A lightning arrester status monitoring device based on partial discharge and leakage current

Abstract

The present invention relates to a lightning arrester state monitoring device based on partial discharge and leakage current, belongs to the field of lightning arrester state detection technology, and solves the problem that the prior art is prone to misjudgment. The device of the present invention includes: a partial discharge acquisition circuit, which acquires the partial discharge signal of the lightning arrester; an SDP image generation module, which generates an SDP image based on the partial discharge signal; a resistive current extraction circuit, which acquires the leakage current of the lightning arrester and extracts the resistive current signal in the leakage current; a resistive current waveform generation module, which generates a waveform image of the resistive current signal; a state detection model, which determines that the lightning arrester is in a normal state, an internal valve damp state, or a structural damage state based on the SDP image; when the lightning arrester is in a structural damage state, the lightning arrester is determined to be in an internal valve aging state or an insulation layer damage state based on the waveform image of the resistive current signal. In the present invention, the working state of the lightning arrester is detected based on the partial discharge signal and the leakage current signal, which improves the detection accuracy and reduces misjudgment.

Description

Technical Field

[0001] The invention relates to the technical field of arrester state detection, and in particular to an arrester state monitoring device based on partial discharge and leakage current. Background Art

[0002] Lightning arresters are important electrical equipment that protect power systems from overvoltage. Lightning arresters use the nonlinear volt-ampere characteristics or gap of their valve body to allow a very small current to flow when working under normal voltage; when working under overvoltage, the resistance of the arrester valve body drops sharply to release the overvoltage energy caused by lightning or operation, thereby reducing the damage caused by overvoltage to power grid equipment and lines, and playing an important role in protecting power systems from overvoltage disasters. Lightning arresters are widely used in lines and equipment at all levels, and can limit the impact of surge overvoltage caused by lightning strikes and operating overvoltage on power grids and related equipment. The quality of their operating performance has a very important impact on the safe and stable operation of primary equipment in the power grid.

[0003] When the arrester is actually operating in the power system, it is exposed to the air for a long time and is affected by factors such as atmospheric temperature and humidity changes and overvoltage caused by various reasons. The arrester structure will age and deteriorate, thus losing its protective effect on power equipment. In addition, the arrester with severe aging and deterioration may explode and cause safety hazards. As an important facility of the smart grid, the arrester status monitoring device monitors the working status of the arrester in real time, which is of great significance to ensure the safe and stable operation of the power system.

[0004] The existing arrester monitoring device usually only collects the leakage current of the arrester, and then judges whether the arrester is in a normal state or a fault state according to the collected leakage current value. However, the existing arrester monitoring device only judges according to the value, which is easy to make a misjudgment, and it is also unable to accurately identify the type of the arrester's fault state. Summary of the invention

[0005] In view of the above analysis, an embodiment of the present invention aims to provide a lightning arrester status monitoring device based on partial discharge and leakage current, so as to solve the technical problem that the existing lightning arrester status detection method is prone to misjudgment and cannot accurately identify the type of lightning arrester fault status.

[0006] An embodiment of the present invention provides a lightning arrester state monitoring device based on partial discharge and leakage current, the device comprising:

[0007] Partial discharge collection circuit, collecting partial discharge signals of lightning arresters;

[0008] An SDP image generation module, generating an SDP image based on the partial discharge signal;

[0009] A resistive current extraction circuit collects the leakage current of the arrester and extracts a resistive current signal from the leakage current;

[0010] A resistive current waveform generating module, generating a waveform image of the resistive current signal;

[0011] The state detection model determines, based on the SDP image, that the arrester is in a normal state, an internal valve is damp, or a structurally damaged state; when the arrester is in a structurally damaged state, the arrester is determined to be in an internal valve aging state or an insulation layer damaged state based on the waveform image of the resistive current signal.

[0012] Based on the further improvement of the above device, the partial discharge acquisition circuit includes: a first current detection device, an autotransformer, a first resistor, an inductor, a first capacitor and a second resistor,

[0013] Among them, the input end of the first current detection device is connected to the grounding wire of the lightning arrester, the output end of the first current detection device is connected to the input end of the autotransformer, one output end of the autotransformer is connected to one end of the first resistor, the other end of the first resistor is connected to one end of the inductor, the other end of the inductor is connected to one end of the first capacitor, the other end of the first capacitor is connected to the other output end of the autotransformer, the two ends of the second resistor are respectively connected to the two ends of the first capacitor, and the two ends of the second resistor output the partial discharge signal.

