RU2720710C1 - Method for relay protection of power facility - Google Patents

Abstract

FIELD: electric power engineering.

SUBSTANCE: method for relay protection of a power facility is carried out by fixing counts of electrical values, converting counts to an intermediate value, converting intermediate values into metering, training of relay protection by measurements from simulation models and determination of characteristics of relay protection operation, intermediate values are formed at each step of observation of short-circuit mode and converted to current measurement of relay protection, training of protection is carried out by current measurements formed during simulation modeling. Results of the preliminary simulation modeling are added to the simulation modeling results storage unit, performing analogue-to-digital conversion of current and voltage signals with their subsequent digital filtration, as well as determination of parameters of currents and voltages in the form of intermediate values, based on the obtained intermediate values by their linear combination in the computing unit, generic measurement features are generated, which are supplied to the comparison circuits, the number of which is selected corresponding to the number of generalized measurement features, the second inputs of comparison circuits with the simulation modeling results storage unit are supplied with protection actuation parameters for each generalized feature, signals from the outputs of the comparison circuits are supplied to a logic unit in which, based on signals from the outputs of the comparison circuits, a protection actuation signal is generated depending on the need to ensure high reliability of operation or sensitivity of the relay protection device, generation of generalized features and protection actuation parameters supplied to comparison circuits, taking into account the protected power object functioning mode, therefore, when changing the mode with signals from the output of the simulation modeling results storage unit, calculating the generalized features in the computing unit and changing the protection operation parameters on the second inputs of the corresponding comparison circuits.

EFFECT: design of a method for relay protection of a power facility having high recognition capability of emergency modes, as well as high-speed and simple technical implementation.

1 cl, 7 dwg, 3 tbl

Description

The invention relates to the electric power industry, namely to relay protection and automation (RZA) of electrical systems.

One of the areas of intellectualization is related to relay protection training. Known methods of relay protection, where training is an essential technical feature [for example, RF Patent No. 2316871, IPC H02H 3/40, 2006; RF patent No. 2316872, IPC H02H 3/40, 2006; RF patent No. 2404499, IPC Н02Н 3/40, 2009; RF patent No. 2450402, IPC H02H 3/40, 2010]. The training is conducted with a teacher, whose role is a simulation model of the protected object, reproducing its modes. The model can be modified in relation to the type of mode. Two types are fundamentally distinguished: controlled modes (α-modes) to which the protection must respond, and alternative modes (β-modes) in which the protection is strictly prohibited.

For example, in the case of remote protection of the power line, the controlled modes include short circuits in the protection zone, and alternative ones include short circuits outside the circuit, short circuits “behind”, as well as all non-emergency modes.

In technical solutions [for example, RF Patent No. 2316871, IPC H02H 3/40, 2006; RF patent No. 2316872, IPC H02H 3/40, 2006; RF patent No. 2404499, IPC Н02Н 3/40, 2009; RF patent No. 2450402, IPC H02H 3/40, 2010] it is assumed that the relay protection and automation device is multidimensional, that is, for its operation several features are involved that characterize the emergency mode and are extracted from the corresponding emergency currents and voltages. The simultaneous use of several features for recognition of emergency conditions against the background of permissible leads to the need to formulate complex decision-making procedures and reduce the performance of relay protection and automation devices.

Thus, the disadvantage of the known methods of relay protection is the complex decision-making procedure for recognizing emergency conditions and the low speed of relay protection and automation.

The closest technical solution is the method of relay protection of an energy object [RF Patent No. 2594361 IPC H02H 3/40, publ. 08/20/2016, bull. No. 23] using its simulation models in controlled short-circuit modes and in alternative modes by fixing samples of electrical quantities, converting samples to an intermediate value, converting intermediate values to measurement, training relay protection with measurements from simulation models and determining the response characteristics on the measurement plane. According to the proposal, intermediate values are formed as current signals at each step of observation of the short circuit mode, starting from two samples of each value, current signals are converted to the current measurement of relay protection, protection training is carried out by current measurements at each observation step, and for each observation step they determine their own response characteristics protection.

