A kind of feeder line fault diagnostic method and system based on PMU data
阅读说明：本技术 一种基于pmu数据的电力线路故障诊断方法及系统 (A kind of feeder line fault diagnostic method and system based on PMU data ) 是由 葛维春 张艳军 苏禹泽 刘爱民 孔剑虹 刘劲松 李斌 谢强 张建 姜枫 刘凯 于 2019-05-31 设计创作，主要内容包括：本发明涉及一种基于PMU数据对配电网线路进行故障选线的方法,通过结合总体平均经验模式分解方法进行高频能量计算,根据PMU数据的特点设计完整的数据提取使用流程和故障选线系统。包括本地子站以“子站-主站”协同架构上传故障时的高频数据；根据高频数据计算所有馈线平均电流幅值差；根据PMU自带的故障录波器提供的故障录波数据做EEMD变换计算所有线路的高频暂态能量；计算所有线路的故障可信度；根据故障可信度确定最终的故障馈线。本发明实现配电网线路故障在线选线,具有自适应分解的优势,消除传统模态分解方法所带来的模态混叠效应,适用于所有故障类型地选线方法,选线结果更加精准,定量的故障可信度更加直观明了。(The present invention relates to a kind of methods for carrying out failure line selection to distribution network line based on PMU data, by combining population mean ensemble empirical mode decomposition method to carry out high-frequency energy calculating, complete data are designed according to the characteristics of PMU data and extract process for using and failure line selection system.High-frequency data when including local substation with " substation-main website " co-architecture upload failure；It is poor that all feeder line average current magnitudes are calculated according to high-frequency data；The fault recorder data that the fault oscillograph carried according to PMU provides does the high frequency transient energy of all routes of EEMD transformation calculations；Calculate the fault credibility of all routes；Final fault feeder is determined according to fault credibility.The present invention realizes the online route selection of distribution network line fault, advantage with adaptive decomposition eliminates modal overlap effect brought by traditional modal decomposition method, suitable for all fault types selection method, route selection result is more accurate, and quantitative fault credibility is more simple and clear.)
1. a kind of feeder line fault diagnostic method and system based on PMU data, it is characterized in that: including:
Step 1: local substation uploads high-frequency data when failure with " substation-main website " co-architecture；
Step 2: it is poor that all feeder line average current magnitudes being calculated according to high-frequency data；
Step 3: the fault recorder data that the fault oscillograph carried according to PMU provides does the height of all routes of EEMD transformation calculations Frequency transient state energy；
Step 4: calculating the fault credibility of all routes；
Step 5: final fault feeder is determined according to fault credibility.
2. a kind of feeder line fault diagnostic method and system based on PMU data according to claim 1, feature Be: " substation-main website " co-architecture includes:
The high frequency sampling in the substation 1:PMU obtains effective value, the phase angle of high frequency；
2nd: based on the phase angle being locally stored, instantaneous frequency f is calculated, shown in the calculating of instantaneous frequency f such as formula (8):
In formulaFor the instantaneous frequency of two neighboring sampled point；T is the high frequency sampling interval；
3rd: each wave period takes the maximum value f of instantaneous frequencymax, fmaxWith low frequency (50HZ) with voltage and current phasor together on Reach main website；
4th: main website detects fmaxWhether abnormal, abnormal judgement is set as whether being greater than twice of fundamental frequency, i.e. fmaxWhether it is greater than 100HZ；
5th: if detecting f in step 4maxOccur it is abnormal, then will be before and after fault moment on the high frequency measurement of 5 wave periods It reaches main website and provides data for subsequent select-line analysis.
