RU2540267C1 - Method for determining homogeneity of electrical quantity - Google Patents

Abstract

FIELD: electricity.

SUBSTANCE: invention relates to the field of electric engineering and may be used in relay protection and automatic equipment. In the method electrical quantity is measured at evenly fixed time periods, an adaptive filter is set to suppression of electrical quantity, output signal of the set filter is generated by processing of electrical quantity measurements made upon setting and is delivered to input of an actuating relay; against return of the actuating relay commencement of a new interval and ending of the previous interval for homogeneity of electrical quantity is fixed. The evenly shifted downconverted signals in time with fixed interval of downconversion are composed out of measurements of electrical quantity so that overlaying of all downconverted signals to the same time axis gives measurements of electrical quantity. The adaptive filter is set to suppression of one of downconverted signals, copies of the set adaptive filter are made with number equal to the number of downconverted signals, output signals of filter copies are defined at processing of downconverted signals and supplied to the actuating relay.

EFFECT: improved sensitivity at processing of electrical quantity with high frequency of measurements and potential detection and correction of electrical quantity measurements with surges.

3 cl, 5 dwg

Description

The invention relates to the field of electrical engineering, namely to relay protection and automation.

A device is known that implements a method for determining the uniformity of an electric quantity [1], according to which a structural analysis of an electric quantity is carried out using an adaptive filter, and the process is homogeneous by the level and ratio of components. The disadvantage of this method is its complexity, because for the purpose of determining the interval of homogeneity of the electrical quantity, the method requires tuning an adaptive filter and determining the structure of the signal at each sample.

Closest to the claimed invention in terms of use, technical nature and technical result achieved is a method for determining the intervals of homogeneity [2], according to which the electrical quantity is measured at uniformly fixed times and the adaptive filter is tuned to suppress the electrical quantity. Then the tuned filter is divided into direct and inverse converters. The signals of the direct and inverse converters are fed to the input of the executive relay, by the return of which they judge the beginning of a new and the end of the previous homogeneity intervals.

The disadvantage of the prototype is to reduce its sensitivity to changes in the structure of the electric quantity at a high measurement frequency (sampling frequency), especially when the signal structure in the new homogeneity interval differs from the structure of the previous interval only by the presence of an aperiodic component. This is caused by the fact that at a high measurement frequency, an adaptive filter tuned to suppress the fundamental harmonic greatly weakens the aperiodic component. Therefore, the signals of the direct and inverse converters will be close in level, and therefore the characteristic point, the coordinates of which are determined by the level of the signals of the direct and inverse converters, and in the new uniformity interval will be in close proximity to the bisector on the plane of the response characteristic of the actuating relay. This worsens the conditions for determining a new homogeneity interval and reduces the sensitivity of the method.

In addition, in the prototype there is no detuning from impulse noise (ejection) in electrical quantity. Any outburst will be perceived as the beginning of a new interval of uniformity.

The aim of the invention is to increase the sensitivity of the method for determining the intervals of uniformity in the processing of electrical quantities with a high measurement frequency and giving the method the ability to detect and adjust measurements of electrical quantities with emissions.

This goal is achieved by the fact that in the known method for determining the intervals of homogeneity of the electrical quantity, according to which the electrical quantity is measured at uniformly fixed time instants, the adaptive filter is configured to suppress the electrical quantity, the output signal of the tuned filter is generated by processing subsequent measurements of the electrical quantity after adjustment, and it is supplied the input of the executive relay and the return of the executive relay record the beginning of a new and the end of the previous int vomited uniformity of electrical quantities, introduce new operation is as follows. Initially, from the measurements of the electric quantity, the decimated signals uniformly shifted in time with a fixed decimation step are made so that the superposition of all the decimated signals on one time axis gives measurements of the electric quantity. Then, the adaptive filter is configured to suppress one of the decimated signals and form copies of the tuned adaptive filter by the number of decimated signals, the output signals of the filter copies are determined when processing their decimated signals, and they are fed to the executive relay.

One of the options for performing the response characteristics of the actuating relay is a star, the rays of which are segments symmetrical about the coordinate origin on the axes in the space of the filter output signals, the size of which is determined by the noise level in the signal. When using an executive relay with such a response characteristic, it is believed that a new homogeneity interval has been reached if the output signals of all filters are successively outside the star.

