CN114563620A

CN114563620A - Circuit and method for identifying signal transmission direction of electric circuit

Info

Publication number: CN114563620A
Application number: CN202210206083.4A
Authority: CN
Inventors: 周亚萍
Original assignee: Individual
Current assignee: Individual
Priority date: 2022-02-28
Filing date: 2022-02-28
Publication date: 2022-05-31

Abstract

The invention discloses a circuit and a method for identifying the signal transmission direction of an electric line, namely, a filter is arranged on the electric line to be identified, transmission signals on the line are collected at the two sides of the filter, the signals are further conditioned and subjected to analog-to-digital conversion, then characteristics are extracted, the characteristics of two groups of digital signals are compared to generate corresponding signal characteristic vectors, and finally the signal characteristic vectors are input into a pre-trained network model, so that the transmission direction of the signals on the line is obtained through calculation and identification. By adopting the invention, the signal transmission direction, namely the direction from which the relevant interested signal comes from the electric circuit, can be effectively distinguished, the misjudgment caused by the environmental bypass crosstalk is eliminated, and the reliability of the product is improved; on the other hand, the invention can effectively distinguish signal sources, and can also greatly improve the sensitivity and the reaction speed of fault identification of the protection product.

Description

Circuit and method for identifying signal transmission direction of electric circuit

Technical Field

The invention belongs to the technical field of electrical protection equipment, and particularly relates to a circuit and a method for identifying the signal transmission direction of an electrical circuit.

Background

The AC low-voltage distribution network is a parallel access system, and phase lines and zero lines of all equipment accessed into the system are respectively connected together in a line mode, so that certain harmonic waves or EMC electromagnetic interference signals generated by the equipment accessed into the system can be transmitted to any other equipment connected into the transmission line or other parts of the system through the phase lines and the zero lines.

Similarly, the dc distribution network is also a parallel line system, the positive and negative poles of the power supply device are respectively connected with the positive and negative poles of all dc consumers, and any harmonic wave or EMC electromagnetic interference signal generated inside the power supply device and consumers can be transmitted to any other device connected to the transmission line or other parts of the system through the positive and negative poles.

On the other hand, general electrical equipment and power distribution systems can detect the operation of the equipment and find faults by detecting signals on power supply lines. For example, by detecting the current signal, the power consumption condition of the device or power distribution system can be obtained; by detecting the zero-sequence current signal of the phase zero line, the leakage fault of the equipment or the power distribution system can be detected; by detecting the harmonic signals, the power quality condition of the equipment or the power distribution system can be detected; by detecting some sort of current transient signal, an arc fault condition or the like of the equipment or power distribution system may be detected.

In particular, when the detection device is used, due to the large application of power electronic circuits and other power equipment containing high-frequency harmonic components in a power distribution network, the power equipment generates a large amount of high-frequency harmonic waves or EMC electromagnetic interference signals during operation, the high-frequency harmonic waves or the EMC electromagnetic interference signals are transmitted into the power distribution network through a power line, and the normal operation of other equipment connected to the power distribution network is influenced. How the detection device detects the source of the electromagnetic interference signal is beneficial to correctly diagnosing the occurrence position of the fault, such as the positioning of high-frequency harmonic generation equipment and the anti-interference design of power grid noise during fault arc detection, and the processing error caused by crosstalk signals generated by environmental harmonic waves or EMC noise is prevented.

Typical applications are arc fault protection appliances, which are generally required by the national standard GB/T31143(AFDD), crosstalk tests in test item 9.9.5.2; in response to the situation of intensified interference in real-world environment, for example, chinese patent publication No. CN1720652A provides a fault arc protection device for a bus device and chinese patent publication No. CN1917320A provides a fault arc protection circuit and a fault arc detection method, but these prior arts do not provide a solution to this problem. In practical tests, the reliability identification problem of the interference cannot be well solved by most market products; particularly, in the practical situation, under the working condition of low current, the electric arc fault recognition with high sensitivity can be realized, no misoperation is kept, and reliable recognition products are not seen so far.

Disclosure of Invention

In view of the above, the present invention provides a circuit and a method for identifying a signal transmission direction of an electrical line, and is particularly used for identifying a signal source as a power distribution network bypass or an equipment main path, so as to avoid misjudgment of a fault arc detection device due to the influence of a bypass signal.

