CN111579056A - Transformer direct-current magnetic bias prediction method and system - Google Patents
Transformer direct-current magnetic bias prediction method and system Download PDFInfo
- Publication number
- CN111579056A CN111579056A CN202010424515.XA CN202010424515A CN111579056A CN 111579056 A CN111579056 A CN 111579056A CN 202010424515 A CN202010424515 A CN 202010424515A CN 111579056 A CN111579056 A CN 111579056A
- Authority
- CN
- China
- Prior art keywords
- spectrum information
- real
- audio signal
- frequency spectrum
- transformer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Complex Calculations (AREA)
Abstract
The application relates to a transformer direct current magnetic biasing prediction method and a system, wherein the transformer direct current magnetic biasing prediction method comprises the steps of collecting audio signals; carrying out feature extraction on the audio signal to obtain frequency spectrum information; inputting the frequency spectrum information into a deep neural network model, and training the deep neural network model; and acquiring real-time frequency spectrum information corresponding to the real-time audio signal, and inputting the real-time frequency spectrum information into the trained deep neural network model to obtain the prediction probability of the direct current magnetic biasing state corresponding to the real-time frequency spectrum information. The transformer is not required to be contacted when data are collected, the working state of the transformer cannot be influenced, the learning effect cannot be influenced due to environment or equipment replacement, and the robustness is stronger.
Description
Technical Field
The application belongs to the technical field of transformer state monitoring, and particularly relates to a transformer direct-current magnetic bias prediction method and system, wherein the monitoring of a transformer direct-current magnetic bias state is realized by utilizing audio analysis.
Background
Power transformers are very important devices in power systems, and their operating conditions directly affect the safety of the power system. In recent years, with more and more domestic direct current transmission projects, when a high-voltage direct current transmission system operates in a single-pole ground loop mode, a transformer with a grounded neutral point at the periphery of an earth electrode can generate a direct current magnetic bias phenomenon, so that the normal operation of a power grid is influenced. The dc bias of the transformer is an abnormal operating state of the transformer, which means that a dc current enters a winding of the transformer due to some reason, and a dc magnetic potential or a dc magnetic flux and a series of electromagnetic effects generated thereby are generated in a core of the transformer. When the transformer generates direct current magnetic biasing, the temperature rise of structural accessories of the transformer is increased, and the noise is increased; on the other hand, a large amount of harmonic waves are generated, reactive loss is increased, system voltage waveform distortion is caused, relay protection misoperation is caused, the reactive compensation device can be overloaded or the system voltage is reduced, and the influences are adverse to the safety of the substation equipment and the stable operation of the power grid.
In the related art, the direct current magnetic bias of the transformer is detected based on a vibration noise acquisition unit. The detection method comprises the steps of collecting vibration noise signals of all vibration noise sensitive point areas in real time, collecting vibration noise in a no-load and load state, and when the vibration noise sensitive point areas exceed a preset value, considering that the problem of direct current magnetic biasing occurs. However, in the method, the vibration sensor needs to contact the transformer so as to acquire a vibration signal, and the occurrence of the magnetic bias state is simply compared with a preset value. If the environment or equipment is changed, the robustness of the method is poor.
Disclosure of Invention
In order to overcome the problem that in the related art, a vibration sensor needs to contact a transformer to acquire a vibration signal in a method for detecting the direct current magnetic bias of the transformer based on a vibration noise acquisition unit, and meanwhile, the occurrence of the magnetic bias state is compared with a preset value simply, and in addition, if the environment or equipment is changed, the robustness of the method is poor, the method and the system for predicting the direct current magnetic bias of the transformer are provided.
In a first aspect, the present application provides a method for predicting dc magnetic bias of a transformer, including:
collecting an audio signal;
extracting the characteristics of the audio signal to obtain frequency spectrum information;
inputting the frequency spectrum information into a deep neural network model, and training the deep neural network model;
and acquiring real-time frequency spectrum information corresponding to the real-time audio signal, and inputting the real-time frequency spectrum information into the trained deep neural network model to obtain the prediction probability of the direct current magnetic biasing state corresponding to the real-time frequency spectrum information.
