KR102690534B1 - Method for Estimating Model Parameters of Generator Based on PMU(Phasor Measurement Unit) Measurement - Google Patents

Abstract

The present invention relates to a method for estimating generator integers based on PMU measurement. The method for estimating generator model integers of the present invention is to: (A) use a measurement device installed on a power system containing one or more generators connected to the power grid through a transformer; Receiving measurements including instantaneous voltage and current for the generator online via a network; (B) For a simulated reduced system for simulating the output of a simulated regenerative generator on the power grid using a predetermined program, simulated input data including a plurality of estimation target integers determined based on the measured values are transmitted to the simulated regenerative generator. generating simulation result data for the output of the simulated regenerative generator by performing a simulation by inputting the input into the; And (C) after modifying one or more of the estimation target integers based on sensitivity analysis of the estimation target integers, re-performing the simulation to determine model constants for the generator according to the error. .

Description

PMU measurement-based generator constant estimation method {Method for Estimating Model Parameters of Generator Based on PMU (Phasor Measurement Unit) Measurement}

The present invention relates to a method for estimating the generator constant, and in particular, to a method for estimating the generator constant of the system based on actual measurement data acquired when a system failure occurs using a PMU (Phasor Measurement Unit) device linked to the actual system. It's about.

Recently, as the size of the power system has grown and the number of new and complex facilities has increased, many different technologies are needed to operate it with precision. Among them, the basic technology for stable system operation and planning is system stability analysis. For accurate and sophisticated stability analysis, the parameters of the stability analysis model must be tuned similar to those of an actual generator. Tuning of existing generator parameters was done when installing a new generator or control system or through periodic performance tests. However, existing generator performance tests have limitations in tracking generator parameters that reflect system characteristics because they use measured data based on test scenarios under specific conditions.

For example, a representative conventional technology uses a method of deriving model constants using a one-stage equivalent model using data acquired through a generator characteristic test. In the current generator characteristic test, when there is no generator model constant, specific test conditions are applied to the generator and the acquired data is used to derive the model constant. However, the existing method is a constant obtained by using the generator response characteristics for specific test conditions. Since the generator is connected to the system and operating, the response to a failure that occurs during operation in connection with the actual system may be very different from the test conditions. there is. Accordingly, when using an existing model, there is a problem that the output may appear more or less than the actual generator output.

Therefore, the present invention was created to solve the above-mentioned problems, and the purpose of the present invention is to use the actual measurement data acquired when a system failure occurs in the PMU installed in the real system, such as at the generator or near the generator in the power system, to calculate the generator model constant. By estimating and adjusting, it is possible to reflect the response characteristics due to system failure of the generator connected to the actual system to the model constant, deriving more accurate and practical model constants, and securing accurate model constants provides the basis for accurate stability analysis. The purpose is to provide a generator constant estimation method that can be used to establish countermeasures for system failure by applying it to the operation of the system and the establishment and operation of planning measures.

First, to summarize the features of the present invention, the generator model constant estimation method according to one aspect of the present invention to achieve the above object is (A) installed on a power system including one or more generators connected to the power grid through a transformer. Receiving measured values including instantaneous voltage and current for each generator using a measuring device; (B) For a simulated reduced system for simulating the output of a simulated regenerative generator on the power grid using a predetermined program, simulated input data including a plurality of estimation target integers determined based on the measured values are transmitted to the simulated regenerative generator. generating simulation result data for the output of the simulated regenerative generator by performing a simulation by inputting the input into the; And (C) after modifying one or more of the estimation target integers based on sensitivity analysis of the estimation target integers, re-performing the simulation to determine model constants for the generator according to the error. .

If the error between the measured value and the simulation result data is smaller than a threshold, the corresponding estimation target constants are determined as model constants for the generator.

In the step of generating the simulation result data, steps (A) and (B) are performed from a predetermined start time to an end time to re-perform the simulation according to the error, or from the start time to the end time. Divide into a plurality of sample sections and perform steps (A) and (B) for each sample section, but re-perform the simulation for all of the plurality of sample sections according to the error after performing the simulation for the entire plurality of sample sections. Use one of the methods.

The measured value is data on the primary or secondary side of the transformer connected to the generator. The measured value includes effective voltage, frequency, active power, and reactive power converted from the instantaneous voltage and current for the generator.

A central server on a network receives the measured values from the measuring device online and performs the simulation, correction of the estimation target constant, and determination of model constants for the generator. Through time synchronization, a plurality of generators can be measured and model constants for the plurality of generators can be determined simultaneously.