An identification circuit for identifying a signal transmission direction of an electrical line, comprising:

the filter is arranged on an electrical circuit needing to be identified and used for filtering a transmission signal on the circuit;

the sensor is also arranged on the electrical circuit to be identified, positioned on two sides of the filter and used for detecting a transmission signal on the circuit;

the conditioning circuit is used for preprocessing the two groups of signals acquired by the sensor;

and the Microcontroller (MCU) is used for carrying out analog-to-digital conversion on the two groups of preprocessed signals, synchronously observing and comparing the frequency spectrum and the intensity of the digital signals, and further analyzing the transmission direction and the source of the signals on the outgoing line according to the signal characteristics.

Furthermore, the electric line can be a phase line or a zero line of a low-voltage electric line, a positive pole or a negative pole of a direct-current power supply line, or other medium-high voltage electric lines, communication lines or sensing signal transmission lines.

Furthermore, the filter is used for effectively attenuating signals passing through two sides of the filter, and the filter can be a low-pass filter, a high-pass filter, a band-pass filter or a band-stop filter according to the characteristics of the required identification signals, or a capacitor device is adopted to be bridged on two lines of an electric loop, or surge protection devices such as a piezoresistor, a gas discharge tube and a TVS tube are adopted.

Furthermore, the sensors are used for effectively sensing signals transmitted on a line, are designed in pairs, are respectively arranged on two sides of the filter, and can be current sensors or voltage sensors; if a current sensor is adopted, signals on the circuit can be coupled in a non-contact manner, or the manganin shunt is connected in series to the circuit to obtain signals; if a voltage sensor is adopted, the voltage sensor needs to be lapped on a phase line and a zero line of an electric circuit, and for a direct current power supply circuit, a communication circuit or a sensing signal transmission circuit, the voltage sensor needs to be lapped on a positive line and a negative line.

Furthermore, the conditioning circuit applies the conversion processing of isolation, bias, amplification, filtering, phase discrimination, amplitude discrimination and the like to the transmission signal on the line, so that the conditioning circuit is suitable for the analog-to-digital conversion processing of the microcontroller.

Further, the microcontroller performs feature extraction on the digital signals, and then performs comparison processing (such as feature value difference and feature value ratio) on the features of the two groups of digital signals to generate corresponding signal feature vectors, and then inputs the signal feature vectors into a pre-trained network model, and finally obtains the transmission direction of the signals on the line through forward calculation and identification.

Further, the features extracted by the microcontroller on the digital signal include: signal peak-to-peak value (maximum value of signal height), signal mean value (average value of signal height), signal effective value (root mean square value of signal height), effective signal quantity (quantity of signal height exceeding specific value), signal duty ratio (proportion of quantity of signal height smaller than specific value to total quantity), long-period signal peak-to-peak value, long-period signal mean value, long-period signal effective value, long-period effective signal quantity and long-period signal duty ratio, wherein the first 5 characteristics are specific to unit time length, and the last 5 characteristics are specific to long period, namely, several unit time lengths.

Further, the pre-training process of the network model is as follows: firstly, a large amount of experimental signal data needs to be collected and subjected to feature extraction and processing to generate a plurality of groups of signal feature vectors; under the condition of predicting the signal transmission direction, labeling each group of signal feature vectors so as to obtain a large number of positive samples and negative samples; and then selecting a multilayer neural network as a model architecture, inputting all positive samples and negative samples into the neural network one by one, and performing supervised machine learning training on the neural network, namely continuously performing back propagation on network parameters by using a gradient descent method according to an error function, and performing multiple iterations until the error function is converged, thus completing the training.

A method for identifying the signal transmission direction of an electric line comprises the steps of installing a filter on the electric line to be identified, collecting transmission signals on the line on two sides of the filter, further conditioning the signals, performing analog-to-digital conversion, extracting characteristics, comparing the characteristics of two groups of digital signals to generate corresponding signal characteristic vectors, and finally inputting the signal characteristic vectors into a pre-trained network model so as to calculate and identify the transmission direction of the signals on the line.