Further, before inputting the spectrum information into the neural network model, the method further includes:
and marking the audio signal according to the working condition category, specifically comprising: and intercepting and separating the audio signal containing the DC magnetic bias state and the audio signal not containing the DC magnetic bias to create an audio library containing two types of audio signals.
Further, the acquiring the audio signal includes: and recording the working state of the transformer by adopting a pickup sensor to obtain the audio signal in the direct current magnetic bias state and the audio signal without the direct current magnetic bias.
Further, training the deep neural network model includes:
establishing a deep neural network model, wherein the deep neural network model comprises a plurality of full-connection layers, spectrum information is used as input of the neural network model, an audio library is used as marking data, a training loss function is two-classification cross entropy, and an Adam optimizer is used for carrying out gradient descent operation to train the neural network model.
Further, the extracting the features of the audio signal to obtain the spectrum information includes:
framing the audio signal, and selecting a plurality of sampling points as a frame;
and carrying out fast Fourier transform operation on the frame to obtain frequency spectrum information.
Further, the performing a fast fourier transform operation on the frame includes:
and inputting the frame into a Hamming window and then carrying out fast Fourier transform operation.
Further, the acquiring real-time spectrum information corresponding to the real-time audio signal includes:
monitoring the transformer in real time for 24 hours to obtain a real-time audio signal;
and performing feature extraction on the real-time audio signal to obtain real-time frequency spectrum information.
Further, the method also comprises the following steps:
smoothing the prediction probability by adopting a moving average processing mode to obtain a final prediction probability, wherein the final prediction probability comprises the following steps:
where i is the frame number, yiFor the probability value, 2n +1 is the number of smoothing frames.
Further, the method also comprises the following steps:
and setting a probability screening threshold, and converting the predicted probability into a binary working condition prediction result according to the screening threshold, wherein the working condition prediction result comprises the phenomenon of DC magnetic biasing and the phenomenon of no DC magnetic biasing.
In a second aspect, the present application provides a transformer dc magnetic bias prediction system, including:
the acquisition module is used for acquiring audio signals;
the characteristic extraction module is used for extracting the characteristics of the audio signal to obtain frequency spectrum information;
the training module is used for inputting the frequency spectrum information into a deep neural network model and training the deep neural network model;
and the prediction module is used for acquiring real-time frequency spectrum information corresponding to the real-time audio signal, inputting the real-time frequency spectrum information into the trained deep neural network model and obtaining the prediction probability of the direct current magnetic biasing state corresponding to the real-time frequency spectrum information.
The technical scheme provided by the embodiment of the application can have the following beneficial effects:
according to the transformer direct-current magnetic bias prediction method and system provided by the embodiment of the invention, audio signals are collected, characteristics of the audio signals are extracted to obtain frequency spectrum information, the frequency spectrum information is input into a deep neural network model, the deep neural network model is trained to obtain real-time frequency spectrum information corresponding to the real-time audio signals, the real-time frequency spectrum information is input into the trained deep neural network model to obtain the prediction probability of a direct-current magnetic bias state corresponding to the real-time frequency spectrum information, and whether the direct-current magnetic bias state occurs or not is diagnosed. Compared with methods such as current analysis and vibration signal analysis for predicting direct current magnetic biasing, the method does not need to contact the transformer when acquiring data, and does not influence the working state of the transformer. Meanwhile, the audio signal analysis adopts a supervised learning method based on a deep neural network, so that the learning effect is not influenced by environment or equipment replacement, and the method has stronger robustness.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and together with the description, serve to explain the principles of the application.
Fig. 1 is a flowchart of a method for predicting dc magnetic bias of a transformer according to an embodiment of the present disclosure.
Fig. 2 is a flowchart of a method for predicting dc magnetic bias of a transformer according to another embodiment of the present disclosure.
