JPH06343226A - Method for monitoring stability of electric power system - Google Patents

Abstract

PURPOSE:To make it possible to monitor the static stability of an electric power system quickly by using the knowledge base, wherein the geographical, electric correlation of a generator are described beforehand, recognizing the correlation of the elements of the characteristic vectors of the intrinsic values of various power fluctuation modes at a high speed, and discriminating the intrinsic value of the power fluctuation mode. CONSTITUTION:In an electric-power-system stability monitoring apparatus 1, the connecting state of generators, loads, and power plants and the system operating state 31 such as the amount of power generation, loads, voltages are observed with a data collecting part 11 from an electric power system 30. The intrinsic value and the characteristic vector are obtained with a characteristic-vector computing part 12 by using the knowledge base. A power-fluctuation-mode-intrinsic-value extracting part 14 extracts the intrinsic value of the power fluctuation mode based on the result of the computation of the characteristic vector. The computed intrinsic value 15 of the power fluctuation mode is displayed to the operators of the electric power system on a display part 16. Therefore, the intrinsic values when the state of the electric power system is changed is estimated to be at a high speed, and the static stability can be quickly judged.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention creates a coefficient matrix of a state equation from the system configuration and the operating state observed from a power system,
The present invention relates to a power system stability monitoring method for calculating an eigenvalue and an eigenvector from a coefficient matrix and monitoring the steady-state stability of the power system based on the eigenvalue and the eigenvector.

[0002]

2. Description of the Related Art A coefficient matrix of a state equation is created from a system configuration and an operating state observed online from a power system, eigenvalues and eigenvectors are calculated from the coefficient matrix, and the steady state stability of the power system is calculated by the eigenvalues and eigenvectors. There is no practical example of the power system stability monitoring device to be monitored.

[0003]

When determining the steady state stability of a power system, a coefficient matrix of a state equation is created from the system configuration and the operating state observed from the power system, and eigenvalues and eigenvectors are calculated from the coefficient matrix. There is a need. Further, from the calculated eigenvalues and eigenvectors, it is necessary to extract eigenvalues (hereinafter, referred to as power oscillation mode eigenvalues) that affect the stationary stability of the power system.

Further, since there are tens or hundreds of generators in the power system, the number of state variables is hundreds to thousands, and the system matrix has dimensions of hundreds to thousands. Then, there are eigenvalues and eigenvectors corresponding to the number of dimensions. Calculation for solving these eigenvalues and eigenvectors (hereinafter referred to as eigenvalue calculation)
Although it takes a considerable amount of time by itself, it takes a lot of labor to extract various power fluctuation mode eigenvalues from the eigenvalue calculation. This is because the extraction of the power fluctuation mode eigenvalues is greatly related to the elements of the eigenvector other than the value of the eigenvalue represented by a complex number. The elements of the eigenvector are represented by complex numbers and are distributed on the complex plane.

Here, the elements of the eigenvector represent each generator, and there are as many generators as there are generators. The angle and distance between each element distributed on the complex plane represent the phase difference and magnitude of the fluctuation between the generators. Each element of the eigenvector of the power fluctuation mode eigenvalue is distributed in a unique pattern on the complex plane, and it is necessary to recognize this unique pattern and extract the power fluctuation mode eigenvalue. In the power system stability monitoring device, in order to monitor the steady state stability of the power system that changes from moment to moment, it is necessary to calculate and extract the power fluctuation mode eigenvalue at high speed. For that purpose, it is necessary to recognize the unique pattern of each element of the eigenvector of the power fluctuation mode eigenvalue at high speed.

Also, each element of the eigenvalues and eigenvectors changes, although it is minute, due to the power system that changes from moment to moment, and at each time, the eigenvalues and eigenvectors of a system matrix consisting of hundreds to thousands of dimensions are calculated, It takes an enormous amount of time to recognize the unique pattern of each element of the eigenvector of the oscillation mode eigenvalue. Further, the main storage area of the computer is largely occupied, and it is not very effective in utilizing computer resources.

The present invention has been made in view of the above circumstances, and extracts various power fluctuation mode eigenvalues at a high speed, estimates these eigenvalues when the power system status changes at a high speed, and promptly estimates the eigenvalues. It is an object of the present invention to provide a power system stability monitoring method having a function of determining steady-state stability.