The disadvantage of the method of relay protection of an energy facility is a complex decision-making procedure for recognizing emergency conditions and reduced speed.

Once again, we note that the application of the information approach in relay protection and automation equipment is accompanied by an increase in the dimension of the attribute space and a corresponding complication of the decision-making procedure taking into account multidimensional settings [for example, Lyamets, Yu.Ya. Multidimensional Relay Protection, Part 2. Relay Recognition Analysis. / Yu.Ya. Lyamets, G.S. Nudelman, D.V. Zinoviev // Electricity. - 2009. - No. 11. - C. 9-15.]. Moreover, the information content of both individual parameters of currents and voltages analyzed by the protection, and their combination can be different. Digital relay protection and automation technology allows you to create complex algorithms using many different recognition features and has, as a rule, increased sensitivity, but often requires large computational costs [for example, Kulikov, A.L. Analysis of the implementation of the information approach in relay protection / A.L. Kulikov, D.I. Soulless // Relay. - 2016. - No. 2. - C. 28-33].

It is advisable to develop relay protection and automation algorithms based on generalized features that make it possible to implement decision-making procedures taking into account "information compression" and requiring significantly less computation.

One of the options for the transition to generalized signs of relay protection and relay protection is the application of the principal component method. The method of principal components, also known as the Karunen-Loev decomposition, is one of the methods for reducing the dimension of a feature space used in recognition and regression problems [for example, Duda, R. Pattern recognition and scene analysis: Per. from English G.G. Weinstein and A.M. Vasilkovsky, ed. V.L. Stefanyuk / R. Duda, P. Hart. - M .: Mir, 1976. - 511 c].

The Karunen-Loev transformation is suitable for use in recognition problems, since it allows you to abandon complex multidimensional classification algorithms and introduce a single generalized attribute or attribute space that has a significantly smaller dimension than the original training sample.

In relation to relay protection and automation of electric power facilities, the transition to generalized signs involves the accumulation of statistics on the parameters of currents and voltages selected for measurement, followed by analysis and selection, based on the main component method, of the most informative signs sufficient to implement the recognition procedure. In particular, you can use only the first main component and build on its basis a one-dimensional relay protection. Such protection will have the simplest implementation, from the point of view of calculations. It is enough to compare with the setpoint a linear combination of primary features.

To illustrate the proposed approach, we consider the embodiment of the relay protection and automation section of the network (Fig. 1) using a one-dimensional generalized feature. Suppose that the task of relay protection and control devices is to disconnect interphase short circuits on the ω1 line. In this case, the parallel line ω2 is not included in the coverage area of the line ω1, therefore, an appropriate detuning of protection against short circuits on ω2 must be provided.

As with any recognition method based on simulation modeling, to construct a relay protection based on the principal component method, it is necessary to obtain the required statistical set for the values of the primary features that reflect the actual behavior of the modeled object in various modes. In this case, as a rule, the Monte Carlo method is used, which consists in the multiple implementation of the model of the process under study with parameters generated randomly from predetermined ranges. In particular, such equivalent circuit parameters as equivalent EMF and system resistances change over time, therefore, when modeling, instead of specific values, one should set the possible ranges of variation of the indicated values. Additionally, we define the place of the short circuit as a relative random variable that varies in the range from 0 to 1. Zero corresponds to a fault at the very beginning of the line, and one corresponds to the end of the line. We believe that the line resistances are unchanged and their values are uniquely determined. Constant and variable parameters of the equivalent circuit are given in table. 1 and 2.

Each experiment to obtain a training sample begins with the generation of variable parameters of the equivalent circuit. For this, a random number generator is used that generates random values independent of each other and having a uniform distribution within given ranges. After all parameters for the current experiment are obtained, the calculation is made by compiling and then solving the system of contour equations that determine the analyzed equivalent circuit. The solution of the system of equations in the form of a set of currents in the branches and voltages in the nodes of the circuit allows us to obtain the values of primary signs with the subsequent formation of a generalized sign.