Calculate the Jacobian matrix ( J ) for the sensitivity based on the difference between the measured value y and the simulation result data according to the correction amount ( δ ) of the estimation target integers ( β ), and when calculating the sensitivity, The estimation target integers ( β ), which are variables of the objective function f, are determined so that the difference between the measured value y and the simulation result data is minimized and the following equation is satisfied,

Here, I is the identity matrix, and μ is an adjustment coefficient to prevent divergence and improve convergence speed during the sensitivity calculation process.

In the step of determining the model constants, in the next estimation, a method of estimating parameters that are the constants to be estimated from a plurality of measurement values while excluding some of the previously used measurement values and adding new measurement values accordingly, or in the next estimation, Excluding some of the previously used parameters and adding as many parameters estimated from new measured values, a method of individually estimating parameters from the measured values and then selecting a representative value from a plurality of estimated parameters is used.

In the step of determining the model constants, in order to select one of the estimated values for the estimation target integers as a representative value, the distribution of the sum of the gravitational fields for the ordered pairs of the estimation target integers is calculated, and the gravitational field is the most The corresponding value of the strong point can be selected as a representative value of the ordered pair of integers to be estimated.

In addition, a recording medium implemented with computer-readable code according to another aspect of the present invention is (A) used to measure each generator using a measuring device installed on a power system that includes one or more generators connected to the power grid through a transformer. Ability to receive measurements including instantaneous voltage and current; (B) For a simulated reduced system for simulating the output of a simulated regenerative generator on the power grid using a predetermined program, simulated input data including a plurality of estimation target integers determined based on the measured values are transmitted to the simulated regenerative generator. A function of generating simulation result data for the output of the simulated regenerative generator by performing a simulation by inputting it into; and (C) performing a function of determining model constants for the generator according to errors by re-performing the simulation after modifying one or more of the estimation target integers based on a sensitivity analysis of the estimation target integers. , the model constants of the generator on the power grid system can be estimated.

According to the generator constant estimation method according to the present invention, the generator model constant can be estimated and adjusted so that the response characteristics of the generator that appear when a system failure occurs in actual system operating conditions and the response characteristics of the generator that appear in simulation are the same. Using this method, the generator model constants derived from existing specific test conditions do not reflect the response characteristics in actual system operation situations by comparing and analyzing the results extracted from PMU data linked to the system and the simulation results. By adjusting the generator model constants to be similar to the PMU data, simulation results closer to the real system can be obtained. System stability simulation is the most basic simulation from the perspective of planning and operating a power system. If the reliability of the simulation is increased and the characteristics of the real system are reflected in the simulation, more accurate and reliable stability analysis results can be derived. As a result, this becomes a foundational technology for stable system planning and operation, allowing it to be accurately and easily utilized in establishing countermeasures against system failures.

Figure 1 is a diagram for explaining a method of measuring actual data at a generator terminal or grid connection point according to an embodiment of the present invention.
Figure 2 is a diagram for explaining a method of measuring actual data to minimize cost according to an embodiment of the present invention.
Figure 3 is a diagram for explaining a method of generating virtual data for dynamic simulation according to an embodiment of the present invention.
Figure 4a is a diagram for explaining an example of the generator model constant estimation method of the present invention.
Figure 4b is a diagram for explaining another example of the generator model constant estimation method of the present invention.
Figure 5 is a diagram for explaining an example of a model integer estimation and representative integer selection method in the present invention.
Figure 6 is a diagram for explaining a method of using statistical techniques when estimating model constants in the present invention.
Figure 7 is an example of a three-dimensional distribution of the sum of the gravitational fields generated by each point when there are 10 random ordered pairs in a two-dimensional plane.
FIG. 8 is a diagram showing the distribution of the sum of the gravitational fields in FIG. 7 using contour lines.
Figure 9 is a diagram showing the distribution of the sum of the gravitational fields for a large number of ordered pairs consisting of two model integer pairs as a contour line.
Figures 10, 11, and 12 show the distribution of estimated model constants K and T5 based on actual measurement data, contour lines, and ordered pairs of reactive power-related constants, respectively, taking a generator as an example.

Hereinafter, the present invention will be described in detail with reference to the attached drawings. At this time, the same components in each drawing are indicated by the same symbols whenever possible. Additionally, detailed descriptions of already known functions and/or configurations will be omitted. The content disclosed below focuses on the parts necessary to understand operations according to various embodiments, and descriptions of elements that may obscure the gist of the explanation are omitted. Additionally, some components in the drawings may be exaggerated, omitted, or shown schematically. The size of each component does not entirely reflect the actual size, and therefore the content described here is not limited by the relative sizes or spacing of the components drawn in each drawing.

