CN107154783B - The method for detecting photovoltaic system failure electric arc using independent component analysis and S-transformation - Google Patents
The method for detecting photovoltaic system failure electric arc using independent component analysis and S-transformation Download PDFInfo
- Publication number
- CN107154783B CN107154783B CN201710254713.4A CN201710254713A CN107154783B CN 107154783 B CN107154783 B CN 107154783B CN 201710254713 A CN201710254713 A CN 201710254713A CN 107154783 B CN107154783 B CN 107154783B
- Authority
- CN
- China
- Prior art keywords
- photovoltaic system
- electric arc
- fault electric
- analysis
- 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
Links
- 238000010891 electric arc Methods 0.000 title claims abstract description 265
- 238000000034 method Methods 0.000 title claims abstract description 91
- 238000012880 independent component analysis Methods 0.000 title claims abstract description 42
- 239000011159 matrix material Substances 0.000 claims abstract description 54
- 230000008878 coupling Effects 0.000 claims abstract description 33
- 238000010168 coupling process Methods 0.000 claims abstract description 33
- 238000005859 coupling reaction Methods 0.000 claims abstract description 33
- 230000009897 systematic effect Effects 0.000 claims abstract description 33
- 238000012545 processing Methods 0.000 claims abstract description 28
- 230000009466 transformation Effects 0.000 claims abstract description 7
- 238000004458 analytical method Methods 0.000 claims description 136
- 238000001514 detection method Methods 0.000 claims description 67
- 238000012937 correction Methods 0.000 claims description 20
- 230000008859 change Effects 0.000 claims description 17
- 230000008569 process Effects 0.000 claims description 15
- 238000005070 sampling Methods 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 12
- 235000013399 edible fruits Nutrition 0.000 claims description 7
- 230000002087 whitening effect Effects 0.000 claims description 7
- 238000012790 confirmation Methods 0.000 claims description 6
- 238000001228 spectrum Methods 0.000 claims description 6
- 230000001186 cumulative effect Effects 0.000 claims description 5
- 230000035945 sensitivity Effects 0.000 claims description 4
- 230000004069 differentiation Effects 0.000 claims description 2
- 108090000623 proteins and genes Proteins 0.000 claims description 2
- 239000000470 constituent Substances 0.000 claims 1
- 238000009412 basement excavation Methods 0.000 abstract 1
- 238000011897 real-time detection Methods 0.000 abstract 1
- 230000005611 electricity Effects 0.000 description 16
- 230000001052 transient effect Effects 0.000 description 8
- 238000010276 construction Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 7
- 230000007257 malfunction Effects 0.000 description 6
- 239000007787 solid Substances 0.000 description 6
- 230000000007 visual effect Effects 0.000 description 6
- 238000007689 inspection Methods 0.000 description 4
- 230000009467 reduction Effects 0.000 description 4
- 230000026676 system process Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 230000006378 damage Effects 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 230000003595 spectral effect Effects 0.000 description 3
- 208000032368 Device malfunction Diseases 0.000 description 2
- 230000009471 action Effects 0.000 description 2
- 210000001367 artery Anatomy 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000000087 stabilizing effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 210000003462 vein Anatomy 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000003556 assay Methods 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004907 flux Effects 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 235000021384 green leafy vegetables Nutrition 0.000 description 1
- 238000009413 insulation Methods 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 230000001172 regenerating effect Effects 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000010792 warming Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02S—GENERATION OF ELECTRIC POWER BY CONVERSION OF INFRARED RADIATION, VISIBLE LIGHT OR ULTRAVIOLET LIGHT, e.g. USING PHOTOVOLTAIC [PV] MODULES
- H02S50/00—Monitoring or testing of PV systems, e.g. load balancing or fault identification
- H02S50/10—Testing of PV devices, e.g. of PV modules or single PV cells
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/50—Photovoltaic [PV] energy
Abstract
The invention discloses a kind of methods detecting photovoltaic system failure electric arc using independent component analysis and S-transformation, based on photovoltaic system output current signal, the main source signal of independence of electric current is obtained by independent component analysis, variance processing is made to the frequency information after the signal Fourier transformation and obtains fisrt feature amount, electric current is handled by S-transformation, second feature amount is formed after the time and high fdrequency component integral to gained time-frequency matrix.The corresponding given threshold of characterizing magnitudes relatively after, reuse the output judgement of two characterizing magnitudes on weight coefficient weighted decision layer as a result, completing the real-time detection of photovoltaic system fault electric arc.The present invention is compared by dynamic threshold, weight coefficient weights two result of decision, apparent excavation betides the essential difference of photovoltaic system fault electric arc among systematic procedure, the photovoltaic system fault electric arc under coupling condition can be quickly and accurately cut off, the safe and stable operation ability of photovoltaic system is improved.
Description
Technical field
The invention belongs to the electrical fault detection technique fields of photovoltaic, and in particular to a kind of application independent component analysis and S become
It gets in return to two characterizing magnitudes, characterizing magnitudes is obtained into two result of decision compared with corresponding dynamic given threshold, use dynamic
The weight coefficient of setting weights the result of decision of two characterizing magnitudes, carries out photovoltaic system fault electric arc and detects in real time, hence it is evident that excavates
The essential difference of photovoltaic system fault electric arc among systematic procedure is betided, photovoltaic system fault electric arc under coupling condition is promoted and examines
The rapidity and reliability of survey, to ensure no matter when photovoltaic system can be stablized, safety, the method for economic output operation.
Background technology
The problems such as global energy crisis and climate warming, is increasingly serious so that the novel greens such as photovoltaic, wind-force, fuel cell
Regenerative resource has been more and more widely used.In recent years, with the continuous reduction of photovoltaic products cost, photovoltaic both domestic and external
Industry grows at top speed.The expansion of photovoltaic generating system scale improves photovoltaic system DC terminal output voltage, generally from tens volts
To several hectovolts, large-scale photo-voltaic power generation station can even reach the high direct voltage of kilovolt, when major photovoltaic plant puts into operation
Between extension also increase insulation ag(e)ing degree so that photovoltaic system failure occur more and more frequently, photovoltaic system DC side
Fault electric arc be exactly one of them.Once photovoltaic system fault electric arc is generated, due to there is no the zero crossing of AC fault electric arc
And seem more dangerous, and photovoltaic system fault electric arc safeguard measure cannot be such as taken in time, it will be to photovoltaic module and power transmission line
Road causes huge damage even to cause fire, leads to the safety problems such as serious economic loss and casualties.Occur earliest
Photovoltaic system fault electric arc can trace back to Sweden's Mont Soleil photovoltaic plants of last century the nineties, especially from
Over 2006, for photovoltaic system fire incident more and more by media report, fire spot includes house photovoltaic facility, commercialization
Photovoltaic facility and large-scale photovoltaic power station.The photovoltaic system fire incident for betiding 2006 is due to being connect in BP solar energy
Caused by photovoltaic system fault electric arc occurs in wire box, BP companies make due to most defective photovoltaic modulies also are substituted because recalling
At huge economic loss.Therefore, fully and effectively photovoltaic system fault electric arc examinations are controlled in rudiment
State, to ensureing that the safe and reliable operation of photovoltaic generating system is of great significance.
Currently, what correlative study object was directed to both at home and abroad is non-coupling photovoltaic system fault electric arc, i.e., in photovoltaic system
It unites when fault electric arc occurs and systematic procedure is not present.However, in actually detected, photovoltaic system frequently experiencings derived from photovoltaic battle array
The transient processes such as changed power, the startup of side or load-side are arranged, so that photovoltaic system electricity experiencings frequent variation temporarily
State.The time of origin of photovoltaic system fault electric arc is uncontrollable, thus photovoltaic system fault electric arc also has certain probability meeting
It is happened among these systematic procedures.For example, in the systematic procedures such as photovoltaic system startup, increased system power, photovoltaic system
System output current constantly increases, and on the other hand, tandem photovoltaic system failure electric arc can then reduce photovoltaic system output current.Cause
This, in the photovoltaic system fault electric arc of systematic procedure coupling, photovoltaic system failure output current macroscopically can't be with
The normal output current of photovoltaic system generates difference, and the requirement to photovoltaic system fault electric arc detection algorithm is more harsh.Existing inspection
If method of determining and calculating can not detect in time from the visual angle for being appropriately determined the normal output current of photovoltaic system, photovoltaic system fault electric arc,
There is tripping in corresponding photovoltaic system dc side fault arc detection device, fails the photovoltaic system fault electric arc of row elimination and can cause
Photovoltaic system fire incident brings life and property loss；If from the visual angle for being appropriately determined photovoltaic system failure output current, light
Volt system normal operation will be judged by accident, and malfunction occurs in corresponding photovoltaic system dc side fault arc detection device, these are wrong
The normal condition sentenced can cause photovoltaic system to stop transport and reduce system generating efficiency.Therefore, detection algorithm must extraction system mistake
Photovoltaic system fault electric arc basic feature under journey coupling accurately identifies photovoltaic system fault electric arc and the moment occurs, to light
Volt system output current state is accurate, it is reliable, rapidly recognize, be achieved in installation photovoltaic system DC side fault electric arc detection
The functional requirement of device.
Invention content
The photovoltaic system failure electricity among systematic procedure is betided it is an object of the invention to accurate, reliable, Fast Identification
Arc provides a kind of method detecting photovoltaic system failure electric arc using independent component analysis and S-transformation.
