Fault arc experiment platform and fault arc detection method
By designing the fault arc experimental platform and the dual-channel convolutional neural network, the problem of the inability to simulate complex electrical faults and load conditions in the existing technology is solved, and high-accurate fault arc detection is achieved, providing sufficient data support and detection model. , improves the reliability of fault arc detection.
Patent Information
- Application Number
- CN202411136918.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-08-19
AI Technical Summary
The existing AC fault arc experimental platform lacks the complexity and diversity of power supply and distribution lines, and cannot simulate the recurrence of faults under serial and parallel connection methods and different load conditions. It also lacks electrical fault devices, resulting in insufficient detection data.
A fault arc experiment platform was designed, including power supply, electrical fault simulator, circuit breaker switch, fault arc generation module and current acquisition and analysis module. The experimental load is connected in series or in parallel, and different types of load combinations are used to replace loads, and use Dual-channel convolutional neural networks combine time and frequency domain features for fault arc detection.
It realizes the series or parallel function of the experimental load and the fault arc device, simulates the fault arc under different line connection methods and load conditions, reproduces, and obtains the current data of multiple fault arcs, improves the reliability of fault arc detection, and passes through the nerves. The network model achieves high accuracy of fault type identification.
Smart Images

Figure CN118937926A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of fault arc experiments, and in particular to a fault arc experiment platform and a fault arc detection method. Background Art
[0002] Electric arc, commonly known as "electric spark", has the characteristics of high temperature, low current and short duration. It is a gas ionization discharge phenomenon caused by excessive electric field strength when gas breaks down. Once breakdown occurs, it will occur frequently. Since arc discharge generates a lot of heat, it can ignite the surrounding flammable and explosive materials, causing fires or even explosions. Therefore, fault arcs are identified as an important cause of electrical fires, and it is necessary to conduct in-depth research on fault arc detection.
[0003] There are three main types of fault arcs, including series fault arcs, parallel fault arcs and ground fault arcs. Most of the existing AC fault arc test platforms are series circuits, and the experimental load is a single load. The complexity of the electrical circuits in the power supply and distribution lines and the diversity of electrical load types are not taken into account. There is a lack of series and parallel connection methods and fault reproduction under different load conditions. In addition, there is no electrical fault detector in the existing AC fault arc test platform, and there is a lack of arc data under different electrical fault conditions. Summary of the invention
[0004] The purpose of the present invention is to provide a fault arc experimental platform and a fault arc detection method to address the above-mentioned problems. By constructing the fault arc experimental platform, the function of connecting the experimental load in series or in parallel with the fault arc device can be realized, and different types of load combinations can be used for load replacement to ensure the reliability of fault arc detection.
[0005] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is as follows:
[0006] According to one aspect of the present invention, a fault arc experimental platform and a fault arc detection method are provided, including a power supply, an electrical fault simulator, a circuit breaker, a fault arc generation module and a current acquisition and analysis module;
[0007] The power supply is connected to the electrical fault simulator;
[0008] The electrical fault simulator, the circuit breaker and the fault arc generation module are connected in series;
[0009] A current sampling point is provided between the circuit breaker and the arc fault generation module, and the current acquisition and analysis module is connected to the current sampling point;
[0010] The fault arc generating module comprises a fault arc generating device and an experimental load, and the fault arc generating device is connected in series or in parallel with the experimental load.
[0011] Preferably, the electrical fault simulator includes a lightning surge generator, a pulse group generator and an electrostatic discharge generator.
[0012] Preferably, the fault arc generating device comprises a movable electrode and a stationary electrode, the movable electrode is fixedly arranged on a movable mechanism, the movable mechanism is used to move the movable electrode, the stationary motor is fixedly arranged on an insulating base, and the movable electrode can be close to or away from the movable electrode.
[0013] Preferably, the experimental load includes an electric kettle, a hair dryer, an induction cooker, an electric fan, an energy-saving lamp, a laptop computer and a microwave oven.
[0014] Preferably, the current acquisition and analysis module comprises a current clamp and an oscilloscope, the current clamp is mounted on the current sampling point, and the current clamp is electrically connected to the oscilloscope.
