JPH10327530A - Ratio-differential relay for protecting transformer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To speedily determine the occurrence of an accident by monitoring the quantity of electricity in the distribution system of a transformer, etc. SOLUTION: From a monitoring signal obtained by an electricity quantity monitoring sensor 21 provided in a distribution system, a preprocessing part 31 obtains the quantity of electricity, including a current value, which is converted into digital data by an A/D converting part 32. Then, the digital data are converted into wavelet data by a wavelet-analyzing part 33. Under a learning supporting means 35 which has learnt from these data in advance, the presence of abnormality is judged by a neural network 34.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the prevention of a malfunction caused by an inrush current caused by a switching operation of a transformer.
Also, the present invention relates to a transformer protection ratio differential relay capable of quickly detecting the presence / absence of an abnormality when an abnormality occurs.

[0002]

2. Description of the Related Art FIG. 4 is a diagram for explaining the principle of a conventional ratio differential relay, in which current detecting coils and the like are installed on the primary side (power side) and the secondary side (load side) of a transformer 1, respectively. Both ends of the constructed sensors CT ₁ and CT ₂ are connected to a ratio differential relay 2.
And current values I ₀ , I ₁ , I ₂ are detected by current detection means 2a, 2b, 2c. One end of each of the sensors CT ₁ and CT ₂ is connected to each other with the current detecting means 2b and 2c interposed in the middle, the other ends are directly connected to each other, and both connection points are connected in the middle. The current detection means 2a is connected with the current detection means 2a interposed therebetween.
_1, I _2, but also the current value I ₀ is a differential current by the current detecting means 2a obtained.

Based on the combination of the current values I ₀ , I ₁ , I ₂ detected by the respective current detecting means 2 a, 2 b, 2 c, the presence / absence of an abnormality and the location of occurrence are determined as shown in FIG. Table 5 shows the current values I ₀ , I ₁ , I ₂ and the ratio I ₂ /
Detecting the occurrence of an accident on the basis of I _1, the secondary-side internal or external fault, the primary internal, or the presence or absence of external accident is adapted to be detected.

However, a transformer has an inrush current,
When the transformer is turned on with no load, a current that reaches 10 times the rated current flows as shown in FIG. 6, and in addition, current is supplied only from the power supply side. In this case, it acts as if an internal fault current occurred, and malfunctions. FIG. 6 is an explanatory diagram showing a comparison between the steady current and the inrush current in the transformer.

As a countermeasure against this, conventionally, in order to distinguish between the inrush current and the internal fault current, a second harmonic component contained in the inrush current is detected and its content is calculated to calculate the inrush current and the internal inrush current. There has been proposed a ratio differential relay that can be classified as an internal fault current.

FIG. 7 is a block diagram showing the configuration of the above-mentioned conventional ratio differential relay. In the figure, reference numeral 1 denotes a transformer, and 2 denotes a ratio differential relay. Sensors CT ₁ and CT _{2 each} having a coil for current detection are provided on both sides of the transformer ₁ , and outputs thereof are input to a pre-processing unit 11 in the ratio differential relay 2. The configuration of the preprocessing unit 11 is substantially the same as that of the principle explanatory diagram shown in FIG.
Three current detection means are provided, and current values I ₀ , I ₁ , and I ₂ , which are respective detection values, are input to the A / D converter 12. A
The / D converter 12 outputs current values I ₀ , which are analog signals,
I ₁ and I ₂ are converted into digital signals and output to the ratio differential detection unit 13 and the second harmonic detection unit 14, respectively.

[0007] The ratio differential detection unit 13 outputs a current value ratio I ₂ /
Together to calculate the I _1, and inputs it to one input terminal of the AND gate 15, the operation timer 16a, recovery timer 17
The signal is input to one input terminal of the inhibit circuit 18 via a.

On the other hand, the second harmonic detecting section 14 detects the second harmonic component and inputs it to the other input terminal of the AND gate 15. The AND gate 15 is a second harmonic detector 14.
The logical product of the output of the differential circuit and the output of the ratio differential detection unit 13 is calculated, and the result is negatively input to the other input terminal of the inhibit circuit 18 via the operation timer 16b and the recovery timer 17b. The inhibit circuit 18 performs an accident determination output when the second harmonic is not included, and does not perform an accident determination output when the second harmonic is included.