3. a kind of feeder line fault diagnostic method and system based on PMU data according to claim 2, feature Be: high-frequency data when uploading failure in the step 1 obtains in the following manner:
(1) EEMD algorithm:
White Gaussian noise signal is added in basis of the EEMD based on EMD algorithm during carrying out EMD to signal:
The EMD decomposition of signal x (t) is exactly to be broken down into one group of IMF component CiWith remainder rnThe sum of, as shown in formula (1):
EEMD algorithm is before carrying out EMD decomposition to signal, using Gauss white noise frequency-flat characteristic and its to random dry Random white noise is added to original signal, then to adding the signal after making an uproar to be iterated analysis, finally in the smoothing effect for disturbing ingredient Iteration obtains IMF component；
The step of EEMD algorithm, is as follows:
1. being initialized to the amplitude coefficient k that EMD executes total degree M and white noise is added, i.e. k, m=1；
2. executing the m times EMD to decompose；
A) the random Gaussian white noise n that amplitude coefficient is k is added in original signal x (t)m(t), the signal to be processed after obtaining plus making an uproar xm(t), such as formula (2)；
xm(t)=x (t)+knm(t) (2)
B) to xm(t) EMD decomposition is carried out, p IMF component C is obtainedj,m(j=1,2 ..., p), Cj,mIndicate that the m times test is decomposed J-th of the IMF component arrived；
C) as m < M, m=m+1, return step is 2.；
3. each IMF decomposed to M times calculates mean value, such as formula (3):
4. exportingAs j-th IMF, j=1,2 ... that EEMD is decomposed, p；
(2) IMF and energy balane are adaptively chosen:
To signal carry out EEMD decomposition after obtain several IMF components, further choose most can faults transient state component IMF Component does feature calculation to it, after the quick spectrum kurtosis map algorithm based on spectrum kurtosis is for realizing overall experience mode decomposition Optimal IMF component selection；
Kurtosis reflects the numerical statistic amount of stochastic variable distribution character, is 4 rank central moments of normalization, asks high and steep to all IMF components Degree, the maximum component of kurtosis is most close to the transient state component of failure；Then IMF component maximum to kurtosis carries out calculating energy, IMF ENERGY E is calculated such as formula (4):
Wherein A is the amplitude of IMF component, t1For sample start times, t2The time is tied for sampling, t is the sampling time；
(3) average current magnitude is poor；
It is poor to define average current magnitude, it is made up into the latter with transient state energy progress integrated application and is only applicable to imbalance fault Defect realizes the faulty comprehensive route selection of institute；Average current magnitude difference DI is defined as follows:
ΔIi=Ifi-Issi, i=A, B, C (5)
In upper two formula: IfiFor the maximum value of failure phase three-phase current amplitude, IssiFor pre-fault stable stage three-phase current amplitude Maximum value；△IA、△IB、△ICCurrent amplitude before and after respectively the failure of A, B, C three-phase occurs is poor；
The route all to power distribution network asks its DI, the DI of faulty line to be significantly greater than non-fault line；
(4) high frequency metric data obtains:
By IEEE Std C37.118.1 and C37.118.2 to PMU device measure and communicate standard it is found that PMU device it is upper The frequency for passing report is 50HZ, and the signal of this low frequency is unable to get enough fault transient features, but the height of the substation PMU Frequency sampling rate can reach 20kHZ or more and be locally stored；On the other hand, PMU carries fault oscillograph, and sample frequency reaches 5kHZ or more is extracted for transient characteristic.
4. a kind of feeder line fault diagnostic method and system based on PMU data according to claim 1, feature It is: EEMD variation is done to the zero-sequence current of all feeder lines in the step 2, and calculate the kurtosis of all IMF components, most to kurtosis Big IMF component seeks its wave type energy.
5. a kind of feeder line fault diagnostic method and system based on PMU data according to claim 1 or 3, special Sign is: the failure front and back route that the data of average current magnitude difference are recorded from fault oscillograph is calculated in the step 2 Three-phase current, then calculated by formula (5) and (6).
6. a kind of feeder line fault diagnostic method and system based on PMU data according to claim 1, feature Be: the data in fault credibility calculated in the step 4 are from step 2 and 3 obtained high frequency transient energy peace Equal current amplitude is poor.
7. a kind of feeder line fault diagnostic method and system based on PMU data according to claim 1 or 3, special Sign is: the fault credibility in the step 4 is the effective combination for realizing transient energy method and average current magnitude method, is finally given One is used to judge the quantitative of faulty line as a result, shown in the calculating such as formula (7) of fault credibility G out:
G in formulaiFor the fault credibility on i-th line road；N is the bus number in power distribution network；E is the IMF obtained after EEMD is decomposed The energy of component, calculation method are shown in formula (4)；DI is that the average current magnitude of route is poor, calculation method such as formula (5) and (6).