[0014] Based on further improvement of the above device, the partial discharge acquisition circuit further includes: a signal amplifier, the input end of the signal amplifier is connected to the two ends of the second resistor, and the signal amplifier amplifies the partial discharge signal and then outputs it.

[0015] Based on the further improvement of the above device, the first current detection device is a current transformer or a Rogowski coil.

[0016] Based on the further improvement of the above device, the resistive current extraction circuit includes: a second current detection device, a capacitive current acquisition module and a differential amplifier,

[0017] The input end of the second current detection device is connected to the grounding wire of the arrester, and is used to collect the leakage current of the arrester, and the output end of the second current detection device is connected to the positive input end of the differential amplifier;

[0018] The capacitive current acquisition module is used to collect the capacitive current in the leakage current, the input end of which is connected between the high-voltage transmission line to which the arrester is connected and the grounding terminal, and the output end is connected to the negative input end of the differential amplifier;

[0019] The output end of the differential amplifier outputs a resistive current signal in the leakage current.

[0020] Based on the further improvement of the above device, the capacitive current acquisition module includes: a voltage transformer, a differential phase shift circuit and a controllable gain amplifier.

[0021] The input end of the voltage transformer is connected between the high-voltage transmission line to which the lightning arrester is connected and the grounding end, the output end of the voltage transformer is connected to the input end of the differential phase shift circuit, the output end of the differential phase shift circuit is connected to the input end of the controllable gain amplifier, and the output end of the controllable gain amplifier is connected to the negative input end of the differential amplifier.

[0022] Based on the further improvement of the above device, the capacitive current acquisition module also includes: a photoelectric isolation module, and the photoelectric isolation module is connected between the voltage transformer and the differential shift circuit.

[0023] Based on the further improvement of the above device, the second current detection device is a current transformer or a Rogowski coil.

[0024] Based on the further improvement of the above device, the resistive current waveform generation module includes:

[0025] A first analog-to-digital conversion module converts the extracted resistive current signal from an analog signal to a digital signal;

[0026] An image drawing module draws a waveform diagram of the resistive current signal according to the resistive current signal converted into a digital signal.

[0027] Based on the further improvement of the above device, the SDP image generation module includes:

[0028] A second analog-to-digital conversion module converts the collected partial discharge signal from an analog signal to a digital signal;

[0029] A signal processing module, performing signal processing on the partial discharge signal converted into a digital signal, and using the processed signal as a state identification signal;

[0030] The image conversion module converts the state recognition signal into an SDP image.

[0031] Based on the further improvement of the above device, the signal processing module uses Hilbert-Huang transform to perform signal processing to obtain the state recognition signal, including the following steps:

[0032] Decomposing the partial discharge signal into a plurality of intrinsic mode components and corresponding residual terms by using an empirical mode decomposition method;

[0033] Performing Hilbert transform on each of the intrinsic modal components to obtain a corresponding analytical signal;

[0034] The analytical signals of all the intrinsic modal components and the residual terms are combined to obtain a state identification signal corresponding to the partial discharge signal.

[0035] Based on the further improvement of the above device, the formula for performing Hilbert transform on each of the intrinsic modal components to obtain the corresponding analytical signal is:

[0036] ;

[0037] ;

[0038] In the formula, is the ith natural mode component, for Hilbert transform, t is the signal time point, is a variable in the integral, representing any time point from the upper limit to the lower limit of the integral. is the Hilport nucleus; yes The analytical signal, Is an imaginary unit.

[0039] Based on the further improvement of the above device, the image conversion module converts the state recognition signal into an SDP image, including the following steps:

[0040] Mapping each discrete point in the state identification signal into a polar coordinate space to obtain the polar coordinates of each discrete point;

[0041] In the polar coordinate space, symmetrizing each discrete point to obtain a corresponding polar coordinate point;

[0042] Each discrete point and the polar coordinate point after symmetry processing are graphically displayed in the polar coordinate space to obtain the SDP image.

