CN108268856A - Variable-step self-adaptive harmonic detecting method based on L2 norms and real tracking error - Google Patents
Variable-step self-adaptive harmonic detecting method based on L2 norms and real tracking error Download PDFInfo
- Publication number
- CN108268856A CN108268856A CN201810112763.3A CN201810112763A CN108268856A CN 108268856 A CN108268856 A CN 108268856A CN 201810112763 A CN201810112763 A CN 201810112763A CN 108268856 A CN108268856 A CN 108268856A
- Authority
- CN
- China
- Prior art keywords
- current
- tracking error
- norms
- error
- fundamental
- 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
Links
Classifications
-
- 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
- G06F2218/04—Denoising
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R23/00—Arrangements for measuring frequencies; Arrangements for analysing frequency spectra
- G01R23/16—Spectrum analysis; Fourier analysis
Abstract
The present invention is based on the variable-step self-adaptive harmonic detecting method of L2 norms and real tracking error, this method is improved in the Harmonic Detecting Algorithm of existing Active noise cancellation principle.The feedback quantity of tracking situation can really be reflected by being found first with running integral device, then updated feedback error is brought into step iteration formula and adjusts step-length.Step iteration formula is introduced using L2 norms to establish error signal, so as to reflect influence of the input signal to performance.This method calculates simple, tracing detection signal is accurate, strong robustness, and faster convergence rate can also be kept while smaller steady output rate is ensured, and when load current mutates also can follow current well variation, monophase system and three-phase system are all suitable for.
Description
Technical field
The invention belongs to the Harmonic currents detection technical fields of electric system, and in particular to one kind is based on L2 norms and really
The variable-step self-adaptive harmonic detecting method of tracking error.
Background technology
The a large amount of of the nonlinear loads such as power electronic equipment are used so that the harmonic pollution of power grid is on the rise, on a large scale newly
The access of the energy has been further exacerbated by this situation, has seriously affected the economy, stabilization and safe operation of electric system, therefore has
Effect is administered harmonic wave and is just particularly important.Active Power Filter-APF (APF) being capable of rapidly dynamic tracing compensation harmonic wave and idle
Power, it is made of harmonic detecting part, current tracking part, driving circuit and main circuit part.Harmonic detecting is as harmonic wave
The primary link of improvement has vital effect to harmonic compensation quality and quality.Existing harmonic detecting method is main
There are Fast Fourier Transform (FFT) (FFT), instantaneous reactive power theory and neural network, wavelet transformation etc., all done for harmonic detecting
Gone out very big contribution, but its respectively the defects of limit further application.
In numerous methods, the Harmonic Detecting Algorithm based on Active noise cancellation principle has very strong adaptive ability,
It is little to system dependence, and algorithm is simple, strong robustness, and tracing detection signal is accurate.But traditional fixed step size is adaptive
Harmonic Detecting Algorithm can not take into account convergence rate and stable state accuracy simultaneously, the bigger expression convergence speed of the algorithm of step-length in algorithm
Faster but Simultaneous Stabilization error can also become larger；On the contrary, the smaller convergence speed of the algorithm of step-length is slow but steady-state error can become smaller.Cause
This, the selection of step-length must do a balance between, both to ensure to have quick convergence rate or to ensure smaller stable state
Error.Numerous scholars have also done variable-step self-adaptive harmonic detecting many research." variable step size adaptive algorithm exists document 1
Application in active filter harmonic detecting " (Li Hui, Wu Zhengguo, Zou Yunping, Liu Fei, Wu Yanfeng, Proceedings of the CSEE
[J].2006(09):99-10 3), this method using error signal e (n) in signal to be detected shared ratio K (n) as adaptive
Feedback amount is answered, and the coherence average for passing through K (n) is estimated to adjust the step-length of algorithm, but when near current zero-crossing point, the algorithm
It will appear the difficulty of numerical computations, and poor to the ability of tracking of load sudden change, it is impossible to eliminate fundamental reactive current to step-length more
New interference；Document 2 " a kind of novel quick self-adapted Harmonic Detecting Algorithm " (He Na, Huang Lina, Wu Jian, Chinese electrical engineering
Journal [J] .2008, (22):124-129) this method propose a kind of fuzzy variable-step self-adaptive detection algorithm, but introduce from
Dynamic gene is a vector, and parameter is excessive, affects convergence rate.