8. a kind of feeder line fault diagnostic method and system based on PMU data according to claim 1, feature It is: the fault credibility of the step 5 all routes according to obtained in step 4, the fault credibility of more all routes, therefore Barrier confidence level is maximum to be determined as faulty line.
9. a kind of feeder line fault diagnostic method and system based on PMU data according to claim 1, feature Be: the step 1 specifically includes the miniature PMU device local high frequency sampled voltage electric current phasor installed on distribution network line, and The real-time low frequency of calculating fault features amount instantaneous frequency maximum value is uploaded to main website；The step 2 specifically includes judgement fault signature Whether amount there is exception, if abnormal, upload local high-frequency current data and provides data source for subsequent route selection, extract simultaneously The data of miniature PMU included fault oscillograph；The step 3 specifically includes the zero-sequence current data for providing fault oscillograph EEMD decomposition is carried out, chooses high frequency transient component by calculating spectrum kurtosis, and further calculate the high frequency transient energy of each route； It is poor that the step 4 specifically includes the average current magnitude that the three-phase current data of local high frequency sampling are calculated each route；It is described Step 5 specifically includes the event that each route is calculated according to the high frequency transient energy and average current magnitude difference of each route being calculated Hinder confidence level, compares fault credibility, the highest route of fault credibility is faulty line.
Technical field
The present invention relates to a kind of methods for carrying out failure line selection to distribution network line based on PMU data, especially by combination Population mean empirical mode decomposition (Ensemble Empirical Mode Decomposition, EEMD) method carries out high frequency Energy balane, and the present invention devises complete data according to the characteristics of PMU data and extracts the process used and failure line selection System.
Background technique
With the installation of synchronous phasor measurement unit (Phasor Measurement Unit, PMU) in the power system, Its measurement data based on GPS is provided convenience for the application such as fault diagnosis, status assessment, Load flow calculation of electric system, but Be traditional PMU volume it is big, it is expensive due to cause it that can not popularize in power distribution network, all grinding both at home and abroad in the recent period Make a kind of miniature synchronous phasor measurement unit (Micro Phasor cheap, small in size suitable for distribution network system Measurement Unit, μ PMU), miniature PMU future will on power distribution network large-area applications.Since domestic power distribution network is grounded Mode is small current neutral grounding mode, and fault current is unobvious, and distribution network fault line selection is always the research hotspot of domestic scholars, choosing The method of line is divided into steady state method and transient method according to data extraction, and wherein transient method needs such as small using the method for frequency-domain analysis Wave conversion, empirical mode decomposition method (Empirical Mode Decomposition, EMD) etc., but wavelet transformation presence needs The shortcomings that choosing suitable small echo, although EMD decomposition can be with adaptive decomposition, there is modal overlap in it.
Summary of the invention
For above-mentioned problems of the prior art, the power circuit event based on PMU data that the present invention provides a kind of Hinder diagnostic method and system, it is intended that in order to overcome
For achieving the above object, the technical solution adopted by the present invention to solve the technical problems is:
A kind of feeder line fault diagnostic method and system based on PMU data, comprising:
Step 1: local substation uploads high-frequency data when failure with " substation-main website " co-architecture；
Step 2: it is poor that all feeder line average current magnitudes being calculated according to high-frequency data；
Step 3: the fault recorder data that the fault oscillograph carried according to PMU provides does all routes of EEMD transformation calculations High frequency transient energy；
Step 4: calculating the fault credibility of all routes；
Step 5: final fault feeder is determined according to fault credibility.
" substation-main website " co-architecture includes:
The high frequency sampling in the substation 1:PMU obtains effective value, the phase angle of high frequency；
2nd: based on the phase angle being locally stored, instantaneous frequency f is calculated, shown in the calculating of instantaneous frequency f such as formula (8):
In formulaFor the instantaneous frequency of two neighboring sampled point；T is the high frequency sampling interval；
3rd: each wave period takes the maximum value f of instantaneous frequencymax, fmaxWith low frequency (50HZ) with voltage and current phasor one It rises and is uploaded to main website；
4th: main website detects fmaxWhether abnormal, abnormal judgement is set as whether being greater than twice of fundamental frequency, i.e. fmaxIt is whether big In 100HZ；
5th: if detecting f in step 4maxThere is exception, then measures the high frequency of 5 wave periods before and after fault moment Amount is uploaded to main website and provides data for subsequent select-line analysis.
