CN111695452A - Parallel reactor internal aging degree evaluation method based on RBF neural network - Google Patents

Parallel reactor internal aging degree evaluation method based on RBF neural network Download PDF

Info

Publication number
CN111695452A
CN111695452A CN202010459935.1A CN202010459935A CN111695452A CN 111695452 A CN111695452 A CN 111695452A CN 202010459935 A CN202010459935 A CN 202010459935A CN 111695452 A CN111695452 A CN 111695452A
Authority
CN
China
Prior art keywords
reactor
neural network
rbf neural
output
age
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.)
Granted
Application number
CN202010459935.1A
Other languages
Chinese (zh)
Other versions
CN111695452B (en
Inventor
孟令明
高树国
张明文
何瑞东
岳国良
乔国华
张克谦
赵芳初
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.)
State Grid Corp of China SGCC
Electric Power Research Institute of State Grid Hebei Electric Power Co Ltd
State Grid Hebei Energy Technology Service Co Ltd
Original Assignee
State Grid Corp of China SGCC
Electric Power Research Institute of State Grid Hebei Electric Power Co Ltd
State Grid Hebei Energy Technology Service 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 State Grid Corp of China SGCC, Electric Power Research Institute of State Grid Hebei Electric Power Co Ltd, State Grid Hebei Energy Technology Service Co Ltd filed Critical State Grid Corp of China SGCC
Priority to CN202010459935.1A priority Critical patent/CN111695452B/en
Publication of CN111695452A publication Critical patent/CN111695452A/en
Application granted granted Critical
Publication of CN111695452B publication Critical patent/CN111695452B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/003—Environmental or reliability tests
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24133—Distances to prototypes
    • G06F18/24137—Distances to cluster centroïds
    • G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Environmental & Geological Engineering (AREA)
  • Monitoring And Testing Of Nuclear Reactors (AREA)

Abstract

The invention relates to a parallel reactor internal aging degree evaluation method based on a RBF neural network, which comprises the steps of collecting noise signals of a high-voltage parallel reactor, simultaneously recording the operation time of the reactor, and extracting fundamental frequency amplitude values in signals of all channels; inputting the fundamental frequency amplitude of the noise signal of each reactor into the RBF neural network to predict the expected operation age of the reactor, comparing the expected operation age with the actual input age of the reactor, judging whether the inside of the reactor has obvious defects or not through errors of the fundamental frequency amplitude and the actual input age, and when the prediction result is far deviated from the actual age, showing that the aging speed inside the reactor is higher than the natural trend, and constructing continuous influence on the internal structure due to too fast aging. The invention belongs to the depth state evaluation of a high-voltage parallel reactor, and the state evaluation is carried out from the internal aging angle of the reactor. And training the RBF neural network through the noise signal data and the commissioning life, thereby obtaining the commissioning life output based on the network model.

