JPH07107667A - Monitoring apparatus for electric power system - Google Patents

Abstract

PURPOSE:To judge stability at a high speed in high accuracy by starting a detail-stability judging/operating part in response to the result of high-speed operation and judgment for the stability of the power flow, voltage, phase and the like of the linking line of an electric power system, which are changing from time to time, with a neural network. CONSTITUTION:The quantities of the states of a system such as the power flow of a linking line 6 between electric power systems A and B, which are connected to a plurality of generators 11, and the terminal voltage voltages and phases of the generators are detected and inputted into a system-state- quantity collecting part 2. The collected data are stored in a system-state- quantity memory part 3a. High-speed stability operation is performed in an operating part 3b by using a neural network. The result of the operation is judged with a judged-result evaluating part 3d. DB Depending on the result of the judgment, a detail-stability judging/operating part 3c is started, and the detailed operation is performed. A judged-result output part 4 selects the output of the judged-result evaluating part 3d or the detail-stability juding/operating part 3c. The result is inputted into a display part 5 and displayed. Thus, the stability is judged by a monitoring person 7 at a high speed in high accuracy.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention takes in a connection state of a power system and a system state quantity that changes from moment to moment (for example, a tidal current of an interconnection line connecting two systems) as inputs, and based on those inputs. The present invention relates to a power system monitoring device that determines the stability of a power system and outputs a stability determination result for stable operation of the system.

[0002]

2. Description of the Related Art In recent years, with the increase in scale of electric power systems,
In order to provide a stable power supply in the operation of a power system, there is an increasing need for online and accurate stability determination. Therefore, high-speed and highly accurate stability determination is required for the power system monitoring device.

A conventional electric power system monitoring device off-line calculates a value of a system state quantity (hereinafter referred to as a stability limit value) at which the electric power system becomes unstable from a predicted connection state of the electric power system and a supply and demand situation. It is a device that compares the system state quantity and the stability limit value that are input moment by moment and determines the stability.

[0004]

Generally, the stability limit value changes depending on the supply and demand situation of the electric power system and the connection state. Therefore, in order to accurately determine the stability, the stability limit value should be adjusted according to the change of the supply and demand situation and the connection state of the electric power system. It is necessary to calculate stability limit values online. However, in order to obtain the stability limit value online, a simplified system model is used to shorten the calculation time, and the calculation accuracy decreases.

In addition, since calculation time increases if a detailed system model is used to improve the calculation accuracy, the conventional device makes a high-speed and highly accurate system stability determination in a large-scale power system. It was impossible. The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a power system monitoring device capable of determining stability at high speed and with high accuracy.

[0006]

According to a first aspect of the present invention, there is provided a power system monitoring device, which is a high-speed stability determination calculation unit for determining stability of a power system using a neural network, and a high-speed stability determination unit. Evaluate the judgment result of the calculation unit with a preset fixed evaluation criterion, and start the detailed stability judgment calculation unit according to the evaluation result.The judgment result evaluation unit and the stability by calculation using a precise system model. And a detailed stability determination calculation unit for determining.

In the electric power system monitoring apparatus according to the second aspect of the present invention, the high speed stability judgment arithmetic unit for judging the stability of the electric power system using the neural network and the judgment result of the high speed stability judgment arithmetic unit are A judgment result output unit that outputs the result as it is when it is stable or unstable, and a judgment result output unit that is provided before the judgment result output unit when the judgment result cannot be judged, and judges the stability using a precise system model. It is composed of a detailed stability judgment calculation unit for carrying out.

An electric power system monitoring apparatus according to [Claim 3] of the present invention provides a high-speed stability judgment calculation unit for judging the stability of the power system using a neural network and a judgment result of the high-speed stability judgment calculation unit. Automatically calculates and changes the evaluation criteria for evaluation. The evaluation criteria automatic change unit and the evaluation result calculated by the evaluation criteria automatic change unit are used to evaluate the judgment results of the high-speed stability judgment calculation unit. The determination result evaluation unit that activates the detailed stability determination calculation unit according to the above, and the detailed stability determination calculation unit that determines the stability by calculation using a precise system model.

