RU125743U1 - TRANSITION MODE MONITORING SYSTEM IN POWER UNION ACCORDING TO DIGITAL RECORDERS - Google Patents

Abstract

A system for monitoring transient conditions in an energy connection according to digital recorders, which contains at least two data generation and processing circuits, each of which includes a digital sensor and a storage device connected in series, characterized in that a block for calculating the cross-correlation function is introduced, and in each from the data generation and processing circuits, a data format preparation and conversion unit has been introduced, the input of which is connected to the output of the storage device, a data verification and control unit, the path of which is connected to the output of the preparation and conversion unit of the data format, the sampling step changing unit, the input of which is connected to the output of the verification and data control unit, the power density spectrum determining unit, the input of which is connected to the output of the sampling step changing unit, the spectrum analyzer, the input of which is connected with the output of the power density spectrum determination unit, a data filtering unit, the first input of which is connected to the output of the spectrum analyzer, and the second input is connected to the output of the step change unit d sampling, autocorrelation function calculation unit and harmonic amplitude and phase determination unit, the input of which is combined with the input of the autocorrelation function calculation unit and connected to the output of the data filtering unit, and the output is the output of the corresponding data generation and processing circuit, while the inputs of the cross-correlation function calculation unit connected to the outputs of at least two chains of the formation and processing of data.

Description

The utility model relates to specialized computing devices and can be used to create systems for processing and analyzing data from digital recorders for monitoring transient conditions in an energy pool, in particular, based on spectral analysis of the results of vector measurements and calculation of cross-correlation functions.

The transient monitoring system (SMR) used in power systems, based on the use of digital recorders, is used to regularly verify their basic dynamic model. Data analysis of digital recorders is performed, as a rule, only in case of technological irregularities, accompanied by significant power imbalances. Obviously, a full-fledged analysis of the vibrational properties should be based not only on the study of the behavior of the energy combination during technological failures, but also on the study of their vibrational properties in quasi-steady-state modes, considered over long time intervals. Such an analysis will reveal the sources and causes of low-frequency oscillations and develop, if necessary, recommendations for their elimination by correcting the operating settings of the system stabilizers of automatic excitation regulators (ARVs) of generators or installing additional system stabilizers on specific generators during their modernization or reconstruction.

A device is known that contains a group of analog sensors, a group of digital sensors, multichannel first analog and second digital switches, a module driver, a zero-organ, a voltage reference, a decoder, first, second and third random access memory, read-only memory, a microcontroller, a timer, first-fourth single-channel analog switches, analog-to-digital converter, first and second analogue comparators, register, first-fifth counters, first-third triggers, eleme T AND-NO, the first to fifth AND gates, the first to fourth OR elements, the first to sixteenth monostable multivibrator, the numerical comparator and a clock pulse generator [RU 2376625, C1, G06F 17/40, 20.12.2009].

The disadvantage of this device is the relatively narrow functionality.

A device is also known for remote monitoring of current and voltage parameters in the high-voltage part of electric power systems, including transient monitoring in these systems, comprising a high-voltage measuring module connected to a high-voltage network, and including a passive mains transducer magnetically coupled to the high-voltage network and / or a passive network voltage converter electrically connected to the high-voltage network, wherein, the high-voltage measuring module contains secondary power supply connected to the secondary power supply unit, magnetically connected to the high-voltage network, a low-voltage current transformer and / or electrically connected to the high-voltage network and included in the passive converter circuit of the mains voltage, a low-voltage voltage transformer with a filter capacitor, shunting the primary winding, and parallel him a damping resistor, an active transducer of measurement information signals connected to a passive transducer with a mains current transformer and / or a passive mains voltage transducer and a secondary power supply unit and having radio-frequency and / or optical outputs for the converted measurement information signals, and the passive mains voltage transformer is made in the form of series-connected high-voltage reference capacitor and low-voltage arm, moreover, all elements of a high-voltage the measuring module, in addition to the high-voltage reference capacitor, is placed in an electric shield connected to the mains cable through the throttle and a damping resistor parallel to it [RU 2143165, C1, H02J 13/00, 12.20.1999].

The disadvantage of this device is the relatively narrow functionality.

The closest in technical essence to the proposed one is a device for recording transient parameters of voltage and current changes in electric networks, containing a group of analog sensors, a group of digital sensors, analog and digital switches, first to fourth counters, a group of random access memory devices, read-only memory, microcontroller, timer, register, analog-to-digital converter, clock, D-flip-flops, first to fourth single vibrators [RU 2402067, C1, G06F 17 / 40, 10.20.2010].

