CN106597178B - A kind of pre- abnormal method of ANFIS number of test device for relay protection LPA - Google Patents

A kind of pre- abnormal method of ANFIS number of test device for relay protection LPA Download PDF

Info

Publication number
CN106597178B
CN106597178B CN201710035791.5A CN201710035791A CN106597178B CN 106597178 B CN106597178 B CN 106597178B CN 201710035791 A CN201710035791 A CN 201710035791A CN 106597178 B CN106597178 B CN 106597178B
Authority
CN
China
Prior art keywords
lpa
anfis
abnormal
relay protection
current
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
CN201710035791.5A
Other languages
Chinese (zh)
Other versions
CN106597178A (en
Inventor
孙晓明
彭炜峰
张明钱
张莉
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.)
Chongqing Water Resources and Electric Engineering College
Original Assignee
Chongqing Water Resources and Electric Engineering College
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 Chongqing Water Resources and Electric Engineering College filed Critical Chongqing Water Resources and Electric Engineering College
Priority to CN201710035791.5A priority Critical patent/CN106597178B/en
Publication of CN106597178A publication Critical patent/CN106597178A/en
Application granted granted Critical
Publication of CN106597178B publication Critical patent/CN106597178B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Geometry (AREA)
  • Computer Hardware Design (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Testing Electric Properties And Detecting Electric Faults (AREA)
  • Emergency Protection Circuit Devices (AREA)

Abstract

The invention discloses a kind of pre- abnormal methods of the ANFIS of test device for relay protection LPA number; after passing through training and study using ANFIS; remove to approach the nonlinear inverse mathematical model of the test device for relay protection LPA in compact set; replace inverse number word model to generate command voltage/electric current corresponding with desired output voltage/electric current, realizes that number is pre- abnormal.Overall thinking of the present invention LPA from digital analog converter to each amplifying stage again to all possible non-linear of load, it is no longer overcritical to the precision of each amplifying stage；Using quasi- online mode, it can be realized on most of original microprocessor, improve cost performance；By means of the adaptive ability of ANFIS, the load of different characteristics and numerical value can be adapted to automatically.The present invention is not only able to satisfy the required precision of fault waveform reproduction, and the limited frequency bandwidth of LPA and output capacity can also be made to be fully used.