According to a fourth aspect of the present invention, there is provided a power system monitoring device which comprises a high-speed stability determination calculation unit for determining stability of a power system using a neural network and a determination result of the high-speed stability determination calculation unit. Automatically calculates and changes the evaluation criteria for evaluation. The evaluation criteria automatic change unit and the evaluation result calculated by the evaluation criteria automatic change unit are used to evaluate the judgment results of the high-speed stability judgment calculation unit. A detailed result evaluation unit that activates the detailed stability judgment unit according to the above, a detailed stability judgment calculation unit that judges the stability by calculation using a precise system model, and a new judgment result from the detailed stability judgment calculation unit. Neural network re-learning that creates training data, adds it to existing training data, trains the neural network again, and saves the weighting coefficient of the neural network after training It was formed from the part.

[0010]

Now, the neural network will be briefly described before the description of the operation. It can be said that a neural network is a realization of a biological neural network on a computer. Although neural networks of various structures have been proposed so far, layered networks are excellent in pattern recognition, and non-layered networks are suitable for optimization problems. The layered neural network, which is suitable for this problem, changes its structure through learning using learning data and outputs a reasonable result even for input data other than learning data.

FIG. 12 shows a three-layer neural network. From the top, each layer is called an output layer, an intermediate layer, and an input layer,
The number of nodes (hereinafter referred to as neurons) in each layer is determined according to the target problem. The mathematical model of each neuron is shown in Fig. 13. One neuron has n inputs, and the output Vj of the neuron j from these inputs is given by the following equation.

Vj = fj {Σ (wijVi) −θj + Ij} (1) i ≠ j where wij is a weight for each input, θj is a threshold,
Ij is an external input, and fj is a non-linear function called a neuron characteristic.

As the learning data, consider a set of an input value and a resulting output value. In the initial state, the weights and thresholds for each neuron of the neural network are set to appropriate values. When this input data is input to the neural network, some value is output from the output layer. This output value is different from the output value of the learning data. Therefore, the weight and the threshold value are changed so that the two match. This method includes backpropagation by Rumelhart. By learning the neural network using a large number of learning data in this way, the neural network outputs an appropriate value even for an input different from the learning data.

The operation of the power system monitoring device according to the first aspect of the present invention will be described below. The system state quantity that changes every moment in the power system is collected by the system state quantity collection unit 2, and the collected system state quantity is stored in the system state quantity storage unit 3a. Next, the system state quantity stored in the system state quantity storage unit 3a is input to the high-speed stability determination calculation unit 3b, the stability determination is performed by the neural network, and the determination result is sent to the determination result evaluation unit 3d. To be The determination result evaluation unit 3d determines whether or not the determination result of the neural network is within the evaluation reference range, and if it is within the evaluation reference range, the determination result of the neural network is output to the determination result output unit 4 as it is.

If the judgment result is outside the evaluation standard range,
The detailed stability determination calculation unit 3c is activated, the detailed stability determination calculation unit 3c performs detailed stability determination using a precise system model, and outputs the determination result to the determination result output unit 4. Further, the determination result is sent to the display device 5 and also a control command for maintaining the power system stably is sent to the power system control unit.

The operation of the power system monitoring device according to the second aspect of the present invention will be described. The system state quantity that is converted every moment in the power system is collected by the system state quantity collection unit 2, and the collected system state quantity is stored in the system state quantity storage unit 3a. Next, the system state quantity stored in the system state quantity storage unit 3a is input to the high-speed stability determination calculation unit 3b, and the stability is determined by the neural network. If the determination result is stable or unstable, the fact is directly output to the determination result output unit. On the other hand, when the determination result is not determinable, the detailed stability determination calculation unit performs detailed calculation and outputs the result.

The operation of the power system monitoring device according to the third aspect of the present invention will be described. In addition to the operation of the power system monitoring device according to [claim 1], the present invention provides a determination result evaluation unit 3d,
Automatic evaluation criteria changing section 8a selected by the changeover switch 8b
Or the evaluation standard set by the evaluation standard external setting unit 8c is used to judge whether the result of the neural network judgment is within the evaluation standard range. Further, the evaluation criterion automatic changing unit 8a uses an evaluation criterion used in the judgment result evaluation unit based on the result of the detailed stability judgment calculation unit 3c and the result of the high-speed stability judgment calculation unit 3b so as to improve the judgment accuracy. To fix.