In conclusion follow current changing capability is not in load sudden change for existing variable-step self-adaptive harmonic detecting method
By force, calculate excessively numerous and diverse, the accuracy of testing result, convergence rate and stable state accuracy are all to be improved.
Invention content
The purpose of the present invention is to solve above-mentioned technical problem, the present invention provides a kind of based on L2 norms and really tracking
The variable-step self-adaptive harmonic detecting method of error.This method is found using running integral device from feedback error can really reflect
The feedback quantity of tracking situation simplifies calculating, and updated feedback quantity, step-length then are updated to the new step based on L2 norms
Realize that step-length is adjusted in long iterative formula.Input signal is introduced into step iteration function to build by using the formula of L2 norms
Vertical error signal, so as to reflect influence of the input signal to performance.Meanwhile smoothing factor is added, make step-length simultaneously by working as
Preceding error amount and previous error amount determine, make it have certain noise resisting ability.The implementation of this technology can either speed inspection
The convergence rate of survey can also make accuracy of detection higher, also with good robustness, also can quickly be tracked in load sudden change
The variation of load current, detects fundamental active current precisely in real time.
In order to achieve the above objectives, the technical solution adopted by the present invention is：
A kind of variable-step self-adaptive harmonic detecting method based on L2 norms and real tracking error, this method include following
Step：
(1) it samples
To periodical nonsinusoidal load current iL(t) it is sampled, obtains the centrifugal pump i of load currentL(n), while make electricity
Net voltage uS(t) fundamental wave reference input unit vector x is obtained by phaselocked loop1(t)=sin (ω t), x2(t)=cos (ω t),
And fundamental wave reference input unit vector is sampled to obtain the centrifugal pump x of fundamental wave reference input unit vector1(n)=sin (ω
nTS)、x2(n)=cos (ω nTS)；Wherein, ω=2 π f, f is frequency, and f=50Hz, n are the discrete point of time t, TSFor sampling
Period；
(2) by the centrifugal pump x of fundamental wave reference input unit vector1(n)、x2(n) it inputsIn, it obtains
To output signalThe as estimated value of fundamental active current；
Wherein, X (n)=[x1(n),x2(n)]T；W (n) is the corresponding weight vectors of X (n), W (n)=[W1(n),W2(n)
]T, initial value zero；
(3) with the centrifugal pump i of the load current in step (1)L(n) fundamental active current obtained in step (2) is subtracted
Estimated value, obtain to compensate the estimated value of electric current, i.e. error signal e (n)；Then
(4) real tracking error is found using running integral device：
The error signal obtained in step (3) is carried out according to discretization Fourier expansion, then is had：
In formula：
By the error signal e (n) obtained in step (3) and synchronized lockin signal sin ω nTSIt is multiplied, and to being multiplied
Result in the sampled data of a cycle integrated, i.e.,：
E is the average real tracking error of each sampled point in one cycle；N is the sampling in a grid cycle T
Number；N0It is newest sampled point；W is the average weight of each sampled point in a cycle；
(5) joined according to the real tracking error E (n) obtained in step (4) in n-th of sampled point with fundamental wave in step (2)
The centrifugal pump X (n) of input unit vector is examined, the value of parameter P (n) is calculated, i.e.,：
P (n)=E (n) × X (n)；
(6) the parameter P (n) obtained to step (5) with formula P (n)=a × P (n-1)+b × E (n) × X (n), is carried out more
Newly, wherein a and b be smoothing factor, 0<a<1,0<b<1；
(7) set step iteration formula as μ (n)=β (1-exp (- α E (n) | × | | P (n) | |2)), after step (6) is updated
The L2 norms of obtained P (n), i.e., | | P (n) | |2It is brought into the step iteration formula of μ (n)；Wherein, α and β is parameter；
The condition of convergence of step iteration formula is 0<μ(n)<λmax, 0<β<1/λmax；λmaxReference input unit vector from
The maximum eigenvalue of correlation matrix；
(8) multiplying according to the centrifugal pump X (n) of real tracking error E (n), step value μ (n) and reference input unit vector
The variation of product constantly regulate weighted vector W (n), i.e.,：
W (n+1)=+ 2 μ (n) E (n) X (n) of W (n)；
(9) formula in iterative step (6), step (7), step (8), finally so that the best weight coefficient of W (n) infinite approachs
Wopt；At this moment output signalFundamental current active component i can be approached in amplitude and phase1p(n), so as to fulfill harmonic wave
The detection of electric current and fundamental reactive current.