To determine and correct emissions in measurements of electrical quantities, it is proposed in the executive relay to control the order in which the filter signals go beyond the limits of the response characteristics. It is believed that there are emissions in the electric quantity if the sequence of the output of the filter signals beyond the limits of the response characteristic is violated. The measurement of the electrical signal with an ejection is replaced by a count predicted by the corresponding filter, the output signal of which is outside the response characteristics.

) и амплитудно-частотную характеристики адаптивного фильтра, настроенного на подавление основной гармоники: 1 - при исходной частоте дискретизации 1200 Гц, 2 - при виртуальном уменьшении частоты дискретизации за счет децимации с шагом ν=3 до 400 Гц. На фиг.4 представлена характеристика срабатывания исполнительного реле в виде трехмерной звезды. Фиг.5 иллюстрирует выявление выброса в измерениях электрической величины с помощью предлагаемого способа.Figure 1 shows the receipt of decimated signals u ₁ (l), u ₂ (l) and u ₃ (l) by the decimation step ν = 3 (• - measurements of the electric quantity, о - samples of decimated signals), as well as output signals of filter copies of the proposed method u _output1 (l), u _output2 (l) and u _output3 (l) signals at the outputs of the direct u _pr (k) and inverse u _inv (k) converters of the prototype, the output signal of the filter of the prototype u _output (k) = u _ol (k) -u _inv (k). Figures 2 and 3 contain exponential (the dependence of the filter output signal on the normalized attenuation coefficient βT _s under the action of an aperiodic component at the input $${\text{e}}^{\text{-}\beta T_s}$$

) and the amplitude-frequency characteristics of an adaptive filter tuned to suppress the fundamental harmonic: 1 - with an initial sampling frequency of 1200 Hz, 2 - with a virtual decrease in the sampling frequency due to decimation with a step of ν = 3 to 400 Hz. Figure 4 presents the response characteristic of the Executive relay in the form of a three-dimensional star. Figure 5 illustrates the detection of outliers in measurements of electrical quantities using the proposed method.

Designations used: k - number of measurement (reference) of the electric quantity u (k) or discrete time; l is the discrete time of decimated signals (enlarged time scale).

First, we consider the work of the prototype, and then explain the principle of the proposed method. We will illustrate the operation of the methods on signal processing of a real emergency process in the electric network [3], which consists of two intervals of homogeneity: segments of the preceding (in Fig. 1, samples with negative numbers) and transition modes (samples with positive numbers). The initial signal sampling frequency f _s = 1200 Hz (sampling period T _s = 1/1200 s).

The adaptive filter of the prototype is configured on the segment of the previous mode by the condition of suppressing the electrical quantity. The signal of the previous mode consists of the main harmonic with a frequency of 50.18 Hz and noise. Assuming that the signal-to-noise ratio is large enough, the signal of the previous mode can be represented by a simple harmonic model

,

,

where a ₁ is the tunable (desired) coefficient. The adaptive filter is adjusted in such a way that, by selecting the coefficient a ₁ , to achieve complete suppression of the fundamental harmonic of the signal. In this case, the filter output

will only contain signal noise. This is achieved at a ₁ = -2cos (2π50,18T _s ). Then the filter (1) is divided into channels of direct and inverse converters, the output signals of which

$$u_{P\text{R}}\left(k\right)=u\left(k\right)+u\left(k-2\right)\text{(2)}$$

and

$$u_{\mathrm{and}ne}\left(k\right)=-a_{\mathrm{one}}u\left(k-\mathrm{one}\right)=2\cos \left(2\pi \mathrm{fifty},\mathrm{eighteen}{T}_s\right)u\left(k-\mathrm{one}\right)\text{(3)}$$

are connected to the inputs of the executive relay.

The response characteristic of the prototype executive relay is two narrow sectors located near the bisector in the first and third quadrants of the coordinate plane with the axes u _pr (k) and u _inv (k). The width of the sectors is determined by the noise level in the electrical quantity.

Filter the interval preceding mode is already set (a 1 ratio is defined), then a low level noise output of the filter (1) u _O (k) ≈0 and output signal levels of the direct and inverse transformers will be close: u _ave (k) ≈u _inv (k). Therefore, the characteristic point with coordinates [u _ol (k), u _inv (k)] will be near the bisector on the plane of the response characteristics of the executive relay (within the response sector). This means that the current samples of the electric quantity u (k) processed by the filter belong to the interval of the previous mode.