The operation of the power system monitoring device according to [claim 4] of the present invention will be described. The present invention provides a learning data addition creation unit in addition to the operation of the power system monitoring device according to [Claim 1].
9a creates new learning data from the calculation result of the detailed stability judgment calculation unit 3c and adds it to the existing learning data. Using this learning data, the neural network re-learning unit
9c performs neural network learning again. Even during re-learning, the high-speed stability determination calculation unit 3b does not adversely affect the stability determination unit by performing the stability determination using the neural network before re-learning.

The weighting coefficient and the evaluation standard of the neural network after the completion of the re-learning are stored in the learning result storage unit 9d.
Further, all the learning data are stored in the learning data storage unit 9b to prepare for the next additional learning. After the completion of the relearning, the high-speed stability determination calculation unit 3b reads the latest weighting coefficient from the learning result storage unit 9d, and performs the stability determination by the relearning neural network. The judgment result evaluation unit also reads the latest evaluation standard from the learning result storage unit 9d and evaluates the judgment result.

[0019]

Embodiments will be described below with reference to the drawings. Figure 1
FIG. 1 is a configuration diagram of an embodiment of a power system monitoring device according to [Claim 1] of the present invention. In FIG. 1, reference numeral 1 denotes a power system monitoring device according to the present invention, which is a system state quantity that changes from moment to moment (power flow of interconnection line 6, terminal voltage phase difference of each generator).
A system state quantity collecting unit 2 for detecting and collecting the power consumption, a stability determination unit 3 for storing the collected power state quantities, and performing stability determination using the stored system state quantities as input data, and a determination result output unit 4 and.

Further, the stability determining section 3 is a system state quantity storage section 3a for storing the input system state quantity, and a high speed for performing a high-speed stability determining calculation by a neural network using the stored system state quantity as an input. Stability determination calculation unit 3b,
Evaluate the judgment results of the detailed stability judgment calculation unit 3c and the high-speed stability judgment calculation unit 3b, which judge the stability by transient stability calculation using a precise system model, and judge the detailed stability according to the evaluation result. It is composed of a determination result evaluation unit 3d that activates the calculation unit 3c. Reference numeral 7 indicates an observer.

FIG. 2 is an example of a diagram showing the terminal voltage phase difference of each generator with respect to the reference generator. The vertical axis represents the value obtained by normalizing the terminal voltage phase difference of the generator to 0 to 1, and the horizontal axis represents the generator number. The input data thus obtained is patterned and stored in the system state quantity storage unit 3a. The system state quantity stored in the system state quantity storage unit 3a is input to the high-speed stability determination calculation unit 3b.

FIG. 3 is a diagram showing the structure of a three-layer neural network used in the high-speed stability judgment calculation unit 3b in the present invention. In the figure, the input layer L of the neural network
3 is n neurons corresponding to the terminal voltage phase difference of n generators stored in the system state quantity storage unit 3a, and 1 neuron is corresponding to the interconnection line power flow value, and the output layer L1 is system stability It consists of one neuron that outputs a value between 0 and 1 that represents degrees, and the closer the value is to 1, the more unstable it is evaluated. The terminal voltage phase difference and the interconnection line power flow value of the n generators stored in the system state quantity storage unit 3a are input to the input layer L3 of this neural network, and the input layer L3 and the intermediate layer L2.
And the output of the neuron of the output layer L1 is the input data,
Alternatively, it is calculated from Eq. (1) using the output of the anterior layer neurons.

The output Y of the output layer neuron takes a continuous value between 0 and 1, and the result is sent to the judgment result evaluation section 3d. The determination result evaluation unit 3d uses the output of the high-speed stability determination calculation unit 3b to evaluate the determination result as follows.

[Equation 2] Unstable if YH ≤ Y ≤ 1 Cannot judge if YL <Y <YH 0 Stable if Y ≤ Y ≤ YL (2) where YH is the evaluation range of the judgment result , YL represents the lower limit.