In order to achieve the above objectives, present invention employs following technical schemes：
1) sample frequency f is pressed to photovoltaic system output current signal by multiple current sensorssIt is sampled, is obtained point by point
To multichannel current signal xi,j, wherein i is that current sensor indicates that serial number, i ∈ N and i > 1, j are the analysis period to indicate serial number, j
∈N+, for arbitrary two differences i values, when j takes same value, xi,jEqual sampling number N is all had, when N reaches analysis
After the requirement of section, goes to step 2) and carry out Fisrt fault arc characteristic analysis；
2) collected multichannel current signal is formed into higher-dimension mixed signal matrix X=[x1,j,x2,j,…,xi,j]T, to institute
It obtains mixed signal matrix and carries out mean value and whitening processing, then matrix W is mixed by the way that solution can be obtained after fast independent component analysis,
Calculate source signal matrix S=WX=[s1,j,s2,j,…,si,j]T, wherein S includes effective source signal and noise signal, and selection has
The independent main source signal s of effect1,j, to s1,jFast Fourier Transform (FFT) is carried out, the variance of one-dimensional frequency matrix in frequency domain is calculated, obtains the
One characterizing magnitudes r1,j, go to step 3)；
3) fisrt feature amount threshold value is set in the present analysis period as A1×μ1,j–A2×σ1,j, wherein μ1,jFor from first point
Analyse the Estimation of Mean of period to present analysis period all fisrt feature magnitudes, σ1,jFor from the first analysis period to present analysis
The standard deviation of period all fisrt feature magnitudes, A1∈ Z, A2∈ Z, by the fisrt feature amount threshold value of fisrt feature magnitude and setting
Compare, exports corresponding electrical level judging result：If r1,j≥A1×μ1,j–A2×σ1,j, then judgement result 0, deposit to first are exported
Fault electric arc trip current out1[j]；If r1,j<A1×μ1,j–A2×σ1,j, then judgement result 1, deposit to Fisrt fault electricity are exported
Arc trip current out1[j] goes to step 4) and carries out the second fault electric arc signature analysis；
4) it selects the signal all the way in multichannel current signal to carry out S-transformation, obtains the two-dimensional complex number time-frequency square in time-frequency domain
Battle array, calculates integral of the high fdrequency component absolute value along the time of frequency dimension, obtains second feature magnitude r2,j, go to step 5)；
5) second feature amount threshold value is set in the present analysis period as A3×μ2,j–A4×σ2,j, wherein μ2,jFor from first point
Analyse the Estimation of Mean of period to present analysis period all second feature magnitudes, σ2,jFor from the first analysis period to present analysis
The standard deviation of period all second feature magnitudes, A3∈ Z, A4∈ Z, by the second feature amount threshold value of second feature magnitude and setting
Compare, exports corresponding electrical level judging result：If r2,j≥A3×μ2,j–A4×σ2,j, then judgement result 0, deposit to second are exported
Fault electric arc trip current out2[j]=0；If r2,j<A3×μ2,j–A4×σ2,j, then judgement result 1, deposit to the second event are exported
Hinder electric arc trip current out2[j]=1 is gone at the output judgement result weighting that step 6) carries out in two characteristic quantity decision-making levels
Reason；
6) output that independent component analysis and S-transformation are weighted using Dynamic Weights coefficient is judged as a result, obtaining weighted results
outtempj=C1,j×out1[j]+C2,j×out2[j] then carries out preliminary state judgement：If outtempj>N, wherein n is
Weighted results threshold value then exports judgement result 1, deposit to preliminary state determination results matrix outt [j]；Otherwise output judgement knot
Fruit 0, deposit to preliminary state determination results matrix outt [j] go to step 7) and carry out photovoltaic system state differentiation；
7) setting judges precision p, judges a photovoltaic system state per p period：Count preliminary state determination results square
The number that battle array outt is 1 to j-th of element from -3p elements of jth to-p elements of jth, from -2p elements of jth, if being counted
The numerical value of number is all higher than p, then confirmation is a to generation photovoltaic system fault electric arc in the-p periods of jth in jth -2p, takes phase
The photovoltaic system fault electric arc safeguard measure answered；Otherwise it is assumed that photovoltaic system is in just in jth -2p to the-p periods of jth
Normal operating status, return to step 1) current signal in next analysis period is analyzed.
The current sensor is not required for same type, but its bandwidth should be greater than 100kHz, should be installed in photovoltaic system
The different location of system is to show the difference between sampled current signals, and considering accurate acquisition electric current, independently reduction is hard simultaneously for main source signal
The value range of the principle of part testing cost, current sensor is 2~4；Sample frequency fsIt should be greater than twice of photovoltaic system
The fault electric arc characteristic spectra upper limit, value range are 200~500kHz；Consider quickly and accurately to obtain photovoltaic system failure electricity
The value range of the principle of arc feature, sampling number N is 8000~12000.
(being based on negentropy maximization) in the fast independent component analysis, parameters foundation quickly obtains betiding system
Depending on photovoltaic system fault electric arc notable feature in the process, g can be selected in nonlinear function1(u)=u3、g2(u)=u2、g3(u)
=arctan (q1×u)、g4(u)=u × e^ (- q2 2×u2/ 2), wherein q1And q2For constant, preferably g1(u)=u3, maximum
The value range of iterations is 950~1050, and the value range of iteration precision is 0.00006~0.00015.
The main source signal number of the independence i.e. way of sampled current signals that the fast independent component analysis obtains, based on letter
The strongest principle of number impact selects an effective independent main source signal to carry out follow-up Fast Fourier Transform (FFT) processing：It calculates each
The difference of peak-to-peak value of the independent main source signal within the analysis period, it is effective independent master that select difference, which be the main source signal of maximum independence,
Source signal；The negative of photovoltaic system fault electric arc under fisrt feature amount detection coupling condition is reduced based on spectral leakage is reduced as far as possible
Face is rung, and the numerical value that points are converted in Fast Fourier Transform (FFT) is chosen to be the corresponding numerical value of sampling number N.
Farthest to find that the photovoltaic system fault electric arc time-frequency difference in systematic procedure is principle from multichannel electric current
The input all the way as S-transformation is chosen in signal：The corresponding current signal of the highest current sensor of sensitivity is preferably selected, when
When this kind of current sensor more than one, preferably selects and the nearest current sensor in position occurs apart from photovoltaic system fault electric arc
Corresponding current signal, when the nearest current sensor more than one in position occurs apart from photovoltaic system fault electric arc, preferably
Select the current sensor with minimum number of components in photovoltaic system fault electric arc to current sensor propagation path corresponding
Current signal；Based on identical principle, window width Dynamic gene is preferably 1 in the S-transformation.
Absolute value processing is made to gained two-dimensional complex number time-frequency matrix element after S-transformation, the frequency dimension based on the time-frequency matrix
The principle of degree component structure second feature amount be presented significant downward trend when photovoltaic system fault electric arc occurs, and with compared with
Small amplitude form shows the difference of photovoltaic system fault electric arc and systematic procedure before, photovoltaic system fault electric arc characteristic spectra
Be selected as 40~100kHz and with sample frequency fsValue it is uncorrelated.
The fisrt feature amount threshold value A1×μ1,j–A2×σ1,jIt is related with all analysis fisrt feature magnitudes of period before
And fisrt feature amount r is followed in real time1Dynamic change, wherein coefficient A1With A2It is related to fisrt feature amount output characteristics, according to logical
Crossing the fisrt feature amount threshold value of setting can be correctly obtained compared with fisrt feature magnitude depending on corresponding photovoltaic system state, A1With
A2Preferably 1；Estimation of Mean μ1,jAnd standard deviation sigma1,jOutput judgement result according to fisrt feature amount is corrected in real time：For
The fisrt feature magnitude r that first analysis period obtains1,1, enable correction amount rtemp1,1=r1,1, Estimation of Mean μ1,1=r1,1, mark
Quasi- difference σ1,1=0；For the fisrt feature magnitude r of j-th of analysis period1,j, wherein j ∈ N and j > 1, if the present analysis period
When interior fisrt feature magnitude is more than or equal to upper analysis period fisrt feature amount threshold value, correction amount rtemp is enabled1,j=r1,j, mean value
Estimate and the calculation formula of standard deviation is
Wherein, k is analysis period expression serial number, k=1,2 ... j, j ∈ N and j > 1 in cumulative process, if when present analysis
When fisrt feature magnitude is less than upper analysis period fisrt feature amount threshold value in section, correction amount rtemp is enabled1,j=μ1,j-1–σ1,j-1,
The calculation formula of Estimation of Mean and standard deviation is
The second feature amount threshold value A3×μ2,j–A4×σ2,jIt is related with all analysis second feature magnitudes of period before
And second feature amount r is followed in real time2Dynamic change, wherein coefficient A3With A4It is related to second feature amount output characteristics, according to logical
Crossing the second feature amount threshold value of setting can be correctly obtained compared with second feature magnitude depending on corresponding photovoltaic system state, A3With
A4Preferably 1；Estimation of Mean μ2,jAnd standard deviation sigma2,jOutput judgement result according to second feature amount is corrected in real time：For
The second feature magnitude r that first analysis period obtains2,1, enable correction amount rtemp2,1=r2,1, Estimation of Mean μ2,1=r2,1, mark
Quasi- difference σ2,1=0；For the second feature magnitude r of j-th of analysis period2,j, wherein j ∈ N and j > 1, if the present analysis period
When interior second feature magnitude is more than or equal to upper analysis period second feature amount threshold value, correction amount rtemp is enabled2,j=r2,j, mean value
Estimate and the calculation formula of standard deviation is
Wherein, k is analysis period expression serial number, k=1,2 ... j, j ∈ N and j > 1 in cumulative process, if when present analysis
When second feature magnitude is less than upper analysis period second feature amount threshold value in section, correction amount rtemp is enabled2,j=μ2,j-1–σ2,j-1,
The calculation formula of Estimation of Mean and standard deviation is
Based on the quick principle for calculating each given threshold in the present analysis period, when obtaining present analysis using recurrence relation
The calculation formula of Estimation of Mean and standard deviation is in section
Wherein, μm,j、σm,jEstimation of Mean and standard deviation respectively in the present analysis period, μm,j-1、σm,j-1Before respectively
Estimation of Mean and standard deviation in one analysis period, rtempm,jFor the correction amount in the present analysis period, wherein the m amounts of being characterized
Indicate that serial number, value are 1 or 2, j ∈ N and j > 1.
Output judgement after weighting two characterizing magnitudes and given threshold relatively using Dynamic Weights coefficient is as a result, corresponding each spy
The weight coefficient of sign amount output judgement result judges that the statistical property of correctness is true according to characteristic quantity to historical analysis period state
It is fixed, i.e., characteristic quantity the historical analysis period is made correct status judgement the analysis period it is more, this feature amount is in present analysis
The weight coefficient that section is obtained is then bigger, specifically, constructs fisrt feature amount and second feature amount institute respectively based on following formula
Belong to weight coefficient C1,jAnd C2,j：
Wherein, σ2 out1And σ2 out2Respectively Fisrt fault electric arc trip current and the second fault electric arc trip current is from
One element to j-th of element variance, i.e.,
Wherein, out1And out2Respectively Fisrt fault electric arc trip current and the second fault electric arc trip current, k is square
The counting serial number of array element element, k=1,2 ... j, j ∈ N and j > 1,WithRespectively Fisrt fault electric arc trip current and
Estimation of Mean of the second fault electric arc trip current from first element to j-th of element；If Fisrt fault electric arc trip current,
Second fault electric arc trip current is 0 from first element to j-th of element, i.e. two characteristic quantities judge all analysis periods
For normal operating condition, indirect assignment C1,j=0, C2,j=0, then be weighted two characteristic quantities and carry out follow-up photovoltaic system failure electricity
Arc detects；If jth -2p be in normal operating condition to photovoltaic system in the-p periods of jth, former under this p period first
Make element exchange processing in the position of barrier electric arc trip current, the second fault electric arc trip current respective element not etc..