The disadvantage of the device is its relatively narrow functionality, since when registering the parameters of transient modes it does not allow, in particular, to carry out such important functions as spectral analysis of the results of vector measurements and to evaluate the cross-correlation function.

The objective of the proposed system for processing and analyzing data from digital recorders for monitoring transient conditions in an energy connection is to expand the functionality. The required technical result consists in expanding the arsenal of technical means ensuring the expansion of the system’s functionality, in particular, to perform such important functions as spectral analysis of the results of vector measurements and evaluate the cross-correlation function.

The required technical result is achieved by the fact that, in a system containing at least two data generation and processing circuits, each of which includes a digital sensor and a storage device connected in series, a cross-correlation function calculation unit is introduced, and in each of the formation circuits and a data processing unit has been introduced, a data format preparation and conversion unit, the input of which is connected to the output of the storage device, a data verification and control unit, the input of which is connected to the output of the preparation unit, and data format transformations, a sampling step change unit, the input of which is connected to the output of the verification and data control unit, a power density spectrum determination unit, the input of which is connected to the output of the sampling step change unit, a spectrum analyzer, the input of which is connected to the output of the power density spectrum determination unit, data filtering unit, the first input of which is connected to the output of the spectrum analyzer, and the second input is connected to the output of the unit for changing the sampling step, the unit for calculating the autocorrelation function and the unit for determining the amplitudes and phases of harmonics, the input of which is combined with the input of the unit for calculating the autocorrelation function and connected to the output of the data filtering unit, and the output is the output of the corresponding circuit for generating and processing data, while the inputs of the unit for calculating the cross-correlation function are connected to the outputs, at least two chains of data formation and processing.

The drawing shows a structural diagram of a system for monitoring transient conditions in an energy connection according to the data of digital recorders for the special case of using two chains of data formation and processing.

The transient monitoring system in the energy connection according to the data of digital recorders (an example of using two data generation and processing circuits) contains two data generation and processing circuits 1-1 and 1-2, each of which includes a digital sensor 2 and a memory 3 connected in series.

In addition, each of the chains 1-1 and 1-2 of the formation and processing of data contains a unit 4 for preparing and converting the data format, the input of which is connected to the output of the storage device 3, a unit 5 for verification and control of data, the input of which is connected to the output of the preparation unit 4 and converting the data format, block 6 changes the sampling step, the input of which is connected to the output of block 5 of verification and control of data, block 7 determining the power density spectrum, the input of which is connected to the output of block 6 changing the sampling step, is analyzed 8 of the spectrum, the input of which is connected to the output of the power density spectrum determination unit 7, a data filtering unit 9, the first input of which is connected to the output of the spectrum analyzer 8, and the second input is connected to the output of the sampling step change unit 6, autocorrelation function calculation unit 10, and block 11 determining the amplitudes and phases of harmonics, the input of which is combined with the input of the autocorrelation function calculation unit 10 and connected to the output of the data filtering unit 9, and the output is the output of the corresponding circuit 1-1 (1-2) of formation and processing At the same time, the inputs of the block 12 for calculating the cross-correlation function are connected to the outputs of at least two circuits 1-1 and 1-2 of data generation and processing.

In a more general case, when the system uses more than two data generation and processing circuits, the inputs of the cross-correlation function calculation unit 12 are connected to the outputs of at least two data formation and processing circuits, or a separate block is used for each pair of circuits 1-1 and 1-2 formation and processing of data, or a switch is used that provides the connection of block 12 to the required pairs of circuits.

The system, in addition to standard elements (digital sensors, storage devices) also contains elements that are characterized at the functional level, and the described form of their implementation involves the use of a programmable (customizable) multifunctional tool, therefore, below is presented information confirming the possibility of such a tool performing the specific instructions This system functions, including computational algorithms and corresponding mathematical expressions.

The system for monitoring transient conditions in the power supply according to the data of digital recorders is operating as follows.

The system belongs to specialized computing devices and can be used to monitor transient conditions in an energy combination, in particular, spectral analysis of the results of vector measurements and the cross-correlation function. Such an analysis will allow to identify the sources and causes of low-frequency oscillations and develop, if necessary, recommendations for their elimination by correcting the operating settings of the system stabilizers of automatic excitation regulators (ARVs) of generators or installing additional system stabilizers on specific generators during their modernization or reconstruction.

The measurement results from digital sensors 2 installed on the controlled objects of energy connections (generators) for a long period of measurement are received in the storage device 3.

Unit 4 for preparing and converting the source data format of each of the data generation and processing circuits reads information from the storage device and converts it into the required binary format used for further processing.