Description

A kind of pre- abnormal method of ANFIS number of test device for relay protection LPA
Technical field
The invention belongs to technical field of relay protection more particularly to a kind of test device for relay protection linear power amplifiers (LPA) the pre- abnormal method of ANFIS number.
Background technique
Test device for relay protection has become adjusting verification relay protection and the indispensable tool of automatic safety device.For The movement details of relay protection and automatic safety device under electric system true fault, modern times test can be meticulously analyzed comprehensively It is required that test device for relay protection should be able to relatively accurately export various false voltage/current waveforms, Wave data can derive from soft Part emulation or fault oscillograph.On the one hand, because relay protection and automatic safety device are connected to the secondary of potential and current transformers Side, therefore voltage and current must be amplified to the secondary voltage level and current transformer of voltage transformer by test device for relay protection Secondary current it is horizontal, the voltage and current at " scene " could be provided for tested device.Voltage transformer secondary voltage rating one As be 100V (virtual value, root mean square value, RMS)；To grassroot project, to reduce loss, current transformer Secondary rated current generally selects 1A (RMS).Under fault transient, the primitive period component of false voltage is up to the 5 of voltage rating Times, the primitive period component of fault current considers aperiodic component and harmonic wave, total voltage and total electricity up to 30 times of rated current Stream can also be higher.Therefore voltage and current should be able to be at least amplified to 500V (RMS) and 30A (RMS) by test device for relay protection, i.e., Voltage dynamic range is at least 0~5 (per unit value, per unit, pu), and current dynamic range is at least 0~30 (pu).Another party Face, for the test for meeting fault location device, harmonic source monitoring positioning device, active filter and Static Var Compensator etc., It is required that the frequency bandwidth of output waveform is at least 0~10kHz.According to sampling thheorem, it is desirable that the frequency band of test device for relay protection Width is at least 0~20kHz.To sum up, test device for relay protection of new generation requires to realize while realizing wide dynamic range Broadband, this is always the technological difficulties of test device for relay protection Designing power amplifier, is caused when fault waveform reappears Wave distortion.Because current design criteria was formulated before 10 years, fault waveform producing contents to be added, therefore think wouldn't be by for existing product Operation constraint, it is difficult for fear of realizing, it is often empty to claim faulty wave recurrence function.To overcome the technological difficulties, most relay protections Test device is tended to using LPA, it is believed that compared with switching power amplifier simplicity, opened loop control precision even can for the control of LPA It matches in excellence or beauty with the closed-loop control precision of switching power amplifier.This is that LPA " linear " misreads caused by designer.Actually " linear " of LPA is a relative concept, and no LPA can completely eliminate non-linear.Reason is to obtain sufficiently large put Big multiple, LPA need to be cascaded by several amplifying stages and be formed.Each amplifying stage can be considered a LPA, generally by a differential amplifier It realizes.One the non-linear of differential amplifier is easy to control, however, after several cascades, it is each it is non-linear can be by step by step Amplify simultaneously it is superposed on one another, eventually become one it is uppity non-linear.Theoretically, non-linear after cascade to reduce, each The hardware of amplifying stage must sophisticated design and adjustment, and between power supply and grade connect circuit must also optimize, following methods can be used: 1. It uses dynamic switch power supply instead to be connected with LPA, be reduced by specific modulation non-linear caused by Traditional DC power-supply fluctuation；2. being every A amplifying stage design simulation feed forward circuit, corrects the non-linear of each amplifying stage in advance；3. by optimizing and revising peak compensation line, simultaneously Join capacitor and switch bias realizes the linearisation of LPA, and improves the linearity to distortion compensation after use；4. using cross-coupling Double deference to and source-degeneration resistance two linearisation measures realize LPA High Linears respond.High line acquired by these methods Property degree is and thus may to bring the reduction of reliability, and the portable of method to increase the complexity of hardware as cost Property is poor.Different from hardware above amendment or compensation method, the characteristics of digital pre- abnormal method is to be no longer limited to the raising each portion LPA The precision of part generally only needs modification software that the promotion of precision can be realized without changing hardware circuit.It is current to realize number Pre- abnormal method mainly has: 1. being modeled using memoryless multinomial to LPA, polynomial coefficient is true by recurrent least square method It is fixed, it is pre- abnormal by model realization number；2. being modeled using Wiener filter to LPA, and using several piecewise linear functions to non- Linearly estimated, it is pre- abnormal that the two is implemented in combination with number；3. using that can be eliminated with the linearization technique in conjunction with iteration noise cancellation technology The influence of additive noise makes the pre- abnormal device of number not depend on the accurate model of LPA.These pre- abnormal methods of number have preferable portable Property, but establish on the basis of the linear maths tool such as multinomial, linear filter and piecewise linear function, it is non-thread for estimating Property, it is possible in theory, and it is practical in can be limited by order, precision is difficult to improve.In addition, in estimation or adjusting parameter and coefficient When, numerical stability is difficult to ensure sometimes, may cause system unstability.In view of being limited by rated capacity, the output of LPA is limited It from the point of view of functional analysis, can be defined as a unknown nonlinear in compact set in a limited range by system Functional.Theoretical proof, ANFIS can infinitely approach any non-linear letter in compact set after the training of finite number of time and study Number.Therefore the more aforementioned linear math tool of ANFIS is more suitable to Nonlinear Modeling.In addition, it is provable, when using gradient descent method When the hybrid learning algorithm combined with least square method replaces single gradient descent method, it there would not be numerical stability and ask Topic.
Summary of the invention
It is an object of the invention to propose the pre- abnormal method of ANFIS number of relay protection test test device LPA a kind of, with Test device for relay protection of new generation is solved to require to realize wide band technical problem while realizing wide dynamic range.
The pre- abnormal method of ANFIS number of relay protection test test device LPA of the present invention is achieved in that utilization After ANFIS is by training and study, remove to approach the nonlinear inverse mathematical model of the test device for relay protection LPA in compact set, Replace inverse number word model to generate command voltage/electric current corresponding with desired output voltage/electric current, realizes that number is pre- abnormal.