Fig. 3 is a functional block diagram of a transformer dc magnetic bias prediction system according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the technical solutions of the present application will be described in detail below. It is to be understood that the embodiments described are only a few embodiments of the present application and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the examples given herein without making any creative effort, shall fall within the protection scope of the present application.
Fig. 1 is a flowchart of a transformer dc bias prediction method according to an embodiment of the present application, and as shown in fig. 1, the transformer dc bias prediction method includes:
s11: collecting an audio signal;
s12: carrying out feature extraction on the audio signal to obtain frequency spectrum information;
s13: inputting the frequency spectrum information into a deep neural network model, and training the deep neural network model;
s14: and acquiring real-time frequency spectrum information corresponding to the real-time audio signal, and inputting the real-time frequency spectrum information into the trained deep neural network model to obtain the prediction probability of the direct current magnetic biasing state corresponding to the real-time frequency spectrum information.
The vibration and noise of the transformer and the harmonics in the transformer current increase with the increase of the dc current through the transformer neutral, while the magnitude of the dc current in the transformer grounded neutral is related to the dc line power delivered in the single pole earth mode, the dc ground distance. The phenomenon of direct current magnetic bias refers to: the direct current flowing into the winding becomes a part of the exciting current of the transformer, the direct current makes the iron core of the transformer generate magnetic bias, the exciting curve working point of the transformer is changed, a part of the original magnetizing curve working area is moved to the iron core magnetic saturation area, the total exciting current becomes spike wave, the problems of local overheating, insulation aging, load performance reduction and the like of the transformer caused by the spike wave cause great influence on the normal operation of the transformer, and the transformer can be damaged and protected when the spike wave is serious. Meanwhile, the transformer after direct current magnetic biasing can generate considerable broadband harmonic waves, and the operating environment of a power grid is polluted. Although researchers have certain knowledge about the negative effects of the alternating current power grid, according to most of the events of aggravation of noise and vibration of the transformer which occur in the power grid with the direct current transmission converter station, it can be found that the negative effects generated after the direct current magnetic biasing of the transformer are larger than expected, and even the transformer with serious magnetic biasing response can not normally operate.
In the related art, the direct current magnetic bias of the transformer is detected based on a vibration noise acquisition unit. The detection method comprises the steps of collecting vibration noise signals of all vibration noise sensitive point areas in real time, collecting vibration noise in a no-load and load state, and when the vibration noise sensitive point areas exceed a preset value, considering that the problem of direct current magnetic biasing occurs. However, in the method, the vibration sensor needs to contact the transformer so as to acquire a vibration signal, and the occurrence of the magnetic bias state is simply compared with a preset value. If the environment or equipment is changed, the robustness of the method is poor.
Some prior art adopt a neutral point current measurement method, which is to measure whether a current of a neutral point has a direct current component, and if the direct current component is measured, it is determined that a transformer has direct current magnetic bias. However, this method is poor in real-time performance and cannot monitor the dc bias state timely and effectively.
In this embodiment, real-time spectrum information corresponding to a real-time audio signal is acquired, the real-time spectrum information is input to a deep neural network model which is trained, a prediction probability p, 0< ═ p < ═ 1 of a direct-current magnetic bias state corresponding to the real-time spectrum information is obtained, the real-time performance of direct-current magnetic bias monitoring can be improved, the direct-current magnetic bias state is diagnosed timely and effectively through the prediction probability p, for example, when the prediction probability p is 0.9, it is determined that a direct-current magnetic bias phenomenon occurs, and when p is 0.2, the direct-current magnetic bias phenomenon does not occur.
In this embodiment, the audio signal is collected, the audio signal is subjected to feature extraction to obtain spectrum information, the spectrum information is input into the deep neural network model, the deep neural network model is trained to obtain real-time spectrum information corresponding to the real-time audio signal, the real-time spectrum information is input into the trained deep neural network model to obtain the prediction probability of the dc magnetic bias state corresponding to the real-time spectrum information, and whether the dc magnetic bias state occurs is diagnosed. Compared with methods such as current analysis and vibration signal analysis for predicting direct current magnetic biasing, the method does not need to contact the transformer when acquiring data, and does not influence the working state of the transformer. Meanwhile, the audio signal analysis adopts a supervised learning method based on a deep neural network, so that the learning effect is not influenced by environment or equipment replacement, and the method has stronger robustness.