Here, the reason why the evaluation formula (2) is used is shown below. Neural network for all system states,
It is impossible to completely learn under the conditions, and even a sufficiently learned neural network may make an erroneous decision. It is known that the stability of the power system shifts from stable to unstable when the load amount reaches a certain value when the system state (eg, load amount in the system) is plotted on the horizontal axis as shown in FIG. However, it is difficult to identify the point at which this transition from stable to unstable depends on the system state and conditions.

For example, if the learning data near the stability limit used for learning the neural network are points A and B, it is difficult for the neural network to make a decision in the systematic state between points A and B. it is conceivable that. In addition,
There are various determination methods for the upper limit value YH and the lower limit value YL of the evaluation range depending on how to select the input data and the teaching signal used in the neural network, but this embodiment will be described later.

Subsequently, the judgment result evaluation unit 3d performs evaluation using the above equation (2). If the evaluation result is stable or unstable, the result of the high speed stability judgment calculation unit 3b is output as it is. Send to Part 4. If the evaluation result cannot be judged, the detailed stability judgment calculation unit 3c performs transient stability calculation using a precise system model, judges the stability, and sends the result to the judgment result output unit 4. The determination result output unit 4 outputs the stability determination result to the determination result display device 5 and transmits the stability of the power system to the monitor 7.

Next, the learning method of the neural network in this embodiment will be described below. As a means for collecting learning data, stability calculation is performed using a model of the power system to be monitored. The learning data is the target interconnection line power flow value obtained as a stability calculation result, input data indicating the target voltage phase difference of n generators, and the generator 1 second after the accident as an index indicating the stability of the power system. It consists of the teaching signal calculated by Eq. (3) using the internal phase angle of.

(3) Yoi = Asi / As max (3) where Asi is the maximum generator internal phase angle (degrees) in the learning data i 1 second after the accident, and As max is 1 second after the accident. Maximum internal phase angle (degree) of the generator in the learning data in. The learning data of a plurality of cases are collected by changing the accident point and the accident mode.

Here, a method of determining the teaching signal represented by the equation (3) and the upper limit value YH and the lower limit value YL of the evaluation range will be described below. In general, the stability of the power system is the transient stability that indicates the stability for 1 to 2 seconds after the occurrence of a large disturbance such as a lightning strike, the small stability such as a fine ground fault, and the stability for a few seconds to a dozen seconds after the occurrence of a medium disturbance. It is roughly classified into stability in the intermediate region, and steady-state stability, which represents the stability for small disturbances such as load fluctuations that constantly occur. This embodiment is intended for transient stability, and the characteristic indicating stability appears in 1 to 2 seconds after the occurrence of disturbance.

A characteristic of this embodiment is that the internal phase angle of the generator is selected. As shown in FIG. 5, when the system is unstable, there is a generator that is out of synchronization (step out), and its internal phase angle is as shown by A in FIG. Even if the system is stable, its internal phase angle has a larger peak as it approaches the unstable side, depending on the magnitude of the disturbance and the state of the system. In FIG. 5, B and C are stable, but it can be said that B is more unstable than C. From the above, the maximum internal phase angle (A in FIG. 5) in the learning case is set as the reference value Asmax, and the maximum internal phase angle in other cases (B and C in FIG. 5).
Is normalized to a value of 0 to 1 to represent the stability of the system.
That is, the closer the teaching signal is to 0, the more stable it is, and conversely, the closer it is to 1, the more unstable it is.

Since the learning data is discrete as described above, there is a region where it is difficult to make a decision by the neural network. Therefore, the upper limit value YH of the evaluation range of the equation (2) of this embodiment is the same value as the teaching signal of the unstable learning data closest to the stable side at the point where the system shifts from stable to unstable (stability limit point). That is, specifically, the smallest internal phase angle Ash (degree) in the unstable case is used to calculate by the equation (4). The lower limit value YL is the same value as the teaching signal of stable learning data that is closest to the unstable side at the point where the system shifts from stable to unstable (stability limit point). The maximum internal phase angle Ash (degree) is used to calculate with equation (5).