First, the generator constant estimation method according to an embodiment of the present invention can be implemented using an electronic device such as an EMS (Energy Management System) and a central server, as described below, and the electronic device is a semiconductor processor. It may be implemented as hardware such as, software such as an application program, or a combination thereof. The electronic device may be a computing system (PC (Personal Computer), large-capacity computer, etc.) and may include at least one processor, memory, user interface input device, user interface output device, storage, and network interface. A processor may be a central processing unit (CPU) or a semiconductor device that executes processing of instructions stored in memory or storage.

<1. Measure >

Figure 1 is a diagram for explaining a method of measuring actual data at a generator terminal or grid connection point according to an embodiment of the present invention.

Referring to FIG. 1, one or more integer estimation target generators are selected among the generators (G ₁ , ... , G _n ) that supply power through a busbar to the main grid, and a transformer connected to the output of each generator is selected. Target generator, such as the instantaneous voltage/current (M1p,..., Mnp) of the system connection point on the primary side or the instantaneous voltage/current (M1s,..., Mns) on the secondary side of the transformer connected to each generator output, which is the generator terminal. The actual measurement data is measured using a measurement device such as a Potential Transformer/Current Transformer (PT)/Current Transformer (MOF), Metering Out Fit (MOF), or Phasor Measurement Unit (PMU). It can be measured in time units (△t). The above EMS (control device linked to the central server) can convert actual measurement data about the measured target generator into effective value voltage, frequency, and active/reactive power and store and manage the converted measured values in memory, etc. In this way, estimation accuracy can be improved by measuring both at the grid connection point and the generator terminal as needed. In addition, in order to accurately reflect the state, the Automatic Generation Control (AGC) signal transmitted from the EMS to the generator can be acquired during measurement and used for synchronization, etc. to improve estimation accuracy.

It is also possible to estimate the generator constant using individual ground truth data (M1s, M1p,..., Mns, Mnp) separately, but at the same time point for the corresponding estimation target generators (G ₁ , ... , G _n ). In order to ensure that the characteristics of each generator are accurately reflected, individual measurements for each generator must be made at the same time. To this end, each individual measurement is made at the same time based on the start of measurement and time synchronization using an AGC signal, etc., and time synchronization between measurement devices is achieved through local communication between measurement devices or GPS (Global Positioning System) time reception. You can. In particular, time synchronization in a wide area, such as a nationwide range, can be mainly achieved through GPS time reception.

Using this measurement method, it is possible to obtain regional actual measurement data at the national level at the same time. The actual measurement data is transmitted online, so that the central server that receives it can simultaneously determine the exact model constants of all generators at key points. As a result, field testing can be minimized and the delay time until the generator model is reflected can be significantly shortened, contributing to securing the accuracy of system analysis. Additionally, costs and manpower due to field testing can be minimized.

Since minimizing cost is an important factor for nationwide measurement, the measurement method that minimizes cost among the measurement methods at various locations as shown in Figure 1 can be shown as shown in Figure 2.

Figure 2 is a diagram for explaining a method of measuring actual data to minimize cost according to an embodiment of the present invention.

Referring to FIG. 2, it shows a method that eliminates redundancy in measurement and can minimize the cost involved in measurement. In other words, a CT can be installed at the grid connection point of the primary side of the transformer connected to the output of each generator to measure the instantaneous current at the grid connection point of each individual generator, and a PT can be installed to measure the instantaneous voltage at the grid connection point of each individual generator. You can. Accordingly, based on the measured instantaneous voltage and current, the EMS (control device linked to the central server) converts the effective voltage, frequency, and active and reactive power input from individual generators to the system and posts the converted measured values on the Internet. , can be transmitted to a central server through a network such as a wireless communication network.

<2. Virtual data creation and verification>

When data is acquired from all points on the primary and secondary sides of each generator as shown in Figure 1, dynamic simulation can be performed without separate virtual data. However, by generating virtual data at another point based on the acquired data and comparing it with the measured data at that point, the accuracy of the measurement can be verified and defective data can be removed.

Virtual data can be generated based on the transformer impedance or line impedance depending on the actual system configuration, and it is possible to generate secondary data using primary data or vice versa. For example, when the transformer secondary voltage, frequency, active and reactive power data are measured at the generator G terminal as shown in Figure 3, the impedance of the transformer between generator G and the PB generator (simulated regenerative generator for estimating system changes) or By reflecting variables that affect data, such as line impedance, it is possible to generate PB (play back) virtual data on the primary side of the opposite transformer, and virtual data can also be generated in the reverse direction. That is, for dynamic simulation, the simulation device of the central server does not directly measure the data at the PB generator location when inputting simulated input data containing the estimation target integer determined based on the measurement data to the PB generator. Even if not, dynamic simulation is possible by generating the above virtual data (data from the primary side of the transformer corresponding to the data from the secondary side of the transformer) and reflecting the simulated input data.