Based on the principle of the photovoltaic system fault electric arc accurately identified in systematic procedure, the weighted results threshold value n's takes
Value ranging from 0.45~0.55；The principle that reliability and quick-action are detected based on photovoltaic system fault electric arc, avoids too small p value from drawing
The photovoltaic system fault electric arc that the systematic procedure nonaction of hair and excessive p value cause acts phenomenon, the judgement essence not in time
The value range for spending p is 2~5.
The present invention has following beneficial technique effect：
1) the method increase photovoltaic systems to the recognition capability of electric current normal state, solves and is exported with photovoltaic system failure
The photovoltaic system DC side fault electric arc inspection that electric current visual angle detection algorithm is generated in face of transient processes such as system power variation, startups
It surveys error action of device problem and substantially extends the normal of photovoltaic system by the way that systematic procedure is correctly determined as normal operating condition
Run time significantly improves the generating efficiency of photovoltaic system, enhances the stabilizing power of photovoltaic system normal operation；
2) this method can accurately catch the photovoltaic system fault electric arc basic feature among betiding systematic procedure, solve
The photovoltaic system fault electric arc production occurred in face of and then systematic procedure with the normal output current visual angle detection algorithm of photovoltaic system
Raw photovoltaic system DC side fault arc detection device tripping problem, by correctly by the photovoltaic system failure under coupling condition
Electric arc is determined as malfunction, has ensured the validity of photovoltaic system fault electric arc detection, eliminates this kind of photovoltaic system event in time
Hinder the harm such as photovoltaic fire incident, the life and property loss that electric arc causes, expands current photovoltaic system fault electric arc detection side
The scope of application of method；
3) this method has wide photovoltaic system fault electric arc inspection range, not photovoltaic system failure under by coupling condition
Electric arc variation tendency direction caused by photovoltaic system output current influences, and the moment occurs in photovoltaic system fault electric arc no matter
Photovoltaic system output current becomes larger or becomes smaller, still remains unchanged, and detection algorithm can reliable, accurately checkout system mistake
Photovoltaic system fault electric arc in journey；
4) this method has very strong quick-action, detect single analysis period of photovoltaic system fault electric arc for 40~
60ms, setting precision are 2~5, that is, it is 0.3s to detect photovoltaic system fault electric arc longest and spend the time, and most bob is out of order electric arc
Circuit cut-out control signal the time it takes is 0.08s, the reliable judgement for detecting photovoltaic system fault electric arc under coupling condition
Duration is much smaller than the 2s standards of existing American Standard UL1699B defineds；
5) this method utilizes the Estimation of Mean and standard deviation construction feature amount threshold value of characteristic quantity, uses characterizing magnitudes and threshold value
Comparison procedure realizes the normalization of each characteristic quantity output, solves different characteristic amount the number of output grade difference and is detected to multi-characteristicquantity quantity
The interference of photovoltaic system fault electric arc is also beneficial to realize subsequent multi-characteristicquantity quantity decision-making level weighting, and threshold value and weight coefficient exist
Setting up for dynamic change processing and photovoltaic system fault electric arc standard is carried out in different analytical cycle, is conducive to detect
What algorithm more reliably provided system mode within each analysis period is appropriately determined result；
6) requirement of this method to the detection hardware change of existing photovoltaic system fault electric arc be not high, it is only necessary in original light
It is rationally laid with required current sensor in volt system, is added in original photovoltaic system DC side fault arc detection device
Signal input port is detected, then the software program of photovoltaic system DC side fault arc detection device need to only calculate two methods
Under characteristic quantity, carry out dynamic threshold setting, dynamic weighting coefficient calculates, the final weighting for realizing two characteristic quantities in decision-making level is real
Existing photovoltaic system fault electric arc detection, Programming Principle is simple, and cost of implementation is cheap.
Further, when assert that photovoltaic system fault electric arc occurs, Estimation of Mean and standard deviation in being set to threshold value
Calculating need to be modified, avoid the fluctuation that threshold value is caused since Feature change is larger；When identification photovoltaic system is normal
It runs and two characteristic quantities exports whens judging that result is not equal, to two fault electric arc trip currents, accordingly not equal elements make exchange processing,
It realizes correct weight coefficient dynamic change, eliminates the normal operation malfunction problem caused by external interference, effectively increase
The reliability of photovoltaic system fault electric arc detection, increases the economic benefit of photovoltaic system operation.
Description of the drawings
Fig. 1 a are the photovoltaic system fault arc detection method flow chart of the present invention；
Fig. 1 b are dynamic threshold setting process figure in the photovoltaic system fault arc detection method of the present invention；
Fig. 2 is the present invention in including the specific of the photovoltaic system DC side fault arc detection device for being integrated in bus bar
Functional block diagram when photovoltaic system application hardware is realized；
Fig. 3 a are to carry out the pointer in fault electric arc that photovoltaic system fault electric arc detects under coupling condition using the present invention
The photovoltaic system output current signal of standby constant trend；
Fig. 3 b be using independent component analysis carry out coupling condition under photovoltaic system fault electric arc detect characteristic quantity and its
Given threshold waveform；
Fig. 3 c are to carry out the characteristic quantity and its given threshold that photovoltaic system fault electric arc detects under coupling condition using S-transformation
Waveform；
Fig. 3 d are to carry out the system mode that photovoltaic system fault electric arc detects under coupling condition using the present invention to judge output
Signal；
Fig. 4 a are to carry out the pointer in fault electric arc that photovoltaic system fault electric arc detects under coupling condition using the present invention
The photovoltaic system output current signal of standby reduction trend；
Fig. 4 b be using independent component analysis carry out coupling condition under photovoltaic system fault electric arc detect characteristic quantity and its
Given threshold waveform；
Fig. 4 c are to carry out the characteristic quantity and its given threshold that photovoltaic system fault electric arc detects under coupling condition using S-transformation
Waveform；
Fig. 4 d are to carry out the system mode that photovoltaic system fault electric arc detects under coupling condition using the present invention to judge output
Signal；
Fig. 5 a are to carry out the pointer in fault electric arc that photovoltaic system fault electric arc detects under coupling condition using the present invention
The photovoltaic system output current signal of standby increase tendency；
Fig. 5 b be using independent component analysis carry out coupling condition under photovoltaic system fault electric arc detect characteristic quantity and its
Given threshold waveform；
Fig. 5 c are to carry out the characteristic quantity and its given threshold that photovoltaic system fault electric arc detects under coupling condition using S-transformation
Waveform；
Fig. 5 d are to carry out the system mode that photovoltaic system fault electric arc detects under coupling condition using the present invention to judge output
Signal；
In figure：1, photovoltaic system；2, photovoltaic system DC side fault arc detection device；3, trip gear；4, breaker；
5, it loads；6, current sensor；7, photovoltaic system fault electric arc；8, photovoltaic module.
Specific implementation mode
Explanation is described in detail to the method for the present invention with reference to the accompanying drawings and examples.
In conjunction with Fig. 1 a, in the case of application independent component analysis of the present invention and S-transformation detecting system PROCESS COUPLING
The step of photovoltaic system fault electric arc method, is specifically described.
Step 1: Parameter Initialization procedure includes sample frequency f of the setting electric current sensor to current signals, analysis when
Section in sampling number N, judge precision p, weighted results threshold value n, reset seek Estimation of Mean and standard deviation each variable, independence at
Parameters etc. in analysis and two kinds of fault electric arc signature analysis tools of S-transformation.Current sensor is according to set sampling
Frequency fsParallel sampling is carried out to the multichannel current signal needed for photovoltaic system DC side fault arc detection device, obtains multichannel
Current signal xi(the expression serial number i ∈ N and i > 1 of current sensor), once the sampling number of these current signals reaches N number of,
Just it is input to photovoltaic system DC side fault arc detection device through multiple ports, goes to step 2 extraction photovoltaic system failure electricity
The multi-party region feature of arc.
Current sensor of the present invention is not required for same type, as long as the current sensor bandwidth parameter selected
More than 100kHz, ensure that it can obtain photovoltaic system fault electric arc characteristic spectra.Current sensor should use it is multiple, pass through
The different location of photovoltaic system is installed on to reflect photovoltaic system fault electric arc to photovoltaic system output current at different sampled points
The influence of signal.After carrying out optimization arrangement of the current sensor in photovoltaic system, accurately obtaining electric current, independently main source is believed
Number while, can also reduce the number that current sensor uses as far as possible, thus reduce a whole set of photovoltaic system fault electric arc hardware
Testing cost.The current sensor that the present embodiment uses is 4.
In the photovoltaic system DC side fault arc detection device course of work, with sample frequency fsElectricity is exported to photovoltaic system
Stream signal samples point by point, and excessively high sample frequency can increase the hardware cost of current sensor, and too low sample frequency can not be contained
The photovoltaic system fault electric arc characteristic frequency that lid current signal is reflected.Therefore, the photovoltaic system failure of interest in the present invention
The arc characteristic frequency range upper limit is under the selection of 100kHz, and to reduce the hardware realization requirement of current sensor, the present embodiment determines
Sample frequency fs=200kHz.It is x to remember that i-th of sensor analyzes period collected current signal at j-thI, j(electric current passes
The expression serial number i ∈ N and i > 1 of sensor；Analyze the expression serial number j ∈ N of period+), for arbitrary two differences i values, when j takes together
When one value, xi,jEqual sampling number N is all had, i.e., photovoltaic system fault electric arc detection algorithm proposed by the invention is to more
Road current signal carries out the analysis of equal periods.Sampling number N values cross conference and increase the analysis of photovoltaic system fault electric arc detection algorithm
The time of operation is unfavorable for the Rapid Detection of photovoltaic system fault electric arc, and sampling number N values are too small to be not sufficient to ensure that in system mistake
The detection result of photovoltaic system fault electric arc is accurately detected among journey.Therefore, the present embodiment consideration quickly and accurately obtains photovoltaic
The principle of system failure arc characteristic, determining sampling number N=10000.