[Formula 4] YH = Ash / As max ………… (4) YL = Asl / As max ………… (5)

The flow of the learning method of the neural network will be described with reference to FIG. First, in S1, the condition of the electric power system, the accident point, and the accident mode are set. The stability calculation is performed under the conditions set in S1 (S2), and the terminal voltage phase difference of the n generators and the target interconnection line power flow value are extracted (S2).
3) The value calculated by the equation (3) is given as a teaching signal from the stability calculation result (S4). The terminal voltage phase difference and the target interconnection line power flow value of the n generators obtained in S2 and the result obtained in S4 are registered together as one learning data (S5). Next, if learning data is to be created, S1
~ Repeat S5.

N of the learning data collected as described above
The terminal voltage phase difference of each generator and the target interconnection line flow value are used as input data of the neural network, and data indicating stable or unstable is given as a teaching signal, and the connection weight and threshold of the neural network by the back propagation method. The value is changed, and the learning is repeated for all learning cases until the output of the neural network and the teaching signal converge within the allowable error range. With the method described above, it is possible to construct a neural network for system monitoring stability determination with sufficient accuracy.

As described above, in the power system monitoring apparatus according to this embodiment, the neural network is characterized by high-speed operation processing and high-accuracy stability determination by handling a plurality of input information. In the system state where it is difficult to determine the network, stability is determined by a detailed stability determination method such as transient stability calculation that solves the state equation of the power system including the dynamic characteristics of the generator by the numerical integration method. High-speed, high-accuracy and highly-reliable stability determination can be performed for the system state quantity that changes from moment to moment. The present invention is not limited to the above embodiment.

FIG. 7 is a diagram showing the configuration of another embodiment according to [Claim 1] of the present invention. In the figure, reference numerals 1, 2 and 4 to 7 are equivalent to those in FIG. 1 and have equivalent functions. Reference numeral 3 denotes a stability determination unit that stores the collected power state quantities and determines the stability using the stored system state quantities as input data. The stability determination unit 3 stores a system state quantity storage unit 3a that stores the input system state quantity, and a high-speed stability determination calculation unit 3b that performs a high-speed stability determination calculation by a neural network using the stored system state quantity as an input. And a detailed stability judgment calculation unit 3c for judging stability by transient stability calculation using a precise system model.

FIG. 8 is a diagram showing the structure of a three-layer neural network used in the high-speed stability judgment calculation unit 3b in the present invention. In the figure, an input layer L3 of the neural network has n neurons corresponding to the terminal voltage phase differences of the n generators stored in the system state quantity storage unit 3a, and 1 corresponding to the interconnection line power flow value. The output layer L1 includes three neurons a, b, and c. If the system stability is stable, the neuron a outputs 1 and the other two neurons output 0. If it is unstable, the neuron of c outputs 1 and the other two neurons output 0. When it is difficult to determine the stability (state where determination is impossible), the neuron of b outputs 1 and the other two neurons output 0.

The terminal voltage phase difference and the interconnection line power flow value of the n generators stored in the system state quantity storage unit 3a are input to the input layer L3 of this neuron network, and the input layer L3, the intermediate layer L2, and the output. The output of the neuron of the layer L1 is calculated by the equation (1) using the input data or the output of the previous layer neuron. Output Ya of output layer neurons a, b, c
, Yb, Yc take continuous values between 0 and 1. Using this result, the judgment result is evaluated as follows.

[Equation 5] If Ya ≧ Yb, Yc, it is stable. If Yb ≧ Ya, Yc, it cannot be judged. If Yc ≧ Ya, Yb, it is unstable ………… (6)

If the evaluation result by the equation (6) is stable or unstable, the result of the high-speed stability judgment calculation unit 3b is sent to the judgment result output unit 4 as it is. If the determination is not possible, the detailed stability determination calculation unit 3c performs transient stability calculation using a precise system model to determine the stability, and sends the result to the determination result output unit 4. The determination result output unit 4 outputs the stability determination result to the determination result display device 5 and transmits the stability of the power system to the monitor 7.

