CN102721885A - Island effect detecting method based on wavelet analysis - Google Patents

Abstract

The invention relates to an island effect detecting method based on wavelet analysis, and belongs to the field of detection of electrical power systems. Problems that in the prior art, harmonics are inevitably injected in a power grid, and the quality of the power grid is damaged when an active detecting method is adopted, a wider undetected range appears, and detecting time is uncertain when a general passive detection method is adopted are solved. An island effect detector is used for detecting an island effect of a power supply system in the method, and a power grid in the power supply system and a distributed grid-connected generation system jointly supply power to an electricity load. The method comprises that a detector samples voltage signals of a common connection point of the power grid, the distributed grid-connected generation system and the electricity load; the voltage signals are decomposed in a four-order wavelet manner; and whether conditions that Ed4 is larger than or equal to A and Ed4 (i) is larger than or equal to B are simultaneously met or not is judged so as to determine generation of the island effect.

Description

A kind of island effect detection method based on wavelet analysis

Technical field

The present invention relates to a kind of island effect detection method, belong to the electric system detection range based on wavelet analysis.

Background technology

When electrical network because of happening suddenly former thereby when breaking away from, distributed grid-connected system is failed in time to detect the disengaging of electrical network and self is cut off, and continuation and load keep the state that moves, is called island effect.Island effect has sizable harmfulness; Relevant criterion requirement according to IEEE; Distributed grid-connected system must have the island detection ability; The control strategy of island effect need detect the disengaging of electrical network apace, so that break off being connected of distributed grid-connected system and electrical network in time, harm is dropped to minimum.

At present, known island effect detection method mainly is divided into active detection and passive type detection two big classes.The active detecting method is sent disturbing signal disturbance electrical networks such as voltage, frequency, phase place or power on one's own initiative, detects the parameters such as voltage, frequency, phase place or harmonic wave of points of common connection (PCC) simultaneously, judges the isolated island situation through the variation of parameter; The passive type detection method detects parameters such as the voltage, frequency, phase place harmonic of the electrical network of points of common connection (PCC) in real time and judges island effect.But island effect commonly used is is now prevented and treated method and all can be run into the problem of some; According to active detection side's ratio juris; Any active detection method all can't can be injected harmonic wave with avoiding in electrical network; Destroy power grid quality, and general passive type detection method can exist the bigger time that can not detect territory and detection uncertain.

Summary of the invention

The present invention seeks to existing to adopt active detection method all can't understand when detecting island effect in electrical network, to inject harmonic wave in order to solve with avoiding; Destroy power grid quality; And adopt general passive type detection method can have bigger uncertain problem of the time that can not detect territory and detection, a kind of island effect detection method based on wavelet analysis is provided.

A kind of island effect detection method according to the invention based on wavelet analysis; The island effect pick-up unit that this method relates to detects the island effect of electric power system; Said electric power system comprises electrical network, distributed grid-connected system, with electric loading, first isolating switch and second isolating switch, electrical network and distributed grid-connected system are the electricity consumption electric jointly, electrical network is by the first isolating switch CS state; Distributed grid-connected system is by the second isolating switch CS state

The island effect pick-up unit that this method relates to comprises detecting device and island effect detecting unit,

Detecting device comprises Hall voltage sensor, operational amplification circuit and A/D convertor circuit;

The island effect detecting unit comprises low-pass filtering module, quadravalence wavelet filter and island effect judge module;

Said a kind of island effect detection method based on wavelet analysis may further comprise the steps:

Step 1, Hall voltage sensor in real time sample detecting electrical network, distributed grid-connected system and with the voltage signal of electric loading three's points of common connection; And this voltage signal handled obtain a series of points of common connection voltage signal u (i), i=1,2; ...; N, wherein: i is a sampling number, n is a positive integer;

Step 2, quadravalence wavelet filter carry out the quadravalence wavelet decomposition to said points of common connection voltage signal u (i), obtain sampling quadravalence wavelet coefficient d4 (i) the i time;