[0008]

The power system stability monitoring method according to [Claim 1] of the present invention is provided in advance for geographically
Using the knowledge base that describes the electrical correlation, the correlation of each element of the eigenvectors of various power fluctuation mode eigenvalues is recognized at high speed, and the power fluctuation mode eigenvalues are discriminated.

In the power system stability monitoring method according to [Claim 2] of the present invention, in [Claim 1], the knowledge base has a hierarchical database and a power fluctuation mode discrimination rule.

The power system stability monitoring method according to [claim 3] of the present invention uses the calculated eigenvalues and eigenvectors as follows:
Extract the eigenvalues and eigenvectors that affect the power system,
Furthermore, by estimating a small change in the power oscillation mode eigenvalue when the power system status changes, the steady state stability can be promptly determined by estimating it with a neural network learned beforehand.

[0011]

The power system stability monitoring method according to [Claim 1] of the present invention uses the knowledge base in which the geographical and electrical correlations of the generators are described in advance to generate eigenvectors of various power fluctuation mode eigenvalues. The power fluctuation mode eigenvalue is determined at high speed by means for recognizing the correlation of each element at high speed.

In the power system stability monitoring method according to [Claim 2] of the present invention, in [Claim 1], since the knowledge base has the hierarchical database and the power fluctuation mode determination rule, the power fluctuation mode is high speed. It can be detected accurately.

According to a third aspect of the present invention, there is provided a power system stability monitoring method, wherein a means for estimating a power fluctuation mode eigenvalue when a power system status is changed by a pre-learned neural network is used to rapidly stabilize the system. Judge the degree.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A power system stability monitoring method according to claim 1 of the present invention will be described with reference to FIG. As shown in the figure, FIG. 1 is an overall configuration diagram, and 1 is a stability monitoring device for a power system.
From 30, data collection unit 11 for observing connection status such as generator, load, substation, and system operation status 31 such as power generation amount, load, voltage, and eigenvalue for solving eigenvalue and eigenvector,
An eigenvector calculation unit 12, a power fluctuation mode eigenvalue extraction unit 14 that extracts a power fluctuation mode eigenvalue based on the eigenvalue and eigenvector calculation result 13, and a display unit that displays the calculated power fluctuation mode eigenvalue 15 to a power system operator. It consists of 16 and.

FIG. 2 is a block diagram of the power fluctuation mode eigenvalue extraction unit 14 shown in FIG. 1, which comprises an eigenvalue / eigenvector calculation result storage unit 141, a knowledge base 142, and an inference engine 143. The knowledge base 142 is a hierarchical database 14
21 and power fluctuation mode discrimination rule 1422. Then, the eigenvalue / eigenvector calculation result 13 is input to the eigenvalue / eigenvector calculation result storage unit 141.

Based on the knowledge of the geographical and electrical correlation of the generators in the knowledge base 142, the power fluctuation mode eigenvalue is determined. In addition, knowledge of the geographical and electrical correlation of the generator is stored in the hierarchical database 1421 and the power fluctuation mode discrimination rule 1422. This part is shown in [Claim 2]. The eigenvalue / eigenvector calculation result storage unit 141, and the knowledge base 142
The inference engine 143 drives the power fluctuation mode discrimination rule 1422 using the geographical and electrical information of the generator of the hierarchical database 1421 in FIG.

FIG. 3 is a diagram showing an example of the structure of the hierarchical database 1421 in the knowledge base 142. The power system is usually hierarchical. FIG. 3 is an example in which the power system is hierarchized at level 1, level 2, level 3, and level 4.
Level 1 is the largest level divided into larger areas. In this example, the north, west, and eastern systems are divided. Next, level 2 is the second largest level. This is often divided by ultra high voltage substation. Level 3 is an individual power plant and level 4 is an individual generator.

Next, an example of the power fluctuation mode discrimination rule 1422 will be shown. This is a group of programs represented by discriminant sentences, and in this example, the wide-range power fluctuation mode and its type can be known. (Example of power fluctuation mode discrimination rule 1422) (Rule 1) When the element with the maximum absolute value of the eigenvector element (representing the fluctuation of the generator) of a certain eigenvalue on the complex plane is standardized to 1, An element with a value of 0.3 or more is a sway generator.

(Rule 2) Oscillation When there is one generator, the mode indicated by this eigenvalue is a mode in which only one generator oscillates, which indicates the oscillation of the control system of the generator.