Compared with prior art, the beneficial effects of the invention are as follows：
1) present invention has found real tracking error E (n), and the AC influence part in former tracking error e (n) is rejected,
Make algorithm that there is faster convergence rate and better accuracy；2) real tracking error being found using running integral device can subtract
Few a large amount of calculating can have better real-time to avoid longer delay effect；3) current error value and input signal are drawn
Enter to the variable step iteration function μ (n) based on L2 norms=β (1-exp (- α E (n) | × | | P (n) | |2)) in, this formula energy and
Shi Fanying error amounts and input signal change the influence to algorithm, make the adjustment of step-length more quick and precisely.And it also solves
Formula is excessively complicated in existing variable-step self-adaptive harmonic detecting method, and when error is close to zero, significant changes can occur for waveform
The problem of；4) present invention when load current mutates also can quick follow current variation, crossing half period, just energy is accurately
In tracking, there is good robustness.
To sum up this method calculates simple, and tracing detection signal is accurate, strong robustness, is ensureing the same of smaller steady output rate
When can also keep faster convergence rate, and when load current mutates also can follow current well variation, it is right
It is all suitable in monophase system and three-phase system.
Description of the drawings
Fig. 1 is the variable-step self-adaptive harmonic detecting method principle based on L2 norms and real tracking error of the present invention
Figure；
Fig. 2 (a) is nonlinear load current waveform；
Fig. 2 (b) is load current spectrogram；
Tradition fixed step algorithm tracking comparison oscillogram when Fig. 3 (a) is μ=0.005；
Tradition fixed step algorithm tracking comparison oscillogram when Fig. 3 (b) is μ=0.08；
Tradition fixed step algorithm tracking comparison oscillogram when Fig. 3 (c) is μ=0.3；
Fig. 4 (a) is the emulation used the present invention is based on L2 norms and the adaptive neural network of real tracking error
The current tracking to be compensated comparison oscillogram of figure；
Fig. 4 (b) is the emulation used the present invention is based on L2 norms and the adaptive neural network of real tracking error
The fundamental active current tracking comparison oscillogram of figure；
Fig. 4 (c) is the partial enlarged view in Fig. 4 (b)；
Fig. 4 (d) is the emulation used the present invention is based on L2 norms and the adaptive neural network of real tracking error
The spectrogram of electric current after the detection of figure；
Fig. 5 (a) is the current waveform figure that load current mutates in t=0.1s；
Current tracking compares oscillogram when Fig. 5 (b) is load sudden change.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples, but is not protected in this, as to the application
Protect the restriction of range.
Assuming that the electric current of periodical nonlinear load is iL(t), the Fourier expansion formula of discretization is：
Wherein：M=2,3 ..., represent overtone order；i1p(t) it is fundamental current power component；i1q(t) it is fundamental current
Reactive component；ih(t) it is harmonic current part.It is i then to need the electric current compensatedd(t)=i1q(t)+ih(t)。
The present invention is based on the variable-step self-adaptive harmonic detecting method of L2 norms and real tracking error (abbreviation method), packets
Include following steps：
(1) it samples
To periodical nonsinusoidal load current iL(t) it is sampled, obtains the centrifugal pump i of load currentL(n), while make electricity
Net voltage uS(t) fundamental wave reference input unit vector x is obtained by phaselocked loop (PLL)1(t)=sin (ω t), x2(t)=cos
(ω t), and fundamental wave reference input unit vector is sampled to obtain the centrifugal pump x of fundamental wave reference input unit vector1(n)=
sin(ωnTS)、
x2(n)=cos (ω nTS)；Wherein, ω=2 π f, f is frequency, and f=50Hz, n are the discrete point of time t, TSTo adopt
The sample period；
(2) by the centrifugal pump x of fundamental wave reference input unit vector1(n)、x2(n) it inputsIn, it obtains
To output signalThe as estimated value of fundamental active current；
Wherein, X (n)=[x1(n),x2(n)]T；W (n) is the corresponding weight vectors of X (n), W (n)=[W1(n),W2(n)
]T, initial value zero；
(3) with the centrifugal pump i of the load current in step (1)L(n) fundamental active current obtained in step (2) is subtracted
Estimated value, obtain to compensate the estimated value of electric current, i.e. error signal e (n)；Then
(4) real tracking error is found using running integral device：
The centrifugal pump i of load currentL(n) discretization Fourier expansion formula is iL(n)=i1p(n)+i1q(n)+ih(n),
Compensate electric current id(n)=i1q(n)+ih(n), wherein i1q(n) it is the fundamental wave reactive power part in load current centrifugal pump, ih(n) it is
Harmonic；
Then the expression formula of error e (n) is：
Wherein A (n) is theoretical real tracking error, so the i that accounting is very heavydIt (n) will strong influence step-length and power
It is worth the speed and accuracy adjusted, finally influences harmonic detecting.