Step 2: forming higher-dimension mixed signal matrix X=[x by collected multichannel current signal1,j, x2,j... xi,j
]T, average value processing, i.e. x' are carried out to gained mixed signal matrix1,j=x1,j–E(x1,j), wherein E (x1,j) indicate x1,jIt is equal
Value estimation；Then whitening processing is carried out to gained zero-mean signal, enables E=(e1, e2... en) it is with covariance matrix C=E
(x1,j xT 1,j) unit norm feature vector be row matrix, enable D=diag (e1, e2... en) it is with covariance matrix C
Characteristic value is the diagonal matrix of diagonal element, and V=D can be obtained after linear transformation-1/2ET, the signal through whitening processing is z=after transformation
Vx'1,j.By being based on that the mixed matrix W of solution can be obtained after the maximum fast independent component analysis of negentropy, source signal matrix S=is calculated
WX=[s1,j, s2,j... si,j]T, wherein S includes effective source signal and noise signal, selects effectively independent main source signal s1,j,
Fast Fourier Transform (FFT) is carried out to it, is calculated the variance of one-dimensional frequency matrix in frequency domain, is obtained fisrt feature magnitude r1,j.Selection
Signal x all the way in multichannel current signali,j, S-transformation is carried out to the signal, obtains the two-dimensional complex number matrix distribution in time-frequency domain,
The high fdrequency component absolute value of frequency dimension is calculated along the integral of time, obtains second feature magnitude r2,j, step 3 is gone to corresponding
Given threshold compares the judgement result for obtaining each characteristic quantity within the present analysis period.
The fast independent component analysis that the present embodiment determines is the fast independent component analysis based on negentropy maximization：Pass through
Negentropy maximization algorithm is sought to mix matrix W to a suitable solution, finally obtains the main source signal of each independence of electric current.In order to as early as possible
Seek to suitable solution to mix matrix to accelerate photovoltaic system fault electric arc detection process, the nonlinear function that the present embodiment is selected is
u3, determine terminate iterative process iteration precision value be 0.0001, maximum iteration 1000.Using quick independent element point
The method of analysis analyzes multichannel current signal, obtains the current signal way that the number of independent main source signal is analyzed, meter
The difference for calculating these peak-to-peak values of the independent main source signal within the analysis period, selects the independence corresponding to the difference of maximum peak-to-peak value main
Source signal is effectively independent main source signal.This effective independent main source signal is subjected to Fast Fourier Transform (FFT) processing, excessively
Transformation points can cause the distortion to primary current spectrum analysis, very few transformation points can then cause serious spectral leakage
Phenomenon, these factors are all unfavorable for fisrt feature amount and accurately detect photovoltaic system fault electric arc.Therefore, fast in the present embodiment
The numerical value that points are converted in fast Fourier transformation is chosen to be 10000.After photovoltaic system fault electric arc occurs in systematic procedure, frequency spectrum
Energy transfer so that spectral matrix is more uniformly spread in the present analysis period, thus its variance is sent out in photovoltaic system fault electric arc
There is spike in the raw moment, entirety has smaller amplitude in fault electric arc state compared with normal operating condition, so it has accurately
It was found that hiding the photovoltaic system fault electric arc among systematic procedure, it is chosen to be fisrt feature amount.
Different from independent component analysis, the input of S-transformation only has current signal all the way.Therefore, compare selection multichannel electric current
That road current signal of most effective reflection photovoltaic system fault electric arc difference is only the most suitable as input signal in signal.
Compare the distance of sensor mounting location and photovoltaic system fault electric arc, current sensor sensitivity grade higher as priority
Selective goal, i.e. certain class high sensitivity current sensor more than one when, preferably select apart from photovoltaic system fault electric arc
Input of the nearest corresponding current signal of current sensor in position as S-transformation occurs.The matching optimization of each parameter in S-transformation
It is also based on the feature of photovoltaic system fault electric arc during maximum piece-rate system, more reliably to identify photovoltaic system
System fault electric arc.After S-transformation, current signal becomes the two-dimensional complex number matrix on time-frequency domain, and real part, imaginary part or phase angle are to photovoltaic
The instruction of system failure electric arc is not so good as the effect of absolute value processing.40~100kHz of the absolute value in frequency dimension possess compared with
Good consistency, amplitude integrally has significant falling after the generation of photovoltaic system fault electric arc, and systematic procedure is to this frequency
The influence of section is often weaker, thus betides the photovoltaic system fault electric arc among systematic procedure and have in this frequency range preferably
Separating effect, be chosen to be second feature amount.To promote the reliability of photovoltaic system fault electric arc detection, to this time-frequency conversion
40~100kHz photovoltaic system fault electric arc characteristic spectras under tool carry out integral along time shaft and make overlap-add procedure, this feature frequency
Section and sample frequency fsValue it is unrelated, thus using technical scheme of the present invention when, sample frequency must not be less than 200kHz.
Step 3: after carrying out analyzing processing to current signal by above two method, characteristic layer in each analysis period
Two characterizing magnitudes of upper acquisition, by the comparison of fisrt feature magnitude, second feature magnitude and respective threshold, by each characteristic quantity
Result normalizing is exported to decision-making level.Threshold value carries out dynamic change processing within the different analysis periods, is analyzed in the period when assert
When photovoltaic system fault electric arc occurs, to carrying out threshold value setting again after the calculating amendment of Estimation of Mean and standard deviation, first is obtained
Fault electric arc trip current out1, the second fault electric arc trip current out2, go to step 4 and be weighted processing.
For the magnitude differences for overcoming each characteristic quantity itself to generate, itself flux matched exclusive given threshold of each feature.
Here the step for illustrating by taking fisrt feature magnitude and its given threshold comparison as an example.By fisrt feature magnitude and the first of setting
Characteristic quantity threshold value comparison exports corresponding electrical level judging result：If fisrt feature magnitude is more than the threshold value of setting, judgement is exported
As a result 0, deposit to Fisrt fault electric arc trip current out1[j]；If fisrt feature magnitude is less than the threshold value of setting, output is sentenced
Determine result 1, deposit to Fisrt fault electric arc trip current out1[j].Therefore, threshold value comparison process, which is equivalent to, makees each characteristic quantity
Normalized enables weighting procedure not generate significant amplitude fluctuations.Second fault electric arc trip current out2It can be to the second spy
Sign amount and second feature amount threshold value are obtained using similar approach.
It is fluctuated to adapt to characteristic quantity caused by the normal assay period, the threshold value of setting is estimated with the mean value of individual features amount
Meter is related to standard deviation, that is, sets in the present analysis period fisrt feature amount threshold value as A1×μ1,j–A2×σ1,j, wherein μ1,jFor
From the Estimation of Mean of first analysis period to present analysis period all fisrt feature magnitudes, σ1,jTo analyze the period extremely from first
The standard deviation of present analysis period all fisrt feature magnitudes, A1∈ Z, A2∈Z.Photovoltaic occurs when assert in the present analysis period
When system failure electric arc, the calculating of Estimation of Mean and standard deviation need to be modified, avoid causing since Feature change is larger
The fluctuation of threshold value eliminates the normal operation malfunction problem caused by ambient systems process is interfered, effectively increases coupling
In the case of photovoltaic system fault electric arc detect reliability, increase photovoltaic system operation economic benefit.
In conjunction with Fig. 1 b, by taking fisrt feature amount threshold setting procedure as an example, to being moved in photovoltaic system fault arc detection method
State threshold setting procedure is specifically described.
The fisrt feature amount threshold value A1×μ1,j–A2×σ1,jIt is related with all analysis fisrt feature magnitudes of period before
And fisrt feature amount r is followed in real time1Dynamic change, wherein coefficient A1With A2It is related to fisrt feature amount output characteristics, according to logical
Crossing the fisrt feature amount threshold value of setting can be correctly obtained compared with fisrt feature magnitude depending on corresponding photovoltaic system state, this reality
It applies example and feature is levied based on the fisrt feature scale that independent component analysis is built, determine coefficient A1=1, A2=1, i.e. fisrt feature amount
Threshold value is μ1,j–σ1,j；Estimation of Mean μ1,jAnd standard deviation sigma1,jOutput judgement result according to fisrt feature amount is corrected in real time：
The fisrt feature amount r obtained for first analysis period1,1, enable correction amount rtemp1,1=r1,1, Estimation of Mean μ1,1=r1,1,
Standard deviation sigma1,1=0, the given threshold accordingly exported is μ1,1–σ1,1；For the fisrt feature amount r of j-th of analysis period1,j,
In, j ∈ N and j > 1, if fisrt feature magnitude is more than or equal to upper analysis period fisrt feature amount threshold value in the present analysis period
When, i.e., when the preliminary judgement present analysis period is normal state, enable correction amount rtemp1,j=r1,j, the meter of Estimation of Mean and standard deviation
Calculating formula is
Wherein, k is analysis period expression serial number, k=1,2 ... j, j ∈ N and j > 1 in cumulative process, if when present analysis
When characterizing magnitudes are less than upper analysis period fisrt feature amount threshold value in section, i.e. fault case is presented in the preliminary judgement present analysis period
When, another set of threshold value need to be used to set scheme to ensure the correct judgement of photovoltaic system fault electric arc detection algorithm, enable correction amount
rtemp1,j=μ1,j-1–σ1,j-1, the calculation formula of Estimation of Mean and standard deviation is
Based on the quick principle for calculating each given threshold in the present analysis period, existing Estimation of Mean and standard deviation are utilized
Initial value μ1,1、σ1,1And correction amount rtemp1,jEach analysis period from the second analysis period is calculated by following recurrence relations
Estimation of Mean and standard deviation, and then the given threshold accordingly exported is μ1,j–σ1,j：
Wherein, μ1,j、σ1,jEstimation of Mean and standard deviation respectively in the present analysis period, μ1,j-1、σ1,j-1Before respectively
Estimation of Mean and standard deviation in one analysis period, rtemp1,jFor the correction amount in the present analysis period, j ∈ N and j > 1.
Step 4: the output judgement result to independent component analysis and S-transformation matches corresponding weight coefficient, to currently dividing
The fault electric arc trip current of two characteristic quantities is weighted under the analysis period, obtains weighted results outtempj=C1,j×out1[j]+
C2,j×out2[j] then carries out preliminary state judgement：If outtempj>N, wherein n is weighted results threshold value, then exports judgement
As a result 1, deposit to preliminary state determination results matrix outt [j]；Otherwise output judgement result 0, deposit to preliminary state judge
Matrix of consequence outt [j] goes to step 5 and is made whether to send out the judgement of fault electric arc cut-out control signal.