An embodiment of the present invention provides another transformer dc magnetic bias prediction method, as shown in a flowchart in fig. 2, where the transformer dc magnetic bias prediction method further includes:
s21: gather audio signal, carry out operating mode classification mark to audio signal, specifically include: and intercepting and separating the audio signal containing the DC magnetic bias state and the audio signal not containing the DC magnetic bias to create an audio library containing two types of audio signals.
In some embodiments, capturing the audio signal comprises: and recording the working state of the transformer by adopting a pickup sensor to obtain an audio signal in a direct current magnetic bias state and an audio signal without the direct current magnetic bias.
It should be noted that the deployment location of the pickup sensor is selected from key locations around the transformer, for example, within several meters of the transformer, and the specific location is determined according to the performance or operating parameters of the pickup sensor. The pickup sensor keeps carrying out audio acquisition to transformer operating condition and surrounding environment for a long time, covers various operating condition of transformer as far as possible, and transformer operating condition includes direct current magnetic biasing operating mode and normal operating mode.
Just can only gather the vibration signal of laminating transformer position for vibration sensor, this application adopts pickup sensor to gather sound signal, can acquire all sounds of transformer whole scope, and information itself is more comprehensive, more make full use of audio signal information to improve the accuracy of prediction result.
S22: carrying out feature extraction on the audio signal to obtain frequency spectrum information;
the method specifically comprises the following steps:
framing the audio signal, and selecting a plurality of sampling points as a frame;
and carrying out fast Fourier transform operation on the frame to obtain frequency spectrum information.
In some embodiments, performing a fast fourier transform operation on the frame includes:
and inputting the frame into a Hamming window and then carrying out fast Fourier transform operation.
Taking the audio sampling frequency of 44.1kHz as an example, the audio is framed, and 1024 sampling points are selected as one frame. A Hamming window is added to the frame, thereby reducing the rectangular window frequency leakage problem when performing a fast Fourier transform.
S23: establishing a deep neural network model, wherein the deep neural network model comprises a plurality of full-connection layers, spectrum information is used as input of the neural network model, an audio library is used as labeled data, a training loss function is a two-classification cross entropy, and an Adam optimizer is used for carrying out gradient descent operation to train the neural network model.
An adaptive moment estimation (Adam) optimizer is a method of calculating the adaptive learning rate for each parameter.
By establishing a deep neural network model and training and verifying through a deep neural network model seed training set and a verification set, the precision ratio and the recall ratio of the direct current magnetic bias prediction result can reach more than 99%.
S24: inputting the frequency spectrum information into a deep neural network model, and training the deep neural network model;
s25: and acquiring real-time frequency spectrum information corresponding to the real-time audio signal, and inputting the real-time frequency spectrum information into the trained deep neural network model to obtain the prediction probability of the direct current magnetic biasing state corresponding to the real-time frequency spectrum information.
In some embodiments, obtaining real-time spectrum information corresponding to the real-time audio signal includes:
monitoring the transformer in real time for 24 hours to obtain a real-time audio signal;
and performing feature extraction on the real-time audio signal to obtain real-time frequency spectrum information.
By monitoring the transformer in real time for 24 hours, the stability of monitoring data can be ensured, the real-time property of direct current magnetic biasing prediction can be improved by acquiring real-time audio signals, whether the direct current magnetic biasing phenomenon occurs or not can be diagnosed in time, and the work of the transformer is prevented from being influenced.
S26: smoothing the prediction probability by adopting a moving average processing mode to obtain a final prediction probability, wherein the final prediction probability comprises the following steps:
where i is the frame number, yiTo predict the probability value, 2n +1 is the number of smoothing frames.