High-frequency data when uploading failure in the step 1 obtains in the following manner:
(1) EEMD algorithm:
White Gaussian noise signal is added in basis of the EEMD based on EMD algorithm during carrying out EMD to signal:
The EMD decomposition of signal x (t) is exactly to be broken down into one group of IMF component CiWith remainder rnThe sum of, as shown in formula (1):
EEMD algorithm be to signal carry out EMD decomposition before, using Gauss white noise frequency-flat characteristic and its to Random white noise is added to original signal in the smoothing effect of machine interference component, then to adding the signal after making an uproar to be iterated analysis, Final iteration obtains IMF component；
The step of EEMD algorithm, is as follows:
1. being initialized to the amplitude coefficient k that EMD executes total degree M and white noise is added, i.e. k, m=1；
2. executing the m times EMD to decompose；
A) the random Gaussian white noise n that amplitude coefficient is k is added in original signal x (t)m(t), to be processed after obtaining plus making an uproar Signal xm(t), such as formula (2)；
xm(t)=x (t)+knm(t) (2)
B) to xm(t) EMD decomposition is carried out, p IMF component C is obtainedj,m(j=1,2 ..., p), Cj,mIndicate the m times test Decompose j-th obtained of IMF component；
C) as m < M, m=m+1, return step is 2.；
3. each IMF decomposed to M times calculates mean value, such as formula (3):
4. exportingAs j-th IMF, j=1,2 ... that EEMD is decomposed, p；
(2) IMF and energy balane are adaptively chosen:
Several IMF components are obtained after carrying out EEMD decomposition to signal, further choosing most can faults transient state component IMF component, feature calculation is done to it, the quick spectrum kurtosis map algorithm based on spectrum kurtosis is for realizing overall experience mode point The selection of optimal IMF component after solution；
Kurtosis reflects the numerical statistic amount of stochastic variable distribution character, is 4 rank central moments of normalization, to all IMF components Kurtosis is sought, the maximum component of kurtosis is most close to the transient state component of failure；Then IMF component maximum to kurtosis carries out calculating energy Amount, IMF ENERGY E are calculated such as formula (4):
Wherein A is the amplitude of IMF component, t1For sample start times, t2The time is tied for sampling, t is the sampling time；
(3) average current magnitude is poor；
It is poor to define average current magnitude, it is made up into the latter with transient state energy progress integrated application and is only applicable to uneven event The defect of barrier realizes the faulty comprehensive route selection of institute；Average current magnitude difference DI is defined as follows:
ΔIi=Ifi-Issi, i=A, B, C (5)
In upper two formula: IfiFor the maximum value of failure phase three-phase current amplitude, IssiFor pre-fault stable stage three-phase current The maximum value of amplitude；△IA、△IB、△ICCurrent amplitude before and after respectively the failure of A, B, C three-phase occurs is poor；
The route all to power distribution network asks its DI, the DI of faulty line to be significantly greater than non-fault line；
(4) high frequency metric data obtains:
The standard for being measured PMU device and being communicated by IEEE Std C37.118.1 and C37.118.2 is it is found that PMU device The frequency of upload report be 50HZ, and the signal of this low frequency is unable to get enough fault transient features, but the substation PMU High frequency sample rate can be up to 20kHZ or more and being locally stored；On the other hand, PMU carries fault oscillograph, and sample frequency reaches To 5kHZ or more, extracted for transient characteristic.
EEMD variation is done to the zero-sequence current of all feeder lines in the step 2, and calculates the kurtosis of all IMF components, it is right The maximum IMF component of kurtosis seeks its wave type energy.
The failure front and back line that the data of average current magnitude difference are recorded from fault oscillograph is calculated in the step 2 Then the three-phase current on road is calculated by formula (5) and (6).
The data in fault credibility calculated in the step 4 are from step 2 and 3 obtained high frequency transient energy It is poor with average current magnitude.