Further, the realization of the pre- abnormal method of ANFIS number of the relay protection test test device LPA includes following step It is rapid:
1) by the input/output data of LPA inverse number word model to the input/output data pair as ANFIS, that is, number is trained According to；
2) type and number of subordinating degree function are chosen；
3) initial configuration of ANFIS is generated；
4) training parameter, i.e., maximum training batch, initial step length and target error are selected；
5) training is completed by hybrid learning algorithm, generates the pre- abnormal device of ANFIS number；
6) pre- abnormal command voltage/electric current is generated through the pre- abnormal device of ANFIS number with desired output voltage/electric current；
7) realize that the amplification of voltage/current exports with pre- abnormal command voltage/current control LPA；
8) back production and actual output voltage/electric current and desired output voltage/electric current RMSE are calculated, confirmation ANFIS number The performance of pre- abnormal device.
Further, the establishment step of the LPA mathematical model is as follows:
1) actual test wiring is pressed, test device for relay protection is connect with tested relay protection and automatic safety device；
2) sinusoidal voltage and sinusoidal current that control LPA output frequency is 10kHz, its amplitude of smooth adjustment are allowed to increase from 0 Maximum value is arrived greatly；
3) sample, record voltage, the current amplitude of gradual change；
4) amplitude of the amplitude of command signal and output voltage, electric current is arranged in input/output data correspondingly It is right.
Further, steps are as follows for the formulation of the hybrid learning algorithm:
1) choosing minimum RMSE is learning objective:
In formula: dmIt is exported for ideal, | | | |2Indicate 2- norm；
2) to the coefficient of network after being adjusted by following learning algorithm:
In formula: k=0,1,2 ... indicates kth time adjustment, and β > 0 is learning rate；
It 3) will after 1 adjustmentIt is temporarily fixed, next adjust the parameter of feedforward network, different subordinating degree functions There is different parameters, without loss of generality, ifThere are 2 parameter aijAnd bij, learning algorithm are as follows:
In formula: l=0,1,2 ... indicates the l times adjustment, and γ > 0 is learning rate；
4) it usesaij(l+1) and bij(l+1) matrix A is formed with input data, uses αjUnknown vector α is formed, is used dmIdeal output vector d is formed, overdetermination matrix equation A α=d is obtained；With Least Square Method optimal solution α, minimum two can be obtained Multiply solution:
αLS=(ATA)-1ATd。
Another object of the present invention is to provide a kind of relay protection test dresses using the pre- abnormal method of the ANFIS number It sets.
The pre- abnormal method of ANFIS number of test device for relay protection LPA proposed by the present invention, overall thinking LPA From digital analog converter to each amplifying stage (including DC power supply) again to all possible non-linear of load, to the essence of each amplifying stage Degree is no longer overcritical, so that only needing modification software, traditional relay protection test device can be upgraded to device of new generation, can not only The required precision for meeting fault waveform reproduction, also makes limited frequency bandwidth and output capacity be fully used；Meanwhile also making Universal LPA is in use to device of new generation from traditional relay protection test device and is possibly realized, because universal LPA technology maturation, High reliablity and cost is relatively low, therefore the cost performance of device of new generation can be kept.The mathematical model of LPA can be filled by relay protection test There are enough for the training and study setting the self testing before exporting false voltage/electric current and recording, and be the pre- abnormal device of ANFIS number Time, this i.e. quasi- online mode；Using quasi- online mode, the pre- abnormal device of ANFIS number can be realized on original microprocessor, therefore No replacement is required high speed microprocessor meets requirement of real-time control, thus further increases the cost performance of device.In addition, by means of The adaptive ability of ANFIS, the present invention can also adapt to load (i.e. different relay protections and the safety of different characteristics and numerical value automatically The input characteristics of automatic device).
Detailed description of the invention
Fig. 1 is the realization stream of the pre- abnormal method of ANFIS number of test device for relay protection LPA provided in an embodiment of the present invention Cheng Tu.
Fig. 2 is the generic structure diagram of ANFIS provided in an embodiment of the present invention.
Fig. 3 is the positive digital model curve figure of current mode LPA provided in an embodiment of the present invention.
Fig. 4 is the inverse number word model curve graph of current mode LPA provided in an embodiment of the present invention.
Fig. 5 is simulated fault current curve diagram provided in an embodiment of the present invention.
The shape graph of subordinating degree function before Fig. 6 (a) is training provided in an embodiment of the present invention and learns.
Fig. 6 (b) is the shape graph of subordinating degree function after training provided in an embodiment of the present invention and study.
Fig. 7 is simulation example provided in an embodiment of the present invention without pre- abnormal actual output current curve graph.
Fig. 8 is the pre- abnormal instruction current curve graph of simulation example provided in an embodiment of the present invention.
Fig. 9 is the true fault current curve diagram of fault oscillograph record provided in an embodiment of the present invention.
Figure 10 is true fault electric current provided in an embodiment of the present invention without pre- abnormal actual output current curve graph.
Figure 11 is the pre- abnormal instruction current curve graph of true fault electric current provided in an embodiment of the present invention.
Specific embodiment
It is clear to be more clear the objectives, technical solutions, and advantages of the present invention, below in conjunction with specific embodiment, to this hair It is bright to be described in detail.It is noted that particular embodiments described herein is not intended to limit this hair only to explain the present invention It is bright.
The invention discloses a kind of test device for relay protection linear power amplifier (linear power Amplifier, LPA) adaptive neural network-fuzzy inference system (adaptive neuro-fuzzy inference System, ANFIS) the pre- abnormal method of number.The present invention using ANFIS by training and study after, go to approach in compact set after The nonlinear inverse mathematical model of electric protection test device LPA replaces inverse number word model to generate opposite with desired output voltage/electric current Command voltage/the electric current answered realizes that number is pre- abnormal.Overall thinking of the present invention LPA is from digital analog converter to each amplifying stage (including DC power supply) arrives all possible non-linear of load again, no longer overcritical to the precision of each amplifying stage.This makes universal LPA is in use to device of new generation from traditional relay protection test device and is possibly realized, because of universal LPA technology maturation, reliability It is high and cost is relatively low, therefore the cost performance of device of new generation can be kept.Using quasi- online mode, the present invention can be most of original It is realized on microprocessor, so there is no need to replace high speed microprocessor to meet requirement of real-time control, thus further increases device Cost performance.By means of the adaptive ability of ANFIS, the present invention can also adapt to different characteristics automatically and the load of numerical value is (i.e. different The input characteristics of relay protection and automatic safety device).The present invention is not only able to satisfy the required precision of fault waveform reproduction, may be used also The limited frequency bandwidth of LPA and output capacity is set to be fully used, so that modification software is only needed, it can be by traditional relay protection Test device upgrades to device of new generation.