S27: setting a probability screening threshold value, and converting the predicted probability into a binary working condition prediction result according to the screening threshold value, wherein the working condition prediction result comprises the phenomenon of DC magnetic biasing and the phenomenon of no DC magnetic biasing.
The prediction output from the deep neural network model is changed into one y per frameiThe prediction probability value of (2) is between 0 and 1, wherein a frame time is about 23ms (1024/44100). And post-processing the prediction probability so as to enable the final prediction result to be more accurate.
The first post-processing mode: smoothing:
the time of each frame is about 23ms, if the frame result is taken as the final result of the type judgment, the prediction result of the previous frame and the next frame is not considered, the prediction result may be unstable, and therefore, the moving average is adopted
And a second post-treatment mode: setting a threshold value:
and obtaining a prediction probability value with the state prediction probability of each frame being 0-1, setting a proper threshold value, converting the probability pi into a binarization judgment result, and obtaining a final required prediction result, namely a judgment result of the occurrence of the direct current magnetic biasing phenomenon or the non-occurrence of the direct current magnetic biasing phenomenon.
The post-processing method includes, but is not limited to, the above two processing methods, and other processing methods for predicting the probability are also within the scope of the present invention.
In the embodiment, the working condition of the transformer is predicted in real time by collecting the sound of the transformer during working and training the deep neural network. In addition, the deep neural network model is established, so that the robustness and the adaptability of the algorithm are stronger, and the robustness of a prediction result cannot be influenced by the change of the environment or equipment. Furthermore, the accuracy of the prediction result is improved by carrying out post-processing such as smoothing and threshold setting on the multi-frame prediction probability.
An embodiment of the present invention provides a transformer dc magnetic bias prediction system, as shown in a functional structure diagram in fig. 3, where the transformer dc magnetic bias prediction system includes:
the acquisition module 31 is used for acquiring audio signals;
the feature extraction module 32 is configured to perform feature extraction on the audio signal to obtain frequency spectrum information;
the training module 33 is used for inputting the frequency spectrum information into the deep neural network model and training the deep neural network model;
the prediction module 34 is configured to obtain real-time frequency spectrum information corresponding to the real-time audio signal, input the real-time frequency spectrum information into the trained deep neural network model, and obtain a prediction probability of a dc magnetic bias state corresponding to the real-time frequency spectrum information.
In some embodiments, the transformer dc bias prediction system further comprises:
the operating condition type labeling module 35 is configured to perform operating condition type labeling on the audio signal, and specifically includes: and intercepting and separating the audio signal containing the DC magnetic bias state and the audio signal not containing the DC magnetic bias to create an audio library containing two types of audio signals.
A smoothing module 36, configured to smooth the prediction probability by using a moving average processing method to obtain a final prediction probability, where the final prediction probability includes:
where i is the frame number, yiFor the probability value, 2n +1 is the number of smoothing frames.
And a probability screening threshold setting module 37, configured to set a probability screening threshold, and convert the prediction probability into a binary working condition prediction result according to the screening threshold, where the working condition prediction result includes a dc magnetic bias phenomenon and a dc magnetic bias phenomenon.
In the embodiment, the audio signal is collected through the collection module, the characteristic extraction module performs characteristic extraction on the audio signal to obtain frequency spectrum information, the training module inputs the frequency spectrum information into the deep neural network model and trains the deep neural network model, the prediction module obtains real-time frequency spectrum information corresponding to the real-time audio signal and inputs the real-time frequency spectrum information into the trained deep neural network model to obtain the prediction probability of the direct current magnetic biasing state corresponding to the real-time frequency spectrum information, and compared with methods such as current analysis and vibration signal analysis, the direct current magnetic biasing state is predicted without contacting a transformer when data are collected, and the working state of the transformer is not influenced. Meanwhile, the audio signal analysis adopts a supervised learning method based on a deep neural network, so that the learning effect is not influenced by environment or equipment replacement, and the method has stronger robustness.
It is understood that the same or similar parts in the above embodiments may be mutually referred to, and the same or similar parts in other embodiments may be referred to for the content which is not described in detail in some embodiments.