<3. Generator model integer estimation>

The abbreviated system for estimating the generator model integer in the simulation device of the central server is the primary side of the transformer connected to the main grid for estimating the model integer for the estimation target generator (G) on the secondary side of the transformer, as shown in Figure 3. It consists of a PB generator (simulated regenerative generator for estimating system changes). This simulated reduction system contributes to minimizing computational resources for model integer estimation.

Simulated input data containing the estimation target constant determined based on the measurement data is input to the PB generator, the output of the PB generator in the power grid (model) on this system is simulated, and according to the simulation results, the generator (G ), the simulation device of the central server can be simulated using a predetermined application program to perform operation simulation of the circuit, and the output voltage, frequency, active and reactive power of the PB generator. Simulation result data can be generated. Using the simulation result data of the PB generator, a model constant of the estimation target generator (G) that can be satisfied to approximate the value measured in the estimation target generator (G) according to the principle of FIG. 3 can be determined.

Figure 4a is a diagram for explaining an example of the generator model constant estimation method of the present invention. Here, a method of determining the model constant of the generator (G) is shown by performing simulation from a predetermined start time (t ₀ ) to an end time (t _f ).

Referring to FIG. 4a, in an example of a method of estimating a generator model constant performed in a simulation device, first, for the system on the power grid (Main Grid), the measured values, which are actual data such as effective value voltage, frequency, active and reactive power, as above. The output of the PB generator in the power grid is simulated once from a predetermined simulation start time (t ₀ ) to the end time (t _f ) along with the simulated input data of the PB generator that includes the estimation target integer determined based on the simulation results (e.g. , voltage, frequency, active and reactive power, etc.) are generated (S10), and then estimated based on the simulation results (e.g., voltage, frequency, active and reactive power, etc.) to simulate once again by modifying some of the estimation target integers. Analyze the sensitivity of the model to the target integer (S11). As the number of integers to be estimated increases, the number of simulations for sensitivity analysis increases linearly.

For example, according to sensitivity analysis, among the estimation target constants (e.g., values related to input/output voltage/current, active/reactive power, gain, impedance, etc.), one or more model constants with high sensitivity to parameter value changes can be selected for modification. You can. Afterwards, the value of the corresponding estimation target integer selected based on the sensitivity calculation is modified (S12), and further system simulation is performed, and the simulation results (e.g., voltage, frequency, active and reactive power, etc.) and the generator (G) are calculated. Calculate the error with the measured value (S13). If the error due to the estimated integer is sufficiently small (smaller than the threshold) (S14), the corresponding estimation target integers are confirmed as model integers of the generator (G) and reflected in the generator model of a given system analysis program (S15). If not, the above sensitivity analysis is repeatedly performed until the error becomes sufficiently small, and the estimated target integers including the final modified model integer are confirmed as the model integer of the generator (G). This is the case for general local optimization techniques, and it seems reasonable to select the method in FIG. 4a because its performance will be superior to that of the method in FIG. 4b.

Figure 4b is a diagram for explaining another example of the generator model constant estimation method of the present invention. Here, we show a method of determining the model constants of the generator (G) by performing system simulation by dividing each sample of the measured value. Tracking techniques such as the Kalman filter correspond to this, but the estimation of generator model constants has very strong nonlinearity, making it very difficult to estimate using general tracking techniques. The ensemble Kalman filter, which has excellent nonlinear analysis capabilities through a probabilistic approach, is applicable.

Referring to FIG. 4b, first, upon receiving measurement values such as the above effective value voltage, frequency, active and reactive power for the system on the power grid (Main Grid), a simulation of the PB generator including the estimation target constant determined based on this is performed. Simulation results (e.g., voltage, frequency, active and reactive power, etc.) are generated for the output of the PB generator in the power grid along with the input data, from a predetermined start time (t ₀ ) to an end time (t _f ). When dividing the time unit (△t) into n sample sections (t _n = t ₀ ~ t _n = t _f ), the simulation begins from the first sample section (S21). For these simulation results, statistical analysis such as sensitivity analysis is performed to confirm the estimation target integers as model integers of the generator (G) using the ensemble Kalman filter method (S22), and the estimation target integers are modified based on this. Do it (S23). This process is repeatedly performed for all sample intervals (t _n = t ₀ to t _n = t _f ) (S24, S25). That is, after performing a systematic simulation for one sample section, a plurality of stochastically possible virtual samples of target integers to be estimated are generated, and a systematic simulation is performed for each. In this process, more than hundreds of virtual samples can be created, and a significant number (e.g., hundreds) of systematic analyzes are involved for each systematic analysis based on the original data received.