The i time sampling quadravalence wavelet coefficient d4 (i) that step 3, island effect judge module obtain according to step 2; Calculate the quadravalence wavelet coefficient energy E of the i time sampling place power frequency period _D4With the i time sampling quadravalence wavelet coefficient amplitude D _{D4 (i)},

The quadravalence wavelet coefficient energy E of the i time sampling place power frequency period _D4By formula

Obtain,

In the formula: f is a power frequency, and k is a SF, and SF k is the integral multiple of public network frequency f,

The i time sampling quadravalence wavelet coefficient amplitude D _{D4 (i)}D by formula _{D4 (i)}=d4 (i)-d4 (i-2) obtains,

In the formula: d4 (i-2) is the i-2 time sampling quadravalence wavelet coefficient;

Step 4, judge whether to satisfy condition simultaneously E _D4>=A and D _{D4 (i)}>=B,

Judged result is then thought not have island effect for not; Judged result then thinks to have island effect for being, island effect judge module output control command lets distributed grid-connected system suspension,

In the formula: A is an energy threshold; wherein a is the energy threshold coefficient; And a gets 1.5～3,

be the mean value of the quadravalence wavelet coefficient energy of 5～10 power frequency periods of sampling under the normal state;

B is an amplitude thresholds;

wherein b is the amplitude thresholds coefficient; B gets 2～5,

be the quadravalence wavelet coefficient amplitude mean value of 10～50 sampled points of sampling under the normal state.

Advantage of the present invention: island effect detection technique of the present invention can not produce harmonic pollution to electrical network; Can reduce effectively to detect the generation that the territory also detects island effect accurately; What reduced the passive type detection method can not detect the territory; Can detect general passive type detection method the island effect that can't discern, and can promptly discern the characteristic of island effect, improve the detection speed that island effect is taken place significantly.

Description of drawings

Fig. 1 is the structural representation of a kind of island effect pick-up unit that relates to based on the island effect detection method of wavelet analysis according to the invention;

Fig. 2 is the general theory diagram of quadravalence wavelet filter;

Fig. 3 utilizes wavelet analysis method analytic signal exemplary plot;

Fig. 4 is the wavelet analysis figure of PCC voltage under-frequency;

Fig. 5 is the superpotential wavelet analysis of PCC voltage.

Embodiment

Embodiment one: this embodiment is described below in conjunction with Fig. 1 to Fig. 5; The said a kind of island effect detection method of this embodiment based on wavelet analysis; The island effect pick-up unit that this method relates to detects the island effect of electric power system; Said electric power system comprises electrical network 1, distributed grid-connected system 2, with electric loading 3, first isolating switch 4 and second isolating switch 5, electrical network 1 is common for supplying power with electric loading 3 with distributed grid-connected system 2, electrical network 1 is by first isolating switch, 4 CS states; Distributed grid-connected system 2 is by second isolating switch, 5 CS states

The island effect pick-up unit that this method relates to comprises detecting device 6 and island effect detecting unit 7,

Detecting device 6 comprises Hall voltage sensor 6-1, operational amplification circuit 6-2 and A/D convertor circuit 6-3;

Island effect detecting unit 7 comprises low-pass filtering module 7-1, quadravalence wavelet filter 7-2 and island effect judge module 7-3;

Said a kind of island effect detection method based on wavelet analysis may further comprise the steps:

Step 1, Hall voltage sensor 6-1 real-time sampling detection of grid 1, distributed grid-connected system 2 and with the voltage signal of electric loading 3 threes' points of common connection; And this voltage signal handled obtain a series of points of common connection voltage signal u (i); I=1,2..., n; Wherein: i is a sampling number, and n is a positive integer;

Step 2, quadravalence wavelet filter 7-2 carry out the quadravalence wavelet decomposition to said points of common connection voltage signal u (i), obtain sampling quadravalence wavelet coefficient d4 (i) the i time;

The i time sampling quadravalence wavelet coefficient d4 (i) that step 3, island effect judge module 7-3 obtain according to step 2; Calculate the quadravalence wavelet coefficient energy E of the i time sampling place power frequency period _D4With the i time sampling quadravalence wavelet coefficient amplitude D _{D4 (i)}'

