model to distinguish; responding to the judging model to distinguish fault data from a generated sample, adjusting the generated model according to the difference between the generated sample and the fault data and returning to the previous step; and responsive to the discrimination model failing to distinguish between fault data and a generated sample, augmenting the fault data with the generated model.
In some embodiments, the online stage performing depth feature extraction on the multi-source data to obtain source domain data and target domain data, respectively, includes: extracting local features of the multi-source data by adopting a variable-scale sliding window method and an RBM network; and fusing the local features by adopting a DBN network model to extract deep features.
In some implementations, the reducing the distribution variance of the source domain data and the target domain data to update the pre-training model to a training model includes: and reducing the distribution difference of the source domain data and the target domain data by adopting MK-MMD characteristic measurement criteria.
In some embodiments, the reducing the distribution variance of the source domain data and the target domain data using MK-MMD feature metric criteria includes: the source domain features and the target domain features are differentially measured using a linear combination of the plurality of base kernels.
In some embodiments, the reducing the distribution variance of the source domain data and the target domain data using MK-MMD feature metric criteria includes: network parameters are updated by back propagation to reduce the difference between the source domain and the target domain.
Fig. 3 is a schematic hardware structure of an embodiment of the above-mentioned computer device for diagnosing power failure according to the present application.
Taking the example of the apparatus shown in fig. 3, the apparatus includes a processor 201 and a memory 202, and may further include: an input device 203 and an output device 204.
The processor 201, memory 202, input devices 203, and output devices 204 may be connected by a bus or other means, for example in fig. 3.
The memory 202 is used as a non-volatile computer readable storage medium for storing non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions/modules corresponding to the method for diagnosing power failure in the embodiments of the present application. The processor 201 executes various functional applications of the server and data processing, that is, implements the method of diagnosing power failure of the above-described method embodiment, by running nonvolatile software programs, instructions, and modules stored in the memory 202.
Memory 202 may include a storage program area that may store an operating system, at least one application program required for functionality, and a storage data area; the storage data area may store data created according to the use of a method of diagnosing power failure, etc. In addition, memory 202 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, memory 202 may optionally include memory located remotely from processor 201, which may be connected to the local module via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input device 203 may receive input information such as a user name and a password. The output device 204 may include a display device such as a display screen.
One or more program instructions/modules corresponding to the method of diagnosing a power failure are stored in the memory 202 that, when executed by the processor 201, perform the method of diagnosing a power failure in any of the method embodiments described above.
Any one of the embodiments of the computer apparatus that performs the above-described method of diagnosing power failure may achieve the same or similar effects as any of the previously-described method embodiments that correspond thereto.
The application also provides a computer readable storage medium storing a computer program which when executed by a processor performs the method as above.
FIG. 4 is a schematic diagram of an embodiment of the above-mentioned computer storage medium for diagnosing power failure according to the present application. Taking a computer storage medium as shown in fig. 4 as an example, the computer readable storage medium 3 stores a computer program 31 that when executed by a processor performs the above method.
Finally, it should be noted that, as will be understood by those skilled in the art, implementing all or part of the above-described embodiments of the method may be implemented by a computer program to instruct related hardware, and the program of the method for diagnosing power failure may be stored in a computer readable storage medium, where the program may include the steps of the embodiments of the methods described above when executed. The storage medium of the program may be a magnetic disk, an optical disk, a read-only memory (ROM), a random-access memory (RAM), or the like. The computer program embodiments described above may achieve the same or similar effects as any of the method embodiments described above.
The foregoing is an exemplary embodiment of the present disclosure, but it should be noted that various changes and modifications could be made herein without departing from the scope of the disclosure as defined by the appended claims. The functions, steps and/or actions of the method claims in accordance with the disclosed embodiments described herein need not be performed in any particular order. Furthermore, although elements of the disclosed embodiments may be described or claimed in the singular, the plural is contemplated unless limitation to the singular is explicitly stated.
It should be understood that as used herein, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly supports the exception. It should also be understood that "and/or" as used herein is meant to include any and all possible combinations of one or more of the associated listed items.
The foregoing embodiment of the present application has been disclosed with reference to the number of embodiments for the purpose of description only, and does not represent the advantages or disadvantages of the embodiments.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program for instructing relevant hardware, and the program may be stored in a computer readable storage medium, where the storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
Those of ordinary skill in the art will appreciate that: the above discussion of any embodiment is merely exemplary and is not intended to imply that the scope of the disclosure of embodiments of the application, including the claims, is limited to such examples; combinations of features of the above embodiments or in different embodiments are also possible within the idea of an embodiment of the application, and many other variations of the different aspects of the embodiments of the application as described above exist, which are not provided in detail for the sake of brevity. Therefore, any omission, modification, equivalent replacement, improvement, etc. of the embodiments should be included in the protection scope of the embodiments of the present application.