(Rule 3) There are two or more vibration generators,
If all Levels 3 are the same, the mode indicated by this eigenvalue is the mode swaying in the same power plant.

(Rule 4) There are two or more vibration generators,
If all Level 2 are the same, the mode indicated by this eigenvalue is a mode swaying in the same substation.

(Rule 5) There are two or more vibration generators,
If all Level 1s are the same, the mode indicated by this eigenvalue is a mode swaying in the same global area.

(Rule 6-1) If there are two or more perturbation generators and the level 1 is different, the mode indicated by this eigenvalue is the wide-range power perturbation mode.

(Rule 6-2) When Rule 6-1 is established, if the level 1 type is the north system or the west system, it is a wide-range power fluctuation mode in which the north system and the west system are swaying. .

(Rule 6-3) When the rule 6-1 is established and the type of level 1 is the eastern system and the western system, it is a wide-area power fluctuation mode in which the eastern system and the western system are oscillating. .

(Rule 6-4) When the rule 6-1 is established and the type of level 1 is the north system or the east system, it is the wide area power fluctuation mode in which the north system and the east system are swaying.

(Rule 6-5) When Rule 6-1 is established, if the type of Level 1 is north, west or east, it is a wide area swaying in north, west or east. It is a power fluctuation mode.

FIG. 4 shows an example of distribution on the complex plane of the eigenvector element 17 of the power fluctuation mode eigenvalue 15. Of the eigenvalues that are usually solved, there are several to several tens of eigenvalues called power fluctuation mode eigenvalues, depending on the scale of the power system.
And, in particular, there are only a few eigenvalues that are said to have a great influence on the steady-state stability of the power system. In monitoring the steady stability of the power system, it is important to extract these several eigenvalues at high speed and display them to the operator of the power system.

In the example of FIG. 4, the generator 18 and the generator 1
9. The position of the generator 20 is unique. The distribution of the eigenvector elements 17 of the power fluctuation mode eigenvalue 15 on the complex plane does not change significantly in the system of the same electric power company, but the system conditions, that is, the magnitude of the generator output, the magnitude of the load, and the transmission It changes slightly depending on the connection status of the electric wire.

FIG. 5 shows an example of changes in the positions of the generator 18, the generator 19, and the generator 20 on the complex plane in the example of FIG. 4 due to the change of the system condition. The dashed circles are the original positions, and the solid circles are the changed positions. If the generator group represented by the generator 18 and the generator 19 belongs to, for example, the western system, and the generator 20 belongs to the eastern system, this is a wide-area power fluctuation mode that is shaking in the western system and the eastern system. . Subtle changes in the power system do not change the wide-area power perturbation mode.

According to this embodiment, the geographical location of the generator is
By using the knowledge base that describes the electrical correlation and recognizing the correlation of each element of the eigenvectors of various power fluctuation mode eigenvalues, the power fluctuation mode eigenvalues can be determined at high speed.

A power system stability monitoring method according to [claim 3] of the present invention will be described with reference to FIG. In FIG.
The same parts as those in FIG. In FIG. 6, reference numeral 22 is a power fluctuation mode eigenvalue estimation unit, which is the calculated power fluctuation mode eigenvalue 15 and system operation state.
The power fluctuation mode eigenvalue estimate 23 is calculated from 31 and. Then, the display unit 16 displays the power fluctuation mode eigenvalue 15 and the power fluctuation mode eigenvalue estimated value 23 to the operator of the power system. Other configurations are the same as those in FIG.

FIG. 7 is a block diagram of an example of the power fluctuation mode eigenvalue extraction unit 14. The eigenvalue / eigenvector calculation result 13 is input to the neural network 144. Then, the neural network 144 learned in advance extracts the power fluctuation mode eigenvalue 15 from the eigenvalue and eigenvector calculation result 13.

FIG. 8 is a block diagram of the power fluctuation mode eigenvalue estimation unit 22. The power fluctuation mode eigenvalue 15 and the system operation state 31 are shown.
To the neural network 221. Then, the neural network 221 learned in advance has the power fluctuation mode eigenvalue 15
Estimated value of power fluctuation mode eigenvalue from
Calculate 23. Here, the neural network is a method of engineeringly simulating nerve cells in the human brain, and its research has recently been advanced in the field of electric power systems, and the application example of the hierarchical neural network is most known in the construction method. ing. The features are excellent in pattern recognition ability and interpolation characteristics.