The simulation results of the equivalent circuit (FIG. 1) are presented in FIG. 2.a and they allow you to analyze the change in the effective value of the short-circuit current Ia in case of damage on the protected and parallel (indicated by different background) lines, depending on the removal of the short-circuit point. For clarity, in FIG. 2.a shows only two-phase short circuits between phases A and B.

To ensure selectivity, the digital RPA algorithm must not allow protection to be activated during short circuit on a parallel line. Analysis of the histogram of the distribution of currents f (Ia) (Fig. 2.b) shows that when using the value of the current of the damaged phase as a recognizing sign, taking into account the detuning of the setpoint from the maximum short-circuit current on the buses of the substation B, most of the short circuits on the line ω1 do not fall into response area. An accurate calculation shows that the protection performed in this way will ensure that only 40.85% of all faults on the ω1 line are switched off.

Approximately the same percentage of trips can be achieved if, instead of current Ia , the value of the active power of the damaged phase Pa , measured at the place of installation of the protection (Fig. 2.c, d) is used. short circuits. Obviously, the use of similar parameters of phase B ( Ib and Pb ) will lead to similar results.

However, combining all four features ( Ia , Ib , Pa , Pb ) into a single four-dimensional space and highlighting the first main component in it, one can achieve a significant increase in the quality of mode recognition. It is advisable to divide the procedure for implementing such a relay protection algorithm into several stages: the formation of a training sample; centering and rationing of the training sample; obtaining a correlation matrix; calculation of eigenvalues and eigenvectors of the correlation matrix; determination of protection operation parameters; calculation of coefficients to obtain a generalized characteristic.

Training sample formation

It is supposed to carry out N necessary model experiments, in each of which the values used to build the defense are fixed - Ia , Ib , Pa and Pb . In the first N / 2 experiments, SCs on the ω1 line (recognizable mode) are simulated, while in the remaining N / 2 experiments, SCs on the ω2 line (alternative mode) are simulated. The dimension N of the necessary model experiments is determined from the requirements of representativeness of the sample.

Subsequently, the results of simulation are recorded in the matrix Q , formed as follows. The line number corresponds to the number of the experiment, and the results of the experiments for recognized and alternative modes are placed one below the other. The column number corresponds to the characteristic number. The principles of filling the matrix Q are illustrated in Table. 3.

Centering and rationing the training sample

The main task of centering is to shift the coordinates of the points of the training set so that the mathematical expectation of each of the signs is 0. Accordingly, standardization includes a procedure that scales the set of points of the training set to ensure equality of 1 variance (standard deviation).

признаков по выражению (1) To implement these tasks, we obtain mathematical expectations

signs by expression (1)

(1)

(1)

- математическое ожидание признака с номером m; q _{n , m} - элемент матрицы Q, представляющий собой значение признака m в эксперименте n. where m is the number of the sign; n is the number of the model experiment ; N is the number of experiments for each of the modes;

- the mathematical expectation of the sign with the number m ; q _{n , m} is the element of the matrix Q , which is the value of the attribute m in experiment n.

Similarly, we obtain the standard deviations for each of the signs (3)

(2)

(2)

и вектор-строку среднеквадратичных отклонений σ согласно выражениям (3) и (4) For further calculations, we introduce a row vector of mathematical expectations of signs

and a row vector of standard deviations σ according to expressions (3) and (4)

(3)

(3)

(4)

(4)

и σ принимают следующий видTaking into account the numerical results (tab. 3), the vectors

and σ take the following form

In order to obtain a normalized and centered sample for each dimension of each feature, we subtract the mathematical expectation of this feature and divide it by its standard deviation.

(5)

(5)

From the obtained results, we form the matrix B , whose components are determined by the expression (7)

(6)

(6)

Getting the correlation matrix

The correlation matrix ( S ) is a square matrix, the main diagonal of which contains units, and the remaining elements are the correlation coefficients between the corresponding random variables (features). In the formation of the correlation matrix S , centered and normalized sample values b _{n , m} (5) are used, and for the considered example, the correlation matrix takes the form

Getting eigenvectors and eigenvalues

The next step in the formation of a generalized feature is to obtain the eigenvectors and eigenvalues of the correlation matrix S. The eigenvector V of the matrix S is a vector whose multiplication by the matrix S gives a vector collinear with the vector V (that is, the vector V multiplied by some scalar value λ). Such a value of λ is called the eigenvalue of the matrix S.