For electrical protection products in different application scenarios, a key problem is interference noise crosstalk of an electrical environment, which leads to misjudgment of the working state of a main circuit. By adopting the technology, the signal transmission direction, namely the direction from which the relevant interested signal comes from the electric circuit, can be effectively distinguished, the misjudgment caused by the environmental bypass crosstalk is eliminated, and the reliability of the product is improved; on the other hand, the invention can effectively distinguish signal sources, and can also greatly improve the sensitivity and the reaction speed of fault identification of the protection product.

Drawings

FIG. 1 is a schematic diagram of an identification circuit structure according to the present invention.

Fig. 2 is a connection schematic diagram of a single capacitor filter circuit.

Fig. 3(a) is a display result of a signal acquired by the sensor 1 when a dimming lamp signal is input from the main circuit side of the device in MATLAB software.

Fig. 3(b) is a display result of the signal acquired by the sensor 2 when the dimming lamp signal is input from the main circuit side of the device in the MATLAB software.

Fig. 4(a) is a display result of a signal acquired by the sensor 1 when a switching power supply signal is input from the main device side in MATLAB software.

Fig. 4(b) is a display result of a signal acquired by the sensor 2 when the switching power supply signal is input from the main device side in MATLAB software.

Fig. 5(a) is a display result of a signal acquired by the sensor 1 when a halogen lamp signal is input from the main device side in MATLAB software.

Fig. 5(b) is a display result of the signal acquired by the sensor 2 when the halogen lamp signal is input from the main device side in the MATLAB software.

Fig. 6(a) is a display result of a signal acquired by the sensor 1 when a dimming lamp signal is input from the side of the power distribution network in MATLAB software.

Fig. 6(b) is a display result of a signal acquired by the sensor 2 when a dimming lamp signal is input from the side of the power distribution network in MATLAB software.

Fig. 7(a) is a display result of a signal acquired by the sensor 1 when a switching power supply signal is input from the side of the power distribution network in MATLAB software.

Fig. 7(b) is a display result of a signal acquired by the sensor 2 when a switching power supply signal is input from the side of the power distribution network in MATLAB software.

Fig. 8(a) is a display result of a signal acquired by the sensor 1 when a halogen lamp signal is input from the side of the power distribution network in MATLAB software.

Fig. 8(b) is a display result of the signal acquired by the sensor 2 when the halogen lamp signal is input from the side of the power distribution network in the MATLAB software.

FIG. 9 is a schematic diagram of the schematic structure of the multi-layer neural network adopted in the present invention.

FIG. 10 is a flow chart of a machine learning algorithm according to the present invention.

Fig. 11 is a general flow chart illustrating a signal transmission direction identification scheme according to the present invention.

Detailed Description

In order to more specifically describe the present invention, the following detailed description is provided for the technical solution of the present invention with reference to the accompanying drawings and the specific embodiments.

The invention connects pairs of sensors in series on an electrical line with filters interposed between the sensors. For transmission signals, as shown in fig. 1, if the signal comes from a bypass of the distribution network, i.e. the signal enters from port 1, it will reach sensor 1 first, after being attenuated by the filter, it will reach sensor 2, and finally it will be output from port 2. If the transmission signal comes from the main device path, that is, the signal enters from the port 2, the signal will reach the sensor 2 first, and after being attenuated by the filter, the signal will reach the sensor 1, and finally the signal is output from the port 1.

Because the signal sensed by the sensor arriving firstly is stronger, and the signal sensed by the sensor arriving subsequently is weaker due to the attenuation effect of the filter. Therefore, for two equivalent sensors, induction quantities with different strengths are obtained, corresponding to the transmission directions of the signals.

In a specific embodiment, devices such as sensors and filters need to be selected specifically according to line characteristics, which is specifically shown in table 1:

TABLE 1

For the sensor in fig. 1, a pair of sensors capable of effectively sensing the signal transmitted by the line needs to be designed according to the frequency band and the intensity of the line signal. The sensors are arranged on either side of the filter, generally pairs of sensors of the same type and having the same parameters, sensors having the same response to the same signal; but not limited to, sensors with the same parameters, such as sensors with two sensing quantities having a fixed proportional relationship, even sensors of different types, can be finally restored to the spectral intensity of the original signal, so that the quantitative relationship of the original signal is obtained from the sensor output. The sensor type can be a current sensor, and can also be a voltage sensor, for example, the current sensor couples signals on a line in a non-contact manner, or for example, a manganin shunt is connected in series to the line to obtain signals; while the voltage sensor needs to be connected to two wires of the electrical line L, N, the dc power supply, communication or sensing signal transmission line needs to be connected to two wires of positive and ground (or positive and negative). The embodiment adopts paired current sensors with the acquisition frequency band of 100 k-50 Mhz.