FIG. 9 shows an example of the structure of a hierarchical neural network. It is composed of neurons 32 corresponding to human brain cells and interneuron connections 33 corresponding to nerves connecting the brain cells. Human brain cells are said to develop the nerves that connect the brain cells through learning, which is simulated in neural networks by weighting the inter-neuron coupling 33. That is, the weighting of the inter-neuron coupling 33 is determined by learning. As shown in the hierarchical neural network of FIG. 9, the neuron is divided into several layers. That is, the input layer 34, the intermediate layer 35, and the output layer 36. The example of FIG. 9 is called a three-layer type neural network, and has an input layer 34 and an intermediate layer 3
5 and the output layer 36 are each one layer.

Then, the phenomenon is taught to the input layer and the judgment result is taught to the output layer. When the phenomenon and the judgment result are taught a proper number of times, the weighting of the inter-neuron coupling 33 is determined (learning), and the neural network can judge even the unlearned phenomenon (the neural network Called pattern recognition and interpolation). Generally, learning of a neural network takes time, but once learning is completed, the determination becomes fast.

The distribution of the eigenvector element 17 of the power fluctuation mode eigenvalue 15 on the complex plane by the neural network is the same as in the case of FIG. As already described with reference to FIG. 4, the distribution of the eigenvector elements 17 of the power fluctuation mode eigenvalue 15 on the complex plane does not change significantly in the same power company system, but the system condition, that is, the output of the generator It varies slightly depending on the size, the load, and the connection status of the power transmission line. Therefore, the distribution on the complex plane of the unique generator is used as the input layer of the neural network, the type of power fluctuation mode fluctuation is used as the output layer, and the learning of the neural network is performed based on the results of the offline simulation. The generated neural network is used to extract the power fluctuation mode eigenvalue 15.

FIG. 10 shows another distribution example on the complex plane of the eigenvector element 17 of the power fluctuation mode eigenvalue 15. In the example of FIG. 10, the positions of the generator 18, the generator 19, and the generator 20 are characteristic. Figure 11 shows a configuration example of the input layer, intermediate layer, and output layer of the neural network used to extract the power fluctuation mode eigenvalue 15. When the example of FIG. 4 is the power oscillation mode 1 and the example of FIG. 10 is the power oscillation mode 2, the generator is set in the power oscillation mode 1.
18, the generator 19, and the generator 20, and in the power fluctuation mode 2, the generator 24, the generator 25, and the generator 26 have features.

Therefore, as shown in FIG. 11, the phases and absolute values on the complex plane of the generator 18, the generator 19, the generator 20, the generator 24, the generator 25, and the generator 26 as the input layer, that is, 12 in total
As the layers and the output layers, there were a total of two layers of the power fluctuation mode 1 and the power fluctuation mode 2. It seems that 6 to 10 layers are appropriate as the intermediate layer. In the example shown in FIG. 11, the generator 18, the generator 19, the generator 20, the generator 24, the generator are obtained from the result of the simulation in each system cross section offline.
25, the phase and absolute value of the generator 26 on the complex plane, and the offline determination results of the power fluctuation mode 1 and the power fluctuation mode 2 are input to the neural network 144, the learning is performed, the weighting factor is determined, and then online. Used for.

FIG. 12 shows a configuration example of the input layer, the intermediate layer, and the output layer of the neural network used in the power fluctuation mode eigenvalue estimation unit 22. As shown in FIG. 6, the power oscillation mode eigenvalue extraction unit 14 inputs the power oscillation mode eigenvalue 15 and the system operation state 31 from the power system 30 to the power oscillation mode eigenvalue estimation unit 22. The power fluctuation mode eigenvalue 15 is classified into several power fluctuation modes by the power fluctuation mode eigenvalue extraction unit 14.

Specifically, the system operation state 31 includes the amount of power transmitted through each transmission line, the voltage of each substation, the generated power of each power plant, and the opening and closing of the busbars of each substation. Further, each power fluctuation mode eigenvalue is closely related to these system operation states 31.
The system operation state 31 changes from moment to moment, and the power oscillation mode eigenvalues are affected by this and change slightly. Furthermore,
System operation state that greatly affects the eigenvalue of each power fluctuation mode
The 31 elements are limited to a few.