Next, the learning method of the neural network will be described below. As a means of collecting learning data,
Stability calculation is performed using the model of the power system to be monitored. The learning data is obtained as the stability calculation result. The target interconnection line power flow value and the terminal voltage phase difference of the n generators. If the stability of the power system at this time is stable, the neuron of a is 1 and other Two neurons are 0, if it is unstable, the neuron of c is 1 and the other two neurons are 0, and if the stability is difficult to determine (when the system state is near the stability limit), the neuron of b is The 1 and the other 2 neurons consist of a teaching signal indicating 0, and learning data of a plurality of cases are collected by changing the accident point or the accident mode.

The above flow will be described with reference to FIG. First, in S11, the condition of the electric power system, the accident point and the accident mode are set. Stability calculation is performed under the conditions set in S11 (S12), the terminal voltage phase difference of N generators and the target interconnection line power flow value are extracted (S13), and if the stability calculation result is stable, a One neuron is 1 and the other two are 0
(S17), if it is unstable, the neuron of c is 1 and the other two neurons are 0 (S18), and if it is difficult to determine the stability (when the system state is near the stability limit), The teaching signal is 1 for the neuron and 0 (S16) for the other two neurons. The terminal voltage phase difference and the target interconnection line power flow value of the n generators obtained in S12 and the results obtained in S16, S17 and S18 are registered together as one learning data (S19). If learning data is to be created subsequently, S11
~ Repeat S19.

N of the learning data collected as described above
The terminal voltage phase difference of each generator and the target interconnection line power flow value are used as input data of the neural network, and data indicating stable, unstable or undecidable is given as a teaching signal, and the connection weight of the neural network by the back propagation method. The threshold value is changed and the learning is repeated until the output of the neural network and the teaching signal converge within the allowable error range for all learning cases. With the method described above, it is possible to construct a neural network for system monitoring stability determination with sufficient accuracy.

As described above, in the power system monitoring apparatus according to the present embodiment, the neural network is characterized by high-speed arithmetic processing, which is a feature of the neural network, and a highly accurate stability determination base that can handle a plurality of input information. In the system state where it is difficult to determine, stability is determined by detailed stability determination methods such as transient stability calculation that solves the state equation of the power system including the dynamic characteristics of the generator by the numerical integration method. For the system state quantity that changes with
High-speed, high-accuracy and highly reliable stability determination is possible.
The present invention is not limited to the above embodiment.

FIG. 10 is a diagram showing the configuration of a power system monitoring device according to [claim 3] of the present invention. In the figure,
Reference numeral 7 is the same as that in FIG. 1 and has the same function. The judgment result evaluation unit 3d uses the evaluation standard of the evaluation standard automatic changing unit 8a selected by the changeover switch 8b or the evaluation standard set by the evaluation standard external setting unit 8c, and the judgment result of the neural network is within the evaluation standard range. To determine if. This changeover switch is operated by the manual changeover by the monitor 7 or by sending a signal from the computer.

The evaluation reference value of the evaluation reference external setting device 8c is set manually by the monitor 7 on the computer panel or by a signal from another computer. When the detailed stability determination is performed by the evaluation criteria automatic changing unit 8a, the stability with respect to the system state at that time becomes clear, so in order to improve the determination accuracy by the neural network, the calculation result of the detailed stability determination unit 3c And high-speed stability judgment calculation unit 3b
The evaluation criteria are modified as follows based on the judgment result Y of (A) If the result of the detailed stability determination unit is large and unstable, change YH = Y. (B) When the result of the detailed stability determination unit is other than (A), the evaluation standard is not changed. The learning method of the neural network is equivalent to [claim 1].

As described above, in the power system monitoring apparatus according to the present embodiment, the neural network is characterized by high-speed arithmetic processing and high-precision stability determination by handling a plurality of input information. In the system state where it is difficult to determine the network, stability is determined by a detailed stability determination method such as transient stability calculation that solves the state equation of the power system including the dynamic characteristics of the generator by the numerical integration method. By correcting the evaluation standard based on the result of the detailed stability judgment unit and the judgment result of the high-speed stability judgment calculation unit, high-speed, high-accuracy and high reliability can be achieved against the system state quantity that changes from moment to moment. Stability can be determined. The present invention is not limited to the above embodiment.