The quadravalence wavelet coefficient energy E of the i time sampling place power frequency period _D4By formula

Obtain,

In the formula: f is a power frequency, and k is a SF, and SF k is the integral multiple of public network frequency f,

The i time sampling quadravalence wavelet coefficient amplitude D _{D4 (i)}D by formula _{D4 (i)}=d4 (i)-d4 (i-2) obtains,

In the formula: d4 (i-2) is the i-2 time sampling quadravalence wavelet coefficient;

Step 4, judge whether to satisfy condition simultaneously E _D4>=A and D _{D4 (i)}>=B,

Judged result is then thought not have island effect for not; Judged result then thinks to have island effect for being, island effect judge module output control command lets distributed grid-connected system suspension,

In the formula: A is an energy threshold;

wherein a is the energy threshold coefficient; And a gets 1.5～3,

be the mean value of the quadravalence wavelet coefficient energy of 5～10 power frequency periods of sampling under the normal state;

B is an amplitude thresholds;

wherein b is the amplitude thresholds coefficient; B gets 2～5,

be the quadravalence wavelet coefficient amplitude mean value of 10～50 sampled points of sampling under the normal state.

The island effect problem can be described to the up-down of PCC point voltage amplitude, the increase and decrease of PCC point voltage frequency, and the variation of PCC point voltage harmonic wave and saltus step of voltage-phase or the like, these all are the characteristics that is used for recognizing the island effect problem.When electrical network breaks away from; Common point PCC voltage will inevitably bring unexpected variation; This variation can be introduced a lot of high fdrequency components, and is to catch these high fdrequency components by the analysis of time domain, and the Fourier analysis method can't satisfy real-time requirement again; Therefore, this method is introduced wavelet analysis and is detected island effect through analyzing PCC place voltage signal.Island effect detection method based on wavelet analysis carries out wavelet analysis through the voltage signal that PCC is ordered, and differentiates island effect.Because when electrical network breaks away from, if need of load power and photovoltaic generating system provided does not match, the amplitude and the frequency of PCC place voltage will change this moment; Be that circuit can run on new steady state point, at new place, working point, photovoltaic generating system provides whole active power of load and reactive power; In this process, if degree of not matching is not high, general passive type detection algorithm will be malfunctioning; Because on time domain, see; Voltage or change of frequency little, but common point PCC voltage will inevitably bring unexpected variation when electrical network breaks away from, and this variation can be introduced a lot of high fdrequency components; And be to catch these high fdrequency components by the analysis of time domain, thereby wavelet analysis just can capture the generation that these information pick out island effect.

Fig. 1 is a structural representation of the present invention; Voltage signal through PCC is ordered detects in real time; The island effect detecting unit carries out the wavelet decomposition analysis to the voltage signal that the detecting device transmission comes then; When island effect takes place, thereby the generation that island effect is told in huge variation can take place down at the yardstick that we set in the wavelet coefficient that wavelet analysis obtains, and detects this device behind the result and to the control section transmission signal of distributed grid-connected system it is quit work.

The wavelet analysis of this method and device adopts the Daub4 small echo to carry out the analysis to signal; Use the quadravalence wavelet decomposition; Detecting device 6 detects the voltage of points of common connection PCC in real time, and converts voltage signal to digital signal transfers to island effect detecting unit 7 through processing of circuit, and digital voltage signal at first passes through low-pass filtering treatment; Eliminate the influence of noise to signal analysis; Then signal is carried out wavelet decomposition, selecting analysis module that the signal after decomposing is analyzed through scale coefficient, judge through the island effect judge module whether island effect takes place and produce to judge signal at last.

Shown in Figure 2 is realizes with wave filter that the theory diagram of wavelet decomposition, resolution decompose signal placed respectively and decomposes under the different scale and draw detailed information.Remainder is a part of describing signal, is decomposed into the approximate information of signal.Each detailed information and approximate information all can from before more high-resolution approximate information obtain, and more approximate information can be through to do convolution resultant with high pass and low-pass filter respectively under the high resolving power.