Description

Parallel reactor internal aging degree evaluation method based on RBF neural network
Technical Field
The invention belongs to a method for predicting and diagnosing the operation life of a high-voltage shunt reactor by using a vibration noise data set of the high-voltage shunt reactor, and particularly relates to a method for evaluating the internal aging degree of the shunt reactor based on a Radial Basis Function (RBF) neural network.
Background
The high-voltage parallel oil-immersed reactor is one of important devices in a power system, and has important significance for stable and safe operation of the power system. At present, a great deal of research is carried out on the noise characteristics of the shunt reactor at home and abroad, and the internal mechanical state of the reactor can be evaluated by utilizing abundant characteristic quantities in an acoustic signal transmitted by the shunt reactor while the surface of an oil tank of the reactor is subjected to vibration reduction and noise reduction. In the process of long-term operation of the reactor, the inside of the reactor gradually generates mechanical aging, thermal aging and other reactions, so that vibration is intensified and acoustic signals are abnormal.
The noise of the high-voltage shunt reactor is the same as the electrical performance and the mechanical performance of the high-voltage shunt reactor, and is an extremely important technical parameter, and the noise level of the oil tank body is one of important indexes for measuring the operation level and the defect degree of the oil tank body. Whether the internal structure of the reactor is good or not, the mixed noise is transmitted to the shell through the internal structure elements of the reactor and is radiated to the periphery, so that the noise on the surface of the oil tank can be used as an important factor for judging the running condition of the reactor.
And processing the acoustic signal on the surface of the reactor oil tank, predicting the operation age of the reactor oil tank through an RBF neural network, and comparing the operation age with the actual investment age. When the prediction result is far away from the actual age, the aging speed inside the reactor is larger than the natural trend, and the internal structure is continuously influenced when the reactor is aged too fast.
Disclosure of Invention
The method is suitable for carrying out operation age prediction fault diagnosis on the high-voltage shunt reactor on the premise of having a complete high-voltage shunt reactor vibration noise data set.
The technical scheme of the invention is as follows:
a parallel reactor internal aging degree evaluation method based on RBF neural network collects high-voltage parallel reactor noise signals, simultaneously records the reactor operation time, and extracts fundamental frequency amplitude values in each channel signal;
inputting the fundamental frequency amplitude of the noise signal of each reactor into the RBF neural network to predict the expected operation age of the reactor, comparing the expected operation age with the actual input age of the reactor, judging whether the inside of the reactor has obvious defects or not through errors of the fundamental frequency amplitude and the actual input age, and when the prediction result is far deviated from the actual age, showing that the aging speed inside the reactor is higher than the natural trend, and constructing continuous influence on the internal structure due to too fast aging.
Further, in the actual operation process of the high-voltage shunt reactor, the iron core and the winding vibrate under the stress action, vibration data of relevant measuring points are obtained by using the vibration sensor and the microphone, and data in the future time are predicted by using historical data, so that state evaluation is performed.
Further, the vibration noise signal acquisition is carried out on the high-voltage shunt reactor by utilizing an online monitoring system, and the process of utilizing the RBF neural network comprises the following steps: 1) collecting noise signals of 4 measuring points on the surface of an oil tank; 2) carrying out fast Fourier transform on the data, and extracting a fundamental frequency amplitude; 3) and forming a large number of fundamental frequency amplitude values into vectors corresponding to the age of the reactor.
Further, the RBF neural network algorithm for the noise signal and the operation life comprises the following steps:
step 1: determining an input vector X: x ═ X1,x2,...,xn]TN is the number of input layer units
Step 2: determining an output vector Y: y ═ Y1,y2,...,yq]TQ is the number of output layer units
And step 3: initializing connection weight W from hidden layer to output layerk＝[wk1,wk2,...,wkp]T(k＝1,2,...,q)
Wherein p is the number of hidden layer units and q is the number of output layer units;
the method for initializing the reference center provides a method for initializing the weight from a hidden layer to an output layer, which comprises the following steps:
Figure BDA0002512804160000021
where mink is the minimum of all expected outputs in the kth output neuron in the training set; maxk is the maximum of all expected outputs in the kth output neuron in the training set;
and 4, step 4: initializing the central parameter C of each neuron of the hidden layerj＝[cj1,cj2,...,cjn]T