It should be noted that, in the description of the present application, the terms "first", "second", etc. are used for descriptive purposes only and are not to be construed as indicating or implying relative importance. Further, in the description of the present application, the meaning of "a plurality" means at least two unless otherwise specified.
Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing specific logical functions or steps of the process, and the scope of the preferred embodiments of the present application includes other implementations in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the present application.
It should be understood that portions of the present application may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.
It will be understood by those skilled in the art that all or part of the steps carried by the method for implementing the above embodiments may be implemented by hardware related to instructions of a program, which may be stored in a computer readable storage medium, and when the program is executed, the program includes one or a combination of the steps of the method embodiments.
In addition, functional units in the embodiments of the present application may be integrated into one processing module, or each unit may exist alone physically, or two or more units are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may also be stored in a computer readable storage medium.
The storage medium mentioned above may be a read-only memory, a magnetic or optical disk, etc.
In the description herein, reference to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the application. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Although embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present application, and that variations, modifications, substitutions and alterations may be made to the above embodiments by those of ordinary skill in the art within the scope of the present application.
It should be noted that the present invention is not limited to the above-mentioned preferred embodiments, and those skilled in the art can obtain other products in various forms without departing from the spirit of the present invention, but any changes in shape or structure can be made within the scope of the present invention with the same or similar technical solutions as those of the present invention.
Claims (10)
1. A method for predicting the DC magnetic bias of a transformer is characterized by comprising the following steps:
collecting an audio signal;
extracting the characteristics of the audio signal to obtain frequency spectrum information;
inputting the frequency spectrum information into a deep neural network model, and training the deep neural network model;
and acquiring real-time frequency spectrum information corresponding to the real-time audio signal, and inputting the real-time frequency spectrum information into the trained deep neural network model to obtain the prediction probability of the direct current magnetic biasing state corresponding to the real-time frequency spectrum information.
2. The method for predicting the direct current magnetic bias of the transformer according to claim 1, wherein before inputting the frequency spectrum information into the neural network model, the method further comprises:
and marking the audio signal according to the working condition category, specifically comprising: and intercepting and separating the audio signal containing the DC magnetic bias state and the audio signal not containing the DC magnetic bias to create an audio library containing two types of audio signals.
3. The transformer dc bias prediction method of claim 2, wherein the capturing the audio signal comprises: and recording the working state of the transformer by adopting a pickup sensor to obtain the audio signal in the direct current magnetic bias state and the audio signal without the direct current magnetic bias.
4. The method for predicting the direct current magnetic bias of the transformer according to claim 2, wherein the training of the deep neural network model comprises:
establishing a deep neural network model, wherein the deep neural network model comprises a plurality of full-connection layers, spectrum information is used as input of the neural network model, an audio library is used as marking data, a training loss function is two-classification cross entropy, and an Adam optimizer is used for carrying out gradient descent operation to train the neural network model.
5. The method for predicting the DC magnetic bias of the transformer according to claim 1, wherein the extracting the features of the audio signal to obtain the spectrum information comprises:
framing the audio signal, and selecting a plurality of sampling points as a frame;
and carrying out fast Fourier transform operation on the frame to obtain frequency spectrum information.
6. The method of claim 1, wherein the performing a fast fourier transform operation on the frame comprises:
and inputting the frame into a Hamming window and then carrying out fast Fourier transform operation.
7. The method for predicting the dc magnetic bias of the transformer according to claim 1, wherein the obtaining of the real-time spectrum information corresponding to the real-time audio signal comprises:
monitoring the transformer in real time for 24 hours to obtain a real-time audio signal;
and performing feature extraction on the real-time audio signal to obtain real-time frequency spectrum information.
8. The method for predicting the DC bias of the transformer according to claim 1, further comprising:
smoothing the prediction probability by adopting a moving average processing mode to obtain a final prediction probability, wherein the final prediction probability comprises the following steps:
where i is the frame number, yi2n +1 is the number of smoothing frames.