The weighted results determined based on the principle of the photovoltaic system fault electric arc accurately identified in systematic procedure, the present embodiment
Threshold value n is 0.5.Two characterizing magnitudes are weighted using Dynamic Weights coefficient to the output after threshold value comparison to judge as a result, corresponding each spy
The weight coefficient of sign amount output judgement result judges that the statistical property of correctness is true according to characteristic quantity to historical analysis period state
It is fixed, i.e., characteristic quantity the historical analysis period is made correct status judgement the analysis period it is more, this feature amount is in present analysis
The weight coefficient that section is obtained is then bigger, specifically, constructs fisrt feature amount and second feature amount institute respectively based on following formula
Belong to weight coefficient C1,jAnd C2,j：
Wherein, σ2 out1And σ2 out2Respectively Fisrt fault electric arc trip current and the second fault electric arc trip current is from
One element to j-th of element variance, i.e.,
Wherein, out1And out2Respectively Fisrt fault electric arc trip current and the second fault electric arc trip current, k is square
The counting serial number of array element element, k=1,2 ... j, j ∈ N and j > 1,WithRespectively Fisrt fault electric arc trip current and
Estimation of Mean of the second fault electric arc trip current from first element to j-th of element.
If Fisrt fault electric arc trip current, the second fault electric arc trip current are equal from first element to j-th of element
It is 0, i.e. two characteristic quantities judge that all analysis periods are normal operating condition, in this case finally judgement photovoltaic system operation
State is necessarily normal.To avoid the complex calculation of weighting coefficient, indirect assignment C1,j=0, C2,j=0, then it is weighted two spies
Sign amount carries out follow-up photovoltaic system failure arc-detection.
Step 5: setting judges precision p, a judgement that photovoltaic system fault electric arc distinguishes result is carried out per p period.
Decision principle is accordingly：Count preliminary state determination results matrix outt from jth -3p elements to-p elements of jth, from the
The number that j -2p elements are 1 to j-th of element, if the numerical value of count number is all higher than p, confirmation is a to the in jth -2p
In-p periods photovoltaic system fault electric arc occurs for j, takes corresponding photovoltaic system fault electric arc safeguard measure；Otherwise at this
Think that photovoltaic system is in normal operating condition in section, the current signal in return to step a pair of next analysis period is sampled
And analysis.If jth -2p is in normal operating condition to photovoltaic system in the-p periods of jth, photovoltaic is thought within the period
System is in normal operating condition, under this p period, if Fisrt fault electric arc trip current, the second fault electric arc judge square
Battle array respective element output is that 0/1 or 1/0 combination exchanges two elements of corresponding position whens that is, respective element value does not wait
Processing, to ensure the validity of weighting coefficient in any case.
If setting judges precision, p settings are too small, may result in and judged by accident when systematic procedure occurs in photovoltaic system, phase
There is malfunction in the photovoltaic system DC side fault arc detection device answered, and unnecessary photovoltaic system electricity generation power is caused to lose；
If p settings are excessive, faulty line cannot can be in time cut off after the generation of photovoltaic system fault electric arc, corresponding photovoltaic system is straight
There is tripping in stream side fault arc detection device, causes great economic loss and serious security threat.Based on photovoltaic system
Fault electric arc detects the principle of reliability and quick-action, and the systematic procedure nonaction and excessive p value for avoiding too small p value from causing draw
The photovoltaic system fault electric arc of hair acts phenomenon not in time, and what the present embodiment determined judges precision p for 5.
The method for being applied to practical photovoltaic system to the present invention is illustrated, as shown in Fig. 2, illustrating the method for the present invention in reality
Operation in the photovoltaic system of border.Dc power is exported by the photovoltaic system 1 that photovoltaic module 8 forms, through over-current sensor
6, breaker 4 is input in load 5.
Multichannel photovoltaic system output current signal is input to photovoltaic system DC side failure electricity by multiple current sensors 6
Arc detection device 2 carries out above-mentioned photovoltaic system fault electric arc identification process.Since the generation position of photovoltaic system fault electric arc has
There are randomness, multiple current sensors 6 to need rationally to be arranged in photovoltaic system, reduces photovoltaic system failure electricity to greatest extent
Arc missing inspection blind area.Just it is used as S-transformation defeated all the way this example demonstrates that being chosen from multichannel photovoltaic system output current signal here
The method entered.Assuming that 6A~6D is selected as same type current sensor, it is suitable all the way to the 7A selections of photovoltaic system fault electric arc
Photovoltaic system output current signal, the principle of distance after the first sensor proposed according to the present invention should choose electric current biography
Input of the collected photovoltaic system output current signals of sensor 6A as S-transformation.The 7B selections of photovoltaic system fault electric arc are closed
Suitable photovoltaic system output current signal all the way, it is assumed that the photovoltaic module 8 where current sensor 6C on photovoltaic string is enough, enables
Photovoltaic system fault electric arc 7B occur position be much larger than with current sensor 6C at a distance from its with current sensor 6B away from
From the principle of distance after the first sensor proposed according to the present invention should choose the collected photovoltaics of current sensor 6B
Input of the system output current signal as S-transformation；Assuming that the photovoltaic module 8 where current sensor 6C on photovoltaic string is enough
It is more, it enables photovoltaic system fault electric arc 7B that position occurs and is much larger than itself and current sensor 6C at a distance from current sensor 6B
Distance, the principle of distance, should choose the collected photovoltaics of sensor 6C after the first sensor proposed according to the present invention
Input of the system output current signal as S-transformation；Assuming that the photovoltaic module 8 where current sensor 6C on photovoltaic string makes just
Photovoltaic system fault electric arc 7B occur position with current sensor 6B at a distance from equal to it with current sensor 6C away from
From the first principle apart from rear wiring complexity proposed according to the present invention should choose the collected photovoltaics of current sensor 6B
Input of the system output current signal as S-transformation.
In actually detected, in addition to other than every morning undergoes photovoltaic system starting transient in photovoltaic system,
Power regulation transient process can also occur because of normal system operation, so that photovoltaic system electricity experiencings frequent variation
Transient state.The time of origin of photovoltaic system fault electric arc is uncontrollable, thus photovoltaic system fault electric arc also has certain probability
It can be happened among these systematic procedures, will then eventually lead to the photovoltaic system fault electric arc under coupling.For example, exist
In the systematic procedures such as photovoltaic system startup, increased system power, photovoltaic system output current constantly increases, and on the other hand,
Tandem photovoltaic system failure electric arc can then reduce photovoltaic system output current.Therefore, the photovoltaic system failure under coupling condition
In electric arc, and then photovoltaic system failure output current the normal output current of photovoltaic system can occur, to photovoltaic system failure electricity
The requirement of arc detection algorithm is more harsh.
Corresponding photovoltaic system fault electric arc detection algorithm must catch photovoltaic system fault electric arc difference in systematic procedure
Basic feature, it is accurate, reliable, rapidly the moment occurs for identification photovoltaic system fault electric arc, thus can complete installation photovoltaic
The functional requirement of system dc side fault arc detection device 2.Correctly, the photovoltaic system failure among reliable checkout system process
The specific requirement of electric arc is：In photovoltaic system normal operation, photovoltaic system DC side fault arc detection device 2 exports low
Level is failure to actuate breaker 4, and photovoltaic system 1 still stablizes output electric energy to load 5；If photovoltaic system DC side fault electric arc
Detection device 2 detects the photovoltaic system fault electric arc 7 among betiding systematic procedure, then can quickly and accurately send out cut-out
Respective branch controls signal to trip gear 3, and the final breaker 4 that controls cut-offs entire photovoltaic system circuit, and load is stopped,
Extinguish photovoltaic system fault electric arc and eliminate it and threatened to the operational safety that photovoltaic system is brought, avoids photovoltaic system fault electric arc
It the problem of caused photovoltaic system DC side fault arc detection device refused action, avoids caused by photovoltaic system normal operation
Photovoltaic system DC side fault arc detection device malfunction the problem of, thus expand photovoltaic system detection algorithm be applicable in model
It encloses, potential tripping may be occurred and photovoltaic system is threatened to stablize by solving the photovoltaic system fault electric arc among systematic procedure
The problem of safe operation.
In conjunction with Fig. 3 a~3d, illustrate that the photovoltaic system fault arc detection method of the present invention is applied to have fault electric arc hair
Photovoltaic system fault electric arc identification effect under the coupling condition of raw moment invariant features.
With sample frequency fs=200kHz obtains multichannel photovoltaic system output current signal, as shown in Figure 3a, with wherein one
Input waveform explanation is carried out for the photovoltaic system output current signal of road.Before 3.53s, current signal is in normal state, this
When photovoltaic system supply electricity to load by closed circuit；After 3.53s, current signal is in fault case, but fault current wave at this time
Not due to the generation of photovoltaic system series fault arc, dynamic reduces shape, but maintains normally at the fault electric arc generation moment
The form of electric current remains constant curent change trend.
After multichannel current signal carries out mean value and whitening processing, by independent component analysis to multichannel current signal into
Row analysis selects an effective independent main source signal, calculates the side of one-dimensional frequency matrix after the signal Fast Fourier Transform (FFT)
Difference obtains the shown in solid of fisrt feature amount such as Fig. 3 b.In order to preferably observe the final judgement of fisrt feature amount as a result, corresponding
Fisrt feature amount threshold value be μ1,j–σ1,jAlso it is shown in fig 3b in the form of dotted line.As seen from the figure, fisrt feature amount is with big arteries and veins
The analysis period that form instruction photovoltaic system fault electric arc occurs is rushed, it is whole that photovoltaic system failure is presented with lower amplitude level
The distinctiveness feature of electric arc and normal operation before, it is shown that fisrt feature amount has this kind of photovoltaic system fault electric arc detection
Effect property.By the threshold value comparison of fisrt feature magnitude and construction gained, corresponding electrical level judging is exported as a result, being stored in Fisrt fault
Electric arc trip current out1In.Current signal all the way is analyzed by the method for S-transformation, the two dimension obtained in time-frequency domain is multiple
Matrix number is distributed, and after carrying out absolute value processing to each element of two-dimensional matrix, calculates 40~100kHz components edge of frequency dimension
The integral of time obtains the shown in solid of second feature amount such as Fig. 3 c.In order to preferably observe the final judgement of second feature amount
As a result, corresponding second feature amount threshold value is μ2,j–σ2,jAlso it is shown in figure 3 c in the form of dotted line.As seen from the figure, second is special
The fluctuation form of bigger is presented compared with fisrt feature amount within each analysis period for sign amount, but it is still in integrally with lower amplitude level
The distinctiveness feature of existing photovoltaic system fault electric arc and normal operation before, also shows second feature amount to this kind of photovoltaic system
The validity of fault electric arc detection.By the threshold value comparison of second feature magnitude and construction gained, corresponding electrical level judging knot is exported
Fruit, deposit to the second fault electric arc trip current out2In.