In the same block, preliminary transformations can be performed, which include the elimination of jumps in the readings of the absolute angle of the direct sequence voltage (the jump here means the difference of two adjacent readings exceeding 180 °) and the removal of a constant bias of 50 Hz in the frequency readings.

The jumps are due to the limited range of readings to 0-360 °. The absolute angle of the direct sequence voltage itself changes continuously, therefore, to prevent distortions in the spectral analysis, it is necessary to exclude jumps in the readings. This is possible because spectral analysis is carried out over a limited time range. For this, pairs of adjacent readings of absolute angles are sequentially scanned. If the difference between the current and previous readings is greater than 180 °, then 360 ° is subtracted from the current and all subsequent readings. On the contrary, if this difference is less than 180 °, the current and all other readings are increased by 360 °.

After the correction, the results are reviewed, and, if necessary, all readings are changed by a multiple of 360 ° so that the difference between the maximum and the minimum module does not exceed 360 °. Thus, the indications of the absolute angle of the voltage of the direct sequence will remain within acceptable limits.

Changes in the frequency readings compared to its average value are very small, which can lead to a loss of accuracy in the calculations. Therefore, it is advisable to switch to deviations from 50 Hz with the replacement of Hz by MHz, i.e. multiplying the result by 1000.

The source data may contain the following parameters of the electric mode transmitted from the digital sensors to the storage device 3:

- the instantaneous value of the absolute angle of the voltage of the direct sequence δ;

is the instantaneous voltage of the direct sequence voltage U;

is the instantaneous value of the direct sequence current I;

- the value of the frequency of the electric current f;

- astronomical time T;

- value of active power P;

- value of reactive power Q.

For each parameter, a task for reading from the storage device is generated, containing:

- an identification number;

- parameter.

In addition to specifying objects and parameters, data is generated about the beginning and end of the sample:

- date of the start of data sampling (in the format day, month, year, hours, minutes, seconds);

- end date of the sample (day, month, year, hours, minutes, seconds).

During cyclic data processing, this range is divided into a number of subbands with a duration equal to the data processing period.

Based on the results of reading from the database of the storage device 3, a list of satisfied requests and the corresponding binary data array is formed, the length of which is proportional to the period of data processing during cyclic processing, or to the sampling range in the absence of cyclic processing. Thus, the parameters for which at the time of the request there was no data are excluded from the list of processed parameters.

Block 5 verification and control of the source data converts the array with the source data into a form suitable for further analysis. In particular, anomalous measurements (slips) are detected in the data array.

An important role is played by statistical criteria designed to highlight abnormal results of measurements (emissions). The system can use the Grubbs criterion, which is used to check for anomalies (to assess abnormalities) of the distinguished measurement results, like most existing criteria for rejecting "suspicious" data, based on the assumption that the observed random variables belong to the normal law. The statistics of the Grubbs criterion provide the possibility of checking for the presence in the sample of either one anomalous measurement result (smallest or largest), or two (two smallest in the sample or two largest).

Using the Grubbs criteria, the observed sample is ordered in increasing order.

Let X ₁ , X ₂ , ..., X _n be the observed sample, X ₍₁₎ ≤X ₍₂₎ ≤ ... X _(n) is the variational series constructed from it.

When checking the outlier of the largest sample value X _{(n), the} statistics of the Grubbs criterion has the form:

G _n = (X _(n) -X) / S

Where:

When checking for the emission of the smallest sample value, the calculated statistics takes the form

The maximum or minimum sample element is considered an outlier if the value of the corresponding statistics exceeds the critical: G _n ≥G _{n, 1-α} or G ₁ ≥G _1,1-α , where α is the specified significance level.

When checking for the emission of the two highest values simultaneously, the statistics of the Grubbs criterion has the form

Where

To test for the emission of simultaneously the two smallest values, the statistic of the criterion takes the form:

Where:

Both values are considered outliers if the value of the relevant statistics is below critical :. G <G _α .

The statistics for checking for anomalies of both the minimum and maximum sample values are generated in accordance with the ratio:

Where:

Both values are considered outliers for a given significance level α, if the statistic value calculated from the sample is below the critical value: G _{1, n} <G _{1, n, α} .

In the system, tabular values can be used for samples with a length from the range n = 5-30 and 5 significance levels, equal in percentage terms: 0.1, 0.5, 1, 5, and 10.

The absence of an abnormal outlier for one statistic does not guarantee its absence for another. If an abnormal outlier (outliers) has been detected, then it is removed from the test section and repeated control is performed for the shortened section.