[0009]

However, in such a conventional ratio differential relay, the operating state of the emergency generator at the time of a power failure has been reduced and the purchase after the power recovery has been completed, in conjunction with the recent complexity of the distribution system. When the transformer is turned on without turning off the load, such as when switching to electric power, the second harmonic component is small, so that a malfunction may occur.
The present invention has been made in view of the above circumstances, and an object of the present invention is to use a neural network to quickly and accurately determine whether or not an accident has occurred. To provide.

[0010]

SUMMARY OF THE INVENTION According to the present invention, there is provided a transformer protection ratio differential relay comprising: a current detecting means for detecting currents on a primary side and a secondary side of a transformer; An A / D conversion unit for A / D converting and outputting the difference current and the A / D
A wavelet analysis unit that obtains a basis function group corresponding to a frequency component by performing a similarity transformation on the output data of the D conversion unit, and normal data provided separately from the basis function group data that is the output of the wavelet analysis unit. It is characterized by comprising a learning means for learning a relationship with a judgment to be made, and a neural network for determining the presence or absence of an abnormality in the transformer with the support of the learning means.

In the present invention, by operating the neural network under the learning support means that has learned repeatedly, it is possible to detect abnormalities quickly and more accurately.

[0012]

FIG. 1 is an explanatory view of the principle of the present invention, in which a monitoring signal, which is an output from a monitoring sensor 21 for an electric quantity installed in a distribution system, is taken into a ratio differential relay 22. The ratio differential relay 22 includes a preprocessing unit 31, an A / D conversion unit 32, a wavelet analysis unit 33, and an abnormality determination unit and a learning support unit 35 configured using a neural network 34. The configurations of the preprocessing unit 31 and the A / D conversion unit 32 are substantially the same as those of the related art. That is, the pre-processing unit 31
Captures a quantity of electricity, which is a monitoring signal from the monitoring sensor 21, and obtains, for example, a current value and the like from the
Output to 2. The A / D converter 32 converts the input current value or the like, which is an analog amount, into a digital value, and outputs the digital value to the wavelet analyzer 33.

The wavelet analysis unit 33 performs a wavelet transform on the input current value and the like, and outputs the result to the neural network 34. When the wavelet-transformed data, for example, the input data is a square wave, it becomes data in which a discontinuous point appears in the converted waveform at the time of rising. Such wavelet-transformed data is continuously output to the neural network 34 and the learning support means 35. The learning support means 35 has a function of determining the connection weight count between the layers of the neural network 34, and gives a waveform after wavelet transform when an abnormality occurs and a regular square wave prepared separately, and The learning about the relationship between the waveform patterns and the judgment to be made based on these (accident or non-accident, that is, sound) is performed.

The neural network 34 usually comprises an input layer, a hidden layer, and an output layer. Each neuron in the input layer receives a signal (0 or 1) subjected to wavelet transform. The neurons in the input layer and the neurons in the intermediate layer are connected to each other by a predetermined connection weight coefficient, and each neuron in the intermediate layer and each neuron in the output layer are connected to each other by a predetermined connection weight coefficient. Each connection weight coefficient is set in accordance with an instruction from the learning support means 35 so that a predetermined output value is obtained from the output layer.

[0015]

Embodiments of the present invention will be described below with reference to the drawings. Figure 2 is a block diagram showing a configuration of a transformer protection ratio differential relay of the present invention, reference numeral 1 is a transformer, CT _1, CT ₂ primary side of the transformer 1, distribution on the secondary side respectively A sensor composed of a current detection coil and the like, 2 indicates a ratio differential relay. The ratio differential relay 2 has a current value I ₀ flowing through the coils CT ₁ and CT ₂ ,
A preprocessing unit 11 including current detection means for obtaining I ₁ and I ₂ ;
An A / D converter 12, a ratio differential detector 13 composed of a CPU or a DSP, a wavelet analyzer 33, a neural network 34 constituting an abnormality determination unit, and a learning support unit 35 are provided.