After simulating the system for each virtual sample, the most likely new virtual sample is statistically obtained, the virtual samples of the integers to be estimated are modified based on this, and then it is repeated on the next sample of the measured value, so the computation required is very high. It becomes more. In addition, if the resulting estimation result is a sufficiently small error compared to the measured value, the estimation target constants estimated to date are confirmed and updated as the model constants of the generator (G) (S27), but like other Kalman filters for nonlinearity, Since the ensemble Kalman filter does not guarantee that the final convergence value is the optimal value, if it is determined that the error is not sufficiently small through comparison of the final results (S26), all previous processes must be repeated (S27).

Estimation of the constants of the generator (G) has very strong nonlinearity, and various constants of the generator (G) exhibit various characteristics by combination. That is, there may be multiple combinations of integers that can accurately express measurement data. Finding the global optimal value of a problem that includes multiple local optimal points is defined as a global optimization problem, and can be solved with global optimization techniques such as Genetic Algorithm, Particle Swarm Optimization, and Gravity Algorithm. The above global optimization technique requires significantly more computational resources than local optimization, and the likelihood that the derived value will become the global optimal value increases as the amount of computational resources input increases. Therefore, it is not suitable for estimating multiple generator models on a national scale, and it is difficult to reflect actual generator characteristics in a fast system analysis program.

The ultimate goal of generator constant estimation is to accurately reflect the characteristics of the actual generator, so if the purpose is satisfied, it is not very important whether the combination of the generator constants is the same as the actual generator. Therefore, it is reasonable to apply local optimization techniques to generator constant estimation. Additionally, in general, the generator constant does not change significantly in a short period of time, so data measured within a certain period of time shows consistent results. In other words, as measurement data accumulates, the characteristics of the generator must satisfy various situations, and each generator constant estimated to satisfy all data approaches the actual generator constant.

To estimate the generator constant, the Levenberg-marquardt algorithm (LM algorithm), one of the most powerful local optimization techniques to date, was first applied. The algorithm operates according to [Equation 1]. That is, the Jacobian matrix ( J ) for sensitivity is calculated based on the difference between the measured value y according to the correction amount ( δ ) of the estimation target integers ( β ) and the simulation result data, and when calculating sensitivity, the measured value y The estimation target integers ( β ), which are variables of the objective function f, can be determined so that the difference between and the simulation result data is minimized and the following equation is satisfied.

[Equation 1]

Here, the matrices J and I are the Jacobian matrix for sensitivity and the identity matrix, respectively. J ^T is the transpose matrix of J. Vector β (β _P , β _Q ) is a group of parameters (e.g., integers to be estimated, values associated with input/output voltage/current, active/reactive power, gain, impedance, etc.) that determine the objective function (f()). , the objective function is optimized by being updated at each repeated calculation by the vector δ consisting of the change amount (correction amount) of the parameters. In other words, the vector β is the difference between the measured value (y) and the simulation result data (e.g., voltage, frequency, active power (P), and reactive power (Q), etc.) when calculating sensitivity ( J ). It is determined to minimize the difference between the simulation result data. A relatively small or large adjustment coefficient μ reduces the error in the LM algorithm using the Gauss-Newton or gradient descent method, and prevents divergence and improves convergence speed during the sensitivity calculation process. Each parameter of the LM algorithm is determined by [Equation 2] to [Equation 5]. That is, by calculating the change in RMSE (Root Mean Square Error) for active power (P) and reactive power (Q) for each parameter β _P and β _Q in active power (P) and reactive power (Q), , each parameter β _P and β _Q related to the generator (G) and its included governor, PSS (Power system stabilizer), exciter, etc. can be estimated.

[Equation 2]

[Equation 3]

[Equation 4]

[Equation 5]

In this way, it is possible to target all values for integer estimation, but in this case, a very large amount of computation is required, and the success of the estimation cannot be assured. Therefore, it is important to appropriately select key parameters. For example, for IEEEG1 among the generator (G) models, the main integers of the governor are K and T5, and the main integers of the exciter (e.g. ESST4B) are Kpr, Kir, and Kp. The main constant of PSS2A (power stabilizer) is Ks1. Based on this, [Equation 2] to [Equation 5] can be re-expressed as [Equation 2] to [Equation 5]. These generator (G) model constants are illustrative, and appropriate constants may be selected depending on the simulation purpose, power generation environment, etc.