[0043] Based on the further improvement of the above device, mapping each discrete point in the state identification signal to a polar coordinate space to obtain the polar coordinates of each discrete point includes:

[0044] The polar coordinate radius, the clockwise rotation angle of the initial line, and the counterclockwise rotation angle of the initial line corresponding to each discrete point in the polar coordinate space are calculated according to the state recognition signal. The formula is:

[0045] ;

[0046] ;

[0047] ;

[0048] In the formula, for The polar coordinate radius of the discrete point at the time, is the clockwise rotation angle of the initial line, is the counterclockwise rotation angle of the initial line, for The signal amplitude at discrete points in time, is the minimum signal amplitude of the state identification signal, is the maximum signal amplitude of the state identification signal, is the lag time, is the initial rotation angle, k is the magnification;

[0049] According to the polar coordinate radius, the clockwise rotation angle of the initial line, and the counterclockwise rotation angle of the initial line, the polar coordinates of the discrete point are determined as .

[0050] Based on the further improvement of the above device, the state detection model includes a first state detection model;

[0051] The SDP image is input into a first state detection model for classification and identification, and according to the classification result output by the first state detection model, it is determined that the arrester is in a normal state, an internal valve is damp, and a structural damage state.

[0052] Based on the further improvement of the above device, the state detection model also includes a second state detection model;

[0053] When the arrester is determined to be in a structural damage state according to the classification result output by the first state detection model, the waveform image of the resistive current signal is input into the second state detection model for classification and identification, and the arrester is determined to be in an internal valve aging state or an insulation layer damage state according to the classification result output by the second state detection model.

[0054] Based on further improvement of the above device, the first state detection model and the second state detection model are both neural network models.

[0055] Based on the further improvement of the above device, the neural network model includes:

[0056] The first convolutional layer is used to extract features from the input image and output a feature map;

[0057] The first pooling layer is used to aggregate the feature maps output by the first convolutional layer and output the feature maps;

[0058] The second convolutional layer is used to extract features from the feature map output by the first pooling layer and output the feature map;

[0059] A second pooling layer, used to aggregate the feature maps output by the second convolutional layer and output the feature maps; and

[0060] The fully connected layer is used to fuse the feature maps output by the second pooling layer and output the working status of the arrester.

[0061] Based on the further improvement of the above device, the fully connected layer includes a flat layer, a hidden layer and an output layer connected in sequence,

[0062] Among them, the flat layer is used to convert the feature map output by the second pooling layer into one-dimensional data, the hidden layer is used to perform feature fusion on the one-dimensional data output by the flat layer and extract complex features, and the output layer is used to convert the output of the hidden layer into the working state of the arrester.

[0063] Based on the further improvement of the above device, the sizes of the convolution kernels of the first convolution layer and the second convolution layer are 3×3.

[0064] Based on the further improvement of the above device, the activation function of the neural network model is a ReLu function, and the formula of the ReLu function is:

[0065] ;

[0066] Where x is the input.

[0067] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0068] 1. The device of the present invention realizes the monitoring of four states of the arrester, namely, the normal state, the damp state of the inner valve, the aging state of the inner valve, or the damaged state of the insulation layer, based on the partial discharge signal and the leakage current. Specifically, considering that the amplitude and frequency of the partial discharge signal waveform of the arrester in different working states have different characteristics, and the partial discharge signals of the arrester in two structural damage states, namely, the aging state of the inner valve and the damaged state of the insulation layer, are relatively similar, therefore, the working state of the arrester is firstly classified and identified based on the partial discharge signal, and the working state of the arrester is initially divided into three states, namely, the normal state, the damp state of the inner valve, and the structural damage state. Then, when it is determined that the arrester is in a structural damage state, the aging state of the inner valve and the damaged state of the insulation layer are further classified and identified based on the leakage current.

[0069] 2. In the device of the present invention, the local discharge signal of the arrester is first collected by the local discharge collection circuit, and then the SDP image generation module generates the SDP image based on the local discharge signal, and finally the SDP image is input into the state detection model for classification and identification, so as to determine whether the working state of the arrester is a normal state, an internal valve damp state or a structural damage state, and realize the preliminary classification and identification of the arrester state. Among them, the SDP image can bring together the waveform characteristics of the local discharge signal, which is more conducive to image recognition, thereby improving the state detection accuracy of the arrester and reducing misjudgment.