[0015] Preferably, the method comprises the following steps:
[0016] S1. Connect the power supply, electrical fault simulator, circuit breaker, arc fault generator and experimental load in sequence, wherein the arc fault generator and the experimental load are connected in series or in parallel, and connect the current clamp to the circuit and the oscilloscope;
[0017] S2, turning on the electrical fault simulator, controlling the electrical fault simulator to generate an electrical fault and inputting it into the fault arc generating device;
[0018] S3. When a simulated electrical fault occurs, a stable burning arc is generated by adjusting the distance between the moving electrode and the stationary electrode;
[0019] S4, collecting and storing data of current signals of normal and fault arc states through an oscilloscope;
[0020] S5, change different loads and repeat steps S2-S4;
[0021] S6, changing the connection mode between the fault arc generating device and the experimental load, and repeating steps S2-S5;
[0022] S7, replace the electrical fault simulator and repeat steps S2-S6;
[0023] S8. Preprocess the collected arc current data, analyze and extract the time domain features of the current waveform, and extract the frequency domain features of different time scales. The time domain features and the frequency domain features are respectively input into a dual-channel convolutional neural network for training to obtain a fused feature vector. The fused feature vector is used to train a fault arc detection model, and the trained fault arc detection model is used to detect the fault arc and output the fault arc type.
[0024] Preferably, in step S8, the dual-channel convolutional neural network includes a one-dimensional convolutional neural network and a two-dimensional convolutional neural network, the frequency domain features are input into the one-dimensional convolutional neural network for training, and the time domain features are input into the two-dimensional convolutional neural network for training.
[0025] Preferably, the frequency domain features are extracted using wavelet transform, and the formula of the wavelet transform is:
[0026]
[0027] Among them, f(t) is the current time domain signal, ψ is the wavelet basis function, a is the scale that controls the expansion and contraction of the wavelet function; τ is the translation amount that controls the translation of the wavelet function, and t is time.
[0028] Preferably, in step S8, the loss function of the arc fault detection model is:
[0029]
[0030] Where N represents the total number of data samples; y i represents the label of the data sample, Represents the label of the sample prediction.
[0031] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:
[0032] 1. The present invention builds a fault arc experimental platform, which can realize the function of connecting the experimental load in series or in parallel with the fault arc device, and can use different types of load combinations for load replacement to ensure the reliability of fault arc detection. It can also simulate fault arc reproduction experiments under different line connection modes and different load conditions, obtain current data of various fault arcs, and provide sufficient and effective data support for fault arc detection.
[0033] 2. The present invention adds an electrical fault device to simulate different electrical fault generators to simulate fault arc experiments under different electrical faults. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] Figure 1 It is a structural schematic diagram of the series experimental platform of the present invention;
[0035] Figure 2 It is a structural schematic diagram of the parallel experimental platform of the present invention;
[0036] Figure 3 is a flow chart of the fault arc detection method of the present invention;
[0037] Figure 4 It is a schematic diagram of the dual-channel convolutional neural network architecture of the present invention. DETAILED DESCRIPTION
[0038] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and preferred embodiments. However, it should be noted that many details listed in the specification are only for the purpose of enabling the reader to have a thorough understanding of one or more aspects of the invention, and these aspects of the present invention can be realized even without these specific details.
[0039] See also Figures 1 to 4 The present invention provides a fault arc experimental platform and a fault arc detection method, and the technical scheme is as follows:
[0040] A fault arc experimental platform includes a power supply, an electrical fault simulator, a circuit breaker, a fault arc generation module and a current acquisition and analysis module. The power supply is a 220V / 50Hz AC power supply. The electrical fault simulator is one of a lightning surge generator, a pulse group generator and an electrostatic discharge generator. The lightning surge generator is used to simulate high-energy transient interference caused by switches and natural lightning strikes; the pulse group generator is used to simulate electrical fast transient pulse group interference; the electrostatic discharge generator is used to simulate electrostatic discharge interference directly from operators and adjacent objects. The AC power supply is connected to the electrical fault simulator. The electrical fault simulator, the circuit breaker and the fault arc generation module are connected in series. The fault arc generation module includes a fault arc generation device and an experimental load. The fault arc generation device includes a moving electrode and a stationary electrode. The moving electrode is a 6.0mm copper rod with a cone at the front end, and the moving electrode can be controlled to move