The error signal obtained in step (3) is carried out according to discretization Fourier expansion, then is had：
In formula：
By the error signal e (n) obtained in step (3) and synchronized lockin signal (i.e. sin ω nTS) be multiplied, and right
The sampled data of a cycle is integrated in the result of multiplication, i.e.,：
E is exactly the average real tracking error of each sampled point in one cycle, wherein：N is in a grid cycle T
Number of samples；N0It is newest sampled point；W is the average weight of each sampled point in a cycle；
(5) joined according to the real tracking error E (n) obtained in step (4) in n-th of sampled point with fundamental wave in step (2)
The centrifugal pump X (n) of input unit vector is examined, the value of parameter P (n) can be calculated, i.e.,：
P (n)=E (n) × X (n)
(6) the parameter P (n) obtained to step (5) with formula P (n)=a × P (n-1)+b × E (n) × X (n), is carried out more
Newly, it can reflect influence of the previous parameter value P (n-1) to algorithm performance, wherein a and b for smoothing factor, 0<a<1,0<b<1；
(7) set step iteration formula as μ (n)=β (1-exp (- α E (n) | × | | P (n) | |2)), after step (6) is updated
The L2 norms of obtained P (n), i.e., | | P (n) | |2It is brought into the step iteration formula of μ (n)；Wherein, α and β is parameter, ginseng
Number β can determine the range of step value, and parameter alpha can determine | E (n) | × | | P (n) | |2Variation size, improve algorithm
Antijamming capability；
The condition of convergence of step iteration formula is 0<μ(n)<λmax, 0<β<1/λmax；λmaxReference input unit vector from
The maximum eigenvalue of correlation matrix；
(8) multiplying according to the centrifugal pump X (n) of real tracking error E (n), step value μ (n) and reference input unit vector
The variation of product constantly regulate weighted vector W (n), i.e.,：
W (n+1)=+ 2 μ (n) E (n) X (n) of W (n)
(9) formula in iterative step (6), step (7), step (8), finally so that the best weight coefficient of W (n) infinite approachs
Wopt；At this moment output signalFundamental current active component i can be approached in amplitude and phase1p(n), so as to fulfill harmonic wave
The detection of electric current and fundamental reactive current.
Embodiment
Adaptive neural network of the present embodiment based on L2 norms and real tracking error, its step are as follows：
(1) it samples
To periodical nonsinusoidal load current iL(t) it is sampled, obtains the centrifugal pump i of load currentL(n), while make electricity
Net voltage uS(t) fundamental wave reference input unit vector x is obtained by phaselocked loop (PLL)1(t)=sin (ω t), x2(t)=cos
(ω t), and it is sampled to obtain the centrifugal pump x of fundamental wave reference input unit vector1(n)=sin (ω nTS)、x2(n)=
cos(ωnTS)；Wherein, ω=2 π f, f is frequency, and f=50Hz, n are the discrete point of time t, TSFor the sampling period；
(2) by the centrifugal pump x of fundamental wave reference input unit vector1(n)、x2(n) it inputsIn, it obtains defeated
Go out signalThe as estimated value of fundamental active current；
Wherein, X (n)=[x1(n),x2(n)]T；W (n) is the corresponding weight vectors of X (n), W (n)=[W1(n),W2(n)
]T, initial value zero；
(3) with the centrifugal pump i of the load current in step (1)L(n) fundamental active current obtained in step (2) is subtracted
Estimated value, obtain to compensate the estimated value of electric current, i.e. error signal e (n)；Then：
(4) real tracking error is found using running integral device：
The centrifugal pump i of load currentL(n) discretization Fourier expansion formula is iL(n)=i1p(n)+i1q(n)+ih(n),
Compensate electric current id(n)=i1q(n)+ih(n), wherein i1q(n) it is the fundamental wave reactive power part in load current centrifugal pump, ih(n) it is
Harmonic；
Then the expression formula of error e (n) is：
Wherein A (n) is theoretical real tracking error, so the i that accounting is very heavydIt (n) will strong influence step-length and power
It is worth the speed and accuracy adjusted, finally influences harmonic detecting.