[Equation 6]

[Equation 7]

[Equation 8]

[Equation 9]

The Gauss-Newton algorithm can be applied in a similar way, and it operates according to [Equation 10].

[Equation 10]

<4. Model constant estimation based on multiple measurement data>

If measurement continues, there will be multiple data acquired at various points in time. Since the generator (G) constant estimated at a specific point in time cannot be considered representative of the characteristics at all points in time, a generator constant representing multiple data points is selected. Should be.

Figure 5 is a diagram for explaining an example of a model integer estimation and representative integer selection method in the present invention.

The simplest way to think of it is to set the RMSE for all data as one objective function and estimate an integer. As shown at 510 in FIG. 5, parameters (estimation target constants/model constants) are estimated from a plurality of measurement values (e.g., 4 in the figure), but in the next estimation, some of the previously used measurement values (e.g., one ), the method of repeating the estimation by adding newly measured values accordingly can be simple in construction. However, there is a disadvantage that as the measurement data increases, the computational resources required for estimation also increase. In addition, when new data is measured, all data used in the existing estimation must be used again for estimation along with the new data, so a large amount of space is required to store the measured data and a large amount of computing resources must be mobilized for each estimation. Even in cases where old data is not estimated for reasons such as storage space issues or generator constant accuracy, a large amount of computational resources are invested as old data must be deleted and re-estimated using the remaining data. Additionally, this method has a fatal problem in that if the generator constants that satisfy each measurement value are placed in opposite directions, the conflicting measurement values cannot all be reflected and an incorrect constant is derived.

As a method of dramatically reducing computational resources and storage space requirements, as shown at 520 in FIG. 5, parameters (estimate target integers/model integers) are individually estimated for each measurement value, and then an appropriate representative is created using only the estimated parameters. There is a method of selecting a value and repeating the estimation by excluding some (e.g. one) of the previously used parameters in the next estimation and adding as many parameters estimated from the newly measured values. The effectiveness of this method depends on how valid data is selected and representative values are selected, and statistical techniques can be used as the simplest method.

Figure 6 is a diagram for explaining a method of using statistical techniques when estimating model constants in the present invention. Figure 6 shows an example of a case in which the estimated values of 10 ordered pairs consisting of two model integer pairs are expressed in relative position relationships.

If the ordered pair representing the average of individual integers is like point 610, this can lead to an odd result that does not represent any of the 10 integer pairs. The ordered pair with the mode of each integer is represented by point 620, which also does not represent any pair of integers. In Figure 6, only 2 of the 10 integer pairs are concentrated in a specific area, and the remaining 8 integer pairs exist sporadically, so the 8 integer pairs are removed and the representative point 630 of the 2 concentrated integer pairs is selected. It would be more reasonable to do so. The rationality of this choice will increase if there are multiple pairs of estimated integers.

When estimated integers exist sporadically, it is very difficult to establish standards for selecting integer pairs to be used in selecting representative values and integer pairs to be removed, so it is realistically impossible for humans to directly classify the data and select representative values. In addition, there are a total of 6 integers selected for estimation previously (K, T5, Kpr, Kir, Kp, Ks1, etc.), and when introducing the concept of Figure 6, ordered pairs are listed in a total of 6-dimensional space, which makes it even more difficult. It becomes work.

In order to select one of the estimated values for the integers to be estimated as a representative value, a method of solving the problem using gravitational field theory is explained as follows.

Figure 7 is an example of a three-dimensional distribution of the sum of the gravitational fields generated by each point when there are 10 random ordered pairs in a two-dimensional plane. As a way to solve very difficult data selection problems, the gravitational field theory that exists in the natural world can be introduced. When there are 10 random ordered pairs (corresponding to integers to be estimated) in a two-dimensional plane, the sum of the gravitational fields generated by each point can be distributed as shown in FIG. 7.

FIG. 8 is a diagram showing the distribution of the sum of the gravitational fields in FIG. 7 using contour lines. In Figure 8, 10 random points in the two-dimensional plane (green point 810), the contour of the gravitational field, each local optimum point (black star), the global optimum point (black cross 830), the average value of the random points ( The blue cross indicates 850), and the actual optimal point (red cross 840).

Since a strong gravitational field is formed in places where many points are concentrated, the value corresponding to the point where the gravitational field is strongest (the value of the integer to be estimated) can be selected as a representative value. In other words, there are six points below the vertical axis 0 in Figure 8, but they exist sporadically and do not create a strong gravitational field. On the other hand, the upper four points are small in absolute number, but they exist concentrated in one place, creating a strong gravitational field. Therefore, this is suitable as a method to solve the problem mentioned in FIG. 6.