Fault credibility in the step 4 is the effective combination for realizing transient energy method and average current magnitude method, most One is provided eventually for judging the quantitative of faulty line as a result, the calculating such as formula (7) of fault credibility G is shown:
G in formulaiFor the fault credibility on i-th line road；N is the bus number in power distribution network；E is to obtain after EEMD is decomposed IMF component energy, calculation method is shown in formula (4)；DI is that the average current magnitude of route is poor, calculation method such as formula (5) (6).
The fault credibility of the step 5 all routes according to obtained in step 4, the failure of more all routes is credible Degree, fault credibility is maximum to be determined as faulty line.
The step 1 specifically includes the miniature PMU device local high frequency sampled voltage electric current phase installed on distribution network line Amount, and the real-time low frequency of calculating fault features amount instantaneous frequency maximum value is uploaded to main website；The step 2 specifically includes judgement failure Whether characteristic quantity there is exception, if abnormal, upload local high-frequency current data and provides data source for subsequent route selection, simultaneously Extract the data of the included fault oscillograph of miniature PMU；The step 3 specifically includes the zero-sequence current for providing fault oscillograph Data carry out EEMD decomposition, choose high frequency transient component by calculating spectrum kurtosis, and further calculate the high frequency transient of each route Energy；It is poor that the step 4 specifically includes the average current magnitude that the three-phase current data of local high frequency sampling are calculated each route； The step 5, which is specifically included, calculates each route according to the high frequency transient energy and average current magnitude difference for each route being calculated Fault credibility, compare fault credibility, the highest route of fault credibility is faulty line.
Compared with existing distribution network fault line selection technology, the invention has the advantages and beneficial effects that:
1. existing power distribution network selection method is all offline inspection judgement mostly, and the present invention is in conjunction with the online of miniature PMU The included fault recorder data of the synchro measure data and PMU of detection carries out failure line selection, and distribution network line event may be implemented Hinder online route selection, provides judgment basis for quickly excision failure, have to the stability of realization intelligent distribution network, safety great Meaning.
2. there are various defects, the present invention to extract zero using population mean Mode Decomposition for power distribution network transient line selection The transient high-frequency component of sequence current signal, this method have the advantage of adaptive decomposition, and eliminate traditional modal decomposition side Modal overlap effect brought by method.
3. since most of distribution network line fault is singlephase earth fault, so existing fault-line selecting method is all needle To singlephase earth fault, and these methods can not detect two-phase, three-phase fault, define a kind of average current magnitude difference herein Method, suitable for all fault types selection method.
4. selection method of the present invention combines two methods of transient energy method and average current magnitude method, all types are used Earth fault, route selection result are more accurate.
5. the present invention define a fault credibility come it is quantitative provide route selection as a result, fault credibility it is highest be therefore Hinder route, quantitative fault credibility is more simple and clear.
Detailed description of the invention
Understand for the ease of those of ordinary skill in the art and implement the present invention, with reference to the accompanying drawing and specific embodiment The present invention is described in further detail, and the following examples are intended to illustrate the invention, it is to be understood that protection model of the invention It encloses and is not limited by the specific implementation.
Fig. 1 is " substation-main website " fault upload framework flow chart
Fig. 2 is population mean Mode Decomposition flow chart
Fig. 3 is the flow chart of transient state energy route selection method
Fig. 4 is the comparison diagram of two methods
Fig. 5 is the selection method overall flow figure that the present invention designs
Specific embodiment
The present invention is a kind of feeder line fault diagnostic method and system based on PMU data, comprising:
Step 1: local substation uploads high-frequency data when failure with " substation-main website " co-architecture；
Step 2: it is poor that all feeder line average current magnitudes being calculated according to high-frequency data；
Step 3: the fault recorder data that the fault oscillograph carried according to PMU provides does all routes of EEMD transformation calculations High frequency transient energy；
Step 4: calculating the fault credibility of all routes；
Step 5: final fault feeder is determined according to fault credibility.
EEMD variation is done to the zero-sequence current of all feeder lines in the step 2, and calculates the kurtosis of all IMF components, it is right The maximum IMF component of kurtosis seeks its wave type energy.