The error signal obtained in step (3) is carried out according to discretization Fourier expansion, then is had：
In formula：
By the error signal e (n) obtained in step (3) and synchronized lockin signal (i.e. sin ω nTS) be multiplied, and right
The sampled data of a cycle is integrated in the result of multiplication, i.e.,：
E is exactly the average real tracking error of each sampled point in one cycle, wherein：N is a grid cycle T
In number of samples；N0It is newest sampled point；W is the average weight of each sampled point in a cycle；
(5) joined according to the real tracking error E (n) obtained in step (4) in n-th of sampled point with fundamental wave in step (2)
The centrifugal pump X (n) of input unit vector is examined, the value of parameter P (n) can be calculated, i.e.,：
P (n)=E (n) × X (n)
(6) the parameter P (n) obtained to step (5) is carried out more with formula P (n)=a × P (n-1)+b × E (n) × X (n)
Newly, it can reflect influence of the previous parameter value P (n-1) to algorithm performance, wherein a and b for smoothing factor, 0<a<1,0<b<1；
(7) set step iteration formula as μ (n)=β (1-exp (- α E (n) | × | | P (n) | |2)), after step (6) is updated
The L2 norms of obtained P (n), i.e., | | P (n) | |2It is brought into the step iteration formula of μ (n)；Wherein, α and β is parameter, ginseng
Number β can determine the range of step value, and parameter alpha can determine | E (n) | × | | P (n) | |2Variation size, improve algorithm
Antijamming capability；
The condition of convergence of step iteration formula is 0<μ(n)<λmax, 0<β<1/λmax；λmaxReference input unit vector from
The maximum eigenvalue of correlation matrix；
Simultaneously in order to be better understood upon influence of each parameter to algorithm, the present invention is analyzed using control variate method in parameter
The situation of change of Steady State Square Error during variation.It can show that parameter beta mainly has an impact algorithm the convergence speed, and α, a, b
Value not only influences the size that convergence rate also determines steady-state error.
(8) multiplying according to the centrifugal pump X (n) of real tracking error E (n), step value μ (n) and reference input unit vector
The variation of product constantly regulate weighted vector W (n), i.e.,：
W (n+1)=+ 2 μ (n) E (n) X (n) of W (n)
(9) formula in iterative step (6), step (7), step (8), finally so that the best weight coefficient of W (n) infinite approachs
Wopt；At this moment output signalFundamental current active component i can be approached in amplitude and phase1p(n), so as to fulfill harmonic wave
The detection of electric current and fundamental reactive current.
Emulation experiment
In order to verify the feasibility of the adaptive neural network based on L2 norms and real tracking error and superiority,
Carry out emulation experiment.In following emulation experiment, reference input unit vector is the sinusoidal signal of power frequency.
When 0s, detection method is put into system, sample frequency Fs is set as 5000Hz, and signal length L is
2000, then every TS=1/Fs=2 × 10-4S samplings are primary, and total sampling time is 0.4s.Reference input unit vector x1=
sin(100πt)、x2=cos (100 π t) weight w1(n)、w2(n) initial value is 0.Gone out according to expected artificial fitting non-thread
Property load current waveform iL, expression formula is：
iL(t)=23.13sin (100 π t)+4.76sin (500 π t+3.056)+3.13sin (700 π t-3.044)+
2.07sin(1100πt-0.004)+1.73sin(1300πt+0.078)
+1.33sin(1700πt+3.154)+1.18sin(1900πt+3.21)+0.99sin(2300πt-0.091)
It can be seen that the harmonic content of load current and expected fitting are consistent from the load current spectrogram of Fig. 2 (b)
, containing 5 times, 7 inferior harmonic waves, overtone order is up to 23 times, wherein 5 subharmonic contents are maximum.