(β=71,95 с^-1 - коэффициент затухания). Следовательно, выходной сигнал фильтра прототипа, уже настроенного на подавление основной гармоники, на данном интервале должен будет содержать только апериодическую составляющую и шум. Поэтому сигналы (2) и (3) на входе исполнительного реле должны будут значительно различаться. Однако, как видно из фиг.1, в выходном сигнале u_вых(k) фильтра прототипа апериодическая составляющая представлена слабо и практически неощутима на фоне шума, а сигналы на входе исполнительного реле u_пр(k) и u_инв(k) близки по уровню. Эта означает, что интервалы однородности, на которых сигналы отличаются друг от друга только наличием в одном из них еще и апериодической составляющей, будут слабо различимы для прототипа.On the interval of the transition mode (short circuit), the signal u (k) contains the fundamental harmonic of the same frequency as in the previous interval, and the aperiodic component with the base $${\text{e}}^{\text{-}\beta T_s}=0,9418$$

(β = 71.95 s ^-1 - attenuation coefficient). Therefore, the output signal of the filter of the prototype, already configured to suppress the fundamental harmonic, at this interval should contain only an aperiodic component and noise. Therefore, the signals (2) and (3) at the input of the executive relay will have to be significantly different. However, as can be seen from figure 1, in the output signal u _output (k) of the prototype filter, the aperiodic component is weak and almost imperceptible against the background of noise, and the signals at the input of the executive relay u _pr (k) and u _inv (k) are close in level . This means that the intervals of uniformity, at which the signals differ from each other only by the presence in one of them also of an aperiodic component, will be poorly distinguishable for the prototype.

The reason for the lack of prototype lies in the features of the exponential (figure 2) and amplitude-frequency (figure 3) characteristics of the filter (1). At the initial sampling frequency, the transfer coefficient of the aperiodic component is small and the practical one does not depend on the attenuation. At the same time, the filter strongly emphasizes the high-frequency components (Fig. 3), worsening the signal-to-noise ratio. Taking into account the suppression of the aperiodic component by the filter will allow us to restore the actual level of the aperiodic term, but will not change the signal-to-noise ratio, which will remain small. Therefore, the recognition of the aperiodic component in the prototype will be very difficult.

As can be seen from figure 2 and 3, the exponential and amplitude-frequency characteristics of the filter (1) can be significantly improved by reducing the sampling frequency. Such a filter suppresses the aperiodic component much less than with a signal with the original sampling frequency: for example, when the sampling frequency is reduced to 400 Hz, the gain of the constant component of the new filter (curve 2 in FIG. 2) is 8.6 times higher than at the initial frequency 1200 Hz (curve 1). Along with this, the amplification of the high-frequency components of the new filter is lower: the ratio of the maximum gain over the entire frequency band to the gain of the constant component (ωT _s = 0) at a sampling frequency of 1200 Hz is 57.3 (curve 1 in Fig. 3), and at 400 Hz, this ratio is an order of magnitude lower and equal to 5.78 (curve 2). Since noise is associated with high-frequency components, a decrease in the sampling frequency follows from here significantly increases the signal-to-noise ratio of the filter, improving the recognition conditions for the aperiodic component. In this case, oversampling of the initial electrical quantity is not required: the frequency is changed virtually due to decimation of the samples. Thanks to the use of these properties of the new filter, the proposed method outperforms the prototype in sensitivity.

In the proposed method, the advantages of a virtual change in the sampling frequency are fully realized due to the multi-channel processing of measurements (readings) of the electrical quantity.

For this, from the readings of the electric quantity, the decimated signals are uniformly shifted in time in such a way that their superposition on one time axis (the k axis in Fig. 1) gives measurements of the electric quantity (Fig. 1 shows the receipt of three decimated signals u ₁ (l) , u ₂ (l) and u ₃ (l) from measurements of the electric quantity u (k), decimation step ν = 3, sampling frequency of decimated signals 400 Hz).

The adaptive filter is then tuned to suppress any of the decimated signals. Since the first interval of the electrical quantity (samples with negative numbers in Fig. 1) consists only of the fundamental harmonic, the adaptive filter will be similar to the filter (1), but it will use the samples of the decimated signal (here the filter is set to the signal u ₁ (l)) :

$$u_{\mathrm{at}sx\mathrm{one}}\left(l\right)=u_{\mathrm{one}}\left(l\right)+b_{\mathrm{one}}{u}_{\mathrm{one}}\left(l-\mathrm{one}\right)+u_{\mathrm{one}}\left(l-2\right).\text{(four)}$$

The new filter (4) operates on a different time scale than the filter (1), therefore, the coefficient b ₁ = -2cos (2π50,18vT _s ) will be determined taking into account the virtual sampling period vT _s .