We obtain the eigenvectors V ₁ - V _{4 of the} correlation matrix S , as well as the corresponding eigenvalues λ ₁ -λ ₄ according to [For example, Dolgopolov, D.V. Methods for finding the eigenvalues and eigenvectors of matrices: guidelines / D.V. Dolgopolov.- SPb .: SPbGTI (TU), 2005. - 39 p.]

The components of the eigenvector corresponding to the largest eigenvalue (the first generalized criterion) are the projection coefficients of the centered and normalized training sample on the axis of the first principal component ( C ₁ ), that is, on the axis along which the greatest dispersion is observed. Obtaining projection data is implemented by vector-matrix transformation

(7)

(7)

WherePC ₁ - a vector whose elements are projections of points on the axis of the first principal component.

Defining protection operation parameters

Let us construct the distribution of the projections of the training sample obtained by expression (7) on the axis of the first principal component (Fig. 3).

A comparative analysis of FIG. 3 from FIG. 2.b, d shows that the generalized characteristic is more suitable for classification than any of the initial characteristics. However, the transition to a generalized feature does not provide a complete separation of modes.

In relay protection, the approach is applied, which consists in a complete detuning from the operation of the protection outside its zone. Despite the decrease in sensitivity, this approach ensures guaranteed selective operation of several relay protection and automation devices operating without time delay. For the considered example, it is advisable to select the setpoint so that not a single short circuit on the line ω2 falls into the operation zone (Fig. 3). The setpoint selected in accordance with the above principles takes the value -1.035.

Taking the first main component as a sign for recognition and choosing the setpoint (Fig. 3) by the condition of complete detuning from alternative modes (short circuit on the ω2 line), you can get a rather high probability of disconnecting the short circuit on the ω1 line. In relation to the conditions analyzed in the example, the protection will disable 80.75% of all faults on the ω1 line, while a similarly implemented protection using one of the initial signs will disable no more than 45% faults (Fig. 2.b, d).

Calculation of coefficients to obtain a generalized characteristic.

The technical implementation of relay protection, based on the use of a generalized characteristic, involves a linear combination of the results of evaluating the parameters of currents and voltages.

If the results of the measurement of initial values of I _a, I _b, P _a and P _b present row vector M, when considering normalization procedures and centering (5), and the transformation (7) that the current value of the generalized characteristic h (vector projection M on the axis C ₁ ) is determined by the equality (8)

(8)

(8)

и коэффициент

вычисляются согласно соотношениямThe matrix

and coefficient

are calculated according to the relations

, (9)

, (9)

, (10)

, (10)

- диагональная матрица, состоящая из элементов вектора среднеквадратичных отклонений σ, полученного по выражению (5).Where

is the diagonal matrix, consisting of elements of the standard deviation vector σ obtained by expression (5).

не зависят от текущего режима работы защищаемого объекта, и, следовательно, могут быть вычислены заранее по результатам имитационного моделирования.Note that K and

do not depend on the current operating mode of the protected object, and, therefore, can be calculated in advance according to the results of simulation.

принимают следующие численные значения:For the considered example (Fig. 1) K and

take the following numerical values:

Thus, for the simplest implementation of the REA algorithm in real time, it is only necessary to obtain a linear combination of the initial features (parameters of currents and voltages) with previously known coefficients and compare it with the setpoint.

It should be noted that the considered approach to the formation of generalized relay protection features is specific for each recognizable situation. It is based on simulation with subsequent statistical processing of the results of simulation experiments. The proposed RP algorithm has a simplified implementation with high recognition ability and is adapted for each specific circuit-mode situation.

The objective of the invention is the creation of a method of relay protection of an energy facility with a high recognition ability of emergency modes, as well as speed and ease of technical implementation.