Experiment one
Traditional fixed step algorithm is emulated first, because in traditional iterative formula：W (n+1)=+ 2 μ E (n) of W (n)
X (n), step factor μ are changeless, thus in emulating each time, when the setting of μ values is excessive, each iteration of weights W
Numerical value can also become larger therewith, can thus make W faster close to best weight coefficient Wopt, that is, accelerate convergence rate.But simultaneously
Also it can make W and WoptDifference become larger so that steady-state error becomes larger；When μ changes are small, then situation is exactly the opposite.This is exactly that tradition is fixed
The shortcomings that step-length, it is impossible to while take into account convergence rate and stable state accuracy.
The tracking comparison of tradition fixed step algorithm when Fig. 3 (a), Fig. 3 (b), Fig. 3 (c) are μ=0.005,0.08,0.3 respectively
Oscillogram can be observed clearly, and convergence rate is very slow during μ=0.005, about five half periods；During μ=0.08 about
Three half periods；About half period during μ=0.3, with the increase of step-length, convergence rate is getting faster, but tracks essence
Degree but become worse and worse, exactly fixed step algorithm above-mentioned the shortcomings that.It can be seen that the actual tracking wave in Fig. 3 (c)
Shape and ideal waveform registration be not high, always exists steady-state error or even tracks distortion current also certain in waveform.
Experiment two
The Variable Step Algorithm of the present invention is emulated, but before this in order to be better understood upon each parameter to algorithm
It influences, utilizes the situation of change of control variate method analysis Steady State Square Error in Parameters variation herein.It can obtain parameter beta master
To have an impact to algorithm the convergence speed, and the value of α, a, b not only influence the size that convergence rate also determines steady-state error.According to point
Analysis result takes α=800, β=0.14, a=0.96, b=0.3.Fig. 4 (a) is variable-step self-adaptive harmonic wave inspection proposed by the present invention
The tracking comparison oscillogram of the electric current to be compensated of the simulation waveform of survey method, it can be seen that waveform approximately passes through half period just
It can accurately track.Shown in corresponding fundamental active current tracking comparison oscillogram such as Fig. 4 (b), partial enlarged view such as Fig. 4
(c) shown in, ideal waveform and actual tracking waveform just coincide substantially after more half periods, and tracking accuracy is very high, in Fig. 4 (d)
In it can be seen that detection after harmonic wave be substantially filtered out, only remaining frequency be 50Hz fundamental active component, it was demonstrated that the method for the present invention energy
Enough requirements for meeting steady-state behaviour and convergence rate simultaneously.
Experiment three
Fig. 5 (a) is the current waveform figure that load current mutates in t=0.1s.From Fig. 5 (a) it can be seen that load electricity
The load current waveform of fitting in t=0.1s, is become original 1.3 times, current amplitude is by pervious by the variation of stream
23.13A increases to 30.07A, is equivalent to and new RL loads are increased to rectifier bridge, then new load current iLFormula be：
iL(t)=1.3 × [23.13sin (100 π t)+4.76sin (500 π t+3.056)+3.13sin (700 π t-
3.044)+2.07sin(1100πt-0.004)+1.73sin(1300πt+0.078)
+1.33sin(1700πt+3.154)+1.18sin(1900πt+3.21)+0.99sin(2300πt-0.091)]
Current tracking compares oscillogram when Fig. 5 (b) is load sudden change, from Fig. 5 (b) it can be seen that in current break, with
Track waveform can trace into mutation current at once again after half period.This, which turns out the method for the present invention, very strong robust
Property, it is capable of the variation of quick follow current, in real time, accurately detects fundamental active current.
Experiment and it is above theoretical analysis shows that, the present invention is based on the adaptive harmonic detectings of L2 norms and real tracking error
Method can improve the adaptive Harmonic Detecting Algorithm of traditional fixed step size cannot take into account convergence rate and the deficiency of stable state accuracy simultaneously,
Fundamental active current can be precisely tracked while convergence rate is shortened to half period again.And it mutates in load current
When can also quick follow current variation, in real time, accurately detect fundamental active current.