After the present invention passes through training and study using ANFIS, go to approach the test device for relay protection LPA's in compact set Nonlinear inverse mathematical model, so that inverse number word model be replaced to generate command voltage/electricity corresponding with desired output voltage/electric current Stream is achieved that number is pre- abnormal.
Application principle of the invention is described in detail with reference to the accompanying drawing.
As shown in Figure 1, the pre- abnormal method of ANFIS number of test device for relay protection LPA provided in an embodiment of the present invention Realize the following steps are included:
S101: by the output of linear power amplifier (linear power amplifier, LPA) inverse number word model/defeated Enter data to as adaptive neural network-fuzzy inference system (adaptive neuro-fuzzy inference system, ANFIS input/output data pair), i.e. training data；
S102: the type and number of subordinating degree function are chosen；
S103: the initial configuration of ANFIS is generated；
S104: selection training parameter, i.e., maximum training batch, initial step length and target error；
S105: completing training by hybrid learning algorithm, generates the pre- abnormal device of ANFIS number；
S106: pre- abnormal command voltage/electric current is generated through the pre- abnormal device of ANFIS number with desired output voltage/electric current；
S107: realize that the amplification of voltage/current exports with pre- abnormal command voltage/current control LPA；
S108: back production simultaneously calculates actual output voltage/electric current and desired output voltage/electric current root-mean-square error (root Mean square error, RMSE), the validity of the pre- abnormal method of confirmation ANFIS number.
Application principle of the invention is further described with reference to the accompanying drawing.
1, the universal architecture of ANFIS is designed.
Because Takagi-Sugeno (T-S) fuzzy model convenient for analysis and calculates, therefore ANFIS is established in T-S fuzzy model more On the basis of.If input vector x=[x1,x2,…,xM]T, wherein xiFor Fuzzy Linguistic Variable, M is the dimension of x, and T indicates that vector turns It sets.If:
In formula:For xiJ-th of fuzzy language value (fuzzy set), NiFor fuzzy language value Number.DefinitionSubordinating degree function beThen T-S fuzzy rule are as follows:
then yj=pj0+pj1x1+…+pjMxM；
In formula: j is expanded, j=1,2 ..., N,pj0,pj1,…,pjMFor xiLinear combination is Number.By defined above, the universal architecture of ANFIS can be designed as shown in Fig. 2, it is by feedforward network and backward network two parts group At.
Feedforward network is formed for matching with the former piece of T-S fuzzy rule by 4 layers.1st layer is input layer, directly will be each xiIt is transmitted to next layer, therefore the layer has M node.2nd layer 1 fuzzy language value of each node on behalf, function is to pass throughCalculate xiIt is rightDegree of membership, which hasA node.The 1 of 3rd layer of 1 article of fuzzy rule of each node on behalf A former piece, function are to calculate the relevance grade of every fuzzy rule:
Or
The layer has Q node.4th layer of function is to calculate the weight of normalized i.e. every fuzzy rule of relevance grade:
The layer also has Q node.
Backward network is formed for matching with the consequent of T-S fuzzy rule by 3 layers.1st layer is input layer, directly will be each xiIt is transmitted to next layer, wherein x0=1 for generating y in following formulamjConstant termThis layer settable N≤M as needed Node.If ANFIS has R output, the output of every fuzzy rule of the 2nd layer of calculating:
In formula: m=1,2 ..., R, the layer have R × N number of node.3rd layer is output layer, function be output to the 2nd layer into Row synthesis:
The layer has R node.
2, hybrid learning algorithm is formulated.
The pre- abnormal method of ANFIS number uses the hybrid learning algorithm with good numerical stability and fast convergence.Choosing Taking minimum RMSE is learning objective:
In formula: dmIt is exported for ideal, | | | |2Indicate 2- norm.By backpropagation theory, following learning algorithm is pressed first To the coefficient of network after adjustment:
In formula: k=0,1,2 ... indicates kth time adjustment, and β > 0 is learning rate.After 1 adjustment,To temporarily it consolidate It is fixed, next adjust the parameter of feedforward network.Different subordinating degree functions has different parameters, without loss of generality, it is assumed thatThere are 2 parameter aijAnd bij, learning algorithm are as follows:
In formula: l=0,1,2 ... indicates the l times adjustment, and γ > 0 is learning rate.Under the classical gradient of the above 3 formula composition Drop method.At this point, withaij(l+1) and bij(l+1) matrix A is formed with input data, uses αjUnknown vector α is formed, is used dmIdeal output vector d is formed, then obtains overdetermination matrix equation A α=d.With Least Square Method optimal solution α, can obtain minimum Two multiply solution:
αLS=(ATA)-1ATd；
Thus hybrid learning algorithm is just realized.
3, the mathematical model of LPA is established.
Mathematical model is different from mathematical model.The former is obtained by test, using input/output data to expression, is most pasted Nearly actual conditions；The latter is obtained by introducing certain approximate condition by theory deduction, can be indicated with mathematical formulae, but may be with There are errors for actual conditions.To obtain high-precision, the mathematical model of Ying Caiyong LPA.
The mathematical model of LPA is established by following steps:
1) actual test wiring is pressed, test device for relay protection is connect with tested relay protection and automatic safety device；
2) sinusoidal voltage and sinusoidal current that control LPA output frequency is 10kHz, its amplitude of smooth adjustment are allowed to increase from 0 It is big that maximum value, (voltage isElectric current is)；
3) sample, record voltage, the current amplitude of gradual change；
4) amplitude of the amplitude of command signal and output voltage, electric current is arranged in input/output data correspondingly It is right.
When one timing of output waveform amplitude, the non-linear of LPA can increase with the increase of output waveform frequency, and failure is electric Pressure/current harmonics component frequency is usually no more than 10kHz, therefore the compatible 10kHz of the mathematical model measured at 10kHz is below Situation.