Figure 9 is a diagram showing the distribution of the sum of the gravitational fields for a large number of ordered pairs consisting of two model integer pairs as a contour line. In the case of FIG. 8, it may be possible for a human to classify directly, but in a case where there are a large number of arbitrary ordered pairs as in FIG. 9, it is very difficult for a human to classify directly. However, with the proposed technique, representative values can be selected without difficulty.

Since the curved surface shown in FIG. 7 has multiple extreme values, there are multiple local optimal points. Therefore, in order to obtain the global optimal point, a global optimization algorithm generally represented by Genetic algorithm, Particle swarm optimization, gravity algorithm, etc. must be applied. However, as mentioned earlier, the global optimization algorithm requires a lot of computational resources and has the problem that computational resource input must be increased to increase the reliability of the results.

The data classification and representative value selection method that adopts this gravitational field theory also solves the problem of applying this global optimization algorithm. In other words, since the gravitational field is strong where each point exists, there are many points around the optimal point. Therefore, if local optimization is performed starting from each of the points that create the gravitational field, each point converges to the local optimal point that exists in the surrounding area. Therefore, more than one point converges to each local optimal point, and the most points converge to the global optimal point (representative value). Therefore, if the strengths of the gravitational fields of the local optimal points are compared and the strongest point is selected as the global optimal point, the global optimal point is selected and this can be used as a representative value.

Local optimization algorithms such as Levenberg-Marquardt (L-M), Gauss-Newton (G-N), and gradient descent can all be applied, and [Equation 11] and [Equation 12] are used for L-M or G-N.

[Equation 11]

[Equation 12]

At this time, the variables exist in a 6-dimensional space, and if the integers for active and reactive power are separated as in [Equation 13] and [Equation 14], they exist in a 2-dimensional plane and a 4-dimensional space, respectively.

[Equation 13]

[Equation 14]

Figures 10, 11, and 12 show the distribution of estimated model constants K and T5 based on actual measurement data, contour lines, and ordered pairs of reactive power-related constants, respectively, taking a generator as an example.

For example, the distribution of model constants K and T5 related to active power among the constants estimated based on actual data from a given generator using a PMU, etc. creates a gravitational field as shown in Figure 10, which is expressed by the contour line in Figure 11. do. In Figure 11, ordered pairs of integers (green dots), each local optimum (black stars), the global optimum (black cross 910), the average value of random points (blue x 920), and the true optimum (red cross). 930).

The green dots are integer ordered pairs estimated from the active power of each measurement data, and the representative value (910) exists where the dots are most concentrated. On the other hand, the average value (920) refers to an odd place where there are no points nearby. Ordered pairs of integers related to reactive power are shown as shown in Figure 12, and representative values shown as red crosses indicate where ordered pairs are concentrated.

In addition, it was confirmed that the generator model constant estimation method based on the actual measured data of the present invention is effectively applied in the simulation estimation results for other power plants.

In the method for estimating the generator model constant in an electronic device such as a computer according to an embodiment of the present invention, the function used for input/output data processing is implemented as a computer-readable code in a recording medium that can be read by a device such as a computer. It is possible, and the combination of such recording media and devices such as computers can be implemented to input, output, and display data or information necessary for performing functions. Computer-readable recording media include all types of recording devices that store data that can be read by a computer system. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, hard disk, and removable storage, as well as carrier wave (e.g., transmission via the Internet). It also includes implementation in the form of. In addition, the computer-readable recording medium can be distributed in a computer system connected to a network (e.g., Internet, mobile communication network, etc.), store computer-readable code in a distributed manner, and can be executed through the network.

As described above, according to the method of estimating the generator model constant according to the present invention, the generator model constant is estimated and adjusted so that the response characteristics of the generator that appear when a system failure occurs in actual system operating conditions and the response characteristics of the generator that appear in simulation are the same. You can. Using this method, the generator model constants derived from existing specific test conditions do not reflect the response characteristics in actual system operation situations by comparing and analyzing the results extracted from PMU data linked to the system and the simulation results. By adjusting the generator model constants to be similar to the PMU data, simulation results closer to the real system can be obtained. System stability simulation is the most basic simulation from the perspective of planning and operating a power system. If the reliability of the simulation is increased and the characteristics of the real system are reflected in the simulation, more accurate and reliable stability analysis results can be derived. As a result, this becomes a foundational technology for stable system planning and operation, allowing it to be accurately and easily utilized in establishing countermeasures against system failures.