Two characterizing magnitudes have obtained independent component analysis and the output judgement knot of S-transformation after carrying out dynamic threshold relatively
Fruit, weight coefficient are analyzed in the period depending on decision-making system state correctness statistical result, then according to each characteristic quantity at first j -1
Outtemp is obtained after being weighted using Dynamic Weights coefficient in decision-making levelj.By corresponding threshold value comparison, two characteristic quantities are weighted
The judgement in each analysis period is obtained as a result, obtaining preliminary state determination results matrix outt.Count preliminary state judgement knot
The number that fruit matrix outt is 1 to j-th of element from -3p elements of jth to-p elements of jth, from -2p elements of jth, if institute
The numerical value of statistics number is all higher than p, then confirmation is a to generation photovoltaic system fault electric arc in the -2p periods of jth in jth-p, defeated
It is 1 to go out final judgement result, takes corresponding photovoltaic system fault electric arc safeguard measure；Otherwise it is assumed that photovoltaic system is in normal
Operating status, it is 0 to export final judgement result.It is as shown in Figure 3d as a result, detection algorithm faces photovoltaic system normal operation energy
Correct low level instruction is enough provided, correctly high electricity can be provided to no photovoltaic system fault electric arc that any change occurs
Flat instruction, thus the detection algorithm can detect this and betide photovoltaic system fault electric arc among systematic procedure faster.
In conjunction with Fig. 4 a~4d, illustrate that the photovoltaic system fault arc detection method of the present invention is applied to have fault electric arc hair
The raw moment become smaller feature coupling condition under photovoltaic system fault electric arc identification effect.
With sample frequency fs=200kHz obtains multichannel photovoltaic system output current signal, as shown in fig. 4 a, with wherein one
Input waveform explanation is carried out for the photovoltaic system output current signal of road.Before 5.86s, current signal is in normal state, this
When photovoltaic system supply electricity to load by closed circuit；After 5.86s, current signal is in fault case, at this time because photovoltaic system is total
Line occurs series fault arc and generates the fault electric arc current waveform that dynamic reduces, and has at the fault electric arc generation moment and reduces
Curent change trend, but this fault electric arc electric current low compared with normal current fails to maintain, at once increased photovoltaic system
Power makes fault current waveform moment increase, and fault current then consistent with normal levels is maintained.
After multichannel current signal carries out mean value and whitening processing, by independent component analysis to multichannel current signal into
Row analysis selects an effective independent main source signal, calculates the side of one-dimensional frequency matrix after the signal Fast Fourier Transform (FFT)
Difference obtains the shown in solid of fisrt feature amount such as Fig. 4 b.In order to preferably observe the final judgement of fisrt feature amount as a result, corresponding
Fisrt feature amount threshold value be μ1,j–σ1,jAlso it is shown in fig. 4b in the form of dotted line.As seen from the figure, fisrt feature amount is with big arteries and veins
It rushes the analysis period of form instruction photovoltaic system fault electric arc generation and follow-up of short duration variation, it is whole to be in lower amplitude level
Now stablize the distinctiveness feature of photovoltaic system fault electric arc and normal operation before, it is shown that fisrt feature amount is to this kind of photovoltaic system
The validity that fault electric arc of uniting detects.By the threshold value comparison of fisrt feature magnitude and construction gained, corresponding electrical level judging is exported
As a result, being stored in Fisrt fault electric arc trip current out1In.Current signal all the way is analyzed by the method for S-transformation,
The two-dimensional complex number matrix distribution in time-frequency domain is obtained, after carrying out absolute value processing to each element of two-dimensional matrix, calculates frequency dimension
40~100kHz components of degree obtain the shown in solid of second feature amount such as Fig. 4 c along the integral of time.In order to preferably observe
The final judgement of second feature amount is as a result, corresponding second feature amount threshold value is μ2,j–σ2,jAlso figure is illustrated in the form of dotted line
In 4c.As seen from the figure, the fluctuation form of bigger, but its entirety are presented compared with fisrt feature amount within each analysis period for second feature amount
The distinctiveness feature that photovoltaic system fault electric arc and normal operation before are still presented with lower amplitude level, also shows second
The validity that characteristic quantity detects this kind of photovoltaic system fault electric arc.By second feature magnitude with construction gained threshold value comparison,
Corresponding electrical level judging is exported as a result, being stored in the second fault electric arc trip current out2In.
Two characterizing magnitudes have obtained independent component analysis and the output judgement knot of S-transformation after carrying out dynamic threshold relatively
Fruit, weight coefficient are analyzed in the period depending on decision-making system state correctness statistical result, then according to each characteristic quantity at first j -1
Outtemp is obtained after being weighted using Dynamic Weights coefficient in decision-making levelj.By corresponding threshold value comparison, two characteristic quantities are weighted
The judgement in each analysis period is obtained as a result, obtaining preliminary state determination results matrix outt.Count preliminary state judgement knot
The number that fruit matrix outt is 1 to j-th of element from -3p elements of jth to-p elements of jth, from -2p elements of jth, if institute
The numerical value of statistics number is all higher than p, then confirmation is a to generation photovoltaic system fault electric arc in the -2p periods of jth in jth-p, defeated
It is 1 to go out final judgement result, takes corresponding photovoltaic system fault electric arc safeguard measure；Otherwise it is assumed that photovoltaic system is in normal
Operating status, it is 0 to export final judgement result.As shown in figure 4d as a result, detection algorithm face photovoltaic system normal operation energy
Correct low level instruction is enough provided, fault electric arc transient state and the follow-up light that any change does not occur are increased to of short duration reduction
Volt system failure electric arc can provide correct high level instruction, thus the detection algorithm can detect this faster and betide and be
Photovoltaic system fault electric arc among system process.
In conjunction with Fig. 5 a~5d, illustrate that the photovoltaic system fault arc detection method of the present invention is applied to have fault electric arc hair
The raw moment become larger feature coupling condition under photovoltaic system fault electric arc identification effect.
With sample frequency fs=200kHz obtains multichannel photovoltaic system output current signal, as shown in Figure 5 a, with wherein one
Input waveform explanation is carried out for the photovoltaic system output current signal of road.Before 1.21s, current signal is in normal state, this
When photovoltaic system supply electricity to load by closed circuit；After 1.21s, current signal is in fault case, but fault electric arc at this time
It betides in systematic procedure, photovoltaic system power regulation promotes the degree of fault electric arc electric current compared with photovoltaic system bus series fault
The degree that fault electric arc electric current occurs to reduce for electric arc is much bigger, at the fault electric arc generation moment there is increased curent change to become
The fault electric arc electric current of higher level is then maintained by gesture.
After multichannel current signal carries out mean value and whitening processing, by independent component analysis to multichannel current signal into
Row analysis selects an effective independent main source signal, calculates the side of one-dimensional frequency matrix after the signal Fast Fourier Transform (FFT)
Difference obtains the shown in solid of fisrt feature amount such as Fig. 5 b.In order to preferably observe the final judgement of fisrt feature amount as a result, corresponding
Fisrt feature amount threshold value be μ1,j–σ1,jAlso it is shown in figure 5b in the form of dotted line.As seen from the figure, fisrt feature amount it is whole with
The distinctiveness feature of all photovoltaic system fault electric arc states and normal operation before is presented in lower amplitude level, and big pulse refers to
Though the increase tendency shown can influence the correct identification of fault electric arc state, the of short duration continual analysis period does not interfere with failure electricity
The whole identification of arc, it is shown that the validity that fisrt feature amount detects photovoltaic system fault electric arc.By fisrt feature magnitude with
The threshold value comparison for constructing gained exports corresponding electrical level judging as a result, being stored in Fisrt fault electric arc trip current out1In.It is logical
The method for crossing S-transformation analyzes current signal all the way, the two-dimensional complex number matrix distribution in time-frequency domain is obtained, to two-dimensional matrix
Each element carry out absolute value processing after, calculate integral of 40~100kHz components along the time of frequency dimension, obtain second spy
Sign amount such as Fig. 5 c's is shown in solid.In order to preferably observe the final judgement of second feature amount as a result, corresponding second feature amount
Threshold value is μ2,j–σ2,jAlso it is shown in fig. 5 c in the form of dotted line.As seen from the figure, second feature amount compared with fisrt feature amount at each point
The fluctuation form that bigger is presented in the period is analysed, but it is whole still with all photovoltaic system fault electric arcs of lower amplitude level presentation
The distinctiveness feature of state and normal operation before so that correction fisrt feature amount will produce erroneous judgement in certain analysis periods to be become
May, highlight the necessity weighted using two characteristic quantity of the present invention.By the threshold value of second feature magnitude and construction gained
Compare, exports corresponding electrical level judging as a result, being stored in the second fault electric arc trip current out2In.
Two characteristic quantities have obtained the judgement of independent component analysis and S-transformation as a result, weights after carrying out dynamic threshold relatively
It is analyzed in the period depending on decision-making system state correctness statistical result, then in decision-making level at first j -1 according to each characteristic quantity
Outtemp is obtained after being weighted using Dynamic Weights coefficientj.By corresponding threshold value comparison, two characteristic quantities of weighting obtain each point
The judgement in the period is analysed as a result, obtaining preliminary state determination results matrix outt.Count preliminary state determination results matrix outt
The number for being 1 to j-th of element from -3p elements of jth to-p elements of jth, from -2p elements of jth, if counted number
Numerical value is all higher than p, then confirmation is a to generation photovoltaic system fault electric arc in the -2p periods of jth in jth-p, exports and finally judges
As a result it is 1, takes corresponding photovoltaic system fault electric arc safeguard measure；Otherwise it is assumed that photovoltaic system is in normal operating condition,
The final judgement result of output is 0.As fig 5d as a result, detection algorithm can provide correctly in face of photovoltaic system normal operation
Low level instruction, to increase reduce increased chromic trouble electric arc transient state and follow-up higher magnitude fault electric arc stable state are able to again
Remain able to provide correct high level instruction, thus the detection algorithm can detect this and betide among systematic procedure faster
Photovoltaic system fault electric arc.