In view of the control method of voltage-type LPA and current mode LPA can be exchanged simply, and current mode LPA it is non-linear compared with Voltage-type LPA is significant, and current precision control controls difficulty compared with voltage accuracy, therefore is only illustrated by taking current mode LPA as an example below. The mathematical model of certain current mode LPA obtained by the above process can be that the longitudinal axis is plotted as using input data as horizontal axis, output data Curve as shown in Figure 3.Although maximum output current is actual up to 42A as it can be seen that calculating by rated current 30A (RMS) Linearity range has been compressed to -30~30A because non-linear, and positive curve and negative sense curve are not origin symmetries, are had lost close The resource of 10A.The mathematical model of current mode LPA is common " input → output " mapping, i.e., just reflects from the point of view of function It penetrates, therefore referred to as positive number word model.
For positive number word model, when the amplification factor of instruction current is more than 30 times, output electric current will just generate non-linear Distortion.Obviously, in the ideal situation, i.e. in the state that LPA remains linear, when amplification factor reaches 40 times of (specified amplifications Multiple) when, non-linear distortion should not also occur.It can be inverse mapping by positive mapping transformation to obtain desired output current, i.e., it is " defeated Positive digital model conversion is correspondingly inverse number word model by out → input " mapping.It is acquired and ideal output electricity by inverse number word model Corresponding instruction current is flowed, which be can be obtained into desired output current by positive number word model.Inverse number word model Can be by the way that input/output data be become input/output data to obtaining to simple switched position, the characterization on curve is exactly By the horizontally and vertically exchange of Fig. 3, as shown in Figure 4.
4, the pre- abnormal device of ANFIS number is realized by training and study.
After passing through training and study using ANFIS, remove to approach the non-linear of the test device for relay protection LPA in compact set Inverse number word model, to replace inverse number word model to generate command voltage/electric current corresponding with desired output voltage/electric current, just It is pre- abnormal to realize number.It is pre- abnormal why to be known as number, is because instruction current corresponding with desired output current is different from Former instruction current is seen seem that former instruction current is distorted in appearance, and is pre- first occurred before entering current mode LPA. Instruction current corresponding with desired output current is also referred to as pre- abnormal instruction current as a result,.By pre- abnormal instruction current input current Type LPA will obtain desired output current.
Steps are as follows for training, study and the use (i.e. the realization of the pre- abnormal method of ANFIS number) of the pre- abnormal device of ANFIS number:
1) by the input/output data of LPA inverse number word model to the input/output data pair as ANFIS, that is, number is trained According to；
2) type and number of subordinating degree function are chosen；
3) initial configuration of ANFIS is generated；
4) training parameter, i.e., maximum training batch, initial step length and target error are selected；
5) training is completed by hybrid learning algorithm, generates the pre- abnormal device of ANFIS number；
6) pre- abnormal command voltage/electric current is generated through the pre- abnormal device of ANFIS number with desired output voltage/electric current；
7) realize that the amplification of voltage/current exports with pre- abnormal command voltage/current control LPA；
8) back production and actual output voltage/electric current and desired output voltage/electric current RMSE are calculated, confirmation ANFIS number The validity of pre- abnormal method.
The mathematical model of LPA self testing and can be recorded before exporting false voltage/electric current by test device for relay protection, And there are enough time, this i.e. quasi- online modes for the training and study for being the pre- abnormal device of ANFIS number.Using quasi- online mode, The pre- abnormal device of ANFIS number can be realized on most of original microprocessor, without upgrading.
5, the parameter of the pre- abnormal device of preferably ANFIS number.
To be optimal the performance of the pre- abnormal device of ANFIS number, the type and number of subordinating degree function, maximum training batch Secondary, initial step length and target error need to carry out preferred through emulation experiment.Below using simulated fault electric current shown in Fig. 5 as example pair The parameter of the pre- abnormal device of ANFIS number carries out preferably, which contains the up to harmonic component of 9.75kHz, can be used for Traveling wave test.
1) type of subordinating degree function.Because simulated fault electric current be it is ambipolar, i.e., it is positive and negative, therefore should select bilateral Subordinating degree function should not select unilateral subordinating degree function, and otherwise training will error.Secondly, in bilateral subordinating degree function, also not Linear section of subordinating degree function can be selected, because ANFIS is not intended to " being/non-(1/0) " judgement, but is used for output waveform The control of precision, therefore should be using continuous curved subordinating degree function.By upper, subordinating degree function (should be divided into double sigmoid function Difference function and Product function), joint Gaussian function, single Gaussian function, select in generalized bell function and Π shape function.Shown in following table (subordinating degree function number 6, maximum training batch 100, initial step length 0.1, target error 0) is fixed, using not for remaining parameter When same type subordinating degree function, the RMSE of actual output current and desired output current.As it can be seen that generalized bell function is corresponding RMSE is minimum, therefore selects the type of subordinating degree function for generalized bell function.
2) number of subordinating degree function.Following table show remaining parameter and fixes (generalized bell membership degree function, maximum training Batch 100, initial step length 0.1, target error 0), different subordinating degree function number when, actual output current and ideal output electricity The RMSE of stream.As it can be seen that RMSE constantly declines with the increase of subordinating degree function number when just having started (2~3), illustrate to increase ANFIS The complexity of structure is conducive to improve precision；But when subordinating degree function number increases to after 4, not only no longer decline is anti-by RMSE And gradually increase, illustrate that " over-fitting " phenomenon, this needs occurred avoids.In addition, when subordinating degree function number is 3 RMSE is minimum, therefore should select subordinating degree function number is 3.Only 3 subordinating degree functions, the structure of ANFIS is very simple, training and Learning time greatly shortens, highly beneficial to real-time control.
3) maximum training batch.Following table show remaining parameter and fixes (generalized bell membership degree function, number 3, initial step Long by 0.1, target error 0), different maximum training batches when, the RMSE of actual output current and desired output current.As it can be seen that simultaneously Non-training batch is the bigger the better, and after training batch reaches 100 times, is further added by trained batch, and the difference of RMSE is unobvious, After illustrating training 100 times, ANFIS reliable conveyance.This is also demonstrated quickly receives using after hybrid learning algorithm, ANFIS has Holding back property.The shorter the training time the better, therefore selecting maximum training batch is 100 optimums.