For the filter in fig. 1, the frequency band and intensity matching design of the line signal is analyzed, the signal passing through two sides of the filter can be effectively attenuated, and the filter can be a low-pass filter, a high-pass filter, a band-stop filter and is not limited to the above type of filter circuit, or surge protection circuits such as a piezoresistor, a gas discharge tube, a TVS tube and the like are adopted. The simplest single capacitor filter is selected in this embodiment, as shown in fig. 2.

In order to test the filter and the direction discrimination effect, a plurality of commonly used loads with high-frequency harmonic waves, such as a dimming lamp, a switching power supply and a halogen lamp, are respectively connected to a port 2, namely a main device path, two sensors acquire high-frequency signals in circuits where the respective positions are located and input the high-frequency signals into a conditioning circuit for processing, the processed analog signals are subjected to analog-to-digital conversion of an MCU to obtain digital signals, waveforms of the digital signals are displayed in MATLAB, and the results are respectively shown in FIG. 3(a) and FIG. 3(b), FIG. 4(a) and FIG. 4(b), and FIG. 5(a) and FIG. 5 (b).

It can be observed that the different types of signals have a certain reduction in intensity after passing through the filter. Considering the signal height, the results are shown in table 2:

TABLE 2

Type of signal	Sensor 2	Sensor 1
			Dimming lamp	The highest signal height exceeds 2800	The maximum signal height is less than 2800
Switching power supply	Maximum signal height exceeding 2200	Maximum signal height less than 2200
			Halogen lamp	The highest signal height exceeds 1600	The highest signal height is less than 1600

The signal is input from the port 2, the sensor 2 acquires an original signal, the sensor 1 acquires a filtered signal, and the height of the signal acquired by the sensor 2 is higher than that of the signal acquired by the sensor 1. The results from table 2 also show that the direction of signal transmission is indeed port 2 to port 1, since the signal collected by sensor 2 is stronger.

And changing the direction of signal input for experiment, respectively selecting a dimming lamp, a switching power supply and a halogen lamp as bypass loads to be connected into the power distribution network, and selecting a 5A pure resistance load as a load main path. The bypass signal is input from the port 1, the two sensors acquire the high-frequency signals in the circuits and input the high-frequency signals into the conditioning circuit for processing, the processed analog signals are subjected to analog-to-digital conversion of the MCU to obtain digital signals, and the waveforms of the digital signals are displayed in MATLAB, and the results are respectively shown in FIG. 6(a) and FIG. 6(b), FIG. 7(a) and FIG. 7(b), and FIG. 8(a) and FIG. 8 (b).

It can be observed that different kinds of signals have a certain reduction in intensity after passing through the filter. Considering only the signal height, the results are shown in table 3:

TABLE 3

Type of signal	Sensor	1	Sensor 2
				Light modulation lamp	Maximum signal height over 2000	Maximum signal height less than 2000
Switching power supply	Maximum signal height over 1500	Maximum signal height less than 1500
			Halogen lamp	The highest signal height exceeds 1100	The highest signal height is less than 1100

Since the signal is input from the port 1, the signal of the sensor 1 is not filtered, and the signal of the sensor 2 is filtered, the signal height collected by the sensor 1 is higher than that collected by the sensor 2. The results from table 3 also show that the direction of signal transmission is indeed port 1 to port 2, since the signal collected by sensor 1 is stronger. It should be noted that, since the signal of the port 1 enters the main path through the bypass of the distribution network by crosstalk, the signal strength is reduced to some extent.

Although the signal height is a very important index for characterizing the signal strength, it is not enough in algorithm accuracy to use the signal height to characterize the signal strength, and if the signal strength is to be described more accurately, a software algorithm for digital signal processing needs to be added in the MCU. According to the digital signals acquired after analog-to-digital conversion, a digital signal processing algorithm is adopted to transform the two paths of signals, extract various characteristics, compare the frequency spectrum distribution and the strength of the signals, isolate and decompose the signals, and identify the transmission direction and the source of the signals.