[0070] 3. In the device of the present invention, the leakage current of the arrester is collected by the resistive current extraction circuit and the resistive current signal in the leakage current is extracted, and then the resistive current waveform generation module generates a waveform image of the resistive current signal, and finally the waveform image is input into the state model for classification and identification. Among them, considering that the resistive current in the arrester leakage current is related to the structural damage inside the arrester, therefore, extracting the resistive current signal from the leakage current and generating a waveform image for state classification and identification is conducive to improving the state detection accuracy of the arrester.

[0071] 4. In the present invention, the SDP image generated based on the partial discharge signal of the arrester and the waveform image of the resistive current signal extracted by the leakage current signal are classified by the neural network model, thereby realizing the detection of the working state of the arrester. Among them, the neural network model uses two sets of convolution layers and pooling layers that are alternately set to perform double extraction and aggregation of the features of the SDP image, thereby improving the classification accuracy of the model.

[0072] 5. In the present invention, the SDP image generation module sequentially performs analog-to-digital conversion, signal processing and image conversion steps on the collected partial discharge signal, thereby generating an SDP image. Among them, the signal processing module in the SDP image generation module uses Hilbert-Huang transform (HHT transform) to process the partial discharge signal of the arrester, which can analyze the local characteristics of the signal more deeply and effectively remove the noise in the partial discharge signal. In addition, the partial discharge signal is a transient signal and a non-stationary signal, and the high flexibility of the HHT transform in local time-frequency analysis is particularly suitable for processing transient signals and non-stationary signals, which is conducive to identifying and suppressing interference, thereby further improving the state detection accuracy of the arrester.

[0073] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not considered to be limiting of the present invention. In the entire drawings, the same reference symbols represent the same components:

[0075] Figure 1 A schematic diagram of the structure of a lightning arrester state monitoring device according to an embodiment of the present invention;

[0076] Figure 2 It is a structural schematic diagram of a resistive current extraction circuit of a lightning arrester according to an embodiment of the present invention;

[0077] Figure 3 It is a structural schematic diagram of a partial discharge acquisition circuit of a lightning arrester according to an embodiment of the present invention;

[0078] Figure 4 Schematic diagram of the structure of a neural network model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0079] The arrester state monitoring device based on partial discharge and leakage current of the present invention is an intelligent monitoring device, which is used to detect the working state of the arrester in the intelligent power grid and is an important facility in the intelligent power distribution system. In the arrester state monitoring device of the present invention, the partial discharge acquisition circuit and the resistive current extraction circuit are hardware facilities, which are arranged on one side of the arrester; the SDP image generation module, the resistive current waveform generation module and the state detection model are software modules, which can be installed in computer equipment such as industrial control computer systems, cloud platforms or Internet of Things platforms of intelligent power distribution systems in practical applications. Among them, the state detection model of the present invention is an artificial intelligence model, such as a machine learning model, a deep learning model, and preferably a deep learning model.

[0080] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0081] The embodiment of the present invention provides a lightning arrester state monitoring device based on partial discharge and leakage current, such as Figure 1 The device comprises:

[0082] Partial discharge collection circuit, collecting partial discharge signals of lightning arresters;

[0083] An SDP image generation module, generating an SDP image based on the partial discharge signal;

[0084] A resistive current extraction circuit collects the leakage current of the arrester and extracts a resistive current signal from the leakage current;

[0085] A resistive current waveform generating module, generating a waveform image of the resistive current signal;

[0086] The state detection model determines, based on the SDP image, that the arrester is in a normal state, an internal valve is damp, or a structurally damaged state; when the arrester is in a structurally damaged state, the arrester is determined to be in an internal valve aging state or an insulation layer damaged state based on the waveform image of the resistive current signal.

[0087] In the embodiment of the present invention, the working state of the lightning arrester includes four states: normal state, internal valve damp state, internal valve aging state and insulation layer damage state, among which the internal valve aging state and insulation layer damage state belong to structural damage state.