The pre-processing unit 11, A / D conversion unit 12, and ratio differential detection unit 13 are substantially the same as those shown in FIG. The output of the wavelet analysis unit 33 is input to a neural network 34 and learning support means 35, and the output of the neural network 34 and the ratio differential detection unit 13
Is output as an accident detection result by passing through the inhibit circuit 18.

The wavelet analysis unit 33 performs similarity conversion of the input function (waveform data) to create a set of basis functions localized in time and frequency. Signals to be analyzed (input data) using these basis functions as integration kernels
And a correlation value between the basis function group (converted data). That is, by calculating a correlation value between a certain basis function of the group of basis functions and the analysis target signal, a component of the target signal at a certain frequency at a time indicated by the basis function is determined.

When an input signal which is a digital value input from the A / D converter 12 is continuously wavelet-converted by the wavelet analyzer 33, a discontinuity appears at a point where an accident or an inrush occurs, and the discontinuity appears. Since the basis functions of the continuous points have different frequency components between the time of the accident and the time of the excitation rush, they are detected as different basis functions. Neural network 34
, The neural network 34 makes a determination under the learning support means 35 that has learned in advance.

FIG. 3 is a flowchart showing a learning process of the neural network 34. After initial setting of weights (step S1), conversion data from the wavelet analysis unit 33 and an external evaluation standard are given (step S2). ). First, a waveform which is an absolute value to be used for determining the occurrence of an accident is given, and "operation", that is, a case where it is determined that an accident has occurred is set as an evaluation standard. Next, a waveform of the inrush current having the same absolute value but including the second harmonic is given, and "inactive", that is, a case where the inrush current is determined as the evaluation criterion is set. Such an operation is performed on various waveform values to adjust the weight (step S3). That is, at the time of learning, the relationship between the waveform after various wavelet transforms at the time of occurrence of an accident and the waveform of a normal square wave is given to the neural network 34 to learn the relationship between the determination to be a waveform pattern, in other words, the weight of the waveform pattern. Make adjustments.

Next, in a state of being installed in an actual distribution system,
An actual inrush current is supplied to give an "inoperative" state as an evaluation criterion. Since the exciting inrush current changes depending on the applied phase and the load condition, the learning is repeatedly performed many times to enhance the reliability.

[0021]

As described above, according to the present invention, the data after the A / D conversion is introduced into the ratio operation detecting means, and at the same time, the wavelet transform is performed and the data is introduced into the neural network. The discrimination between the inrush current and the fault current can be made accurately and at a high speed. In addition, the operation timer, which was conventionally required, becomes unnecessary, and the damage of the accident can be minimized.
The present invention has excellent effects.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the principle of the present invention.

FIG. 2 is a block diagram showing a configuration of an embodiment according to the present invention.

FIG. 3 is a flowchart showing a learning process of the neural network.

FIG. 4 is a diagram illustrating the principle of a ratio differential relay.

FIG. 5 is a chart showing a relationship between respective current values for each accident case.

FIG. 6 is an explanatory diagram showing a comparison between a steady current and an exciting inrush current in a transformer.

FIG. 7 is a block diagram showing a configuration of a conventional ratio differential relay.

[Explanation of symbols]

1 transformer, 2,22 ratio differential relay, 11,31
Preprocessing unit, 12, 32 A / D conversion unit, 13 ratio differential detection unit, 21 monitoring sensor, 33 wavelet analysis unit, 34 neural network, 35 learning support means.

Claims (1)

[Claims] A current detecting means for detecting currents on a primary side and a secondary side of a transformer and a current difference between the currents; an A / D converter for A / D converting and outputting the currents and the current difference; A wavelet analysis unit that obtains a basis function group corresponding to a frequency component by performing similarity conversion on the output data of the A / D conversion unit, and normal data provided separately from the basis function group data output from the wavelet analysis unit. A learning means for learning a relationship with a judgment to be made based on the above, and a neural network for judging the presence or absence of an abnormality in the transformer with the aid of the learning means. Differential relay.