After receiving the analog signals of the conditioning circuit 1 and the conditioning circuit 2, the MCU respectively performs analog-to-digital conversion to obtain digital signals, and then extracts characteristic values for the two paths of digital signals. The eigenvalues are mainly used to indicate signal strength, and different eigenvalues can be extracted according to different understandings, where the extracted eigenvalues are mainly as shown in table 4:

TABLE 4

The calculation of the obtained signal characteristic value and the identification of the signal direction are complex problems, and the characteristic value is polluted because the signal is interfered by various environmental factors and sometimes the signals in two directions are mixed. The existing large amount of experimental data are utilized to combine a feature vector set for effective synthesis, so that the prediction of the signal in actual occurrence is made, and the method is an effective idea. For example, using the above feature vector set, using a machine learning algorithm, an inference mechanism of the input vector and the recognition result can be established, and the flow is shown in fig. 10:

according to the eigenvalues acquired by the above-mentioned dual sensor channels, comparison processing is performed, for example, differences between the signal input port and the signal output port, such as eigenvalue differences and eigenvalue ratios, and through a large number of experiments, a main path direction signal and bypass direction signal eigenvector set is established, that is, operation 1 in fig. 10. Since the direction of the signal is known in advance in the above experiment, it is possible to set a flag for each feature vector, i.e. to indicate the direction of the signal corresponding to the vector, for example, to set the flag bit of the signal input from port 1 to 1 and the flag bit of the signal input from port 2 to 0, and to combine the flag bit as a target value with the feature value, the input feature vector of the supervised machine learning algorithm, i.e. operation 2 in fig. 10, is formed.

After experimental data exist, machine learning models need to be built, the number of the machine learning models is large, in the embodiment, a multilayer neural network is used, and the structural principle of the multilayer neural network is shown in fig. 9:

the Input layer is an Input layer, the corresponding characteristic values extracted in the front are corresponding, and each characteristic value can correspond to a neuron of the Input layer; the Hidden layers can be set with different layers and the number of neurons of each layer according to the selection of the Hidden layers; the Output layer is the Output layer, the corresponding result is the result of the input characteristic value passing through the multilayer neural network, and the corresponding result is the signal transmission direction.

The flow of a complete neural network is divided into a forward process and a reverse process, the flow of the forward process is Input layer Input, then output is obtained through the first layer of neuron operation, then the output of the first layer is used as the Input of the second layer, operation is continued to obtain the output of the second layer, and the operation is repeated until the output layer is operated, so that a final network operation result is obtained, namely operation 3 in fig. 10.

The reverse process is to compare the result obtained from the forward process with the preset flag bit to find the deviation between the two, and to make the final calculation result of the network approach to the target value of the flag bit by reasonably adjusting the multi-layer neural network, i.e. operation 4 in fig. 10.

After many training iterations, the output result of the model is closer to the expected result, and when the training times reach the preset value or the deviation between the network output value and the target value is smaller than the preset precision, the model is considered to be trained completely, i.e. operation 5 in fig. 10.

After training a machine learning model through a large amount of experimental data, a complete signal direction determination flow is shown in fig. 11:

the two sensors respectively acquire high-frequency signals in circuit current, obtain two paths of analog signals after being processed by a conditioning circuit, and input the analog signals into the MCU for analog-to-digital conversion to form two paths of digital signals; and respectively extracting characteristic values representing the strength from the digital signals, sending the characteristic values into a multi-layer neural network model trained in advance, outputting a final result through forward calculation, and judging the signal transmission direction according to the final result.

The foregoing description of the embodiments is provided to enable one of ordinary skill in the art to make and use the invention, and it is to be understood that other modifications of the embodiments, and the generic principles defined herein may be applied to other embodiments without the use of inventive faculty, as will be readily apparent to those skilled in the art. Therefore, the present invention is not limited to the above embodiments, and those skilled in the art should make improvements and modifications to the present invention based on the disclosure of the present invention within the protection scope of the present invention.

Claims

1. An identification circuit for identifying a signal transmission direction of an electrical line, comprising:

and the microcontroller is used for performing analog-to-digital conversion on the two groups of preprocessed signals, synchronously observing and comparing the frequency spectrum and the intensity of the digital signals, and further analyzing the transmission direction and the source of the signals on the line according to the signal characteristics.