The choice of the length of the tested segments and the significance level is carried out experimentally.

When measuring absolute angle, digital sensors are characterized by the presence of jumps associated with the readings beyond 360 °. Since the lengths of the processed arrays are finite, jumps can be eliminated by adding or subtracting a multiple of 360 ° from the readings. If the correction is carried out before the elimination of mistakes, it is necessary to analyze the behavior of their environment. At a sampling frequency of digital sensors of 50 Hz, the range of transmitted frequencies is 25 Hz. The studied frequencies do not exceed 1 Hz. To increase the speed of the system blocks, it may be appropriate to increase the sampling time step, i.e. to decimate the initial samples, combined with smoothing to prevent frequency swapping during data thinning.

The volume of the generated input sample is determined by the period of the analysis results. The volume of the processed sample may be larger. Therefore, in the general case, old data, when new information is received, are not completely replaced, but are shifted by the corresponding amount to free up space for new data. At the beginning of the work, several steps may be required to fill the arrays until they are ready for further processing.

The parameters measured by modern digital sensors 2 have a time sampling step of 20 ms. Such a step allows one to study the spectral properties of signals up to 25 Hz. Significant frequencies of low-frequency oscillations in the power system do not exceed several hertz. To increase the speed of operation of the nodes and components of the proposed device and the device as a whole, it may be appropriate to increase the time discretization step, i.e. to decimate the original samples. For example, one additional sampling step of 100 ms can be used. The transition to this step in block 6 changes the sampling step is easily implemented - the sequence of five measurements is replaced by one average value, which is assigned to the midpoint of the five. So, for example, a sample for 100 ms will be formed as the average of samples for 60, 80, 100, 120 and 140 ms. To save the first value in a dataset with a zero time offset, you can accept its original value. The limiting frequency for arrays with a step of 100 ms is 5 Hz.

The determination (calculation) of the power density spectrum in block 7 is performed using a discrete Fourier transform. To reduce the influence of the "side lobes" associated with the limited length of the processed arrays of information, windows are used. Windows reduce the distortion of the power density spectrum, but on the other hand, degrade the resolution of the method. The more effective the window suppresses the "side lobes", the lower the resolution. The system provides a choice of several types of windows - a rectangular window, a raised cosine (Hamming window) and weighted cosines (Nuttall window).

Fast Fourier Transform (FFT) algorithms can be used for sequences of arbitrary lengths, but the most effective algorithms remain when the length is a multiple of 2. Arrays of this length can be obtained by adding zeros to the main sequence. In this case, the frequency step of the obtained power density spectrum also changes.

Block 7 determining the power spectrum searches for frequencies with increased density of the spectral characteristics, as a rule, 2-3 values.

A discrete Fourier transform is used:

where Δt is the time quantization step; x (i) = x (i · Δt) - samples of the initial process multiplied by the values of the window function w (i); T is the duration of the initial process; ΔF = 1 / T is the quantization step in frequency; N = T / Δt is the number of samples of the initial process; X (k) = X (k · ΔF) are the Fourier transform points for the process in the frequency domain.

The calculation uses the fast Fourier transform. In the algorithm, instead of considering the real function as complex with the zero imaginary part, the speed is doubled by reducing the problem to a problem with two times fewer values, but a nonzero imaginary part. The algorithm works with sequences whose length is a multiple of 2. There are FFT methods for any length of sequences, but their speed is less. To bring the original sequence to the desired length, it can be padded with zeros.

Three types of windows are considered: rectangular, raised cosine (Hamming window) and weighted cosines (Nuttall window). A rectangular window essentially means transforming the original process without averaging, and is given here for comparison.

Hamming Window:

w (n) = 0.54 + 0.46cos (2π [n]),

Nuttall window:

w (n) = 0.3635819 + 0.4891775cos (2πt [n]) + 0.1365995cos (4πt [n]) + 0.0106411cos (6πt [n]),

where: t [n] = (n- (N-1) / 2) / (N-1), 0≤n≤N-l.

If the sequence was continued by zeros, then the length of the initial sequence should be used as the length N when calculating the window weighting coefficients.

From the obtained spectrum, the power spectral density P (k) can be calculated by the formula:

The spectrum of the signal can be linear in nature, in it numerous local maxima alternate with minima, so in order to highlight areas of the spectrum of interest, it is necessary to pre-process the obtained dependence. This process is carried out in the spectrum analyzer 8.

Oscillations whose effective amplitude is less than the critical specified value are removed from the filter list. Critical amplitudes are indicated for each parameter in the configuration file.