The initial values of the central parameters of the RBF neural network are as follows:
Figure BDA0002512804160000022
and 5: initializing width vector Dj＝[dj1,dj2,...,djn]。
Figure BDA0002512804160000023
Step (ii) of6: computing output value of jth neuron of hidden layer
Figure BDA0002512804160000024
Figure BDA0002512804160000025
And 7: computing output Y ═ Y for neurons in the output layer1,y2,...,yq]TWherein
Figure BDA0002512804160000026
Figure BDA0002512804160000027
And 8: iteratively calculating a weight parameter
Figure BDA0002512804160000031
Figure BDA0002512804160000032
Figure BDA0002512804160000033
wkj(t) is the tuning weight between the kth output neuron and the jth hidden layer neuron at the time of the t iteration. c. Cji(t) is the central component of the jth hidden layer neuron for the ith input neuron at the time of the tth iterative computation; dji(t) is the sum of center cji(t) a corresponding width;
and step 9: calculating the value of the root mean square error RMS output by the network according to the following formula, if the RMS is less than or equal to the training end, otherwise, turning to the step 8;
Figure BDA0002512804160000034
step 10: and comparing the predicted value with the true value to obtain a final diagnosis conclusion.
Furthermore, a noise signal of the high-voltage shunt reactor is collected through an online monitoring system.
Further, when extracting the fundamental frequency amplitude in each channel signal, the FFT transformation is performed on the vibration noise signal, and the time domain signal is converted into the frequency domain signal.
Further, the operation time of each high-voltage shunt reactor is recorded and is in one-to-one correspondence with the fundamental frequency vector of each noise signal.
Further, a training set and a test set are divided from the collected data, relevant parameters of the neural network are determined, 8 neurons are input into the RBF neural network, 30 neurons are hidden layer neurons, 1 neuron is output layer neurons, and the network is trained.
Further, an RBF neural network model structure is established, and the RBF neural network is composed of an input layer, a hidden layer and an output layer.
Furthermore, the input layer is formed by an array formed by all channel fundamental frequency amplitude values of a single reactor, the hidden layer converts low-dimensional space signals in the input layer into high-dimensional space, and the output layer is the actual operation year of the reactor and is determined by linear weighting output by the hidden layer unit.
The invention has the beneficial effects that:
the invention belongs to the depth state evaluation of a high-voltage parallel reactor, and the state evaluation is carried out from the internal aging angle of the reactor. And training the RBF neural network through the noise signal data and the commissioning life, thereby obtaining the commissioning life output based on the network model. The invention has accurate and reliable conclusion and convenient use.
Drawings
FIG. 1 is a diagram of an RBF neural network weight value updating process.
Fig. 2 acoustic signal processing diagram.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the application, its application, or uses. 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 application.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
The relative arrangement of the components and steps, the numerical expressions, and numerical values set forth in these embodiments do not limit the scope of the present application unless specifically stated otherwise. Meanwhile, it should be understood that the sizes of the respective portions shown in the drawings are not drawn in an actual proportional relationship for the convenience of description. Techniques, methods, and apparatus known to those of ordinary skill in the relevant art may not be discussed in detail but are intended to be part of the specification where appropriate. In all examples shown and discussed herein, any particular value should be construed as merely illustrative, and not limiting. Thus, other examples of the exemplary embodiments may have different values. It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, further discussion thereof is not required in subsequent figures.
In the description of the present application, it is to be understood that the orientation or positional relationship indicated by the directional terms such as "front, rear, upper, lower, left, right", "lateral, vertical, horizontal" and "top, bottom", etc., are generally based on the orientation or positional relationship shown in the drawings, and are used for convenience of description and simplicity of description only, and in the case of not making a reverse description, these directional terms do not indicate and imply that the device or element being referred to must have a particular orientation or be constructed and operated in a particular orientation, and therefore, should not be considered as limiting the scope of the present application; the terms "inner and outer" refer to the inner and outer relative to the profile of the respective component itself.