FIG. 11 is a diagram showing the configuration of a power system monitoring device according to [Claim 4] of the present invention. In the figure,
Reference numeral 7 is the same as that in FIG. 1 and has the same function. The learning data addition creation unit 9a creates new learning data from the calculation result of the detailed stability determination unit 3c, and adds this to the existing learning data. Using this learning data, the neural network relearning unit 9c re-learns the neural network. High-speed stability determination unit even during re-learning
3b does not adversely affect the stability determination unit by performing stability determination using the neural network before re-learning.

The weighting coefficient and the evaluation standard of the neural network after the completion of the re-learning are stored in the learning result storage unit 9d.
Further, all the learning data are stored in the learning data storage unit 9b to prepare for the next additional learning. After re-learning, high-speed stability judgment unit
3b reads the latest weighting coefficient from the learning result storage unit 9d,
After re-learning, the stability is judged by the neural network. The judgment result evaluation unit also reads the latest evaluation standard from the learning result storage unit 9d and evaluates the judgment result. If re-learning is not possible, no additional learning data is added,
Neither the weighting factor nor the evaluation criteria are updated. The learning method of the neural network is equivalent to [claim 1].

As described above, in the power system monitoring apparatus according to the present embodiment, the neural network is characterized by high-speed arithmetic processing, which is a feature of the neural network, and highly accurate stability determination by handling a plurality of input information. In the system state where it is difficult to determine the network, stability is determined by a detailed stability determination method such as transient stability calculation that solves the state equation of the power system including the dynamic characteristics of the generator by the numerical integration method. , By re-learning the neural network using the results of the detailed stability determination unit, it is possible to reduce the system states that are difficult to determine by the high-speed stability determination unit, so the system state quantity that changes from moment to moment can be reduced. On the other hand, high-speed, high-accuracy and highly reliable stability determination is possible.

[0048]

As described above, according to the power system monitoring apparatus of [Claim 1] of the present invention, the high speed of the arithmetic processing, which is a feature of the neural network, with respect to the system state quantity which changes momentarily. In addition, based on highly accurate stability judgment by handling multiple input information, in a system state where neural network judgment is difficult, a transient that solves the state equation of the power system including the dynamic characteristics of the generator by numerical integration method By performing stability determination with a detailed stability determination method such as stability calculation, high-speed, high-accuracy and highly reliable stability determination can be performed. Further, according to the electric power system monitoring apparatus of [Claim 2] of the present invention, when the determination result is stable or unstable, the fact is output as it is, and when the determination is impossible, detailed stability determination calculation is performed. Is output.

Further, according to the electric power system monitoring apparatus of the present invention [claim 3], in addition to the above-mentioned effect, the judgment accuracy of the neural network can be improved by changing the evaluation criterion of the stability judgment result manually or automatically. Can be improved.
Further, according to the power system monitoring device of [Claim 4] of the present invention, in addition to the above effect, the high-speed stability determination unit is performed by re-learning the neural network using the result of the detailed stability determination unit. It is possible to improve the determination accuracy of the neural network by reducing the system state that is difficult to determine by.

[Brief description of drawings]

FIG. 1 is a diagram showing the configuration of an embodiment of a power system monitoring device according to [claim 1] of the present invention.

FIG. 2 is a diagram showing normalized terminal voltage phase differences of n generators.

FIG. 3 is a diagram showing a configuration of a neural network used in [claim 1], [claim 2], [claim 3] and [claim 4] of the present invention.

FIG. 4 is a diagram showing a change in stability of a power system.

FIG. 5 is a diagram showing a method of collecting learning data of a neural network.

FIG. 6 is a diagram illustrating the content of a teaching signal of a neural network.

FIG. 7 is a diagram showing a configuration of an embodiment of a power system monitoring device according to [claim 2] of the present invention.

FIG. 8 is a diagram showing a configuration of a neural network used in [claim 1] of the present invention.

FIG. 9 is a diagram showing a method of collecting learning data of a neural network used in [claim 1] of the present invention.