Since all decimated signals are a copy of the electrical quantity, but only in a new time scale, there is no need to configure a filter for each of them. The filter (4) tuned to the signal u ₁ (l) is completely adequate for the remaining decimated signals. Therefore, copies of the tuned adaptive filter (4) are formed for them:

$$u_{\mathrm{at}sx\text{i}}\left(l\right)=u_i\left(l\right)+b_{\mathrm{one}}{u}_i\left(l-\mathrm{one}\right)+u_i\left(l-2\right).\text{(5)}$$

, обрабатывающих децимированные сигналы u₂(l) и u₃(l).For the example of Fig. 1, u _i (l) output signals of the filters of the 2nd and 3rd channels $$i=\overline{2,3}$$

processing decimated signals u ₂ (l) and u ₃ (l).

$$\text{ i}=\overline{\text{1,3}}$$

. Так будет продолжаться до тех пор, пока не наступит граница нового интервала однородности (отрезок сигнала переходного режима - отсчеты с положительными номерами на фиг.1). Как уже отмечалось, второй интервал сигнала фиг.1 содержит основную гармонику и апериодическую составляющую. Так как фильтры (4) и (5) на предыдущем интервале однородности были настроены на подавление основной гармоники, то и на новом интервале они пропустят на свой выход все, кроме основной гармоники. Поэтому на выходе всех фильтров на границе интервалов однородности появится апериодическая составляющая, правда, по истечении собственного переходного процесса фильтров (в течение 3 отсчетов децимированного сигнала). Но собственный переходный процесс фильтров в предлагаемом способе, в отличие от прототипа, не влияет на процесс выявления границы интервала, если, на самом деле, достигнута граница интервала однородности. Как уже говорилось выше, при появлении в отсчетах электрической величины любого выброса прототип ложно определит границу интервала именно из-за собственного переходного процесса в фильтре (в прямом и инверсном преобразователях фильтра). В этом случае характеристическая точка, координаты которой на плоскости срабатывания определяются уровнем сигналов на выходе каналов прямого и инверсного преобразователей, покинет сектор срабатывания, что и приведет к ложному возврату исполнительного реле прототипа. Как будет показано ниже, предлагаемый способ свободен от этого недостатка.After setting the filter (4) in the first interval of signal uniformity, all filter output signals will be small (at the noise level): $$u_{\mathrm{at}sx\text{i}}\left(l\right)\approx 0,$$

$$\text{i}=\overline{\text{1.3}}$$

. This will continue until the boundary of a new interval of homogeneity arrives (segment of the transition mode signal — readings with positive numbers in FIG. 1). As already noted, the second signal interval of figure 1 contains the main harmonic and the aperiodic component. Since the filters (4) and (5) in the previous homogeneity interval were tuned to suppress the fundamental harmonic, then in the new interval they will miss everything except the fundamental harmonic at their output. Therefore, an aperiodic component will appear at the output of all filters at the boundary of the homogeneity intervals, however, after the expiration of the filter's own transient process (within 3 samples of the decimated signal). But the filter’s own transient in the proposed method, unlike the prototype, does not affect the process of identifying the boundary of the interval, if, in fact, the boundary of the interval of uniformity is reached. As mentioned above, when the electrical magnitude of any outlier appears in the samples, the prototype falsely determines the boundary of the interval precisely because of its own transient process in the filter (in direct and inverse filter converters). In this case, the characteristic point, the coordinates of which on the response plane are determined by the level of signals at the output of the channels of the direct and inverse converters, will leave the response sector, which will lead to a false return of the prototype executive relay. As will be shown below, the proposed method is free from this disadvantage.

If the actual boundary of the intervals of homogeneity is reached (k = 0), then, as can be seen from figure 1, in all filters of the proposed method there is a transient process (each on its axis at the moment l = 0). The actuation characteristic of the executive relay must be chosen so that it returns only when all the filters in the output successively appear at a signal that significantly exceeds the noise level. The filter, whose output signal first goes beyond the response characteristics, determines the reference number of the boundary of the homogeneity interval. The sequence of the output of the filter signals plays an important role, since it is a sign that provides a detuning from false triggering during emissions. In Fig.1, the first signal appears at the output of the first channel, then the second and third. In reality, the first signal may appear at the output of any of the filters - it all depends on the location of the interval boundary reference point on the k axis. But it is important that the order of the filters is maintained: for example, if at first the signal appeared at the output of the filter of the 2nd channel, then the signals at the output of the 3rd and 1st filter should appear sequentially, if the 3rd filter is the first, then the sequence signals at the output of the filters should be equal to 3-1-2.