Spatially relative terms, such as "above … …," "above … …," "above … …," "above," and the like, may be used herein for ease of description to describe one device or feature's spatial relationship to another device or feature as illustrated in the figures. It will be understood that the spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. For example, if a device in the figures is turned over, devices described as "above" or "on" other devices or configurations would then be oriented "below" or "under" the other devices or configurations. Thus, the exemplary term "above … …" can include both an orientation of "above … …" and "below … …". The device may be otherwise variously oriented (rotated 90 degrees or at other orientations) and the spatially relative descriptors used herein interpreted accordingly.
It should be noted that the terms "first", "second", and the like are used to define the components, and are only used for convenience of distinguishing the corresponding components, and the terms have no special meanings unless otherwise stated, and therefore, the scope of protection of the present application is not to be construed as being limited.
The technical solution and structure of the present invention will be described in further detail with reference to the accompanying drawings.
The invention discloses a state self-evaluation method of a high-voltage parallel reactor based on an RBF neural network, which is characterized by comprising the following steps:
1. collecting sound pressure signals of multiple parallel reactors
And respectively extracting the fundamental frequency of each channel signal as an input vector of the RBF neural network. And recording the operation period of each high-voltage shunt reactor, and respectively corresponding to the fundamental frequency vector of each noise signal one by one.
Dividing a training set and a test set from the collected data, determining relevant parameters of the neural network, inputting 8 neurons in an input layer, 30 neurons in a hidden layer and 1 neuron in an output layer, inputting the training set into the RBF neural network, and training the network. And inputting the test set into a network for verification, obtaining the predicted age limit of the reactor to be verified, comparing the predicted result with the actual operation age limit, and diagnosing the fault inside the reactor through the error between the two age limits.
Calculating relative errors according to the predicted operation years, considering the influence of environmental factors, enabling the prediction results to have slight errors, and when the error result of the prediction result of a certain reactor relative to the actual year is larger than a certain threshold value, indicating that certain potential faults and defects exist in the reactor.
2. Algorithm flow
The model structure of the RBF neural network is established as shown in figure 1, and the network mainly comprises an input layer, a hidden layer and an output layer. The input layer is formed by an array formed by all channel fundamental frequency amplitude values of a single reactor, the hidden layer converts low-dimensional space signals in the input layer into high-dimensional space, and the output layer is the actual operation year of the reactor and is determined by linear weighting output by the hidden layer unit;
Figure BDA0002512804160000061
the sample is the pth input sample, P is 1,2, …, N is the total number of samples; omega is the connection weight between the output layer and the hidden layer;
step 1: network initialization is carried out, and l training samples are randomly selected as initial clustering centers ci(i ═ 1,2, …, l), calculate xpAnd initial clustering center ciAnd performing clustering grouping according to a nearest neighbor rule by using the Euclidean distance between the two groups.
Step 2: readjusting the clustering centers, calculating the average value of the samples in each clustering set to obtain a new clustering center, stopping the calculation if the new clustering center does not change any more, otherwise returning to the previous step to continuously determine the center of the basis function, wherein the basis function of the RBF neural network is a Gaussian function, and the variance is solved into
Figure BDA0002512804160000062
Wherein i is 1,2, …, l; c. CmaxIs the maximum distance between the centers of the selected basis functions.
And step 3: the connection weight from the hidden layer to the output layer is directly calculated by using a least square method, and the calculation formula is as follows
Figure BDA0002512804160000063
Wherein i is 1,2,. and l; p1, 2, P is the total number of samples.
And 4, step 4: RBF neural network adjusts parameter center c by input and output errorsiAnd adjusting the internal coefficient of the network according to the weight omega, and stopping the calculation and outputting a predicted value by the network through repeated iterative calculation until the output mean square error of the network reaches the preset precision requirement.
An RBF neural network-based shunt reactor internal aging degree evaluation method. In the process of long-term operation, the effect of various internal stresses and the external environment generate comprehensive influence on the device, the structural performance of the device is aged, and then defects and even faults are generated. The operation condition, the environmental influence, the monitoring result and other information make the state change and the fault evolution process of the equipment operation increasingly perfect. The noise is used as one of the information, a reliable and accurate reference basis is provided for the age estimation of the reactor, and the running state and the development trend of the reactor are accurately mastered through the training of the neural network, so that the state management level of the reactor is further improved.