4) initial step length.Following table show remaining parameter and fixes (generalized bell membership degree function, number 3, maximum training batch Secondary 100, target error 0), different initial step lengths when, the RMSE of actual output current and desired output current.As it can be seen that initial step Length is too big or too small all bad, and optimal values are between 0.01~0.1, and for simplicity, can use initial step length is 0.1.
5) target error.In, it is desirable to target error RMSE is the smaller the better, therefore should set target error as 0, and in practice It is only capable of that 0 can not be equal to close to 0.
To sum up, the preferred parameter of the pre- abnormal device of ANFIS number are as follows: generalized bell membership degree function, number 3, maximum training batch Secondary 100, initial step length 0.1, target error 0.
6, embodiment.
Application effect of the invention is described in detail below with reference to emulation and test example.
1) emulation embodiment.
The simulation experiment result using example shown in Fig. 5 when preferred parameter is provided herein.It is pre- that Fig. 6 show ANFIS number Inverse number word model training and study front and back of the abnormal device to current mode LPA, the situation of change of subordinating degree function.As it can be seen that generalized bell Subordinating degree function stretches in bilateral range or to compress more other type subordinating degree functions continuously gentle, therefore precision is higher.
By the simulated fault electric current of example divided by ideal amplification factor 40, normalized instruction current (i.e. input electricity is obtained Flow per unit value), it is passed through to the positive number word model of current mode LPA, can be obtained without the pre- abnormal actual output current of number, such as scheme Shown in 7.Comparison diagram 5 illustrates because of times magnification as it can be seen that the positive waveform of the actual output current occurs cutting top and burr is smoothed out Number more than 30 times and contains higher hamonic wave, and current mode LPA's is non-linear very significant, causes amplitude distortion and frequency distortion. Quantitatively calculation shows that, the RMSE of actual output current and desired output current is up to 11.0199.
For eliminate current mode LPA it is non-linear caused by be distorted, it would be desirable to export electric current and obtained by the digital pre- abnormal device of ANFIS Pre- abnormal instruction current, as shown in Figure 8.By pre- abnormal instruction current by the positive number word model of current mode LPA to obtain the final product to through digital pre- Abnormal actual output current has only been difficult to tell difference from visual angle, therefore has no longer provided song because it is almost the same with Fig. 5 Line chart.It is computed, at this moment the RMSE of actual output current and desired output current only has 0.0311, quantitatively illustrates that the two is non- Very close to.
The simulation experiment result shows that the pre- abnormal method of ANFIS number preferably eliminates the non-linear of current mode LPA, guarantees The precision of actual output current, utilizes the non-linear section of current mode LPA also.
2) test example.
The mathematical model of the lifted current mode LPA of the present invention derives from a universal test device for relay protection product, skill Art parameter are as follows: rated capacity 2kVA, 0~30A of output-current rating (RMS)；When exporting electric current less than 0.2A (RMS), absolutely accidentally Difference is less than 0.4mA (RMS)；Electric current is exported at 0.2~30A (RMS), relative error is less than 0.2%；Rated output frequency 0~ 2.5kHz；Specified amplification factor 40.It may be noted that these technical parameters are only able to satisfy current design criteria, " DL/T 995-2006 relay is protected Shield and power grid security automatic device inspection procedure " as defined in traditional test requirement.And before the regulation is published on 10 years, it lays particular emphasis on Power frequency quantity test needs to be supplemented current new test content and requirement.Therefore the above technical parameter be to power frequency quantity i.e. 50Hz just For string electric current.So surface is seen, precision is very high, but this is only the precision under 50Hz sinusoidal current, if using it Output harmonic wave complicated component, amplification factor are more than 30 times of fault current, then gained precision will be far from each other.In addition, the above essence Degree index (absolute error and relative error) is only capable of the virtual value of simply estimation of sinusoidal electric current, cannot be deep into inside waveform and go Point-to-point error is assessed, it is completely not applicable to the fault current containing non-persistent DC component and harmonic component.Therefore, The present invention is also assessed using that can count and the RMSE of point-to-point error is the pre- abnormal device of target training ANFIS number with RMSE practical Export the precision of electric current.
The pre- abnormal method of ANFIS number is applied to the current mode LPA of above-mentioned test device for relay protection below, passes through test Verify its validity.The true fault electric current shown in Fig. 9 recorded by fault oscillograph is selected in test.By the true fault Electric current obtains normalized instruction current (i.e. input current per unit value) divided by ideal amplification factor 40, controls current mode with it LPA amplification reappears, and obtains without the pre- abnormal actual output current of number, as shown in Figure 10.Comparison diagram 9 is as it can be seen that the reality output There is apparent distortion in failure of the current section, this is because contain non-persistent DC component and harmonic component in faulty section, Show current mode LPA significant non-linear, amplification factor changes correspondingly.It is quantitative calculation shows that, the actual output current with The RMSE of desired output current is 10.7955.
Desired output current is obtained into pre- abnormal instruction current by the pre- abnormal device of ANFIS number below, as shown in figure 11.By pre- The control current mode LPA amplification of abnormal instruction current reappears, and obtains through the pre- abnormal actual output current of number, because its with Fig. 9 almost phase Together, therefore it is similar to emulation embodiment, no longer provides curve graph.It is computed, the RMSE of actual output current and desired output current Only 0.0372, quantitatively illustrate that the precision of actual output current is very high, the pre- abnormal device of ANFIS number is same for actual device Effectively.
Finally by emulation embodiment and test example through number it is pre- it is abnormal after RMSE do a comparison, value is respectively 0.0311 and 0.0372, it is seen that the two is relatively.Although this illustrates to emulate and test that model and practical dress is respectively adopted It sets and is carried out for different fault currents, but its precision has preferable consistency.
It is noted that emulation embodiment provided above and test example are only used for stating application effect of the invention, not To limit the present invention, any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should all It is included within protection scope of the present invention.