Unaccomplished matter of the present invention is known technology.
Claims (2)
1. a kind of variable-step self-adaptive harmonic detecting method based on L2 norms and real tracking error, this method includes following step
Suddenly：
(1) it samples
To periodical nonsinusoidal load current iL(t) it is sampled, obtains the centrifugal pump i of load currentL(n), while make power grid electric
Press uS(t) fundamental wave reference input unit vector x is obtained by phaselocked loop1(t)=sin (ω t), x2(t)=cos (ω t), and it is right
Fundamental wave reference input unit vector is sampled to obtain the centrifugal pump x of fundamental wave reference input unit vector1(n)=sin (ω nTS)、
x2(n)=cos (ω nTS)；Wherein, ω=2 π f, f is frequency, and f=50Hz, n are the discrete point of time t, TSFor the sampling period；
(2) by the centrifugal pump x of fundamental wave reference input unit vector1(n)、x2(n) i ' is inputted1p(n)=WT(n) it in X (n), obtains defeated
Go out signal i '1p(n), i '1p(n) be fundamental active current estimated value；
Wherein, X (n)=[x1(n),x2(n)]T；W (n) is the corresponding weight vectors of X (n), W (n)=[W1(n),W2(n)]T,
Initial value is zero；
(3) with the centrifugal pump i of the load current in step (1)L(n) estimation of fundamental active current obtained in step (2) is subtracted
Value i '1p(n), obtain to compensate the estimated value i ' of electric currentd(n), i.e. error signal e (n)；Then
i′d(n)=e (n)=iL(n)-i′1p(n)；
(4) real tracking error is found using running integral device：
The error signal obtained in step (3) is carried out according to discretization Fourier expansion, then is had：
In formula：
By the error signal e (n) obtained in step (3) and synchronized lockin signal sin ω nTSIt is multiplied, and to the result of multiplication
The sampled data of middle a cycle is integrated, i.e.,：
E is the average real tracking error of each sampled point in one cycle；N is the number of samples in a grid cycle T；
N0It is newest sampled point；W is the average weight of each sampled point in a cycle；
(5) according to obtaining in the real tracking error E (n) of n-th of sampled point and step (2) fundamental wave in step (4) with reference to defeated
Enter the centrifugal pump X (n) of unit vector, the value of parameter P (n) is calculated, i.e.,：
P (n)=E (n) × X (n)
(6) the parameter P (n) obtained to step (5) with formula P (n)=a × P (n-1)+b × E (n) × X (n), is updated,
Wherein a and b be smoothing factor, 0<a<1,0<b<1；
(7) set step iteration formula as μ (n)=β (1-exp (and-α | E (n) | × | | P (n) | |2)), it is obtained after step (6) is updated
P (n) L2 norms, i.e., | | P (n) | |2It is brought into the step iteration formula of μ (n)；Wherein, α and β is parameter；
The condition of convergence of step iteration formula is 0<μ(n)<λmax, 0<β<1/λmax；λmaxIt is reference input unit vector auto-correlation
The maximum eigenvalue of matrix；
(8) according to the product of the centrifugal pump X (n) of real tracking error E (n), step value μ (n) and reference input unit vector not
The disconnected variation for adjusting weighted vector W (n), i.e.,：
W (n+1)=+ 2 μ (n) E (n) X (n) of W (n)；
(9) formula in iterative step (6), step (7), step (8), finally so that the best weight coefficient W of W (n) infinite approachsopt；
At this moment output signal i '1p(n) fundamental current active component i can be approached in amplitude and phase1p(n), so as to fulfill harmonic wave electricity
The detection of stream and fundamental reactive current.