[0088] It should be noted that partial discharge refers to the local discharge phenomenon that occurs in the insulating medium of electrical equipment. Under normal conditions, the partial discharge signal of the arrester is usually weak. When the arrester fails, such as when the internal valve is damp, the internal valve is aged, and the insulation layer is damaged, the partial discharge signal will become more significant and may be accompanied by changes in the frequency, amplitude, and duration of the discharge signal. The leakage current of the arrester refers to the current passing through the valve inside the arrester. Under normal rated power frequency voltage, the arrester can be regarded as an insulator under normal conditions. With long-term use, the insulation performance of the arrester decreases. For example, when the arrester is in a state where the internal valve is damp, the internal valve is aged, and the insulation layer is damaged, a weak current will continue to pass through the valve inside the arrester under normal rated power frequency voltage.

[0089] Compared with the prior art, the device of the embodiment of the present invention realizes monitoring of four states of the arrester, namely, the normal state of the lightning arrester, the damp state of the internal valve, the aging state of the internal valve or the damage state of the insulation layer, based on the partial discharge signal and the leakage current.

[0090] Specifically, considering that the amplitude and frequency of the partial discharge signal waveform of the arrester in different working states have different characteristics, and the partial discharge signals of the arrester in two structural damage states, namely the internal valve aging state and the insulation layer damage state, are relatively similar, therefore, the working state of the arrester is firstly classified and identified based on the partial discharge signal, and the working state of the arrester is preliminarily divided into three states, namely the normal state, the internal valve moisture state and the structural damage state. Then, when it is determined that the arrester is in a structural damage state, the internal valve aging state and the insulation layer damage state are further classified and identified based on the leakage current.

[0091] In the device of the embodiment of the present invention, the local discharge signal of the arrester is first collected by the local discharge collection circuit, and then the SDP image generation module generates the SDP image based on the local discharge signal, and finally the SDP image is input into the state detection model for classification and identification, so as to determine whether the working state of the arrester is a normal state, an internal valve damp state or a structural damage state, and realize the preliminary classification and identification of the arrester state. Among them, the SDP image can bring together the waveform characteristics of the local discharge signal, which is more conducive to image recognition, thereby improving the state detection accuracy of the arrester and reducing misjudgment.

[0092] In the device of the present invention, the leakage current of the arrester is collected by a resistive current extraction circuit and the resistive current signal in the leakage current is extracted, and then the resistive current waveform generation module generates a waveform image of the resistive current signal, and finally the waveform image is input into the state model for classification and identification. Among them, considering that the resistive current in the arrester leakage current is related to the structural damage inside the arrester, therefore, extracting the resistive current signal from the leakage current and generating a waveform image for state classification and identification is conducive to improving the state detection accuracy of the arrester.

[0093] In a specific embodiment, Figure 3 As shown in , the partial discharge acquisition circuit includes: a first current detection device, an autotransformer, a first resistor, an inductor, a first capacitor, and a second resistor. The first current detection device is arranged on the grounding wire of the arrester to sense the current on the grounding wire; the output end of the first current detection device is connected to the input end of the autotransformer, one output end of the autotransformer is connected to one end of the first resistor, the other end of the first resistor is connected to one end of the inductor, the other end of the inductor is connected to one end of the first capacitor, the other end of the first capacitor is connected to the other output end of the autotransformer, the two ends of the second resistor are respectively connected to the two ends of the first capacitor, and the two ends of the second resistor output the partial discharge signal.

[0094] Specifically, the first current detection device is a current transformer or a Rogowski coil.

[0095] The first current detection device is preferably a high frequency current transformer (HFCT). This instrument has the advantages of ultra-wideband, high frequency, and fast response capability. The PD signal of the lightning arrester collected by it has the advantages of good linearity and high accuracy.

[0096] During implementation, the bandwidth range of the high-frequency current transformer is 1 MHz to 60 MHz.