Claims (5)

1. the pre- abnormal method of ANFIS number of test device for relay protection LPA a kind of, which is characterized in that the relay protection test After the pre- abnormal method of ANFIS number of device LPA passes through training and study using ANFIS, go to approach the relay protection in compact set The nonlinear inverse mathematical model of test device LPA replaces inverse number word model to generate finger corresponding with desired output voltage/electric current Voltage/current is enabled, realizes that number is pre- abnormal；
The realization of the pre- abnormal method of ANFIS number of the test device for relay protection LPA the following steps are included:
1) by the input/output data of the nonlinear inverse mathematical model of test device for relay protection LPA to as the defeated of ANFIS Enter/output data pair, i.e. training data；
2) type and number of subordinating degree function are chosen；
3) initial configuration of ANFIS is generated；
4) training parameter, i.e., maximum training batch, initial step length and target error are selected；
5) training is completed by hybrid learning algorithm, generates the pre- abnormal device of ANFIS number；
6) pre- abnormal command voltage/electric current is generated through the pre- abnormal device of ANFIS number with desired output voltage/electric current；
7) realize that the amplification of voltage/current exports with pre- abnormal command voltage/current control LPA；
8) back production and actual output voltage/electric current and desired output voltage/electric current RMSE are calculated, confirmation ANFIS number is pre- abnormal The validity of method；
2. the pre- abnormal method of ANFIS number of test device for relay protection LPA as described in claim 1, which is characterized in that described The establishment step of the nonlinear inverse mathematical model of test device for relay protection LPA is as follows:
1) actual test wiring is pressed, test device for relay protection LPA is connect with tested relay protection and automatic safety device；
2) sinusoidal voltage and sinusoidal current that control test device for relay protection LPA output frequency is 10kHz, its width of smooth adjustment Value, is allowed to increase to maximum value from 0；
3) sample, record voltage, the current amplitude of gradual change；
4) amplitude of output voltage, the amplitude of electric current and command signal is arranged in input/output data pair correspondingly.
3. the pre- abnormal method of ANFIS number of test device for relay protection LPA as described in claim 1, which is characterized in that described Steps are as follows for the formulation of hybrid learning algorithm:
1) choosing minimum RMSE is learning objective:
In formula: dmIt is exported for ideal, | | | |2Indicate 2- norm；
2) to the coefficient of network after being adjusted by following learning algorithm:
In formula: k=0,1,2 ... indicates kth time adjustment, and β > 0 is learning rate；
It 3) will after 1 adjustmentIt is temporarily fixed, the parameter of feedforward network is next adjusted, different subordinating degree functions is different Parameter,There are 2 parameter aijAnd bij, learning algorithm are as follows:
In formula: l=0,1,2 ... indicates the l times adjustment, and γ > 0 is learning rate；
4) it usesaij(l+1) and bij(l+1) matrix A is formed with input data, uses αjUnknown vector α is formed, d is usedmComposition Ideal output vector d obtains overdetermination matrix equation A α=d；With Least Square Method optimal solution α, least square solution can be obtained:
αLS=(ATA)-1ATd。
4. a kind of pre- abnormal method of ANFIS number using test device for relay protection LPA described in 3 any one of claims 1 to 3 The pre- abnormal device of ANFIS number of foundation.
5. a kind of test device for relay protection for being equipped with the pre- abnormal device of the number of ANFIS described in claim 4.
CN201710035791.5A 2017-01-18 2017-01-18 A kind of pre- abnormal method of ANFIS number of test device for relay protection LPA Active CN106597178B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710035791.5A CN106597178B (en) 2017-01-18 2017-01-18 A kind of pre- abnormal method of ANFIS number of test device for relay protection LPA