The small echo that we adopt among the present invention is based on the little wave structure of Daub4, and Daub4 is the compactest small echo and can simply realizes.The decomposable process of small echo is as shown in Figure 2.Realize in DSP that column count is down found the solution and realized that the small echo quadravalence decomposes, obtain the coefficient of dissociation under following each yardstick of quadravalence of small echo.

$a1={(u\left(n\right)*g\left(n\right))}_{2\↓}=\underset{n}{\mathrm{\Σ}}g(2n-k)u\left(k\right)$

$\mathrm{<figure-callout\; id="1"\; label="d"\; filenames="HDA00001817150500011.png,HDA00001817150500031.png"\; state="\{\{state\}\}">d</figure-callout>}1={(u\left(n\right)*h\left(n\right))}_{2\&DownArrow;}=\underset{n}{\mathrm{\&Sigma;}}h(2n-k)u\left(k\right)$

$a2={(a1\left(n\right)*g\left(n\right))}_{2\&DownArrow;}=\underset{n}{\mathrm{\&Sigma;}}g(2n-k)a1\left(k\right)$

$\mathrm{<figure-callout\; id="2"\; label="d"\; filenames="HDA00001817150500011.png,HDA00001817150500031.png"\; state="\{\{state\}\}">d</figure-callout>}2={(a1\left(n\right)*h\left(n\right))}_{2\&DownArrow;}=\underset{n}{\mathrm{\&Sigma;}}h(2n-k)a1\left(k\right)$

$a3={(a2\left(n\right)*g\left(n\right))}_{2\&DownArrow;}=\underset{n}{\mathrm{\&Sigma;}}g(2n-k)a2\left(k\right)$

$d3={(a2\left(n\right)*h\left(n\right))}_{2\&DownArrow;}=\underset{n}{\mathrm{\&Sigma;}}h(2n-k)a2\left(k\right)$

$a4={(a3\left(n\right)*g\left(n\right))}_{2\&DownArrow;}=\underset{n}{\mathrm{\&Sigma;}}g(2n-k)a3\left(k\right)$

$\mathrm{<figure-callout\; id="4"\; label="d"\; filenames="HDA00001817150500032.png"\; state="\{\{state\}\}">d</figure-callout>}4={(a3\left(n\right)*h\left(n\right))}_{2\&DownArrow;}=\underset{n}{\mathrm{\&Sigma;}}h(2n-k)a3\left(k\right)$

Wherein a1, a2, a3 and a4 are the scale coefficient of wavelet decomposition; D1, d2, d3 and d4 are the wavelet coefficient of wavelet decomposition, and wherein: d1 is the single order wavelet coefficient, and d2 is the second order wavelet coefficient, and d3 is three rank wavelet coefficients, and d4 is the quadravalence wavelet coefficient; 2 ↓ be two extractions, the serve as reasons Hi-pass filter of equal value of the discrete little wave structure of Daub4 of the serve as reasons low-pass filter of equal value of the discrete little wave structure of Daub4 of g (n), h (n).

Four groups of discrete wavelets that obtained u (n) through this series formula divide solution sequence, yardstick sequence and little wave train under the corresponding different scale of difference, the characteristic of the voltage signal of reflection under different yardsticks.

Like Fig. 3 is an example that utilizes the wavelet analysis method analytic signal.Among the figure first row waveform is represented the signal waveform that needs are analyzed; The four lines of back is respectively the coefficient of wavelet decomposition; Wherein second row is represented scale coefficient; The third line is represented wavelet coefficient to fifth line, and original signal is made up of the wavelet coefficient under scale coefficient and the different scale, can find out the signal coefficient that on behalf of different scale, the third line observe down to fifth line; The as can be seen from the figure wavelet decomposition of signal, thus can be to different needs to the analyzing and processing in addition of the wavelet coefficient under the different yardsticks.