[0097] When a lightning arrester has a partial discharge, a high-frequency current signal will be generated. The partial discharge acquisition circuit of this embodiment utilizes the high-frequency characteristics of the partial discharge signal, and through the amplification effect of the resonant circuit, the partial discharge signal can be effectively detected in a large number of background noises. Among them, the first current detection device is used to sense the current on the ground of the lightning arrester, the autotransformer is used to reduce the voltage and provide electrical isolation, the first resistor is used to limit the current and play a role in protecting the circuit, the inductor and the capacitor constitute a resonant circuit to selectively amplify the high-frequency signal of the partial discharge, and the signal output from both ends of the second resistor is the partial discharge signal, so that the subsequent detection equipment can collect and analyze it.

[0098] Preferably, the partial discharge acquisition circuit further comprises: a signal amplifier, the input end of the signal amplifier is connected to the two ends of the second resistor, and the signal amplifier amplifies the partial discharge signal and outputs it. Specifically, the signal amplifier is a high-impedance signal amplifier, which is used to amplify the partial discharge signal without causing a load effect.

[0099] In a specific embodiment, Figure 2 As shown in , the resistive current extraction circuit includes: a second current detection device, a capacitive current acquisition module and a differential amplifier. The input end of the second current detection device is connected to the grounding wire of the arrester, and is used to collect the leakage current of the arrester. The output end of the second current detection device is connected to the positive input end of the differential amplifier; the capacitive current acquisition module is used to collect the capacitive current in the leakage current, and its input end is connected between the high-voltage transmission line connected to the arrester and the grounding end, and the output end is connected to the negative input end of the differential amplifier; the output end of the differential amplifier outputs the resistive current signal in the leakage current.

[0100] In this embodiment, a capacitive current compensation method is used to extract a resistive current signal from the leakage current signal. The capacitive current compensation method is a technology used in a power system to extract a resistive current signal from a measured total current. The second current detection device is used to collect leakage current, and the capacitive current collection module estimates the capacitive current component of the leakage current using system parameters and a known voltage waveform, and then subtracts the estimated capacitive current component from the leakage current signal, thereby obtaining a resistive current signal of the leakage current.

[0101] Specifically, the capacitive current acquisition module includes: a voltage transformer, a differential phase shift circuit and a controllable gain amplifier. The input end of the voltage transformer is connected between the high-voltage transmission line to which the arrester is mounted and the ground terminal, the output end of the voltage transformer is connected to the input end of the differential phase shift circuit, the output end of the differential phase shift circuit is connected to the input end of the controllable gain amplifier, and the output end of the controllable gain amplifier is connected to the negative input end of the differential amplifier.

[0102] Preferably, the capacitive current acquisition module further includes: a photoelectric isolation module, and the photoelectric isolation module is connected between the voltage transformer and the differential shift circuit.

[0103] In the resistive current extraction circuit of the embodiment of the present invention, capacitive current compensation is performed in real time so as to be able to quickly respond to changes in the system. First, the voltage waveform at both ends of the arrester is measured by a voltage transformer, and the voltage is stepped down and output to the differential phase shift circuit. During this period, a photoelectric isolation switch is used to protect the entire circuit. The other end measures the leakage current of the arrester, that is, the total current, through a Rogowski coil. Then, the lower end shifts the phase of the voltage ES forward by 90 degrees through the differential phase shift circuit to obtain a new voltage. The phase-shifted voltage is in phase with the capacitive current component in the leakage current. The gain G0 of the controllable gain amplifier is adjusted so that the output of the controllable gain amplifier is in phase with the capacitive current and has the same size. In this way, the resistive current can be obtained through the differential amplifier circuit.

[0104] Specifically, the second current detection device is a current transformer or a Rogowski coil, which is arranged on the grounding wire of the lightning arrester to obtain the leakage current signal of the lightning arrester.

[0105] Preferably, filter capacitors are connected in parallel at both ends of the second current detection device.

[0106] It should be noted that the current detection device (current transformer or Rogowski coil) connected to the arrester grounding wire can measure both leakage current signals and partial discharge signals. The specific signal to be measured can be distinguished by setting the technical specifications of the current detection device or the configuration of the measurement system (such as additional filters, amplifiers, or signal processors, etc.).

[0107] In a specific embodiment, the resistive current waveform generation module includes: a first analog-to-digital conversion module, which converts the extracted resistive current signal from an analog signal to a digital signal; and an image drawing module, which draws a waveform diagram of the resistive current signal based on the resistive current signal converted into a digital signal.