forward or backward. The movable electrode is fixedly mounted on a movable mechanism, and the movable mechanism is used to move the movable electrode. It should be noted that the movable mechanism may be an electrically adjustable mechanism, such as a cylinder driving the movable electrode to move, or an electric telescopic rod driving the movable electrode to move; it may also be a manually adjustable mechanism, such as a slide groove, a slider in the slide groove, and the movable electrode is arranged on the slider, and the movement of the movable electrode can be manually controlled. As long as the mechanism can move the movable electrode, the movement operation of the movable electrode can be realized, and this application will not elaborate on it. The stationary electrode is a 6.0 mm carbon rod with a flat front end, fixedly mounted on an insulating base. When a simulated electrical fault occurs, a stable burning arc is generated by adjusting the distance between the movable electrode and the stationary electrode. Among them, if Figure 1-2As shown, the fault arc generating device is connected in series or in parallel with the experimental load. When the fault arc generating device is connected in series with the experimental load, the movable electrode is connected in series with the circuit breaker, and the stationary electrode is connected in series with the experimental load, or the stationary electrode is connected in series with the circuit breaker, and the movable electrode is connected in series with the experimental load. When the fault arc generating device is connected in parallel with the experimental load, the fault arc generating device is first connected in series with a fixed load and then connected in parallel with the experimental load. In this embodiment, the experimental load is divided into a series load and a parallel load, as shown in the following table:
[0041] Table 1 Series test load
[0042]
[0043]
[0044] Table 2 Parallel test load
[0045] Serial number Load Name Serial number Load Name 1 Electric kettle 9 Induction cooker + electric fan 2 Hair dryer 10 Electric kettle+Electric kettle 3 induction cooker 11 Electric kettle + energy-saving lamp 4 Electric fan 12 Hair dryer + electric fan 5 Energy-saving lamps 13 Induction cooker + energy-saving lamp 6 Laptop 14 Hair dryer + energy-saving lamp 7 Micro-wave oven 15 Hair dryer + laptop 8 Electric kettle + electric fan 16 Energy-saving lamp + microwave oven
[0046] As shown in Table 1, the series experimental loads include electric kettles, hair dryers, induction cookers, electric fans, energy-saving lamps, laptop computers and microwave ovens. When performing the fault arc test of the experimental loads in series connection, the fault arc data under different loads can be obtained by replacing different experimental loads.
[0047] As shown in Table 2, the parallel test loads include electric kettles, hair dryers, induction cookers, electric fans, energy-saving lamps, laptop computers and microwave ovens, or two of them are combined as parallel loads. When performing the fault arc test of the test load under parallel connection, the fault arc data under different loads can be obtained by replacing different test loads.
[0048] In order to collect the current data generated when the fault arc occurs, a current sampling point is set between the wire connected in series between the circuit breaker and the fault arc generation module, and the current acquisition and analysis module is connected to the current sampling point. The current acquisition and analysis module includes a current clamp and an oscilloscope. The current clamp is set on the wire of the current sampling point, and the current clamp is electrically connected to the oscilloscope. The current signal of the normal and fault arc states is collected and stored by the oscilloscope.
[0049] The built fault arc experimental platform can realize the function of connecting the experimental load in series or in parallel with the fault arc device. Since the arc current presents complex nonlinear characteristics, typical household loads are selected and different types of load combinations are fully considered to ensure the reliability of fault arc detection. The fault arc reproduction experiment under different line connection methods and different load conditions can be simulated.
[0050] At present, in the fault arc detection method, many studies focus on the time-frequency domain feature analysis of the arc current signal, and use tools such as discrete wavelet transform and empirical mode decomposition to extract the time-frequency feature components in the current signal. However, whether it is based on the time domain, frequency domain or time-frequency domain fusion method, it is necessary to manually calculate and adjust the threshold, which is difficult to adapt to complex situations. There are also methods that integrate machine learning to identify fault arcs, such as using the least squares support vector machine (LS-SVM) to identify the current feature vector to detect the fault arc; using the BP neural network based on the wavelet transform component to identify the series fault arc of the common load; collecting the high-frequency coupling signal of the normal and fault arc of the typical load for bandpass filtering, and using the TDV-CNN algorithm to extract the high-dimensional features of the grayscale image; using the parallel deep convolutional neural network of AlexNet and parallel AlexNet to detect the fault arc. However, the current fault arc detection combined with neural network often extracts a single feature value, ignores the connection between the time domain and the frequency domain, and cannot fully integrate the characteristics of the fault arc current signal.