Under system's normal condition; Disturbance in any transmission or noise; The capital brings high fdrequency component into for the PCC voltage signal, although in step 2 the voltage signal of PCC has been carried out the LPF noise reduction process, is still risky if simple application is provided with threshold values according to the amplitude of wavelet coefficient; Therefore can cause judging by accident disconnected generation, use energy according to wavelet coefficient and threshold values is set threshold values is set island effect is judged

The present invention finds to adopt the quadravalence wavelet coefficient can judge the generation of island effect fast and accurately through the lot of test experience.Therefore the present invention is the principal character parameter with the energy settings of quadravalence wavelet coefficient, i.e. 2 norms of quadravalence wavelet coefficient.Each power frequency period of the energy of quadravalence wavelet coefficient once calculates, and formula is following:

${\mathrm{<figure-callout\; id="4"\; label="E\; d"\; filenames="HDA00001817150500032.png"\; state="\{\{state\}\}">E</figure-callout>}}_d=\sqrt{\frac{\underset{i=1}{\overset{k/\mathrm{<figure-callout\; id="4"\; label="f\; d"\; filenames="HDA00001817150500032.png"\; state="\{\{state\}\}">f</figure-callout>}}{\mathrm{\&Sigma;}}}{d}^{}{4}^2\left(i\right)}{k/f}},$

Calculate the quadravalence wavelet coefficient energy E of power frequency period _D4,

The present invention is set at the supplemental characteristic parameter with the difference of quadravalence wavelet coefficient amplitude, and formula is following:

D _d4(i)＝d4(i)-d4(i-2)

Rule of judgment according to step 5 determines whether to take place island effect then.

Several kinds of typical situation when taking place in the face of island effect are are down analyzed and researched, and the signal of detection is the voltage signal that PCC is ordered.When electrical network broke away from, power did not match hour (active power＜20% or reactive power＜5%), can cause the voltage magnitude of common point PCC or electric voltage frequency can't exceed normal territory and can't find the island effect phenomenon.Among this zone, passive type detection method commonly used will lose efficacy, and this method is just studied in this zone.

The reactive power that first kind of situation supplied with when electrical network is non-vanishing; But degree of not matching hour; When non-vanishing (be reactive power suspension) electrical network can't be supplied with the variation that can cause PCC point voltage frequency because after the Δ Q suspension, and the frequency that is caused PCC voltage when suppose the island effect generation by Δ Q has descended 5%.

The wavelet analysis of PCC voltage under-frequency shown in Figure 4.Mains frequency is 50Hz, and in the time of 0.2 second, electrical network breaks away from, and this moment, PCC point voltage frequency descended 5%.As can be seen from the figure; In the time of 0.2 second; Huge fluctuation has just taken place in the several sampling periods after island effect takes place in the wavelet coefficient under the high resolving power, recovers later and work as to break away from, and this is because disengaging is that common point PCC voltage signal will inevitably bring unexpected variation; This variation can be introduced a lot of high fdrequency components; Therefore can find promptly just that electrical network breaks away from the high fdrequency component of bringing, and then find the generation of the island effect that common passive detection method can't be discerned, it quit work for then the control section transmission signal of distributed grid-connected system 2.

Another kind of situation is that the active power of supplying with when electrical network is non-vanishing; But do not match hour; The variation of PCC voltage magnitude can take place; The situation that the amplitude of supposing when island effect takes place to be caused by active power PCC voltage has risen, for the variation that makes amplitude continuous, the mode that we take the e index to rise.After electrical network broke away from, PCC voltage raise with the e exponential form, and the PCC voltage signal is continuous at dispersal point like this, this situation and met in practice, and when finishing to emulation, voltage raises 5% altogether.

Fig. 5 is the superpotential wavelet analysis of PCC voltage, and mains frequency is 50Hz, and in the time of 0.2 second, electrical network breaks away from, and this moment, PCC voltage raise with the e exponential form, and when finishing to emulation, voltage raises 5% altogether.As can be seen from the figure: in the time of 0.2 second, huge fluctuation has also taken place in the several sampling periods after island effect takes place in high-resolution wavelet coefficient, and this process is the same when changing with frequency.Through detection to the high-frequency wavelet coefficient, just can find the generation of the island effect that common passive detection method can't be discerned rapidly, it is quit work for then the control section transmission signal of distributed grid-connected system 2.

Embodiment two: this embodiment is described further embodiment one, and the 28335 type digital signal processors that island effect detecting unit 7 adopts TI company to produce are realized.