[0108] During implementation, the first analog-to-digital conversion module may use a data acquisition card. The image drawing module may use a drawing software tool, such as the Matplotlib library in Python, or MATLAB.

[0109] In a specific embodiment, the SDP image generation module includes:

[0110] A second analog-to-digital conversion module converts the collected partial discharge signal from an analog signal to a digital signal;

[0111] A signal processing module, performing signal processing on the partial discharge signal converted into a digital signal, and using the processed signal as a state identification signal;

[0112] The image conversion module converts the state recognition signal into an SDP image.

[0113] In the present invention, the SDP image generation module sequentially performs analog-to-digital conversion, signal processing and image conversion steps on the collected partial discharge signal, thereby generating an SDP image.

[0114] Preferably, the signal processing module uses Hilbert-Huang transform to perform signal processing to obtain the state recognition signal, including the following steps:

[0115] Decomposing the partial discharge signal into a plurality of intrinsic mode components and corresponding residual terms by using an empirical mode decomposition method;

[0116] Performing Hilbert transform on each of the intrinsic modal components to obtain a corresponding analytical signal;

[0117] The analytical signals of all the intrinsic modal components and the residual terms are combined to obtain a state identification signal corresponding to the partial discharge signal.

[0118] Among them, the signal processing module in the SDP image generation module uses Hilbert-Huang transform (HHT transform) to process the local discharge signal of the arrester, which can analyze the local characteristics of the signal more deeply and effectively remove the noise in the local discharge signal. In addition, the local discharge signal is a transient signal and a non-stationary signal, and the high flexibility of the HHT transform in local time-frequency analysis is particularly suitable for processing transient signals and non-stationary signals, which is conducive to identifying and suppressing interference, thereby further improving the state detection accuracy of the arrester.

[0119] Specifically, the empirical mode decomposition method (EMD method) is an adaptive data decomposition method that decomposes complex time series into a series of intrinsic mode components (IMF components). EMD has no universal mathematical formula, and its basic steps are as follows:

[0120] Assume that the initial signal (i.e., partial discharge signal) is , first determine Then, using cubic spline interpolation, the upper and lower envelopes are fitted to the extreme points obtained, and the signal The upper and lower envelope curves correspond to , , whose average envelope curve is set to .

[0121] ;

[0122] use minus After that, the remaining part is set as ,Right now

[0123] ;

[0124] Judge at this time Whether the signal meets the IMF condition. The IMF condition is:

[0125] 1. In the entire data set, the sum of the number of local maxima and local minima must be equal to the number of zero crossings or differ by at most one;

[0126] 2. At any time point, the mean of the upper envelope defined by the local maximum and the lower envelope defined by the local minimum is zero.

[0127] At this time If the signal satisfies the IMF condition, it does not need to be decomposed; if it does not meet the IMF condition, repeat the above steps, that is,

[0128] ;

[0129] Eventually, you will get a signal that tends to be stable and can no longer be decomposed. , is considered to be the first intrinsic mode component (IMF) after the initial signal is decomposed, then As the first IMF component, the remaining signal can be expressed as ,Right now:

[0130] ;

[0131] According to the above decomposition process, the EMD decomposition is continuously performed, and the second IMF component can be obtained in turn, which is set as , and the cycle continues until the final decomposed signal is less than a given threshold, then the decomposition can be stopped. At this time, the initial signal is decomposed into a series of IMF components and the final residual term, that is:

[0132] .

[0133] Then, each of the intrinsic modal components is subjected to Hilbert transform to obtain the corresponding analytical signal, and the formula is:

[0134] ;

[0135] ;

[0136] In the formula, is the ith natural mode component, for Hilbert transform, t is the signal time point, is a variable in the integral, representing any time point from the upper limit to the lower limit of the integral. is the Hilport nucleus; yes The analytical signal, Is an imaginary unit.

[0137] Finally, the analytical signals of all the intrinsic modal components and the residual terms are combined to obtain the partial discharge signal Corresponding status identification signal , the formula is:

[0138] .