2. the variable-step self-adaptive harmonic detecting method according to claim 1 based on L2 norms and real tracking error,
It is characterized in that the α=800, β=0.14, a=0.96, b=0.3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810112763.3A CN108268856B (en) | 2018-02-05 | 2018-02-05 | Variable step size self-adaptive harmonic detection method based on L2 norm and real tracking error |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810112763.3A CN108268856B (en) | 2018-02-05 | 2018-02-05 | Variable step size self-adaptive harmonic detection method based on L2 norm and real tracking error |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108268856A true CN108268856A (en) | 2018-07-10 |
| CN108268856B CN108268856B (en) | 2020-01-21 |
Family
ID=62773736
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810112763.3A Expired - Fee Related CN108268856B (en) | 2018-02-05 | 2018-02-05 | Variable step size self-adaptive harmonic detection method based on L2 norm and real tracking error |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108268856B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109490625A (en) * | 2018-11-14 | 2019-03-19 | 国网江苏省电力有限公司徐州供电分公司 | A kind of harmonic signal analysis method based on sliding window and Semidefinite Programming |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8994363B2 (en) * | 2008-11-07 | 2015-03-31 | Vestas Wind Systems A/S | Grid monitoring system and related method |
| CN105137174A (en) * | 2015-07-24 | 2015-12-09 | 南京理工大学 | Variable-step LMS (Least Mean Square) adaptive harmonic detection method applied to APF (Active Power Filter) |
-
2018
- 2018-02-05 CN CN201810112763.3A patent/CN108268856B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8994363B2 (en) * | 2008-11-07 | 2015-03-31 | Vestas Wind Systems A/S | Grid monitoring system and related method |
| CN105137174A (en) * | 2015-07-24 | 2015-12-09 | 南京理工大学 | Variable-step LMS (Least Mean Square) adaptive harmonic detection method applied to APF (Active Power Filter) |
Non-Patent Citations (2)
| Title |
|---|
| 何娜 等: ""一种新型快速自适应谐波检测算法"", 《中国电机工程学报》 * |
| 还秋云 等: ""引用范数的双曲正切函数变步长LMS算法"", 《信号处理》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109490625A (en) * | 2018-11-14 | 2019-03-19 | 国网江苏省电力有限公司徐州供电分公司 | A kind of harmonic signal analysis method based on sliding window and Semidefinite Programming |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108268856B (en) | 2020-01-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107171584B (en) | The model predictive control method and device of NPC three-phase tri-level gird-connected inverter | |
| CN107102189B (en) | Variable step- size LMS harmonic current detecting method based on S function | |
| CN103472282B (en) | A kind of FBD harmonic current detecting method based on adaptive principle | |
| CN103401238B (en) | A kind of power load modelling approach based on Measurement-based approach | |
| CN104535890A (en) | Rapid network voltage drop detection method | |
| CN111239491B (en) | Generalized impedance real-time experimental measurement method adopting physical controller disturbance injection | |
| CN107294527A (en) | Synchronous rotating frame phaselocked loop and its method of testing, device | |
| CN109188078A (en) | A kind of variable step- size LMS adaptive harmonic current detection method | |
| CN103308747B (en) | A kind of weighting LMS harmonic current detecting method | |
| CN108155643A (en) | A kind of robust estimation method of the single-phase mains voltage parameter based on sliding mode observer | |
| CN103323651B (en) | Based on the variable step affine projection harmonic current detecting method that time coherence is average | |
| Karimi-Ghartemani et al. | Extraction of signals for harmonics, reactive current and network-unbalance compensation | |
| CN103424621A (en) | Artificial neural network detecting method of harmonic current | |
| CN108268856A (en) | Variable-step self-adaptive harmonic detecting method based on L2 norms and real tracking error | |
| Fang et al. | Response time of reactive power based on different definitions and algorithms | |
| CN110365038A (en) | Control method, control device and the microgrid inverter of microgrid inverter | |
| Che et al. | Random harmonic detection and compensation based on synchronous reference frame | |
| Chen et al. | Research on improved harmonic detection method based on instantaneous reactive power theory | |
| Zhou et al. | Improvement Research on Harmonic DetectingBased on Instantaneous Reactive Power Theory | |
| Chen et al. | A algorithm of VMD for the detection of APF harmonics | |
| Peng et al. | Improved harmonic detection algorithm applied to APF | |
| Hu | Optimal Power Quality Control Model of Magnetic Latching Relay for Charging Station Based on Adaline | |
| Zhao et al. | Phase delay solution based on instantaneous reactive power theory | |
| JenoPaul et al. | Design and simulation of phase locked loop controller based three phase unified power quality conditioner for nonlinear and voltage sensitive loads | |
| CN106033477B (en) | Equivalent modeling method for comprehensive load |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200121 Termination date: 20220205 |