The specific implementation comprises the following steps:
in the actual operation process of the high-voltage shunt reactor, the iron core and the winding vibrate under the stress action, vibration data of relevant measuring points are obtained by using a vibration sensor and a microphone, and data in a short time in the future are predicted by using historical data, so that state evaluation is performed.
The method comprises the following steps of utilizing an online monitoring system to collect vibration noise signals of the high-voltage shunt reactor, and utilizing an RBF neural network to mainly collect vibration noise signals of the high-voltage shunt reactor, wherein the process mainly comprises the following steps: 1) and collecting noise signals of 4 measuring points on the surface of the oil tank. 2) And carrying out fast Fourier transform on the data, and extracting the fundamental frequency amplitude. 3) And forming a large number of fundamental frequency amplitude values into vectors corresponding to the age of the reactor.
The RBF neural network algorithm aiming at the noise signals and the operation period comprises the following steps:
step 1: determining an input vector X: x ═ X1,x2,...,xn]TN is the number of input layer units
Step 2: determining an output vector Y: y ═ Y1,y2,...,yq]TQ is the number of output layer units
And step 3: initializing connection weight W from hidden layer to output layerk＝[wk1,wk2,...,wkp]T(k＝1,2,...,q)
Where p is the number of hidden layer elements and q is the number of output layer elements.
The method for initializing the reference center provides a method for initializing the weight from a hidden layer to an output layer, which comprises the following steps:
Figure BDA0002512804160000071
where mink is the minimum of all expected outputs in the kth output neuron in the training set; maxk is the maximum of all expected outputs in the kth output neuron in the training set.
And 4, step 4: initializing the central parameter C of each neuron of the hidden layerj＝[cj1,cj2,...,cjn]TThe initial values of the central parameters of the RBF neural network are as follows:
Figure BDA0002512804160000072
and 5: initializing width vectors
Figure BDA0002512804160000073
Step 6: computing output value of jth neuron of hidden layer
Figure BDA0002512804160000074
Figure BDA0002512804160000075
And 7: computing output Y ═ Y for neurons in the output layer1,y2,...,yq]TWherein
Figure BDA0002512804160000081
Figure BDA0002512804160000082
And 8: iteratively calculating a weight parameter
Figure BDA0002512804160000083
Figure BDA0002512804160000084
Figure BDA0002512804160000085
wkj(t) is the tuning weight between the kth output neuron and the jth hidden layer neuron at the time of the t iteration. c. Cji(t) is the central component of the jth hidden layer neuron for the ith input neuron at the time of the tth iterative computation; dji(t) is the sum of center cji(t) corresponding width.
And step 9: and calculating the value of the root mean square error RMS of the network output according to the following formula, finishing the training if the RMS is less than or equal to the RMS, otherwise, turning to the step 8.
Figure BDA0002512804160000086
Step 10: and comparing the predicted value with the true value to obtain a final diagnosis conclusion.
As shown in fig. 2, the operation condition of the high-voltage shunt reactor of a certain substation
The measurement and comparison of the reactor noise are to distinguish different internal defect degrees of a plurality of reactors, the reactor operation time can visually indicate the internal aging degree, a plurality of groups of noise signals and reactor operation age data are trained by using a neural network algorithm, the critical state is closely watched by correctly estimating the service life of the reactor to be researched, the necessity of maintenance is determined, the conversion of the fault state can be effectively prevented, the loss caused by the fault development of the reactor is reduced or avoided, and the operation reliability of the reactor is enhanced.
The noise signal of the high-voltage parallel reactor is collected by adopting an online monitoring device, the operation time of the reactor is recorded, the internal aging degree of the reactor can be reflected to a certain degree by the characteristics of the noise signal, and the fundamental frequency amplitude in each channel signal is extracted.
And inputting the fundamental frequency amplitude of the noise signal of each reactor into the RBF neural network to predict the expected operation age of the reactor, and comparing the expected operation age with the actual investment age of the reactor. Whether the inside of the reactor has obvious defects or not is judged through errors of the two, and when the prediction result is far away from the actual age, the aging speed inside the reactor is larger than the natural trend, and the internal structure is continuously influenced due to the fact that the aging speed is too fast.