The task is achieved by the method of relay protection of an energy object using its simulation models in controlled short circuit modes and in alternative modes by fixing the readings of electrical quantities, converting the samples to an intermediate value, converting intermediate values to measurement, training relay protection with measurements from simulation models and determining the response characteristics relay protection, intermediate values are formed at each step of observation of the short circuit mode and converted into the current measurement of relay protection, protection training is carried out by current measurements formed during simulation. According to the proposal, various parameters of electrical quantities are selected as intermediate values, as well as their ratios formed from fixed samples, during simulation they form a training sample, center it and normalize it, followed by obtaining a correlation matrix of intermediate measurement values and calculating eigenvectors and eigenvalues of the correlation matrices of intermediate measurement values, select the generalized measurement features according to eigenvectors corresponding to the maximum eigenvalues of the correlation matrix of intermediate measurement values, generate generalized measurement indicators by a linear combination of intermediate values, and linear combination coefficients are selected in accordance with the values of the corresponding eigenvector of the measurement correlation matrix , determine the characteristics of the relay protection, based on its detuning from the acceptable modes of operation, by analyzing the statistical distributions of the generalized measurement attributes obtained from the simulation results, the general measurement attributes are generated from the current measurement and provide relay protection triggering when the generalized characteristics exceed the protection response characteristics.

In FIG. 1 is a diagram of a portion of an electrical network adopted for simulation.

FIG. 2.a characterizes phase A currents in case of faults of different distances on the protected and parallel lines for a section of the electric network of FIG. 1.

FIG. 2.b illustrates the statistical distribution of currents during short-circuit on lines ω1 and ω2 for a section of the electric network of FIG. 1.

FIG. 2.c characterizes the values of the active power of phase A during short-circuit of different distances on the protected and parallel lines for the section of the electric network of FIG. 1.

FIG. 2.d illustrates the statistical distribution of active capacities during short-circuit on the lines ω1 and ω2 for a section of the electric network of FIG. 1.

FIG. 3 illustrates the statistical distribution of the values of the first generalized measurement characteristic, corresponding to the maximum eigenvalue, for a fault on lines ω1 and ω2 for a section of the electric network of FIG. 1.

In FIG. 4 presents a structural diagram of a device that implements the proposed method of relay protection of an energy facility.

The device (Fig. 4) contains: analog-to-digital converters 1 and 2; converters (digital filters) 3 and 4; computing unit 5; executive modules (comparison schemes) 6 ₁ ... 6 _M , where M is the number of generalized indicators of measurement; logic block 7; unit for storing the results of simulation 8.

The device operates as follows.

When implementing the method of relay protection of an energy object, simulation simulation is carried out and relay protection training is carried out in controlled fault modes and alternative modes characteristic of the energy object. When learning in various modes of operation, the following stages are implemented for the formation of generalized measurement attributes and relay protection parameters: formation of a training sample; centering and rationing of the training sample; obtaining a correlation matrix of intermediate measurement values; calculation of eigenvalues and eigenvectors of the correlation matrix of intermediate measurement values; determination of protection operation parameters; calculation of coefficients to obtain generalized measurement features. Thus, as a result of the training, sets of coefficients are formed to obtain generalized measurement features and the parameters of relay protection and relay protection are determined. Moreover, in order to increase the recognition ability of the relay protection and automation device, it is advisable to implement the formation of the specified coefficients and response parameters using a set of predetermined (permissible) modes of operation of the energy object.

Note that in the simplest case, the only generalized measurement criterion is selected that corresponds to the maximum eigenvalue of the correlation matrix of intermediate measurement values. At the same time, relay protection and automation equipment is adopted one-dimensional and assuming a simplified procedure for comparing the only received generalized characteristic with the set value.

The results of simulation corresponding to the specified modes of functioning of the energy object are recorded in block 8 for storing the results of simulation.

At the inputs of analog-to-digital converters 1 and 2, analog values of currents and voltages i ( t ) and u ( t ) are supplied, and discrete values i ( k ) and u ( k ) are formed at the output of the ADC 1 and 2. Converters (digital filters) 3 and 4 fix the samples of electrical quantities i ( k ) and u ( k ) on the observation window from n samples, and also form complexes of current and voltage I ( n ) and U ( n ) involved in obtaining intermediate values .