The failure front and back line that the data of average current magnitude difference are recorded from fault oscillograph is calculated in the step 2 Then the three-phase current on road is calculated by formula (5) and (6).
The data in fault credibility calculated in the step 4 are from step 2 and 3 obtained high frequency transient energy It is poor with average current magnitude.
The fault credibility of the step 5 all routes according to obtained in step 4, the failure of more all routes is credible Degree, fault credibility is maximum to be determined as faulty line.
The present invention improves EMD decomposition, and the same of the advantage for remaining EMD decomposition adaptive decomposition is decomposed using EEMD When eliminate the defect of modal overlap.In addition, to avoid single selection method bring defect, the present invention is using high frequency transient While energy route selection, all fault types may be implemented accurately selects in conjunction with a kind of method of the average current magnitude difference of definition Line.It is limited by communication speed, the high frequency sampled data of the substation PMU can not all be uploaded to main website, and the present invention devises one kind " substation-main website " synergetic structure, the structure only upload high-frequency current phasor after failure occurs, do fault diagnosis for the present invention and provide High-frequency data support.
(1) EEMD algorithm.
White Gaussian noise signal is added, due to height in basis of the EEMD based on EMD algorithm during carrying out EMD to signal This white noise signal has the decomposition scale of frequency-flat distribution；Meanwhile for anomalous events such as pulses present in signal It can effectively handle.There is two above characteristic based on EEMD, can effectively solve lacking for the modal overlap of EMD algorithm generation It falls into.
The EMD decomposition of signal x (t) is exactly to be broken down into one group of IMF component CiWith remainder rnThe sum of, as shown in formula (1).
The essence of EEMD algorithm be exactly before carrying out EMD decomposition to signal, using Gauss white noise frequency-flat characteristic and Random white noise is added in the smoothing effect to random disturbances ingredient, to original signal in it, then to adding the signal after making an uproar to carry out Iterative analysis, final iteration obtain IMF component.
Specific step is as follows for EEMD algorithm:
(1) the amplitude coefficient k that EMD executes total degree M and white noise is added is initialized, i.e. k, m=1；
(2) the m times EMD is executed to decompose；
A) the random Gaussian white noise n that amplitude coefficient is k is added in original signal x (t)m(t), to be processed after obtaining plus making an uproar Signal xm(t), such as formula (2).
xm(t)=x (t)+knm(t) (2)
B) to xm(t) EMD decomposition is carried out, p IMF component C is obtainedj,m(j=1,2 ..., p), Cj,mIndicate the m times test Decompose j-th obtained of IMF component；
C) when m < M, m=m+1, return step (2)；
(3) each IMF decomposed to M times calculates mean value, such as formula (3).
(4) it exportsAs j-th IMF, j=1,2 ... that EEMD is decomposed, p.
(2) IMF and energy balane are adaptively chosen.
Several IMF components are obtained after carrying out EEMD decomposition to signal, needing further to choose most can faults transient state Then the IMF component of component does feature calculation to it, the selection of usual IMF component is chosen dependent on the experience of user, The present invention is based on the quick spectrum kurtosis map algorithms of spectrum kurtosis for realizing the optimal IMF component after overall experience mode decomposition It chooses.
Kurtosis is to reflect the numerical statistic amount of stochastic variable distribution character, is 4 rank central moments of normalization, to all IMF points Amount seeks kurtosis, and the maximum component of kurtosis is most close to the transient state component of failure.Then IMF component maximum to kurtosis carries out calculating energy Amount, IMF ENERGY E are calculated such as formula (4).
Wherein A is the amplitude of IMF component, t1For sample start times, t2For sample end time, t is the sampling time.
(3) average current magnitude is poor.
It can be changed based on three-phase current before and after failure, and the variation of faulty line is more obvious than non-fault line, the present invention Second characteristic quantity is defined using this fault signature, i.e. average current magnitude is poor, it is carried out integrated application with transient state energy The defect that the latter is only applicable to imbalance fault can be made up, realizes the faulty comprehensive route selection of institute.
Average current magnitude difference DI is defined as follows:
ΔIi=Ifi-Issi, i=A, B, C (5)
In upper two formula: IfiFor the maximum value of failure phase three-phase current amplitude, IssiFor pre-fault stable stage three-phase current The maximum value of amplitude, △ IA、△IB、△ICCurrent amplitude before and after respectively the failure of A, B, C three-phase occurs is poor.