As shown in Fig. 1 a~1b, photovoltaic fault arc detection method utilizes characteristic quantity under coupling condition provided by the present invention
Estimation of Mean and standard deviation construction feature amount threshold value, threshold value dynamic change processing is carried out in different analytical cycles, when recognizing
When determining the generation of photovoltaic system fault electric arc, the calculating of Estimation of Mean and standard deviation is both needed to be modified.Use characteristic value and threshold
Value comparison procedure realizes the normalization of each characteristic quantity output, solves different characteristic amount the number of output grade difference to weighting multiple features
The interference of amount detection fault electric arc, the multiple features weighting being advantageously implemented in decision-making level.Weights by each characteristic quantity, correctly go through by identification
Depending on the statistical law of history system mode, when assert that photovoltaic system normal operation and two characteristic quantities output judgement result do not wait,
To two fault electric arc trip currents, accordingly not equal elements do not make exchange processing, be conducive within each analysis period more reliably to
Go out system mode be appropriately determined as a result, effectively increase photovoltaic system fault electric arc detection reliability, increase photovoltaic system
The economic benefit of system operation.
As shown in Fig. 3 a~5d, photovoltaic fault arc detection method passes through two features under coupling condition provided by the present invention
The mode of weight coefficient weighting has grasped the statistical law and core feature of photovoltaic system fault electric arc in amount decision-making level, improves
Photovoltaic system solves with photovoltaic system failure output current visual angle detection algorithm in face of being the recognition capability of electric current normal state
The photovoltaic system DC side fault arc detection device malfunction problem that the transient processes such as power regulation, startup of uniting generate, by just
Systematic procedure is really determined as normal operating condition, the uptime of photovoltaic system is substantially extended, significantly improves light
The generating efficiency of volt system enhances the stabilizing power of photovoltaic system normal operation.The present invention also accurately can catch to betide
Photovoltaic system fault electric arc basic feature among system process, accurately identifies the photovoltaic system failure among betiding systematic procedure
Electric arc, photovoltaic system fault electric arc variation tendency direction caused by photovoltaic system output current is not influenced under by coupling condition,
The scope of application for expanding current photovoltaic system fault arc detection method is solved with the normal output current visual angle of photovoltaic system
The photovoltaic system DC side fault electric arc that the photovoltaic system fault electric arc that detection algorithm occurs in face of and then systematic procedure generates
Detection device tripping problem is ensured by the way that the photovoltaic system fault electric arc under coupling condition is correctly determined as malfunction
The validity of photovoltaic system fault electric arc detection, eliminate in time photovoltaic fire incident that this kind of photovoltaic system fault electric arc causes,
The harm such as life and property loss.
Claims (10)
1. a kind of method detecting photovoltaic system failure electric arc using independent component analysis and S-transformation, it is characterised in that：The detection
The method of photovoltaic system fault electric arc includes the following steps under systematic procedure coupling condition：
1) sample frequency f is pressed to photovoltaic system output current signal by multiple current sensorssIt is sampled, is obtained more point by point
Road current signal xi,j, wherein i is that current sensor indicates that serial number, i ∈ N and i > 1, j are the analysis period to indicate serial number, j ∈
N+, for arbitrary two differences i values, when j takes same value, xi,jEqual sampling number N is all had, when N reaches the analysis period
Requirement after, go to step 2) carry out Fisrt fault arc characteristic analysis；
2) collected multichannel current signal is formed into higher-dimension mixed signal matrix X=[x1,j,x2,j,…,xi,j]T, mixed to gained
It closes signal matrix and carries out mean value and whitening processing, then matrix W is mixed by the way that solution can be obtained after fast independent component analysis, calculate
Source signal matrix S=WX=[s1,j,s2,j,…,si,j]T, select effectively independent main source signal s1,j, to s1,jIt carries out in quick Fu
Leaf transformation calculates the variance of one-dimensional frequency matrix in frequency domain, obtains fisrt feature magnitude r1,j, go to step 3)；
3) fisrt feature amount threshold value is set in the present analysis period as A1×μ1,j–A2×σ1,j, wherein coefficient A1With A2Foundation passes through
The fisrt feature amount threshold value of setting can be correctly obtained compared with fisrt feature magnitude depending on corresponding photovoltaic system state, μ1,jFor
From the Estimation of Mean of first analysis period to present analysis period all fisrt feature magnitudes, σ1,jTo analyze the period extremely from first
The standard deviation of present analysis period all fisrt feature magnitudes, A1∈ Z, A2∈ Z, by the first spy of fisrt feature magnitude and setting
Sign amount threshold value comparison exports corresponding electrical level judging result：If r1,j≥A1×μ1,j–A2×σ1,j, then judgement result 0 is exported, is deposited
Enter to Fisrt fault electric arc trip current out1[j]；If r1,j<A1×μ1,j–A2×σ1,j, then judgement result 1, deposit to the are exported
One fault electric arc trip current out1[j] goes to step 4) and carries out the second fault electric arc signature analysis；
4) it selects the signal all the way in multichannel current signal to carry out S-transformation, obtains the two-dimensional complex number time-frequency matrix in time-frequency domain, count
The high fdrequency component absolute value of frequency dimension is calculated along the integral of time, obtains second feature magnitude r2,j, go to step 5)；
5) second feature amount threshold value is set in the present analysis period as A3×μ2,j–A4×σ2,j, wherein coefficient A3With A4Foundation passes through
The second feature amount threshold value of setting can be correctly obtained compared with second feature magnitude depending on corresponding photovoltaic system state, μ2,jFor
From the Estimation of Mean of first analysis period to present analysis period all second feature magnitudes, σ2,jTo analyze the period extremely from first
The standard deviation of present analysis period all second feature magnitudes, A3∈ Z, A4∈ Z, by the second spy of second feature magnitude and setting
Sign amount threshold value comparison exports corresponding electrical level judging result：If r2,j≥A3×μ2,j–A4×σ2,j, then judgement result 0 is exported, is deposited
Enter to the second fault electric arc trip current out2[j]；If r2,j<A3×μ2,j–A4×σ2,j, then judgement result 1, deposit to the are exported
Two fault electric arc trip current out2[j] is gone at the output judgement result weighting that step 6) carries out in two characteristic quantity decision-making levels
Reason；
6) output that independent component analysis and S-transformation are weighted using Dynamic Weights coefficient is judged as a result, obtaining weighted results
outtempj=C1,j×out1[j]+C2,j×out2[j] then carries out preliminary state judgement：If outtempj>N, wherein n is
Weighted results threshold value then exports judgement result 1, deposit to preliminary state determination results matrix outt [j]；Otherwise output judgement knot
Fruit 0, deposit to preliminary state determination results matrix outt [j] go to step 7) and carry out photovoltaic system state differentiation, C1,jAnd C2,j
For fisrt feature amount and the affiliated weight coefficient of second feature amount；
7) setting judges precision p, judges a photovoltaic system state per p period：Count preliminary state determination results matrix
Outt is 1 number from -3p elements of jth to-p elements of jth, from -2p elements of jth to j-th of element, if counting a
Several numerical value is all higher than p, then confirmation is a to generation photovoltaic system fault electric arc in the-p periods of jth in jth -2p, takes corresponding
Photovoltaic system fault electric arc safeguard measure；Otherwise it is assumed that photovoltaic system is in normal in jth -2p to the-p periods of jth
Operating status, return to step 1) current signal in next analysis period is analyzed.
2. a kind of method detecting photovoltaic system failure electric arc using independent component analysis and S-transformation according to claim 1,
It is characterized in that：The band of the current sensor is wider than 100kHz, and the different location for being installed in photovoltaic system is adopted with showing
The value range of difference between sample current signal, current sensor is 2~4；The sample frequency fsValue range be 200
~500kHz；The value range of the sampling number N is 8000~12000.
3. a kind of method detecting photovoltaic system failure electric arc using independent component analysis and S-transformation according to claim 1,
It is characterized in that：The fast independent component analysis is preferably based on the fast independent component analysis of negentropy maximization, quickly solely
G can be selected in nonlinear function in vertical constituent analysis1(u)=u3、g2(u)=u2、g3(u)=arctan (q1×u)、g4(u)=u
×e^(-q2 2×u2/ 2) one kind in, wherein q1And q2Value range for constant, maximum iteration is 950~1050, repeatedly
Value range for precision is 0.00006~0.00015.
4. a kind of method detecting photovoltaic system failure electric arc using independent component analysis and S-transformation according to claim 1,
It is characterized in that：The main source signal number of the independence i.e. way of sampled current signals that the fast independent component analysis obtains, base
An effective independent main source signal is selected to carry out follow-up Fast Fourier Transform (FFT) processing in the strongest principle of signal impact, i.e.,
The difference for calculating peak-to-peak value of the main source signal of each independence within the analysis period, it is effective to select the main source signal of the maximum independence of difference
Independent main source signal；The transformation point value of the Fast Fourier Transform (FFT) is chosen to be the corresponding numerical value of sampling number N.
5. a kind of method detecting photovoltaic system failure electric arc using independent component analysis and S-transformation according to claim 1,
It is characterized in that：Select the method inputted all the way as S-transformation in multichannel current signal for：Preferential selection sensitivity is highest
The corresponding current signal of current sensor；When this kind of current sensor more than one, preferential chosen distance photovoltaic system failure
The nearest corresponding current signal of current sensor in position occurs for electric arc；It is nearest when position occurs apart from photovoltaic system fault electric arc
Current sensor more than one when, it is preferential select to have in photovoltaic system fault electric arc to current sensor propagation path it is minimum
The corresponding current signal of current sensor of number of components；The window width Dynamic gene of the S-transformation is preferably 1.
6. a kind of method detecting photovoltaic system failure electric arc using independent component analysis and S-transformation according to claim 1,
It is characterized in that：Absolute value processing is made to the two-dimensional complex number time-frequency matrix element obtained by S-transformation, builds the time-frequency of second feature amount
Matrix frequency component is selected as 40~100kHz, photovoltaic system fault electric arc characteristic spectra and sample frequency fsValue not phase
It closes.