As intermediate values, various parameters of currents and voltages can be selected. For example, for the previously considered example, the intermediate values included the effective values of currents and active powers ( Ia , Ib , Pa , Pb ) of phases A and B. Intermediate values are formed by calculation in computing unit 5 from the complexes of currents and voltages I ( n ) and U ( n ), and the set of intermediate values represents the current measurement. According to the current measurement and the coefficients coming from block 8 for storing the results of simulation, in the computing block 5, generalized signs of measurement are formed.

The obtained generalized signs of the current measurement from the outputs of the computing unit 5 are supplied to the inputs of the executive modules (comparison circuits) 6 _{1 ...} 6 _M , where M is the number of generalized measurement signs. The other inputs of the executive modules (comparison circuits) 6 _{1 ...} 6 _M receive the response parameters (settings) of the protection from the outputs of block 8. The results of comparing the generalized protection signs with the reference values (response parameters) are fed to the inputs of the logic unit 7.

The logic unit 7 generates a protection trip signal in accordance with the signals at the outputs of the executive modules. In the simplest case, the logic of the operation of block 7 can be implemented according to the group “And” if it is required to ensure high reliability of the relay protection device (Fig. 4), when the response signal at the output of block 7 is formed when all the executive modules 6 ₁ ... 6 _{M are} triggered. If it is required to ensure high response sensitivity of the relay protection device (Fig. 4), it is advisable to implement the logic of the operation of unit 7 using the group "OR" when the operation signal at the output of unit 7 is generated when at least one actuator module 6 ₁ ... 6 _M is triggered.

It should be noted that since the formation of the coefficients for obtaining generalized signs and parameters of the protection operation depends on the specified (current) mode of operation of the energy object, a control signal is received at the input of the simulation simulation results unit 8, characterizing the mode and determining the choice of data protection required for functioning. The control signal can be supplied manually by the operational (relay) personnel of the power facility, or automatically from the software of the operational information complex (SCADA system) that monitors the power facility mode.

In conclusion, it should be noted:

1. The ability of modern relay protection devices to accumulate and process a large amount of information regarding the modes of the protected object, allows you to implement new methods of relay protection, adaptive to a specific operational situation.

2. In the proposed method, the transition to a small number of generalized features (compared with the current measurement), providing the operation of relay protection, while maximally preserving the information contained in them, is proposed. The transition to generalized features reduces the number of processing operations required to implement protection, and ensures its high speed.

3. It is advisable to use a single generalized attribute and the construction of one-dimensional protection. The considered example illustrated the implementation of relay protection and automation with high recognition ability with a simple technical design.

Claims (1)

The method of relay protection of an energy facility using its simulation models in controlled short circuit modes and in alternative modes by fixing the readings of electrical quantities, converting the samples to an intermediate value, converting intermediate values to measurement, training relay protection with measurements from simulation models and determining the characteristics of the operation of relay protection, intermediate values are formed at each step of the observation of the short circuit mode and converted to the current measurement of relay protection, protection training is carried out by current measurements formed during simulation, characterized in that the results of preliminary simulation are entered into the storage unit of the results of simulation, perform analog-to-digital conversion signals of currents and voltages with their subsequent digital filtering, as well as determining the parameters of currents and voltages in the form of intermediate values, based on the obtained intermediate x values by their linear combination in the computing unit form the generalized measurement characteristics, which are fed to the comparison schemes, the number of which is selected according to the number of generalized measurement indicators, the protection response parameters for each generalized characteristic are supplied to the second inputs of the comparison schemes from the storage unit of the simulation results, s the outputs of the comparison circuits, the signals are fed to the logic block, in which, based on the signals from the outputs of the comparison circuits, a protection response signal is generated depending on the need to ensure high reliability of the response or sensitivity of the relay protection device, the formation of generalized characteristics and protection response parameters applied to the comparison circuit is carried out taking into account the functioning mode of the protected energy object, therefore, when changing the mode with signals from the output of the storage unit of the simulation results, the calculation of generalized attributes is corrected in the computing unit and change of protection operation parameters at the second inputs of the corresponding comparison circuits.

Images