The route all to power distribution network asks its DI, the DI of faulty line to be significantly greater than non-fault line.
(4) high frequency metric data obtains.
The standard for being measured PMU device and being communicated by IEEE Std C37.118.1 and C37.118.2 is it is found that country PMU The frequency of the upload report of device is 50HZ, and the signal of this low frequency is unable to get enough fault transient features, but PMU The high frequency sample rate of substation can reach 20kHZ or more and can be locally stored, and on the other hand, PMU carries fault oscillograph, adopts Sample frequency can reach 5kHZ or more, can be used for transient characteristic extraction.
It is recorded for the PMU data of feature extraction from the included failure of the high frequency sampling and PMU of the substation PMU in the present invention Wave device, wherein describing a kind of " substation-main website " co-architecture in later application content for the limitation for overcoming signal high-frequency transmission For uploading the short duration high frequency sampled data of failure front and back substation.
China's power distribution network generallys use small current neutral grounding mode, although fault current very little when breaking down, power distribution network is also A very long time can be run in case of a fault, but if long-term failure operation can match to power transformer or even entirely Electric network damages, and in order to realize safety, the stability of power distribution network operation, needs and its is purged to failure, so Distribution network fault line selection is always the difficult point and emphasis of fault diagnosis for a long time.Firstly, currently invention addresses future sync phases Phasor measurement unit large area can be installed on power distribution network, not only combine the data characteristics of PMU, while to traditional failure line selection Transient method is improved, and the overall experience Mode Decomposition that the present invention uses can eliminate mould brought by empirical mode decomposition State aliasing.Secondly, the present invention not only uses a kind of selection method of transient state energy, but combine one defined in the present invention Kind average current magnitude difference carries out comprehensive route selection judgement, more all-sidedly and accurately carries out route selection judgement, and the result of quantitative. Finally, the present invention proceeds from the reality, the requirement of data needed for the characteristics of being measured according to PMU data and selection method is clearly provided The source of data and devise the acquisition that the process that a kind of " substation-main website " collaboration reports realizes data.The content of present invention It with very high engineering application value, can be widely used in the power distribution network equipped with miniature PMU, realize the accurate choosing of failure Line.
A fault credibility G is also defined in the present invention, the purpose of definition is to realize the transient state proposed in the present invention Effective combination of energy method and average current magnitude method finally provides one for judging the quantitative of faulty line as a result, event Shown in the calculating such as formula (7) for hindering confidence level G.
G in formulaiFor the fault credibility on i-th line road；N is the bus number in power distribution network；E is to obtain after EEMD is decomposed IMF component energy, calculation method is shown in formula (4)；DI is that the average current magnitude of route is poor, and calculation method is shown in formula (5) (6).
The height that high-frequency data needed for the present invention is uploaded from the fault recorder data and the substation PMU carried derived from PMU Frequency sampling data, fault recorder data can upload in real time, and local high frequency sampled data is due to being limited to signal transmission rate It can not upload in real time, computing capability and data storage capacities be had based on the substation PMU, the present invention devises a kind of reasonable extraction " substation-main website " co-architecture of local high frequency sampled data, concrete operations process are as follows:
(1) the high frequency sampling in the substation PMU obtains effective value, the phase angle of high frequency.
(2) based on the phase angle being locally stored, instantaneous frequency f is calculated, shown in the calculating of instantaneous frequency f such as formula (8).
In formulaFor the instantaneous frequency of two neighboring sampled point；T is the high frequency sampling interval.
(3) each wave period takes the maximum value f of instantaneous frequencymax, fmaxWith low frequency (50HZ) with voltage and current phasor one It rises and is uploaded to main website.
(4) main website detects fmaxWhether abnormal, abnormal judgement is set as whether being greater than twice of fundamental frequency, i.e. fmaxWhether it is greater than 100HZ。
(5) if detecting f in step (4)maxThere is exception, then measures the high frequency of 5 wave periods before and after fault moment Amount is uploaded to main website and provides data for subsequent select-line analysis.