In the example of FIG. 12, the power fluctuation mode, the power transmission amount of some transmission lines, and the switching of bus lines of some substations are used as the input layer. In this example, the power oscillation mode 1, the power oscillation mode 2, the transmission amount of the transmission line A, the transmission amount of the transmission line B, the transmission amount of the transmission line C, the switching of the bus 1 of the substation A, and the switching of the bus 2 of the substation B. There are 7 layers, which are mixed. As the output layer, there are two layers of the real part and the imaginary part of the eigenvalue. The real part of the eigenvalue represents the vibration damping, and the larger the absolute value on the negative side, the more stable the value becomes. or,
The imaginary part represents the agitation system. When only looking at the stability, the real part of the output layer is sufficient, but when looking at the behavior of the sway, the imaginary part is also necessary.

Power fluctuation mode eigenvalue estimation unit 22 shown in FIG.
Neural nets require learning. This is done by off-line simulation. The estimation of the power fluctuation mode eigenvalue by the learned neural network is fast, and this is used online. According to the above-described embodiment, it is possible to determine the steady-state stability at high speed by estimating a small change in the power oscillation mode eigenvalue when the power system status changes by the neural network that has been learned in advance.

[0044]

As described above, according to the present invention, the inference engine or the neural network is used to extract, from the calculated eigenvalues and eigenvectors, those that affect the steady stability of the power system. Since the power fluctuation mode eigenvalue can be estimated at high speed, it is possible to quickly monitor the steady state stability of the power system, which changes from moment to moment.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment for explaining a power system stability monitoring method according to the present invention.

FIG. 2 is a configuration diagram of a power fluctuation mode eigenvalue extraction unit shown in FIG.

FIG. 3 is a diagram showing a configuration example of a hierarchical database in a knowledge base.

FIG. 4 is a distribution example of eigenvector elements of power fluctuation mode eigenvalues on a complex plane.

5 is an example of a change in the position of the generator on the complex plane in the example of FIG. 4 due to a change in the system status.

FIG. 6 is a configuration diagram of another embodiment of the present invention.

7 is a diagram showing a configuration example of a power fluctuation mode eigenvalue extraction unit shown in FIG.

8 is a diagram showing an example of the configuration of a power fluctuation mode eigenvalue estimation unit shown in FIG.

FIG. 9 is a configuration example of a hierarchical neural network.

FIG. 10 shows another example of distribution of eigenvector elements of power fluctuation mode eigenvalues on the complex plane.

FIG. 11 is a configuration example of an input layer, an intermediate layer, and an output layer of the neural network used for extracting the power fluctuation mode eigenvalue.

FIG. 12 is a configuration example of an input layer, an intermediate layer, and an output layer of a neural network used for estimating the power fluctuation mode eigenvalue.

[Explanation of symbols]

 1 stability monitoring device 11 data collection unit 12 eigenvalue, eigenvector calculation unit 13 eigenvalue, eigenvector calculation result 14 power fluctuation mode eigenvalue extraction unit 15 power fluctuation mode eigenvalue 16 display unit 22 power fluctuation mode eigenvalue estimation unit 23 power fluctuation mode eigenvalue estimation value 30 Power system 31 System operating status 144,221 Neural network

Claims (3)

[Claims] 1. A coefficient matrix (system matrix) of a state equation is created from a system configuration and an operating state observed from a power system, eigenvalues and eigenvectors are calculated from the coefficient matrix, and the systematic stability of the system is calculated from the eigenvalues and eigenvectors. In the power system stability monitoring method for monitoring the power system, a knowledge base that describes the geographical and electrical correlations of generators is added in advance, and this knowledge base is used to calculate the steady state stability of the power system from the calculated eigenvalues and eigenvectors. A power system stability monitoring method characterized by extracting eigenvalues and eigenvectors that affect the stability and determining the steady-state stability at high speed. 2. The power system stability monitoring method according to claim 1, wherein the knowledge base includes a hierarchical database and a power fluctuation mode discrimination rule. 3. A coefficient matrix (system matrix) of a state equation is created from a system configuration and an operating state observed from a power system, eigenvalues and eigenvectors are calculated from the coefficient matrix, and the systematic stability of the system is determined by the eigenvalues and eigenvectors. In the power system stability monitoring method for monitoring the eigenvalues, eigenvalues and eigenvectors that affect the steady stability of the power system are extracted from the calculated eigenvalues and eigenvectors, and these eigenvalues when the power system status changes are extracted. A power system stability monitoring method characterized by estimating by a neural network and determining the steady state stability at high speed by these.