[0051] Therefore, in order to solve the above problems, Figure 3 As shown, in this embodiment, based on the above-mentioned arc fault experimental platform, a method for detecting arc fault is disclosed, and the method includes the following steps:
[0052] S1. Electrical circuit connection: Connect the power supply, electrical fault simulator, circuit breaker, fault arc generator and experimental load in sequence. The fault arc generator and experimental load are connected in series or in parallel. The series circuit is as follows: Figure 1 As shown, the parallel circuit is Figure 2 Finally, connect the current clamp to the circuit and connect it to the oscilloscope.
[0053] S2. Simulate electrical fault: turn on the electrical fault simulator, and control the electrical fault simulator to generate an electrical fault and input it into the fault arc generating device.
[0054] S3. Pull open the arc: When a simulated electrical fault occurs, manually adjust the thousandth knob of the fault arc generating device to slowly move the moving electrode back to a certain electrode spacing to generate a stable burning arc.
[0055] S4. Data acquisition: The current signals of normal and fault arc states are collected and stored by an oscilloscope. The sampling frequency of the oscilloscope is set to 62.5kHz and the sampling time is set to 4s.
[0056] S5, repeat the experiment: change different loads and repeat steps S2-S4;
[0057] S6. Repeat the experiment: change the connection mode between the fault arc generating device and the experimental load, and repeat steps S2-S5;
[0058] S7. Repeat the experiment: Replace the electrical fault simulator and repeat steps S2-S6.
[0059] By replacing different loads, changing the series and parallel connection methods of the fault arc generating device and the experimental load, and replacing the electrical fault simulator, various types of fault arcs can be simulated, and current data under various conditions can be obtained for subsequent detection and analysis of the current data.
[0060] S8. Fault arc detection: pre-process the arc current data collected as mentioned above, analyze and extract the time domain features of the current waveform, and extract the frequency domain features of different time scales. The time domain features and the frequency domain features are respectively input into a dual-channel convolutional neural network for training to obtain a fused feature vector. The fused feature vector is used to train a fault arc detection model, and the trained fault arc detection model is used to detect the fault arc, and the fault arc type is output.
[0061] Specifically, the dual-channel convolutional neural network DCNN includes a one-dimensional convolutional neural network 1D-CNN and a two-dimensional convolutional neural network 2D-CNN. Figure 4 As shown, the dual-channel convolutional neural network DCNN divides the time-frequency domain features extracted by wavelet transform into low-frequency and high-frequency components and inputs them into 1D-CNN training, where the formula of wavelet transform is as follows:
[0062]
[0063] Among them, f(t) is the current time domain signal, ψ is the wavelet basis function, a is the scale of controlling the expansion and contraction of the wavelet function, and the scale corresponds to the frequency; τ is the translation amount of controlling the translation of the wavelet function, and the translation amount corresponds to time, and t is time.
[0064] 1D-CNN mainly includes 3 convolutional layers, 3 maximum pooling layers and 1 fully connected layer. The output of the last pooling layer is flattened. To avoid gradient explosion and gradient disappearance problems, a nonlinear activation function ReLU is added after each convolutional layer. The ReLU function is shown in the following formula:
[0065] ReLU(x)=Max(0,x)
[0066] Add the sigmoid function after the fully connected layer. The sigmoid function is shown as follows:
[0067]
[0068] The collected current time domain signal is input into 2D-CNN training. 2D-CNN mainly includes 5 convolutional layers, 3 maximum pooling layers and 3 fully connected layers. ReLU activation function is applied before each maximum pooling layer.
[0069] The features extracted from the two channels are input to the fully connected layer for label classification. Each neuron in the fully connected layer is connected to all neurons with learnable weights in the previous layer. The feature vectors extracted from the two channels are spliced for output discrimination to obtain a fused feature vector. The fused feature vector is used to train the fault arc detection model, which is an artificial neural network model. The trained fault arc detection model is then used to detect the fault arc and identify the type of the fault arc.
[0070] In order to make the fault arc detection model have better generalization ability and improve the feature recognition ability of the convolutional neural network, and make it converge faster, the fault arc detection model sets the number of iterations to 120, the learning rate to 0.0001, and the loss function uses the binary cross entropy. The formula of the loss function is as follows:
[0071]
[0072] Where N represents the total number of data samples; y i represents the label of the data sample, Indicates the label of sample prediction. In the model optimization process, the Adam optimization algorithm is used, Dropout is introduced and the Dropout rate is set to 0.5.