As described above, the present invention has been described with specific details such as specific components and limited embodiments and drawings, but this is only provided to facilitate a more general understanding of the present invention, and the present invention is not limited to the above embodiments. , those of ordinary skill in the field to which the present invention pertains will be able to make various modifications and variations without departing from the essential characteristics of the present invention. Therefore, the spirit of the present invention should not be limited to the described embodiments, and the scope of the patent claims described later as well as all technical ideas that are equivalent or equivalent to the scope of this patent claim are included in the scope of the rights of the present invention. It should be interpreted as

Claims (11)

(A) receiving measurements including instantaneous voltage and current for each generator using a measuring device installed on a power grid that includes one or more generators connected to the power grid through a transformer;
(B) For a simulated reduced system for simulating the output of a simulated regenerative generator on the power grid using a predetermined program, simulated input data including a plurality of estimation target integers determined based on the measured values are transmitted to the simulated regenerative generator. generating simulation result data for the output of the simulated regenerative generator by performing a simulation by inputting the input into the; and
(C) After modifying one or more of the estimation target integers based on sensitivity analysis of the estimation target integers, re-performing the simulation to determine model constants for the generator according to the error.
A generator model integer estimation method comprising: According to paragraph 1,
If the error between the measured value and the simulation result data is less than a threshold, the corresponding estimation target integers are determined as model constants for the generator. According to paragraph 1,
In the step of generating the simulation result data, steps (A) and (B) are performed from a predetermined start time to an end time to re-perform the simulation according to the error, or
The period from the start time to the end time is divided into a plurality of sample sections, and steps (A) and (B) are performed for each sample section, and the plurality of samples are divided according to the error after performing simulation for the entire plurality of sample sections. Method of re-performing simulation for the entire section
A generator model constant estimation method characterized by using any one of the following methods. According to paragraph 1,
A generator model constant estimation method, characterized in that the measured value is data on the primary side or secondary side of the transformer connected to the generator. According to paragraph 1,
The measured value includes an effective value voltage, frequency, active power, and reactive power converted from the instantaneous voltage and current for the generator. According to paragraph 1,
Generator model constant estimation, characterized in that a central server on a network receives the measured value from the measuring device online, and performs the simulation, corrects the estimation target constant, and determines model constants for the generator. method. According to paragraph 1,
A generator model constant estimation method characterized by measuring a plurality of generators through time synchronization and simultaneously determining model constants for the plurality of generators. According to paragraph 1,
Calculate the Jacobian matrix (J) for the sensitivity based on the difference between the measured value y according to the correction amount ( δ ) of the estimation target integers ( β ) and the simulation result data, and when calculating the sensitivity, the measurement The estimation target integers ( β ), which are variables of the objective function f, are determined so that the difference between the value y and the simulation result data is minimized and the following equation is satisfied,

Here, I is an identity matrix, and μ is an adjustment coefficient for preventing divergence and improving convergence speed in the sensitivity calculation process. A generator model constant estimation method. According to paragraph 1,
In the step of determining the model constants,
In the next estimation, a method of estimating parameters that are the integers to be estimated from a plurality of measurement values while excluding some of the previously used measurement values and adding as many new measurement values, or
In the next estimation, a method of estimating parameters individually from the measured values and selecting a representative value from the plurality of estimated parameters, excluding some of the previously used parameters and adding as many parameters estimated from new measured values.
A generator model integer estimation method characterized by using. According to paragraph 1,
In the step of determining the model constants,
In order to select one of the estimated values for the estimation target integers as a representative value, the distribution of the sum of the gravitational fields for the ordered pairs of the estimation target integers is calculated, and the corresponding value at the point where the gravitational field is the strongest is selected as the estimation target. A generator model integer estimation method characterized by selecting a representative value of an ordered pair of integers. (A) The ability to receive measurements, including instantaneous voltage and current for each generator, using a measuring device installed on a power grid that includes one or more generators connected to the grid through a transformer;
(B) For a simulated reduced system for simulating the output of a simulated regenerative generator on the power grid using a predetermined program, simulated input data including a plurality of estimation target integers determined based on the measured values are transmitted to the simulated regenerative generator. A function of generating simulation result data for the output of the simulated regenerative generator by performing a simulation by inputting it into; and
(C) After modifying one or more of the estimation target integers based on sensitivity analysis of the estimation target integers, re-performing the simulation to determine model constants for the generator according to the error
A recording medium implemented with computer-readable code for performing and estimating model constants of generators on a power grid system.