7. a kind of method detecting photovoltaic system failure electric arc using independent component analysis and S-transformation according to claim 1,
It is characterized in that：The fisrt feature amount threshold value A1×μ1,j–A2×σ1,jHave with the fisrt feature magnitude of all analysis periods before
It closes and follows fisrt feature amount r in real time1Dynamic change, wherein coefficient A1With A2According to by the fisrt feature amount threshold value of setting with
Fisrt feature magnitude, which compares, can be correctly obtained depending on corresponding photovoltaic system state；Estimation of Mean μ1,jAnd standard deviation sigma1,jAccording to the
The output judgement result of one characteristic quantity is corrected in real time：The fisrt feature magnitude r obtained for first analysis period1,1, enable
Correction amount rtemp1,1=r1,1, Estimation of Mean μ1,1=r1,1, standard deviation sigma1,1=0；For the fisrt feature of j-th of analysis period
Magnitude r1,j, wherein j ∈ N and j > 1, if fisrt feature magnitude was more than or equal to a upper analysis period first in the present analysis period
When characteristic quantity threshold value, correction amount rtemp is enabled1,j=r1,j, the calculation formula of Estimation of Mean and standard deviation is
Wherein, k is analysis period expression serial number, k=1,2 ... j, j ∈ N and j > 1 in cumulative process, if in the present analysis period
When fisrt feature magnitude is less than upper analysis period fisrt feature amount threshold value, correction amount rtemp is enabled1,j=A1×μ1,j-1–A2×
σ1,j-1, the calculation formula of Estimation of Mean and standard deviation is
The second feature amount threshold value A3×μ2,j–A4×σ2,jIt is related and real with the second feature magnitude of all analysis periods before
When follow second feature amount r2Dynamic change, wherein coefficient A3With A4Foundation is special by the second feature amount threshold value of setting and second
Sign magnitude, which compares, can be correctly obtained depending on corresponding photovoltaic system state；Estimation of Mean μ2,jAnd standard deviation sigma2,jAccording to second feature
The output judgement result of amount is corrected in real time：The second feature magnitude r obtained for first analysis period2,1, enable correction amount
rtemp2,1=r2,1, Estimation of Mean μ2,1=r2,1, standard deviation sigma2,1=0；For the second feature magnitude of j-th of analysis period
r2,j, wherein j ∈ N and j > 1, if second feature magnitude is more than or equal to upper analysis period second feature in the present analysis period
When measuring threshold value, correction amount rtemp is enabled2,j=r2,j, the calculation formula of Estimation of Mean and standard deviation is
Wherein, k is analysis period expression serial number, k=1,2 ... j, j ∈ N and j > 1 in cumulative process, if in the present analysis period
When second feature magnitude is less than upper analysis period second feature amount threshold value, correction amount rtemp is enabled2,j=A3×μ2,j-1–A4×
σ2,j-1, the calculation formula of Estimation of Mean and standard deviation is
。
8. a kind of method detecting photovoltaic system failure electric arc using independent component analysis and S-transformation according to claim 7,
It is characterized in that：Obtaining the calculation formula of Estimation of Mean and standard deviation in the present analysis period using recurrence relation is
Wherein, μm,j、σm,jEstimation of Mean and standard deviation respectively in the present analysis period, μm,j-1、σm,j-1Respectively previous point
Analyse the Estimation of Mean and standard deviation in the period, rtempm,jFor the correction amount in the present analysis period, wherein the m amounts of being characterized indicate
Serial number, value are 1 or 2, j ∈ N and j > 1.
9. a kind of method detecting photovoltaic system failure electric arc using independent component analysis and S-transformation according to claim 1,
It is characterized in that：When the output using Dynamic Weights coefficient weighting independent component analysis and S-transformation judges result, two characteristic quantities
The weight coefficient of output judgement result judges that the statistical property of correctness is true according to individual features amount to historical analysis period state
It is fixed, i.e., characteristic quantity the historical analysis period is made correct status judgement the analysis period it is more, this feature amount is in present analysis
The weight coefficient that section is obtained is then bigger, specifically, constructs fisrt feature amount and second feature amount institute respectively based on following formula
Belong to weight coefficient C1,jAnd C2,j：
Wherein, σ2 out1And σ2 out2Respectively Fisrt fault electric arc trip current and the second fault electric arc trip current is from first member
Element to j-th of element variance, i.e.,
Wherein, out1And out2Respectively Fisrt fault electric arc trip current and the second fault electric arc trip current, k is matrix element
The counting serial number of element, k=1,2 ... j, j ∈ N and j > 1,WithRespectively Fisrt fault electric arc trip current and the second event
Hinder Estimation of Mean of the electric arc trip current from first element to j-th of element；If Fisrt fault electric arc trip current, the second event
It is 0 to hinder electric arc trip current from first element to j-th of element, i.e. two characteristic quantities judge that all analysis periods are normal
Operating status, indirect assignment C1,j=0, C2,j=0；If jth -2p are in normal operation to photovoltaic system in the-p periods of jth
State, to the position of Fisrt fault electric arc trip current, the second fault electric arc trip current respective element under this p period not etc.
Make element exchange processing.
10. a kind of side for detecting photovoltaic system failure electric arc using independent component analysis and S-transformation according to claim 1
Method, it is characterised in that：The value range of the weighted results threshold value n is 0.45~0.55；The value range for judging precision p
It is 2~5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710254713.4A CN107154783B (en) | 2017-04-18 | 2017-04-18 | The method for detecting photovoltaic system failure electric arc using independent component analysis and S-transformation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710254713.4A CN107154783B (en) | 2017-04-18 | 2017-04-18 | The method for detecting photovoltaic system failure electric arc using independent component analysis and S-transformation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107154783A CN107154783A (en) | 2017-09-12 |
| CN107154783B true CN107154783B (en) | 2018-10-30 |
Family
ID=59794220
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710254713.4A Active CN107154783B (en) | 2017-04-18 | 2017-04-18 | The method for detecting photovoltaic system failure electric arc using independent component analysis and S-transformation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107154783B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2019158674A (en) * | 2018-03-14 | 2019-09-19 | オムロン株式会社 | Arc detection device, control method therefor, control program, and DC power supply system |
| CN110456234B (en) * | 2018-05-07 | 2020-11-10 | 珠海格力电器股份有限公司 | Fault arc detection method, device and system |
| CN109560770B (en) * | 2018-10-29 | 2020-03-13 | 西安交通大学 | Photovoltaic system fault arc detection method based on self-adaptive kernel function and instantaneous frequency estimation |
| CN109239558B (en) * | 2018-11-08 | 2020-10-02 | 重庆大学 | Direct current fault arc detection and protection device |
| CN109556649B (en) * | 2018-11-30 | 2020-12-22 | 安徽云汉智能科技有限公司 | Signal acquisition monitoring method of intelligent sensor |
| CN109596955B (en) * | 2018-12-30 | 2021-06-22 | 国网北京市电力公司 | Partial discharge state determination method and device |
| CN112034273A (en) * | 2019-05-16 | 2020-12-04 | 北京新能源汽车股份有限公司 | High-voltage interlocking fault detection method and device and automobile |
| CN110677125B (en) * | 2019-10-11 | 2022-07-01 | 国网冀北电力有限公司秦皇岛供电公司 | Arc fault detection method and device |
| CN111313833B (en) * | 2020-03-06 | 2023-04-18 | 燕山大学 | Method for detecting series arc fault on direct current side of series photovoltaic system |
| CN111812562B (en) * | 2020-06-01 | 2024-01-30 | 国网辽宁省电力有限公司电力科学研究院 | Quench detection method and quench detection device for high-temperature superconductive ring magnet |
| CN113376474A (en) * | 2021-04-26 | 2021-09-10 | 国网天津市电力公司城南供电分公司 | Neural network fault arc identification system and method based on generalized S transformation |
| CN113297786B (en) * | 2021-05-06 | 2023-04-07 | 西安理工大学 | Low-voltage fault arc sensing method based on semi-supervised machine learning |
| CN117007896B (en) * | 2023-10-07 | 2023-12-12 | 深圳市森瑞普电子有限公司 | Data processing method applied to conductive slip ring fault detection |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9008978B2 (en) * | 2011-06-17 | 2015-04-14 | Sunfield Semiconductor, Inc. | System and method for arc detection in solar power arrays |
| DE102013018950A1 (en) * | 2013-11-12 | 2015-05-13 | Anton Naebauer | Generator connection box for increased electrical safety and reduced maintenance costs for PV power plants |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5888972B2 (en) * | 2011-12-22 | 2016-03-22 | 三菱電機株式会社 | Solar power system |
-
2017
- 2017-04-18 CN CN201710254713.4A patent/CN107154783B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9008978B2 (en) * | 2011-06-17 | 2015-04-14 | Sunfield Semiconductor, Inc. | System and method for arc detection in solar power arrays |
| DE102013018950A1 (en) * | 2013-11-12 | 2015-05-13 | Anton Naebauer | Generator connection box for increased electrical safety and reduced maintenance costs for PV power plants |
Non-Patent Citations (1)
| Title |
|---|
| 直流故障电弧研究综述;陈思磊，李兴文，屈建宁;《电器与能效管理技术》;20150815;第1-5页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107154783A (en) | 2017-09-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107154783B (en) | The method for detecting photovoltaic system failure electric arc using independent component analysis and S-transformation | |
| US11489490B2 (en) | Arc fault detection method for photovoltaic system based on adaptive kernel function and instantaneous frequency estimation | |
| Bakdi et al. | Real-time fault detection in PV systems under MPPT using PMU and high-frequency multi-sensor data through online PCA-KDE-based multivariate KL divergence | |
| Anderson et al. | Event detection for non intrusive load monitoring | |
| US20150309092A1 (en) | Current Pattern Matching Method for Non-Intrusive Power Load Monitoring and Disaggregation | |
| CN107025365B (en) | A kind of non-intruding load discrimination method for user side | |
| Sou et al. | Data attack isolation in power networks using secure voltage magnitude measurements | |
| Ahmadi et al. | A new method for detecting series arc fault in photovoltaic systems based on the blind-source separation | |
| US11693062B2 (en) | Method for processing direct current electric arc and apparatus | |
| WO2016138750A1 (en) | Power quality disturbance source locating system and locating method | |
| Buddha et al. | Signal processing for photovoltaic applications | |
| CN105490641A (en) | Photovoltaic system fault electric arc detection method based on comprehensive multiple characteristic quantities | |
| Dutta et al. | Topology tracking for active distribution networks | |
| Choi et al. | Impact analysis of locational marginal price subject to power system topology errors | |
| Wang et al. | An accurate false data detection in smart grid based on residual recurrent neural network and adaptive threshold | |
| Wu et al. | Online identification of bad synchrophasor measurements via spatio-temporal correlations | |
| Onaolapo et al. | Simulation experiments for faults location in smart distribution networks using ieee 13 node test feeder and artificial neural network | |
| Guzel et al. | Principal components null space analysis based non-intrusive load monitoring | |
| Kokoszka et al. | Detection and Localization of Faults in a Regional Power Grid: Faults in a Regional Power Grid | |
| Regev et al. | Hybrid ai-based anomaly detection model using phasor measurement unit data | |
| Giannini | Improving cyber-security of power system state estimators | |
| Cavraro et al. | Learning power grid topologies | |
| Stamatescu et al. | Detection of anomalies in power profiles using data analytics | |
| Xiong et al. | Detection of false data injection attack based on improved distortion index method | |
| Abniki et al. | High impedance fault detection in power transmission lines using Hilbert transform and instantaneous frequency |
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 |