[0073] By training with different types of loads in series circuits and parallel circuits, the accuracy of the fault arc can reach up to 96.6%, which verifies the effectiveness of this method in fault arc detection. By training the artificial neural network model with fused feature vectors, fault arc detection can be achieved for different simulated electrical faults, different load conditions, and different connection methods. Wavelet analysis is used to perform multi-scale decomposition to extract the time-frequency domain characteristics of the current and enhance the connection between the time-frequency domain of the fault arc signal.
[0074] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
Claims
1. A fault arc test platform, characterized in that: It includes power supply, electrical fault simulator, circuit breaker, arc fault generation module and current acquisition and analysis module; The power supply is connected to the electrical fault simulator; The electrical fault simulator, the circuit breaker and the fault arc generation module are connected in series; A current sampling point is provided between the circuit breaker and the arc fault generation module, and the current acquisition and analysis module is connected to the current sampling point; The fault arc generating module comprises a fault arc generating device and an experimental load, and the fault arc generating device is connected in series or in parallel with the experimental load.
2. The arc fault experimental platform according to claim 1, characterized in that: The electrical fault simulator comprises a lightning surge generator, a pulse group generator and an electrostatic discharge generator.
3. The arc fault experimental platform according to claim 1, characterized in that: The fault arc generating device comprises a movable electrode and a stationary electrode. The movable electrode is fixedly arranged on a movable mechanism, and the movable mechanism is used to move the movable electrode. The stationary motor is fixedly arranged on an insulating base, and the movable electrode can be close to or away from the movable electrode.
4. The arc fault experimental platform according to claim 1, characterized in that: The experimental loads include an electric kettle, a hair dryer, an induction cooker, an electric fan, an energy-saving lamp, a laptop computer and a microwave oven.
5. The arc fault experimental platform according to claim 1, characterized in that: The current acquisition and analysis module comprises a current clamp and an oscilloscope. The current clamp is mounted on the current sampling point, and the current clamp is electrically connected to the oscilloscope.
6. A fault arc detection method for a fault arc experimental platform, characterized in that: The following steps are involved: S1. Connect the power supply, electrical fault simulator, circuit breaker, arc fault generator and experimental load in sequence, wherein the arc fault generator and the experimental load are connected in series or in parallel, and connect the current clamp to the circuit and the oscilloscope; S2, turning on the electrical fault simulator, controlling the electrical fault simulator to generate an electrical fault and inputting it into the fault arc generating device; S3. When a simulated electrical fault occurs, a stable burning arc is generated by adjusting the distance between the moving electrode and the stationary electrode; S4, collecting and storing data of current signals of normal and fault arc states through an oscilloscope; S5, change different loads and repeat steps S2-S4; S6, changing the connection mode between the fault arc generating device and the experimental load, and repeating steps S2-S5; S7, replace the electrical fault simulator and repeat steps S2-S6; S8. Preprocess the collected arc current data, analyze and extract the time domain features of the current waveform, and extract the frequency domain features of different time scales. The time domain features and the frequency domain features are respectively input into a dual-channel convolutional neural network for training to obtain a fused feature vector. The fused feature vector is used to train a fault arc detection model, and the trained fault arc detection model is used to detect the fault arc and output the fault arc type.
7. The arc fault detection method of the arc fault experimental platform according to claim 6, characterized in that: In step S8, the dual-channel convolutional neural network includes a one-dimensional convolutional neural network and a two-dimensional convolutional neural network, the frequency domain features are input into the one-dimensional convolutional neural network for training, and the time domain features are input into the two-dimensional convolutional neural network for training.
8. The arc fault experimental platform and arc fault detection method according to claim 6, characterized in that: The frequency domain features are extracted using wavelet transform, and the formula of the wavelet transform is: f(t) is the current time domain signal, ψ is the wavelet basis function, a is the scale that controls the expansion and contraction of the wavelet function; τ is the translation amount that controls the translation of the wavelet function, and t is time.
9. The arc fault detection method of the arc fault test platform according to claim 6, characterized in that: In step S8, the loss function of the arc fault detection model is: Where N represents the total number of data samples; y i represents the label of the data sample, Represents the label of the sample prediction.
Citation Information
Patent Citations
Fault arc signal detection method adopting convolutional neural network
CN113049922A
Three-phase inverter fault identification method based on hybrid convolutional neural network
CN113159077A
Refrigeration equipment fault diagnosis method based on multi-modal feature fusion neural network
CN114970605A
Analog circuit fault diagnosis method based on multi-scale convolutional neural network
CN115327354A
Lightning stroke broken strand simulation test method and processor for power transmission line wire
CN115932504A