9. The method for predicting the DC bias of the transformer according to claim 1, further comprising:
and setting a probability screening threshold, and converting the predicted probability into a binary working condition prediction result according to the screening threshold, wherein the working condition prediction result comprises the phenomenon of DC magnetic biasing and the phenomenon of no DC magnetic biasing.
10. A transformer DC magnetic bias prediction system is characterized by comprising:
the acquisition module is used for acquiring audio signals;
the characteristic extraction module is used for extracting the characteristics of the audio signal to obtain frequency spectrum information;
the training module is used for inputting the frequency spectrum information into a deep neural network model and training the deep neural network model;
and the prediction module is used for acquiring real-time frequency spectrum information corresponding to the real-time audio signal, inputting the real-time frequency spectrum information into the trained deep neural network model and obtaining the prediction probability of the direct current magnetic biasing state corresponding to the real-time frequency spectrum information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010424515.XA CN111579056A (en) | 2020-05-19 | 2020-05-19 | Transformer direct-current magnetic bias prediction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010424515.XA CN111579056A (en) | 2020-05-19 | 2020-05-19 | Transformer direct-current magnetic bias prediction method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111579056A true CN111579056A (en) | 2020-08-25 |
Family
ID=72119091
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010424515.XA Pending CN111579056A (en) | 2020-05-19 | 2020-05-19 | Transformer direct-current magnetic bias prediction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111579056A (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107918813A (en) * | 2017-12-14 | 2018-04-17 | 上海宽全智能科技有限公司 | Trend prediction analysis method, equipment and storage medium |
| CN108169583A (en) * | 2017-11-17 | 2018-06-15 | 国网湖南省电力有限公司 | Auto-transformer D.C. magnetic biasing method of discrimination and system of the neutral point through capacity earth |
| CN109740523A (en) * | 2018-12-29 | 2019-05-10 | 国网陕西省电力公司电力科学研究院 | A kind of method for diagnosing fault of power transformer based on acoustic feature and neural network |
| CN109829916A (en) * | 2019-03-04 | 2019-05-31 | 国网内蒙古东部电力有限公司 | A kind of Diagnosis Method of Transformer Faults based on CNN |
| CN110415709A (en) * | 2019-06-26 | 2019-11-05 | 深圳供电局有限公司 | Transformer working condition recognition methods based on Application on Voiceprint Recognition model |
| CN110534118A (en) * | 2019-07-29 | 2019-12-03 | 安徽继远软件有限公司 | Transformer/reactor method for diagnosing faults based on Application on Voiceprint Recognition and neural network |
| CN110632369A (en) * | 2019-09-23 | 2019-12-31 | 贵州电网有限责任公司 | Online acquisition method for transformer exciting current characteristic quantity |
| CN110634493A (en) * | 2019-09-09 | 2019-12-31 | 国网湖南省电力有限公司 | Transformer state identification method, system and medium based on voiceprint image characteristics |
| CN110767231A (en) * | 2019-09-19 | 2020-02-07 | 平安科技（深圳）有限公司 | Voice control equipment awakening word identification method and device based on time delay neural network |
| CN110827798A (en) * | 2019-11-12 | 2020-02-21 | 广州欢聊网络科技有限公司 | Audio signal processing method and device |
| CN111044814A (en) * | 2019-11-28 | 2020-04-21 | 中国电力科学研究院有限公司 | Method and system for identifying transformer direct-current magnetic bias abnormality |
-
2020
- 2020-05-19 CN CN202010424515.XA patent/CN111579056A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108169583A (en) * | 2017-11-17 | 2018-06-15 | 国网湖南省电力有限公司 | Auto-transformer D.C. magnetic biasing method of discrimination and system of the neutral point through capacity earth |
| CN107918813A (en) * | 2017-12-14 | 2018-04-17 | 上海宽全智能科技有限公司 | Trend prediction analysis method, equipment and storage medium |
| CN109740523A (en) * | 2018-12-29 | 2019-05-10 | 国网陕西省电力公司电力科学研究院 | A kind of method for diagnosing fault of power transformer based on acoustic feature and neural network |
| CN109829916A (en) * | 2019-03-04 | 2019-05-31 | 国网内蒙古东部电力有限公司 | A kind of Diagnosis Method of Transformer Faults based on CNN |
| CN110415709A (en) * | 2019-06-26 | 2019-11-05 | 深圳供电局有限公司 | Transformer working condition recognition methods based on Application on Voiceprint Recognition model |
| CN110534118A (en) * | 2019-07-29 | 2019-12-03 | 安徽继远软件有限公司 | Transformer/reactor method for diagnosing faults based on Application on Voiceprint Recognition and neural network |
| CN110634493A (en) * | 2019-09-09 | 2019-12-31 | 国网湖南省电力有限公司 | Transformer state identification method, system and medium based on voiceprint image characteristics |
| CN110767231A (en) * | 2019-09-19 | 2020-02-07 | 平安科技（深圳）有限公司 | Voice control equipment awakening word identification method and device based on time delay neural network |
| CN110632369A (en) * | 2019-09-23 | 2019-12-31 | 贵州电网有限责任公司 | Online acquisition method for transformer exciting current characteristic quantity |
| CN110827798A (en) * | 2019-11-12 | 2020-02-21 | 广州欢聊网络科技有限公司 | Audio signal processing method and device |
| CN111044814A (en) * | 2019-11-28 | 2020-04-21 | 中国电力科学研究院有限公司 | Method and system for identifying transformer direct-current magnetic bias abnormality |
Non-Patent Citations (1)
| Title |
|---|
| 王小川 等: "《MATLAB神经网络43个案例分析》", 31 August 2013 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Cui et al. | A novel event detection method using PMU data with high precision | |
| US10281504B2 (en) | Method and apparatus for analyzing waveform signals of a power system | |
| Gargoom et al. | Automatic classification and characterization of power quality events | |
| US7672812B2 (en) | Cable fault detection | |
| KR101352204B1 (en) | Apparatus and method for classification of power quality disturbances at power grids | |
| CN110880225B (en) | Residual current type electric fire intelligent monitoring analysis method and device | |
| CN109029960B (en) | Method for detecting mechanical state of circuit breaker | |
| Dong et al. | Frequency prediction of power systems in FNET based on state-space approach and uncertain basis functions | |
| Bhui et al. | Application of recurrence quantification analysis to power system dynamic studies | |
| CN110618314B (en) | Harmonic wave responsibility division method for resisting short-circuit fault interference of power distribution system | |
| CN105510760A (en) | Method for detecting short circuit fault data based on wavelet analysis | |
| CN110618366A (en) | Direct current arc detection method and device | |
| CN115687969A (en) | Low-voltage transformer fault diagnosis method based on sound characteristic analysis | |
| EP3012643B1 (en) | Method and apparatus for identifying causes for cable overcurrent | |
| CN111751645A (en) | Transformer direct-current magnetic bias monitoring method, device, equipment and system | |
| CN108594156B (en) | Improved current transformer saturation characteristic identification method | |
| CN111579056A (en) | Transformer direct-current magnetic bias prediction method and system | |
| CN109324268B (en) | Power distribution network early fault detection method and device based on Bayesian inference | |
| CN115343579B (en) | Power grid fault analysis method and device and electronic equipment | |
| CN109581051A (en) | Adaptive full frequency-domain Wave record method | |
| Wilson et al. | Uncertainty Quantification of Capacitor Switching Transient Location Using Machine Learning | |
| CN114609435A (en) | Voltage sag detection and classification identification method | |
| CN1097732C (en) | Method for testing and evaluating lightning strike and equipment thereof | |
| CN116151511B (en) | Intelligent diagnosis management method and system for distribution feeder and transformer area based on data processing | |
| You et al. | Build Smart Grids on Artificial Intelligence–A Real-world Example |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200825 |
|
| RJ01 | Rejection of invention patent application after publication |