FIG. 10 is a diagram showing the configuration of an embodiment of a power system monitoring device according to [claim 3] of the present invention.

FIG. 11 is a diagram showing a configuration of an embodiment of a power system monitoring device according to claim 4 of the present invention.

FIG. 12 is a diagram showing a general configuration of a neural network.

FIG. 13 is a diagram illustrating a mathematical model of a neuron.

[Explanation of symbols]

 1 power system monitoring device 2 system state quantity collection unit 3 stability determination unit 3a system state storage unit 3b high-speed stability determination calculation unit 3c detailed stability determination calculation unit 3d determination result evaluation unit 4 determination result output unit 5 display device 6 Interconnection line 7 Observer 8a Evaluation standard automatic change unit 8c Changeover switch 8c Evaluation standard external setter 9 Additional learning unit 9a Learning data addition creation unit 9b Learning data storage unit 9c Neural network re-learning unit 9a Learning result storage unit

Claims (4)

[Claims] 1. A connection state of a power system, a bus line voltage, which is a state quantity of the power system that changes from moment to moment, a grid line power flow, etc., are taken in, and whether or not the state of the power system is stable based on these inputs. An electric power system monitoring device having the following configuration, which judges the degree of the above and outputs a stability judgment result for stable operation of the electric power system. A high-speed stability determination calculator that determines the stability of the power system using a neural network. A determination result evaluation unit that evaluates the determination result of the high-speed stability determination calculation unit according to a preset fixed evaluation criterion, and activates the detailed stability determination calculation unit according to the evaluation result. Detailed stability determination calculation unit that determines stability by calculation using a precise system model. 2. The connection state of the electric power system, the bus line voltage, which is the state quantity of the electric power system that changes from moment to moment, the interconnection line power flow, etc. are taken in, and whether or not the state of the electric power system is stable based on these inputs. An electric power system monitoring device having the following configuration, which judges the degree of the above and outputs a stability judgment result for stable operation of the electric power system. A high-speed stability determination calculator that determines the stability of the power system using a neural network. A determination result output unit that outputs the result as it is when the determination result of the high-speed stability determination calculation unit is stable or unstable. A detailed stability determination calculation unit, which is provided in the preceding stage of the determination result output unit when the determination result is undeterminable, for performing stability determination using a precise system model. 3. The connection state of the power system, the bus line voltage, which is the state quantity of the power system, which changes momentarily, the interconnection line power flow, etc. are taken in, and whether the state of the power system is stable or not based on these inputs. An electric power system monitoring device having the following configuration, which judges the degree of the above and outputs a stability judgment result for stable operation of the electric power system. A high-speed stability determination calculator that determines the stability of the power system using a neural network. An automatic evaluation standard changing unit that automatically calculates and changes an evaluation standard for evaluating the determination result of the high-speed stability determination calculation unit. A judgment result evaluation unit that evaluates the judgment result of the high-speed stability judgment calculation unit according to the evaluation standard calculated by the evaluation standard automatic changing unit, and activates the detailed stability judgment calculation unit according to the evaluation result. Detailed stability determination calculation unit that determines stability by calculation using a precise system model. 4. The connection state of the power system, the bus line voltage, which is the state quantity of the power system, which changes momentarily, the interconnection line power flow, etc. are taken in, and whether the state of the power system is stable or not is based on those inputs. An electric power system monitoring device having the following configuration, which judges the degree of the above and outputs a stability judgment result for stable operation of the electric power system. A high-speed stability determination calculator that determines the stability of the power system using a neural network. An automatic evaluation standard changing unit that automatically calculates and changes an evaluation standard for evaluating the determination result of the high-speed stability determination calculation unit. A judgment result evaluation unit that evaluates the judgment result of the high-speed stability judgment calculation unit according to the evaluation standard calculated by the evaluation standard automatic changing unit, and activates the detailed stability judgment calculation unit according to the evaluation result. Detailed stability determination calculation unit that determines stability by calculation using a precise system model. A neural network re-learning unit that creates new learning data from the determination result of the detailed stability determination calculation unit, adds it to existing learning data, performs learning of the neural network again, and saves the weighting coefficient of the neural network after the learning is completed. .