Claims (10)

1. A parallel reactor internal aging degree evaluation method based on RBF neural network is characterized in that noise signals of a high-voltage parallel reactor are collected, the reactor operation time is recorded at the same time, and fundamental frequency amplitude values in signals of each channel are extracted;
inputting the fundamental frequency amplitude of the noise signal of each reactor into the RBF neural network to predict the expected operation age of the reactor, comparing the expected operation age with the actual input age of the reactor, judging whether the inside of the reactor has obvious defects or not through errors of the fundamental frequency amplitude and the actual input age, and when the prediction result deviates from the actual age, showing that the aging speed inside the reactor is higher than the natural trend, and constructing a continuous influence on the internal structure due to the fact that the reactor is aged too fast.
2. The method for evaluating the internal aging degree of the parallel reactor based on the RBF neural network as claimed in claim 1, wherein in the actual operation process of the high-voltage parallel reactor, the iron core and the winding vibrate due to stress, vibration data of relevant measuring points are obtained by using the vibration sensor and the microphone, and data in the future time are predicted by using historical data, so that the state evaluation is performed.
3. The method for evaluating the internal aging degree of the parallel reactor based on the RBF neural network as claimed in claim 2, wherein the vibration noise signal of the high-voltage parallel reactor is collected by using an online monitoring system, and the process of using the RBF neural network comprises the following steps: 1) collecting noise signals of 4 measuring points on the surface of an oil tank; 2) carrying out fast Fourier transform on the data, and extracting a fundamental frequency amplitude; 3) and forming a vector by the fundamental frequency amplitude values to correspond to the age limit of the reactor.
4. The method for evaluating the internal aging degree of the shunt reactor based on the RBF neural network as claimed in claim 3, wherein the RBF neural network algorithm aiming at the noise signal and the operation life comprises the following steps:
step 1: determining an input vector X: x ═ X1,x2,...,xn]TN is the number of input layer units;
step 2: determining an output vector Y: y ═ Y1,y2,...,yq]TQ is the number of output layer units;
and step 3: initializing connection weight W from hidden layer to output layerk＝[wk1,wk2,...,wkp]T(k＝1,2,...,q)；
Wherein p is the number of hidden layer units and q is the number of output layer units;
the method for initializing the reference center provides a method for initializing the weight from a hidden layer to an output layer, which comprises the following steps:
Figure FDA0002512804150000011
where mink is the minimum of all expected outputs in the kth output neuron in the training set; maxk is the maximum of all expected outputs in the kth output neuron in the training set;
and 4, step 4: initializing the central parameter C of each neuron of the hidden layerj＝[cj1,cj2,...,cjn]TThe initial values of the central parameters of the RBF neural network are as follows:
Figure FDA0002512804150000021
and 5: initializing width vector Dj＝[dj1,dj2,...,djn]。
Figure FDA0002512804150000022
Step 6: computing output value of jth neuron of hidden layer
Figure FDA0002512804150000023
j＝1,2,...,p；
And 7: computing output Y ═ Y for neurons in the output layer1,y2,...,yq]TWherein
Figure FDA0002512804150000024
k＝1,2,...,q；
And 8: iteratively calculating a weight parameter
Figure FDA0002512804150000025
Figure FDA0002512804150000026
Figure FDA0002512804150000027
wkj(t) is the tuning weight between the kth output neuron and the jth hidden layer neuron at the time of the t iteration. c. Cji(t) is the central component of the jth hidden layer neuron for the ith input neuron at the time of the tth iterative computation; dji(t) is the sum of center cji(t) a corresponding width;
and step 9: calculating the value of the root mean square error RMS of the network output according to the following formula, if the RMS is less than or equal to the training is ended, otherwise, turning to the step 8
Figure FDA0002512804150000028
Step 10: and comparing the predicted value with the true value to obtain a final diagnosis conclusion.
5. The method for evaluating the internal aging degree of the parallel reactor based on the RBF neural network as claimed in claim 1, wherein the noise signal of the high-voltage parallel reactor is collected by an online monitoring system.
6. The method for evaluating the internal aging degree of the shunt reactor based on the RBF neural network as claimed in claim 1, wherein when the fundamental frequency amplitude in each channel signal is extracted, the vibration noise signal is subjected to FFT conversion, and the time domain signal is converted into the frequency domain signal.
7. The method for evaluating the internal aging degree of the paralleling reactors based on the RBF neural network as claimed in claim 1, wherein the operation time of each high-voltage paralleling reactor is recorded and is in one-to-one correspondence with each noise signal fundamental frequency vector.
8. The method for evaluating the internal aging degree of the shunt reactor based on the RBF neural network as claimed in claim 1, wherein a training set and a test set are divided from collected data, neural network-related parameters are determined, the number of input layer neurons is 8, the number of hidden layer neurons is 30, the number of output layer neurons is 1, the training set is input into the RBF neural network, and the network is trained.
9. The method for evaluating the internal aging degree of the shunt reactor based on the RBF neural network as claimed in claim 1, wherein a model structure of the RBF neural network is established, and the RBF neural network is composed of an input layer, a hidden layer and an output layer.
10. The method for evaluating the internal aging degree of the parallel reactor based on the RBF neural network as claimed in claim 9, wherein the input layer is formed by an array consisting of fundamental frequency amplitudes of all channels of a single reactor, the hidden layer is used for converting a low-dimensional space signal in the input layer into a high-dimensional space, the output layer is the actual operation year of the reactor, and the output layer is determined by linear weighting of the output of the hidden layer unit.
CN202010459935.1A 2020-05-28 2020-05-28 RBF neural network-based parallel reactor internal aging degree assessment method Active CN111695452B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010459935.1A CN111695452B (en) 2020-05-28 2020-05-28 RBF neural network-based parallel reactor internal aging degree assessment method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010459935.1A CN111695452B (en) 2020-05-28 2020-05-28 RBF neural network-based parallel reactor internal aging degree assessment method