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710035791.5A CN106597178B (en) 2017-01-18 2017-01-18 A kind of pre- abnormal method of ANFIS number of test device for relay protection LPA

Publications (2)

Publication Number Publication Date
CN106597178A CN106597178A (en) 2017-04-26
CN106597178B true CN106597178B (en) 2019-07-19

Family

ID=58586491

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710035791.5A Active CN106597178B (en) 2017-01-18 2017-01-18 A kind of pre- abnormal method of ANFIS number of test device for relay protection LPA

Country Status (1)

Country Link
CN (1) CN106597178B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106990312B (en) * 2017-05-25 2023-06-16 贵州电网有限责任公司电力科学研究院 Testing device and testing method for digital relay protection device
CN108107294B (en) * 2017-12-27 2020-02-07 国网冀北电力有限公司张家口供电公司 Microcomputer relay protection test system based on measurement modularization
CN108195704A (en) * 2018-03-05 2018-06-22 重庆水利电力职业技术学院 A kind of grinding parameter test bed component automatically
CN110231586B (en) * 2019-04-11 2024-10-08 浙江三辰电器股份有限公司 Use method of full-range self-calibration device of transformer
CN110108957B (en) * 2019-05-07 2020-06-02 武汉理工大学 Tractor electrical fault diagnosis method based on structural analysis method

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR100471650B1 (en) * 2003-05-30 2005-03-14 국방과학연구소 Method for estimating an inductance profile of srm using adaptive neuro-fuzzy inference system
CN103001591A (en) * 2012-07-16 2013-03-27 北京邮电大学 Method for eliminating nonlinear distortion
CN104155574B (en) * 2014-07-31 2017-12-15 国网湖北省电力公司武汉供电公司 Distribution network failure sorting technique based on Adaptive Neuro-fuzzy Inference
CN104536435B (en) * 2014-12-18 2017-06-30 中国科学院电工研究所 A kind of line control system network inline diagnosis method
CN105912806B (en) * 2016-04-28 2018-12-18 湖南大学 A kind of small overlapping impact air bag control method based on Adaptive Neural-fuzzy Inference
CN106291354A (en) * 2016-07-18 2017-01-04 河海大学 Electric automobile permagnetic synchronous motor Fault Classification based on ANFIS

Also Published As

Publication number Publication date
CN106597178A (en) 2017-04-26

Similar Documents

Publication Publication Date Title
CN106597178B (en) A kind of pre- abnormal method of ANFIS number of test device for relay protection LPA
Roinila et al. Hardware-in-the-loop methods for real-time frequency-response measurements of on-board power distribution systems
CN111541262B (en) MMC frequency coupling impedance modeling method under control of model pre-determination alternating voltage
Serra et al. IDA-PBC control of a DC–AC converter for sinusoidal three-phase voltage generation
CN106021768B (en) The power distribution network of the access containing distributed generation resource simplifies modeling method
Tran et al. Harmonic and unbalanced voltage compensation with VOC‐based three‐phase four‐leg inverters in islanded microgrids
Vu et al. Adaptive backstepping approach for dc‐side controllers of Z‐source inverters in grid‐tied PV system applications
CN111695221B (en) Robust controller design method for ensuring stable operation of direct current bus voltage
Abdelkrim et al. Robust stability power in the transmission line with the use of a UPFC system and neural controllers based adaptive control
Miret et al. PI‐based controller for low‐power distributed inverters to maximise reactive current injection while avoiding over voltage during voltage sags
Santiprapan et al. Adaptive repetitive controller for an active power filter in three‐phase four‐wire systems
Islam et al. Distributed secondary controller to ensure proportional sharing of reactive power in AC microgrid
Firdaus et al. Dynamic power flow based simplified transfer function model to study instability of low‐frequency modes in inverter‐based microgrids
CN108964089B (en) Power system load frequency control method and related product
CN109787234B (en) Method for acquiring ultrahigh harmonic stable mode of distributed power supply containing VSC (voltage source converter) interface
CN113315122B (en) Fault transient current analysis method considering nonlinear characteristics of inverter power supply control system
Zhu et al. Effect of load modeling on voltage stability
Hernández-Ramírez et al. Linearization of periodic power electronic-based power systems for small-signal analysis
Zenk et al. Multi zone power systems load-frequency stability using fuzzy logic controllers
Kataoka A smooth power flow model of electric power system with generator reactive power limits taken into consideration
Oleskovicz et al. Estimation of harmonic currents injected by nonlinear loads for a distorted power supply scenario using artificial neural networks
Junior et al. Development of a Multivariable Deadbeat Controller in dq Coordinates for the Current Loop of a Grid-Connected VSI
Brandl et al. HIL methods supporting the development process from simulations to real environment testing
Liasi et al. Current Harmonic Compensation by Active Power Filter Using Neural Network-Based Recognition and Controller
Feng et al. Data-Driven Time-Frequency-Domain Equivalent Modeling of Wind Farms for Wideband Oscillations Analysis

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