The local maximum for the lowest frequency corresponds to the trend, and can be removed from the list. Oscillations whose effective amplitude is less than a given value are removed from the list.

Block 9 filtering the spectrum performs the selection of signals at frequencies with high spectral density, a list of which was determined by the analyzer 8 spectrum. For each frequency, the parameters of the corresponding symmetric non-recursive bandpass filter are calculated.

For the calculation, the bandwidth, the width of the transition zone and the amplitude of the allowable ripple in the passband and delay are set. The bandwidth is selected so that the bandwidths of different filters do not overlap.

The unit 10 for calculating the autocorrelation functions calculates the autocorrelation functions based on the signals generated by the 9 data filtering unit. For a stationary random process, the values of the autocorrelation function tend to zero. If the signal has an undamped oscillatory component, the values of the autocorrelation function become periodic. The autocorrelation function has a power dimension, so that by the amplitude of its oscillations, one can judge the amplitude of the oscillatory component of the signal. The oscillation frequency of the autocorrelation function corresponds to the frequency of the oscillatory component. In reality, both the amplitude and frequency of the oscillations change, so that only their average values can be used.

The correlation function characterizes the relationship of the values of functions in time. When comparing a function with itself, a function is called an autocorrelation function, when comparing different functions, a cross-correlation function. When comparing not the values of the function themselves, but deviations from the average value, we speak of covariance. Since in this case the correlation function is calculated for functions filtered by a band-pass filter having a zero mean value, this difference is not significant.

The correlation function is defined as the mathematical expectation of the product of the values of the functions. As an estimate of the correlation of sequences of the same length x (i), y (i), i = 0, 1. ... N-1, use the formula:

Here i is the delay expressed in steps of the grid. Multiplying by the grid step, we can obtain the delay in units of time τ. In the case of autocorrelation, the expression s _xx (0) coincides with the dispersion of the signals.

The delay may be negative. At the same time, it is easy to see that for negative i there will appear terms that go beyond the given sequences. Therefore, for this case it is more convenient to modify the formula as follows:

For calculations, it is convenient to use another function:

It is assumed that the functions x () and y () in the last two relations are equal to zero outside the interval [0, N-1]. As in the previous case of signal convolution, FFT can be used. Since the correlation spectrum of functions can be obtained by component-wise multiplication of the spectrum of the signal x (p) by the conjugate spectrum of the signal y (p). This property allows one to perform an FFT (in a time of the order of N · log ₂ N), component-wise multiply the obtained sets of complex numbers, and then perform the inverse transformation to obtain the required correlation function.

After calculating corr (x, y) _i, we can determine s _xy (i) by the formula:

In block 11, the amplitudes and phases of the harmonics extracted by block 9 of the filtering are determined.

Block 12 for calculating the cross-correlation functions for given pairs of power connection control objects on which digital sensors 2 are located, the measurement results of which are supplied to their corresponding memory devices 3, in block 12 the cross-correlation functions are calculated.This allows the direction of the disturbance source to be determined by shifting the maximum availability.

Thus, by expanding the arsenal of technical means, it is possible to expand the system's capabilities when processing and analyzing data from digital recorders, since it provides an additional opportunity to perform spectral analysis of the results of vector measurements and evaluate the cross-correlation function when monitoring transient conditions in an energy connection. Moreover, due to the use of the introduced block for changing the sampling step, it becomes possible to control the efficiency and accuracy of the spectral analysis of the results of vector measurements.

Claims (1)

A system for monitoring transient conditions in an energy connection according to digital recorders, which contains at least two data generation and processing circuits, each of which includes a digital sensor and a storage device connected in series, characterized in that a block for calculating the cross-correlation function is introduced, and in each from the data generation and processing circuits, a data format preparation and conversion unit has been introduced, the input of which is connected to the output of the storage device, a data verification and control unit, the path of which is connected to the output of the preparation and conversion unit of the data format, the sampling step changing unit, the input of which is connected to the output of the verification and data control unit, the power density spectrum determining unit, the input of which is connected to the output of the sampling step changing unit, the spectrum analyzer, the input of which is connected with the output of the power density spectrum determination unit, a data filtering unit, the first input of which is connected to the output of the spectrum analyzer, and the second input is connected to the output of the step change unit d sampling, autocorrelation function calculation unit and harmonic amplitude and phase determination unit, the input of which is combined with the input of the autocorrelation function calculation unit and connected to the output of the data filtering unit, and the output is the output of the corresponding data generation and processing circuit, while the inputs of the cross-correlation function calculation unit connected to the outputs of at least two circuits of the formation and processing of data.

Images