Publications (2)

Publication Number Publication Date
CN111695452A true CN111695452A (en) 2020-09-22
CN111695452B CN111695452B (en) 2023-08-18

Family

ID=72478519

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010459935.1A Active CN111695452B (en) 2020-05-28 2020-05-28 RBF neural network-based parallel reactor internal aging degree assessment method

Country Status (1)

Country Link
CN (1) CN111695452B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113435843A (en) * 2021-06-28 2021-09-24 平安信托有限责任公司 Batch file generation method and device, electronic equipment and storage medium
CN116151169A (en) * 2022-11-30 2023-05-23 西安西电变压器有限责任公司 Method and device for predicting noise intensity of reactor
CN116861164A (en) * 2023-05-08 2023-10-10 华电电力科学研究院有限公司 Turbine operation fault monitoring system
CN117761442A (en) * 2024-02-19 2024-03-26 深圳市鼎泰佳创科技有限公司 Electronic load aging a degree test method System and method for controlling a system test equipment

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1274849A (en) * 1999-05-24 2000-11-29 三菱电机株式会社 Abnormal detector of electric rotating machinery
CN101576443A (en) * 2009-06-16 2009-11-11 北京航空航天大学 Life prediction method of accelerated life test based on grey RBF neural network
CN104504260A (en) * 2014-12-16 2015-04-08 西北工业大学 Fault prediction method of multi-failure mode system
CN107101828A (en) * 2017-05-27 2017-08-29 辽宁大学 The bearing residual life Forecasting Methodology that a kind of encoder health degree is evaluated
CN108181580A (en) * 2018-01-25 2018-06-19 江苏德春电力科技股份有限公司 Solid-state insulated switchgear intelligent diagnostics device and lifetime estimation method
EP3343428A1 (en) * 2016-12-31 2018-07-04 INTEL Corporation Architecture for telemetry and adaptive lifetime control of integrated circuits
CN108445382A (en) * 2018-04-23 2018-08-24 温州大学 A kind of rapid failure detection method of high-voltage circuitbreaker
CN108776818A (en) * 2018-06-05 2018-11-09 北京航空航天大学 Bearing features extracting method, bearing residual life prediction technique and device
CN110377984A (en) * 2019-07-02 2019-10-25 中国科学院深圳先进技术研究院 A kind of industrial equipment remaining useful life prediction technique, system and electronic equipment

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1274849A (en) * 1999-05-24 2000-11-29 三菱电机株式会社 Abnormal detector of electric rotating machinery
CN101576443A (en) * 2009-06-16 2009-11-11 北京航空航天大学 Life prediction method of accelerated life test based on grey RBF neural network
CN104504260A (en) * 2014-12-16 2015-04-08 西北工业大学 Fault prediction method of multi-failure mode system
EP3343428A1 (en) * 2016-12-31 2018-07-04 INTEL Corporation Architecture for telemetry and adaptive lifetime control of integrated circuits
US20180189522A1 (en) * 2016-12-31 2018-07-05 Intel Corporation Architecture for telemetry and adaptive lifetime control of integrated circuits
CN107101828A (en) * 2017-05-27 2017-08-29 辽宁大学 The bearing residual life Forecasting Methodology that a kind of encoder health degree is evaluated
CN108181580A (en) * 2018-01-25 2018-06-19 江苏德春电力科技股份有限公司 Solid-state insulated switchgear intelligent diagnostics device and lifetime estimation method
CN108445382A (en) * 2018-04-23 2018-08-24 温州大学 A kind of rapid failure detection method of high-voltage circuitbreaker
CN108776818A (en) * 2018-06-05 2018-11-09 北京航空航天大学 Bearing features extracting method, bearing residual life prediction technique and device
CN110377984A (en) * 2019-07-02 2019-10-25 中国科学院深圳先进技术研究院 A kind of industrial equipment remaining useful life prediction technique, system and electronic equipment