2. The identification circuit of claim 1, wherein: the electric line can be a phase line or a zero line of a low-voltage electric line, can also be a positive pole or a negative pole of a direct-current power supply line, and can also be other medium-high voltage electric lines, communication lines or sensing signal transmission lines.

3. The identification circuit of claim 1, wherein: the filter is used for effectively attenuating signals passing through two sides of the filter, and can be a low-pass filter, a high-pass filter, a band-pass filter and a band-stop filter according to the characteristics of the signals to be identified, or a capacitor device is adopted to be bridged on two lines of an electric loop, or surge protection devices such as a piezoresistor, a gas discharge tube, a TVS tube and the like are adopted.

4. The identification circuit of claim 1, wherein: the sensors are used for effectively sensing signals transmitted on a line, are designed in pairs, are respectively arranged on two sides of the filter, and can be current sensors or voltage sensors; if a current sensor is adopted, signals on the circuit can be coupled in a non-contact manner, or the manganin shunt is connected in series to the circuit to obtain signals; if a voltage sensor is adopted, the voltage sensor needs to be lapped on a phase line and a zero line of an electric circuit, and for a direct current power supply circuit, a communication circuit or a sensing signal transmission circuit, the voltage sensor needs to be lapped on a positive line and a negative line.

5. The identification circuit of claim 1, wherein: the conditioning circuit is suitable for analog-to-digital conversion processing of the microcontroller by carrying out conversion processing such as isolation, bias, amplification, filtering, phase discrimination, amplitude discrimination and the like on transmission signals on a line.

6. The identification circuit of claim 1, wherein: the microcontroller extracts the characteristics of the digital signals, compares the characteristics of the two groups of digital signals to generate corresponding signal characteristic vectors, inputs the signal characteristic vectors into a pre-trained network model, and finally obtains the transmission direction of the signals on the line through forward calculation and identification.

7. The identification circuit of claim 6, wherein: the characteristics of the microcontroller for extracting the digital signals comprise: signal peak-to-peak value, signal mean value, signal effective value, effective signal quantity, signal duty ratio, long-period signal peak-to-peak value, long-period signal mean value, long-period signal effective value, long-period effective signal quantity, and long-period signal duty ratio, wherein the first 5 characteristics are specific to a unit time length, and the last 5 characteristics are specific to a long period, namely a plurality of unit time lengths.

8. The identification circuit of claim 6, wherein: the pre-training process of the network model comprises the following steps: firstly, acquiring a large amount of experimental signal data, extracting and processing features, and generating a plurality of groups of signal feature vectors; under the condition of predicting the signal transmission direction, labeling each group of signal feature vectors so as to obtain a large number of positive samples and negative samples; and then selecting a multilayer neural network as a model architecture, inputting all positive samples and negative samples into the neural network one by one, and performing supervised machine learning training on the neural network, namely continuously performing back propagation on network parameters by using a gradient descent method according to an error function, and performing multiple iterations until the error function is converged, thus completing the training.

9. A method for identifying the signal transmission direction of an electric circuit is characterized in that: firstly, a filter is installed on an electric circuit to be identified, transmission signals on the circuit are collected on two sides of the filter, then the signals are conditioned and subjected to analog-to-digital conversion, characteristics are extracted, then the characteristics of two groups of digital signals are compared to generate corresponding signal characteristic vectors, and finally the signal characteristic vectors are input into a pre-trained network model, so that the transmission direction of the signals on the circuit is obtained through calculation and identification.

10. The identification method according to claim 9, wherein: the pre-training process of the network model comprises the following steps: firstly, acquiring a large amount of experimental signal data, extracting and processing features, and generating a plurality of groups of signal feature vectors; under the condition of predicting the signal transmission direction, labeling each group of signal feature vectors so as to obtain a large number of positive samples and negative samples; and then selecting a multilayer neural network as a model architecture, inputting all positive samples and negative samples into the neural network one by one, and performing supervised machine learning training on the neural network, namely continuously performing back propagation on network parameters by using a gradient descent method according to an error function, and performing multiple iterations until the error function is converged, thus completing the training.