The island effect detecting unit 7 of this device is made up of digital signal processor system; Adopt TI company 28335 type digital signal processors, the function of processor mainly comprises four modules: low-pass filtering module 7-1, quadravalence wavelet filter 7-2 and island effect judge module 7-3.Three modules are coordinated to be integrated in the digital signal processing chip to realize through software.The test section is made up of Hall voltage sensor 6-1, operational amplification circuit 6-2 and A/D convertor circuit 6-3.

Claims (3)

1. island effect detection method based on wavelet analysis; It is characterized in that; The island effect pick-up unit that this method relates to is put the island effect that detects electric power system; Said electric power system comprises electrical network (1), distributed grid-connected system (2), with electric loading (3), first isolating switch (4) and second isolating switch (5), for supplying power with electric loading (3), electrical network (1) is by first isolating switch (4) CS state jointly for electrical network (1) and distributed grid-connected system (2); Distributed grid-connected system (2) is by second isolating switch (5) CS state

The island effect pick-up unit that this method relates to is put and is comprised detecting device (6) and island effect detecting unit (7), and detecting device (6) comprises Hall voltage sensor (6-1), operational amplification circuit (6-2) and A/D convertor circuit (6-3);

Island effect detecting unit (7) comprises low-pass filtering module (7-1), quadravalence wavelet filter (7-2) and island effect judge module (7-3);

Said a kind of island effect detection method based on wavelet analysis may further comprise the steps:

Step 1, Hall voltage sensor (6-1) real-time sampling detection of grid (1), distributed grid-connected system (2) and with the voltage signal of electric loading (3) three's points of common connection; And this voltage signal handled obtain a series of points of common connection voltage signal u (i), i=1,2; ...; N, wherein: i is a sampling number, n is a positive integer;

Step 2, quadravalence wavelet filter (7-2) carry out the quadravalence wavelet decomposition to said points of common connection voltage signal u (i), obtain sampling quadravalence wavelet coefficient d4 (i) the i time;

The i time sampling quadravalence wavelet coefficient d4 (i) that step 3, island effect judge module (7-3) obtain according to step 2; Calculate the quadravalence wavelet coefficient energy E of the i time sampling place power frequency period _D4With the i time sampling quadravalence wavelet coefficient amplitude D _{D4 (i)}'

The quadravalence wavelet coefficient energy E of the i time sampling place power frequency period _D4By formula

Obtain,

In the formula: f is a power frequency, and k is a SF, and SF k is the integral multiple of public network frequency f, the i time sampling quadravalence wavelet coefficient amplitude D _{D4 (i)}D by formula _{D4 (i)}=d4 (i)-d4 (i-2) obtains,

In the formula: d4 (i-2) is the i-2 time sampling quadravalence wavelet coefficient;

Step 4, judge whether to satisfy condition simultaneously E _D4>=A and E _{D4 (i)}>=B,

Judged result is then thought not have island effect for not; Judged result then thinks to have island effect for being, island effect judge module output control command lets distributed grid-connected system suspension,

In the formula: A is an energy threshold; wherein a is the energy threshold coefficient; And a gets 1.5～3,

be the mean value of the quadravalence wavelet coefficient energy of 5～10 power frequency periods of sampling under the normal state;

B is an amplitude thresholds;

wherein b is the amplitude thresholds coefficient; B gets 2～5,

be the quadravalence wavelet coefficient amplitude mean value of 10～50 sampled points of sampling under the normal state.

2. a kind of island effect detection method based on wavelet analysis according to claim 1 is characterized in that, the 28335 type digital signal processors that island effect detecting unit (7) adopts TI company to produce are realized.

3. a kind of island effect detection method according to claim 1 and 2 based on wavelet analysis; It is characterized in that; The process of in the step 1 this voltage signal being handled comprises: utilize operational amplification circuit (6-2) to carry out signal and amplify; Utilizing A/D convertor circuit (6-3) again is digital signal with analog signal conversion, utilizes low-pass filtering module (7-1) to carry out LPF and noise reduction again.

Images