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
张伟: "基于模糊理论和RBF网络的低压开关设备故障预测诊断研究", pages 40 - 60 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113435843A (en) * 2021-06-28 2021-09-24 平安信托有限责任公司 Batch file generation method and device, electronic equipment and storage medium
CN116151169A (en) * 2022-11-30 2023-05-23 西安西电变压器有限责任公司 Method and device for predicting noise intensity of reactor
CN116151169B (en) * 2022-11-30 2024-03-12 西安西电变压器有限责任公司 Method and device for predicting noise intensity of reactor
CN116861164A (en) * 2023-05-08 2023-10-10 华电电力科学研究院有限公司 Turbine operation fault monitoring system
CN117761442A (en) * 2024-02-19 2024-03-26 深圳市鼎泰佳创科技有限公司 Electronic load aging a degree test method System and method for controlling a system test equipment
CN117761442B (en) * 2024-02-19 2024-05-03 深圳市鼎泰佳创科技有限公司 Electronic load aging degree testing method, system and testing equipment

Also Published As

Publication number Publication date
CN111695452B (en) 2023-08-18

Similar Documents

Publication Publication Date Title
CN111695452B (en) RBF neural network-based parallel reactor internal aging degree assessment method
WO2021232655A1 (en) Method for evaluating mechanical state of high-voltage shunt reactor on the basis of vibration feature
CN113255848B (en) Water turbine cavitation sound signal identification method based on big data learning
CN109033612B (en) Transformer fault diagnosis method based on vibration noise and BP neural network
CN111914883B (en) Spindle bearing state evaluation method and device based on deep fusion network
Gao et al. A newly designed diagnostic method for mechanical faults of high-voltage circuit breakers via SSAE and IELM
CN108073158A (en) Based on PCA and KNN density algorithm Wind turbines Method for Bearing Fault Diagnosis
CN110132554B (en) Rotary machine fault diagnosis method based on deep Laplace self-coding
CN107725283A (en) A kind of fan trouble detection method based on depth belief network model
CN110969096A (en) Motor fault mode diagnosis method based on particle swarm optimization support vector machine
CN112418277A (en) Method, system, medium, and apparatus for predicting remaining life of rotating machine component
CN116380445B (en) Equipment state diagnosis method and related device based on vibration waveform
CN115618732B (en) Nuclear reactor digital twin key parameter autonomous optimization data inversion method
CN112507479B (en) Oil drilling machine health state assessment method based on manifold learning and softmax
CN114429152A (en) Rolling bearing fault diagnosis method based on dynamic index antagonism self-adaption
CN115877205A (en) Intelligent fault diagnosis system and method for servo motor
CN114091525A (en) Rolling bearing degradation trend prediction method
Zhang et al. Complementary ensemble adaptive local iterative filtering and its application to rolling bearing fault diagnosis
CN113237619B (en) Fault early warning method, device, equipment and storage medium for variable-speed rotating machinery vibration
CN112990627A (en) Electric energy quality evaluation method
CN115128345A (en) Power grid safety early warning method and system based on harmonic monitoring
Li et al. A novel application of intelligent algorithms in fault detection of rudder system
CN113554229A (en) Three-phase voltage unbalance abnormality detection method and device
CN116681186A (en) Power quality analysis method and device based on intelligent terminal
CN116884432A (en) VMD-JS divergence-based power transformer fault voiceprint diagnosis 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