, на осях в пространстве выходных сигналов фильтров (фиг.4). Число осей (фильтров) равно шагу децимации ν. В этом случае полагают, что достигнуто начало нового интервала однородности, если выходные сигналы всех фильтров последовательно окажутся за пределами звезды.One of the possible characteristics of the actuation of the actuating relay is a characteristic in the form of a star, the rays of which are segments [-U _i , U _i ] symmetrical with respect to the coordinate origin, $$\left(i=\overline{\mathrm{one},\nu }\right)$$

, on the axes in the space of the output signals of the filters (figure 4). The number of axes (filters) is equal to the decimation step ν. In this case, it is believed that a new homogeneity interval has been reached if the output signals of all filters are successively outside the star.

To detect outliers in measurements of an electrical quantity, the order of the output of the filter signals beyond the response characteristics is monitored. And it is believed that in the electric quantity there is a pulse interference (outlier) if the order of the output of the filter signals does not correspond to the sequence described above. As can be seen from figure 5, the outlier distorts only the samples of the 2nd decimated signal u ₂ (l), and therefore the transient occurs only in the second filter, the signal u _out2 (l) of which goes beyond the response characteristics (beyond the interval [-U ₂ , U ₂ ] in figure 4). However, the output signals of the remaining filters will remain within the response characteristics (within the segments [-U ₁ , U ₁ ] and [-U ₃ , U ₃ ]). Since the order of the output of the filter signals is violated, already at the next sample after the ejection, a decision is made to replace the previous sample with an estimate of the sample by the second filter.

The necessary estimate of the reference is determined from (5), assuming that u _{out i} (0) = 0 (for figure 5, the filter number is i = 2):

$${\widehat{u}}_i\left(0\right)=-b_{\mathrm{one}}{u}_i\left(-\mathrm{one}\right)-u_i\left(-2\right)\text{. (6)}$$

After replacing the outlier reference with estimate (6), the signal u _{subtract 2} (l) of the 2nd filter will decrease and will also return to the limits of the interval [-U ₂ , U ₂ ].

On whatever axis of the decimated signals, the ejection count would be located, the output signal of the corresponding filter, which is outside the response characteristics, will indicate the ejection location and will allow determining the measurement estimate according to (6).

Thus, the proposed method for determining the intervals of homogeneity of the electrical quantity has increased sensitivity in the processing of electrical quantities with a high frequency of measurements and the ability to detect and correct measurements of electrical quantities with emissions.

Literature

1. RF patent No. 2082270, class. HO2H 3/28, HO2H 7/045, 1994.

2. RF patent №2308137, class. HO2H 3/28, 2006.

3. Antonov V.I., Naumov V.A., Fomin A.I. Effective structural models of input signals of digital relay protection and automation // Electricity. 2012. No. 11. C.2-8.

Claims (3)

1. The method of determining the intervals of uniformity of the electrical quantity, according to which the electrical quantity is measured at uniformly fixed time instants, the adaptive filter is suppressed to suppress the electrical quantity, the output signal of the tuned filter is generated by processing the following electric quantity measurements after tuning, and it is fed to the input of the executive relay and on the return of the executive relay, the beginning of the new and the end of the previous intervals of uniformity of the electric quantity are fixed, ex which consists of the fact that the measurements of the electric quantity comprise the decimated signals uniformly shifted in time with a fixed decimation step so that the superposition of all the decimated signals on one time axis gives measurements of the electric quantity, adjust the adaptive filter to suppress one of the decimated signals, form copies of the tuned adaptive filter by the number of decimated signals, determine the output signals of filter copies when processing their decimated signals and ute them to the inputs of the executive relay. 2. The method according to claim 1, characterized in that the actuation characteristic of the actuating relay is performed in the form of a star, the rays of which are segments symmetrical with respect to the coordinate origin on the axes in the space of the filter output signals; and it is believed that a new homogeneity interval has been reached if the output signals of all the filters are successively outside the star. 3. The method according to claim 1 or 2, characterized in that it is believed that if the sequence of the output of the filter signals outside the response characteristics is violated, then there is a pulse interference (surge) in the electrical quantity, and the measurement of the electrical signal with the surge is replaced by the reference predicted by the corresponding filter, the output signal of which was outside the response characteristics.

Images