Claims (6)
1. A method of diagnosing a power failure comprising the steps of:
collecting original multi-source data of a power supply in a normal state and a fault state, and preprocessing the original multi-source data to obtain multi-source data;
performing depth feature extraction on the multi-source data in an offline stage to obtain source domain data, and obtaining a pre-training model according to the source domain data;
performing depth feature extraction on the multi-source data in an online stage to obtain source domain data and target domain data respectively, and reducing distribution difference of the source domain data and the target domain data to update the pre-training model into a training model; and
predicting the fault type of the power supply according to the training model;
the performing depth feature extraction on the multi-source data in the online stage to obtain source domain data and target domain data respectively includes: extracting local features of the multi-source data by adopting a variable-scale sliding window method and an RBM network; fusing the local features by adopting a DBN network model to extract deep features;
the reducing the distribution difference of the source domain data and the target domain data to update the pre-training model to a training model includes: reducing the distribution difference of the source domain data and the target domain data by adopting MK-MMD characteristic measurement criteria; wherein the reducing the distribution difference of the source domain data and the target domain data by using MK-MMD feature metric criteria comprises: performing differential measurement on the source domain features and the target domain features by using linear combination of a plurality of basic cores; network parameters are updated by back propagation to reduce the difference between the source domain and the target domain.
2. The method of claim 1, wherein preprocessing the raw multi-source data to obtain multi-source data comprises:
and amplifying the original multi-source data in the fault state to obtain first multi-source data, and combining the first multi-source data with the original multi-source data in the normal state to obtain the multi-source data.
3. The method of claim 2, wherein the augmenting the original multi-source data in the fault state to obtain first multi-source data comprises:
constructing a generating model and a judging model;
inputting noise data in the original multi-source data under the fault state into the generation model to obtain a generation sample, and inputting the generation sample and the fault data in the original multi-source data under the fault state into the discrimination model to distinguish;
responding to the judging model to distinguish fault data from a generated sample, adjusting the generated model according to the difference between the generated sample and the fault data and returning to the previous step; and
and in response to the discrimination model failing to distinguish fault data from the generated samples, augmenting the fault data by the generated model.
4. A system for diagnosing a power failure, comprising:
the preprocessing module is configured to collect original multi-source data of the power supply in a normal state and a fault state, and preprocess the original multi-source data to obtain multi-source data;
the creation module is configured to perform depth feature extraction on the multi-source data in an offline stage to obtain source domain data, and obtain a pre-training model according to the source domain data;
the updating module is configured to perform depth feature extraction on the multi-source data in an online stage to respectively obtain source domain data and target domain data, and reduce distribution difference of the source domain data and the target domain data so as to update the pre-training model into a training model; and
the execution module is configured to predict the fault type of the power supply according to the training model;
the preprocessing module is further configured to extract local features of the multi-source data by adopting a variable scale sliding window method and an RBM network; fusing the local features by adopting a DBN network model to extract deep features;
the updating module is further configured to reduce a distribution difference of the source domain data and the target domain data by adopting an MK-MMD characteristic measurement criterion; wherein the reducing the distribution difference of the source domain data and the target domain data by using MK-MMD feature metric criteria comprises: performing differential measurement on the source domain features and the target domain features by using linear combination of a plurality of basic cores; network parameters are updated by back propagation to reduce the difference between the source domain and the target domain.
5. A computer device, comprising:
at least one processor; and
a memory storing computer instructions executable on the processor, which instructions when executed by the processor implement the steps of the method of any one of claims 1-3.
6. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the method of any of claims 1-3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110501950.2A CN113204280B (en) | 2021-05-08 | 2021-05-08 | Method, system, equipment and medium for diagnosing power failure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110501950.2A CN113204280B (en) | 2021-05-08 | 2021-05-08 | Method, system, equipment and medium for diagnosing power failure |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113204280A CN113204280A (en) | 2021-08-03 |
| CN113204280B true CN113204280B (en) | 2023-09-26 |
Family
ID=77030710
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110501950.2A Active CN113204280B (en) | 2021-05-08 | 2021-05-08 | Method, system, equipment and medium for diagnosing power failure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113204280B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114660500B (en) * | 2022-05-19 | 2022-08-19 | 秦皇岛广控科技有限公司 | Three-phase monitoring protection system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109947086A (en) * | 2019-04-11 | 2019-06-28 | 清华大学 | Mechanical breakdown migration diagnostic method and system based on confrontation study |
| CN111523785A (en) * | 2020-04-16 | 2020-08-11 | 三峡大学 | Power system dynamic security assessment method based on generation countermeasure network |
| CN111898095A (en) * | 2020-07-10 | 2020-11-06 | 佛山科学技术学院 | Deep migration learning intelligent fault diagnosis method and device, storage medium and equipment |
| CN111898634A (en) * | 2020-06-22 | 2020-11-06 | 西安交通大学 | Intelligent fault diagnosis method based on depth-to-reactance-domain self-adaption |
| CN112101085A (en) * | 2020-07-22 | 2020-12-18 | 西安交通大学 | Adaptive intelligent fault diagnosis method based on importance weighted domain impedance |
| CN112328588A (en) * | 2020-11-27 | 2021-02-05 | 哈尔滨工程大学 | Industrial fault diagnosis unbalanced time sequence data expansion method |
-
2021
- 2021-05-08 CN CN202110501950.2A patent/CN113204280B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109947086A (en) * | 2019-04-11 | 2019-06-28 | 清华大学 | Mechanical breakdown migration diagnostic method and system based on confrontation study |
| CN111523785A (en) * | 2020-04-16 | 2020-08-11 | 三峡大学 | Power system dynamic security assessment method based on generation countermeasure network |
| CN111898634A (en) * | 2020-06-22 | 2020-11-06 | 西安交通大学 | Intelligent fault diagnosis method based on depth-to-reactance-domain self-adaption |
| CN111898095A (en) * | 2020-07-10 | 2020-11-06 | 佛山科学技术学院 | Deep migration learning intelligent fault diagnosis method and device, storage medium and equipment |
| CN112101085A (en) * | 2020-07-22 | 2020-12-18 | 西安交通大学 | Adaptive intelligent fault diagnosis method based on importance weighted domain impedance |
| CN112328588A (en) * | 2020-11-27 | 2021-02-05 | 哈尔滨工程大学 | Industrial fault diagnosis unbalanced time sequence data expansion method |
Non-Patent Citations (2)
| Title |
|---|
| 基于深度-迁移学习的输电线路故障选相模型及其可迁移性研究;杨毅 等;《电力自动化设备》;20201031;第40卷(第10期);第165-171页 * |
| 杨毅 等.基于深度-迁移学习的输电线路故障选相模型及其可迁移性研究.《电力自动化设备》.2020,第40卷(第10期),第165-171页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113204280A (en) | 2021-08-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Appiah et al. | Long short-term memory networks based automatic feature extraction for photovoltaic array fault diagnosis | |
| Zheng et al. | A novel equivalent model of active distribution networks based on LSTM | |
| Bai et al. | A novel parameter identification approach via hybrid learning for aggregate load modeling | |
| US11656589B2 (en) | Systems and methods for automatic power topology discovery | |
| Mukherjee et al. | Development of an ensemble decision tree-based power system dynamic security state predictor | |
| CN110988563B (en) | UPS (uninterrupted Power supply) fault detection method, device, equipment and storage medium | |
| CN108847686A (en) | A kind of photovoltaic DC-to-AC converter failure prediction method | |
| Choi et al. | Impact analysis of locational marginal price subject to power system topology errors | |
| CN106600145A (en) | Power distribution network reliability analysis method and device | |
| CN113204280B (en) | Method, system, equipment and medium for diagnosing power failure | |
| CN111091141B (en) | Photovoltaic backboard fault diagnosis method based on layered Softmax | |
| CN108334427A (en) | Method for diagnosing faults in storage system and device | |
| US11657148B2 (en) | Event analysis in an electric power system | |
| CN109587145B (en) | False data intrusion detection method, device and equipment in power network | |
| CN114997566A (en) | Power grid blocking risk assessment method and system considering node connectivity loss | |
| CN115878992A (en) | Monitoring method and monitoring system for comprehensive pipe rack power supply system | |
| Zhang et al. | Attention mechanism-driven potential fault cause identification in optical networks | |
| Moreira et al. | Evaluation of pattern recognition algorithms for applications on power factor compensation | |
| CN117114458A (en) | Active power distribution network power quality assessment method, device, terminal and storage medium | |
| Bhatnagar et al. | LSTM-based low-impedance fault and high-impedance fault detection and classification | |
| Sawhney et al. | On-line transient stability assessment using artificial neural network | |
| Kulikov et al. | Statistical Sampling for Power-Quality Monitoring in Modern Power-Supply Systems | |
| Thapa et al. | A machine learning-based approach to detection of fault locations in transmission networks | |
| Al-Gubri et al. | Performance analysis of machine learning algorithms for power system dynamic security assessment | |
| Ramadoss et al. | Machine learning approach to differentiate excitation failure in synchronous generators from power swing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |