CN113011096B - Current transformer saturation waveform recovery method based on model and data hybrid driving - Google Patents

Current transformer saturation waveform recovery method based on model and data hybrid driving Download PDF

Info

Publication number
CN113011096B
CN113011096B CN202110321839.5A CN202110321839A CN113011096B CN 113011096 B CN113011096 B CN 113011096B CN 202110321839 A CN202110321839 A CN 202110321839A CN 113011096 B CN113011096 B CN 113011096B
Authority
CN
China
Prior art keywords
current
fault current
training
model
fault
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
Application number
CN202110321839.5A
Other languages
Chinese (zh)
Other versions
CN113011096A (en
Inventor
刘志远
吴建云
郝治国
于晓军
杨松浩
蒙金有
罗美玲
黄伟兵
蔡乾
赫嘉楠
张宇博
史磊
林泽暄
叶涛
王小立
于小艳
沙云
尹琦云
陆洪建
杨晨
安燕杰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Jiaotong University
State Grid Ningxia Electric Power Co Ltd
Electric Power Research Institute of State Grid Ningxia Electric Power Co Ltd
Original Assignee
Xian Jiaotong University
State Grid Ningxia Electric Power Co Ltd
Electric Power Research Institute of State Grid Ningxia Electric Power Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian Jiaotong University, State Grid Ningxia Electric Power Co Ltd, Electric Power Research Institute of State Grid Ningxia Electric Power Co Ltd filed Critical Xian Jiaotong University
Priority to CN202110321839.5A priority Critical patent/CN113011096B/en
Publication of CN113011096A publication Critical patent/CN113011096A/en
Application granted granted Critical
Publication of CN113011096B publication Critical patent/CN113011096B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/22—Source localisation; Inverse modelling

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Testing Of Short-Circuits, Discontinuities, Leakage, Or Incorrect Line Connections (AREA)

Abstract

The invention discloses a current transformer saturation waveform recovery method based on model and data hybrid driving, which comprises the following steps of: s1, constructing a database; s2, building and training a fault current model key parameter identification network based on the long-term and short-term memory network; s3, calculating the real fault current based on the identification network of the step S2; and S4, using the real fault current obtained in the step S3 as the action information of the protection element, and waiting for the next fault. The fault current model key parameter identification network is established by using a physical model of the fault current as prior knowledge through a data driving method to realize waveform recovery, and the fault current model key parameter identification network has the advantages of low sampling rate requirement, strong noise resistance, no need of setting a threshold value, convenience in online/offline deployment and the like.

Description

Current transformer saturation waveform recovery method based on model and data hybrid driving
Technical Field
The invention relates to the technical field of saturation identification and waveform recovery of a current transformer, in particular to a current transformer saturation waveform recovery method based on model and data hybrid driving.
Background
At present, a current transformer is an important measuring element in a power system, and the working performance of the current transformer directly influences the reliability of the action of a relay protection element. The P-level current transformer widely applied to systems of 220kV and below has poor tolerance to direct current components in fault current due to the magnetic saturation characteristic of an iron core, and the saturation problem is difficult to avoid. When the current transformer is saturated, the transmission characteristic of the current transformer is changed from linear to nonlinear, so that a measured value cannot accurately reflect real fault current, false operation and rejection of a protection element can be caused, and normal operation of a power system is influenced. Therefore, the saturation identification and waveform recovery technology of the current transformer is an important guarantee for the correct action of the protection element, and has high engineering application value. The current transformer saturation identification and waveform recovery method widely applied in the current engineering is generally realized by detecting the sudden change information of the waveform, and the basic principle is that the sudden change information of the waveform is utilized to distinguish a saturated section and a non-saturated section in an output waveform, and then the non-saturated section data is utilized to recover a correct waveform which is used as the action information of a protection element.
However, methods based on abrupt information are susceptible to noise and the requirements on the sampling rate are generally high. In order to improve the noise resistance, the threshold value of the abrupt change information detection can be properly increased, but the problem of insufficient sensitivity when the saturation degree of the current transformer is shallow can be caused.
Therefore, there is a contradiction between the sensitivity and reliability of the mutation information-based method, making it difficult to select an appropriate threshold value a problem that needs to be solved by those skilled in the art.
Disclosure of Invention
In view of the above, the invention provides a current transformer saturated waveform recovery method based on model and data hybrid driving, that is, a fault current physical model is used as prior knowledge, a fault current model key parameter identification network is constructed by a data driving method to realize waveform recovery, and the method has the advantages of low sampling rate requirement, strong noise resistance, no need of setting a threshold value, convenience in online/offline deployment and the like.
In order to achieve the purpose, the invention adopts the following technical scheme:
a current transformer saturation waveform recovery method based on model and data hybrid driving comprises the following steps:
s1, constructing a database;
s2, building and training a fault current model key parameter identification network based on the long-term and short-term memory network;
s3, calculating the real fault current based on the identification network of the step S2;
and S4, using the real fault current obtained in the step S3 as the action information of the protection element, and waiting for the next fault.
Preferably, the step S1 specifically includes: a current transformer with a magnetic saturation effect is built, fault current simulation measurement is carried out, and real fault current i for training of one cycle after fault is obtainedfAnd measuring the current imNormalized by the following equation to obtain a normalized fault current i 'for training'fAnd measuring current i'm
Figure BDA0002993174710000031
The fault current being represented as a superposition of the power frequency component and the attenuated DC component, i.e.
Figure BDA0002993174710000032
Wherein, the amplitude A and phase angle of the power frequency component
Figure BDA0002993174710000033
The maximum value B of the decaying direct current component and the decay time constant tau are taken as the reasonKey parameters in the barrier current model are real fault current i 'used for training after normalization'fConversion into key parameters
Figure BDA0002993174710000034
I.e. the normalized measurement current i 'for training'mAnd fault current i'fThe corresponding relation between the measured current and the measured current is converted into normalized measured current i 'for training'mAnd the corresponding relation between the key parameter kp and the database is realized.
Preferably, the step S2 specifically includes: the system comprises an identification parameter submodule and an attention module, wherein the attention module is used for simulating a human cognitive process, namely, attention is flown to a non-saturation section;
the attention module adopts a supervised pre-training mode for training, and the output targets are as follows:
Figure BDA0002993174710000035
where k denotes the sampling point, wkRepresenting the assigned attention to the sampling point k, the greater the value, the more the assigned attention;
with normalized post-training measurement current i'mFor input, wkAnd carrying out supervised training on the attention module for reference output, and embedding the trained attention module into a parameter recognition submodule to construct a complete network model.
Preferably, the step S2 further includes: then normalizing the measured current i 'for post-training'mFor input, kp is reference output to train the complete network model, freeze the attention module parameters, train only the parameter recognition sub-module or adjust the parameters of the attention module.
Preferably, the step S3 specifically includes: deploying the trained current transformer saturated waveform recovery network model, continuously acquiring current data of one cycle after the fault, and acquiring fault current i for testingMNormalizing and recording the normalization factor
Figure BDA0002993174710000041
The normalized test fault current i'MAs an input to the network model, the network outputs an estimate of a key parameter of the true fault current
Figure BDA0002993174710000042
And substituting the fault current into the formula to calculate the converted fault current i'FObtaining the recovered real fault current i after inverse normalization of formula (4)F
iF＝NF×i'F (5)。
According to the technical scheme, compared with the prior art, the method for recovering the saturated waveform of the current transformer based on the model and data hybrid driving is disclosed, namely, the recovery of the waveform is realized by using the physical model of the fault current as priori knowledge and constructing the key parameter identification network of the fault current model through the data driving method, and the method has the advantages of low sampling rate requirement, strong anti-noise capability, no need of setting a threshold value, convenience in online/offline deployment and the like.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a diagram showing a relationship between a real current and a measured current when a current transformer provided by the invention is saturated.
Fig. 2 is a schematic diagram of a fault current model key parameter identification network provided by the invention.
FIG. 3 is a diagram of an equivalent circuit of a simulation model according to the present invention.
Fig. 4 is a diagram illustrating the effect of recovering the saturation waveform provided by the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment of the invention discloses a current transformer saturation waveform recovery method based on model and data hybrid driving, which comprises the following steps:
s1, constructing a database:
a current transformer with a magnetic saturation effect is built, fault current simulation measurement is carried out, and real fault current i for training of one cycle after fault is obtainedfAnd measuring the current imNormalized by the following equation to obtain a normalized fault current i 'for training'fAnd measuring current i'm
Figure BDA0002993174710000061
The fault current being represented as a superposition of the power frequency component and the attenuated DC component, i.e.
Figure BDA0002993174710000062
Wherein, the amplitude A and phase angle of the power frequency component
Figure BDA0002993174710000063
The maximum value B of the attenuation direct current component and the attenuation time constant tau are key parameters in the fault current model, and the normalized real fault current i 'is used for training'fConversion into key parameters
Figure BDA0002993174710000064
I.e. after normalizationMeasured current i 'for training'mAnd fault current i'fThe corresponding relation between the measured current and the measured current is converted into normalized measured current i 'for training'mAnd the corresponding relation between the key parameter kp and the database is realized.
S2, building and training a fault current model key parameter identification network based on the long-term and short-term memory network:
the system comprises an identification parameter submodule and an attention module, wherein the attention module is used for simulating a human cognitive process, namely, attention is flown to a non-saturation section;
the attention module adopts a supervised pre-training mode for training, and the output targets are as follows:
Figure BDA0002993174710000065
where k denotes the sampling point, wkRepresenting the attention allocated to the sampling point k, wherein the larger the numerical value is, the more attention is allocated;
with normalized post-training measurement current i'mFor input, wkSupervised training is carried out on the attention module for reference output, the trained attention module is embedded into a parameter identification submodule to construct a complete network model, and then the measured current i 'for training is normalized'mFor input, kp is reference output to train the complete network model, freeze the attention module parameters, train only the parameter recognition sub-module or adjust the parameters of the attention module.
S3, calculating the real fault current based on the identification network of the step S2:
deploying the trained current transformer saturated waveform recovery network model, continuously acquiring current data of one cycle after the fault, and acquiring fault current i for testingMNormalizing and recording the normalization factor
Figure BDA0002993174710000071
Measure after normalizationTrial fault current i'MAs an input to the network model, the network outputs an estimate of a key parameter of the true fault current
Figure BDA0002993174710000072
And substituting the fault current into the formula to calculate converted fault current i'FObtaining the recovered real fault current i after inverse normalization of formula (4)F
iF＝NF×i'F (5)。
And S4, using the real fault current obtained in the step S3 as the action information of the protection element, and waiting for the next fault.
A simulation model shown in FIG. 3 is built, and a large number of simulation samples are obtained by setting different fault conditions (fault initial phase angle, fault resistance, fault position and the like) and different current transformer operating conditions (load level, initial magnetic flux and the like). The network model is trained by using the processed simulation sample, and the effect of the trained network model is shown in fig. 4, which shows that: the current transformer saturated waveform recovery method can accurately, effectively and quickly recover real fault current from the saturated waveform, can well solve the problems of the failure and misoperation of a protection device caused by the saturation of the current transformer, and ensures the normal operation of a power system; meanwhile, the attention module really allocates more weight to the unsaturated segment, and the interpretability and generalization capability of the model are improved.
The application has the advantages that:
1. the method realizes the ingenious combination of a fault current physical model and a data driving method, has better physical significance and characteristic mining capability, does not need to carry out threshold setting, and simplifies the design flow.
2. Based on data driving, the influence of noise can be fully considered in the network training phase, so that the noise resistance of the model is improved.
3. The LSTM network can improve the utilization rate of information by mining the strong characteristics of the time sequence signals, reduce the requirement on the sampling rate and reduce the hardware cost.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (4)

1. A current transformer saturation waveform recovery method based on model and data hybrid driving is characterized by comprising the following steps:
s1, constructing a database, specifically:
a current transformer with a magnetic saturation effect is built, fault current simulation measurement is carried out, and real fault current i for training of one cycle after fault is obtainedfAnd measuring the current imNormalized by the following equation to obtain a normalized fault current i 'for training'fAnd measuring current i'm
Figure FDA0003649499620000011
The fault current being represented as a superposition of the power frequency component and the attenuated DC component, i.e.
Figure FDA0003649499620000012
WhereinAmplitude A and phase angle of power frequency component
Figure FDA0003649499620000013
The maximum value B of the attenuation direct current component and the attenuation time constant tau are key parameters in the fault current model, and the normalized real fault current i 'is used for training'fConversion into key parameters
Figure FDA0003649499620000014
I.e. the normalized measurement current i 'for training'mAnd fault current i'fThe corresponding relation between the measured current and the measured current is converted into normalized measured current i 'for training'mThe corresponding relation between the key parameter kp and the database is established;
s2, building and training a fault current model key parameter identification network based on the long-term and short-term memory network;
s3, calculating the real fault current based on the identification network of the step S2;
and S4, using the real fault current obtained in the step S3 as the action information of the protection element, and waiting for the next fault.
2. The method for recovering the saturation waveform of the current transformer based on model and data hybrid driving according to claim 1, wherein the step S2 specifically includes: identifying a parameter sub-module and an attention module for simulating a human cognitive process, i.e. assigning attention to the unsaturated segment;
the attention module adopts a supervised pre-training mode for training, and the output targets are as follows:
Figure FDA0003649499620000021
where k denotes the sampling point, wkRepresenting the assigned attention to the sampling point k, the greater the value, the more the assigned attention;
with normalized post-training measurement current i'mFor input, wkAnd carrying out supervised training on the attention module for reference output, and embedding the trained attention module into a parameter recognition submodule to construct a complete network model.
3. The method for recovering the saturation waveform of the current transformer based on the model and data hybrid driving as claimed in claim 2, wherein the step S2 further comprises: then normalizing the measured current i 'for post-training'mFor input, kp is reference output to train the complete network model, freeze the attention module parameters, train only the parameter recognition sub-module or fine tune the parameters of the attention module.
4. The method for recovering the saturation waveform of the current transformer based on model and data hybrid driving according to claim 1, wherein the step S3 specifically includes: deploying the trained current transformer saturated waveform recovery network model, continuously acquiring current data of one cycle after the fault, and acquiring fault current i for testingMNormalizing and recording the normalization factor
Figure FDA0003649499620000022
The normalized test fault current i'MAs an input to the network model, the network outputs an estimate of a key parameter of the true fault current
Figure FDA0003649499620000031
And substituting the fault current into the formula to calculate converted fault current i'FObtaining the recovered real fault current i after inverse normalization of formula (4)F
iF＝NF×i'F (5)。
CN202110321839.5A 2021-03-25 2021-03-25 Current transformer saturation waveform recovery method based on model and data hybrid driving Active CN113011096B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110321839.5A CN113011096B (en) 2021-03-25 2021-03-25 Current transformer saturation waveform recovery method based on model and data hybrid driving

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110321839.5A CN113011096B (en) 2021-03-25 2021-03-25 Current transformer saturation waveform recovery method based on model and data hybrid driving

Publications (2)

Publication Number Publication Date
CN113011096A CN113011096A (en) 2021-06-22
CN113011096B true CN113011096B (en) 2022-07-05

Family

ID=76407324

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110321839.5A Active CN113011096B (en) 2021-03-25 2021-03-25 Current transformer saturation waveform recovery method based on model and data hybrid driving

Country Status (1)

Country Link
CN (1) CN113011096B (en)

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111123048A (en) * 2019-12-23 2020-05-08 温州大学 Series fault arc detection device and method based on convolutional neural network

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106099850B (en) * 2016-06-06 2018-10-12 南京理工大学 CT saturation based on transient current feature identifies improved method

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111123048A (en) * 2019-12-23 2020-05-08 温州大学 Series fault arc detection device and method based on convolutional neural network

Also Published As

Publication number Publication date
CN113011096A (en) 2021-06-22

Similar Documents

Publication Publication Date Title
Kang et al. Novel leakage detection by ensemble CNN-SVM and graph-based localization in water distribution systems
CN103336189B (en) A kind of solenoid valve fault diagnostic method based on current detecting
CN112052796A (en) Permanent magnet synchronous motor fault diagnosis method based on deep learning
CN101533058B (en) Power abnormal failure data analyzing device and diagnosing method
CN109738776A (en) Fan converter open-circuit fault recognition methods based on LSTM
CN101614786B (en) Online intelligent fault diagnosis method of power electronic circuit based on FRFT and IFSVC
CN112033463B (en) Nuclear power equipment state evaluation and prediction integrated method and system
CN110596506A (en) Converter fault diagnosis method based on time convolution network
CN103995237A (en) Satellite power supply system online fault diagnosis method
CN108139432A (en) The method of saturation effect and its intelligent electronic device in correcting current mutual inductor
CN110907826A (en) Motor fault diagnosis method and system based on convolutional neural network filtering
CN110263944A (en) A kind of multivariable failure prediction method and device
Sá et al. A multilayer perceptron-based approach for stator fault detection in permanent magnet wind generators
CN116012681A (en) Method and system for diagnosing motor faults of pipeline robot based on sound vibration signal fusion
CN116401532A (en) Method and system for recognizing frequency instability of power system after disturbance
Yang et al. Change detection in rotational speed of industrial machinery using Bag-of-Words based feature extraction from vibration signals
CN113011096B (en) Current transformer saturation waveform recovery method based on model and data hybrid driving
Dang et al. Cochlear filter cepstral coefficients of acoustic signals for mechanical faults identification of power transformer
Hafizi et al. Model-based fault diagnosis of a DC-DC boost converters using hidden Markov model
Zhu et al. Aiming to Complex Power Quality Disturbances: A Novel Decomposition and Detection Framework
CN115728588B (en) Electromagnetic compatibility detection system and method based on big data
CN112083275A (en) Distribution network fault type identification method and system
Kahlen et al. Improving Machine Learning Diagnostic Systems with Model-Based Data Augmentation―Part B: Application
CN110135021B (en) ATRU system fault grading diagnosis method based on multi-source signals and RBF neural network
CN113625054A (en) Power grid impedance detection device, inverter and power grid impedance detection method

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