[0139] In addition, the signal processing method also includes: discrete wavelet transform, Fourier transform or short-time Fourier transform. Among them, by performing signal processing on the partial discharge signal of the lightning arrester through the above-mentioned discrete wavelet transform, Fourier transform or short-time Fourier transform, the noise component in the signal can be located and removed to obtain a low-noise waveform. Compared with Fourier transform or short-time Fourier transform, discrete wavelet transform can provide localized information in time and frequency at the same time, which is conducive to the analysis of transient signals and non-stationary signals. At the same time, discrete wavelet transform has the characteristics of multi-resolution.

[0140] It should be noted that the HHT transform not only absorbs the multi-resolution advantage of the wavelet transform, but also overcomes the difficulty of selecting the wavelet basis in the wavelet transform. That is, the HHT transform does not rely on the pre-defined basis function, but directly extracts features from the data, and is more suitable for analyzing and processing partial discharge signals. Therefore, in the embodiment of the present invention, the HHT transform is used to process the partial discharge signal and then generate the SDP image to improve the state detection accuracy of the arrester.

[0141] Specifically, the image conversion module converts the state recognition signal into an SDP image, including the following steps:

[0142] Mapping each discrete point in the state identification signal into a polar coordinate space to obtain the polar coordinates of each discrete point;

[0143] In the polar coordinate space, symmetrizing each discrete point to obtain a corresponding polar coordinate point;

[0144] Each discrete point and the polar coordinate point after symmetry processing are graphically displayed in the polar coordinate space to obtain the SDP image.

[0145] More specifically, mapping each discrete point in the state recognition signal to a polar coordinate space to obtain the polar coordinates of each discrete point includes:

[0146] The polar coordinate radius, the clockwise rotation angle of the initial line, and the counterclockwise rotation angle of the initial line corresponding to each discrete point in the polar coordinate space are calculated according to the state recognition signal. The formula is:

[0147] ;

[0148] ;

[0149] ;

[0150] In the formula, for The polar coordinate radius of the discrete point at the time, is the clockwise rotation angle of the initial line, is the counterclockwise rotation angle of the initial line, for The signal amplitude at discrete points in time, is the minimum signal amplitude of the state identification signal, is the maximum signal amplitude of the state identification signal, is the lag time, is the initial rotation angle, k is the magnification;

[0151] According to the polar coordinate radius, the clockwise rotation angle of the initial line, and the counterclockwise rotation angle of the initial line, the polar coordinates of the discrete point are determined as .

[0152] When implemented, the initial rotation angle is a preset fixed value, such as Lag time The selection will affect the offset of the image. The selection of the magnification factor k will affect the magnification degree of the image, and k=30 can be selected specifically.

[0153] Preferably, the SDP image generation module further includes: an image processing module, and the image processing module performs binarization, denoising or enhancement processing on the SDP image.

[0154] In a specific embodiment, the state detection model includes a first state detection model and a second state detection model.

[0155] The SDP image is input into the first state detection model for classification and identification, and the arrester is determined to be in a normal state, an internal valve damp state, or a structural damage state according to the classification result output by the first state detection model. When the arrester is determined to be in a structural damage state according to the classification result output by the first state detection model, the waveform image of the resistive current signal is input into the second state detection model for classification and identification, and the arrester is determined to be in an internal valve aging state or an insulation layer damage state according to the classification result output by the second state detection model.

[0156] Specifically, the first state detection model and the second state detection model are both neural network models.

[0157] Specifically, Figure 4 As shown in , the neural network model includes:

[0158] The first convolutional layer is used to extract features from the input image and output a feature map;

[0159] The first pooling layer is used to aggregate the feature maps output by the first convolutional layer and output the feature maps;

[0160] The second convolutional layer is used to extract features from the feature map output by the first pooling layer and output the feature map;

[0161] A second pooling layer, used to aggregate the feature maps output by the second convolutional layer and output the feature maps; and

[0162] The fully connected layer is used to fuse the feature maps output by the second pooling layer and output the working status of the arrester.

[0163] In the embodiment of the present invention, the SDP image generated based on the partial discharge signal of the arrester and the waveform image of the resistive current signal extracted by the leakage current signal are classified by the neural network model, thereby realizing the detection of the working state of the arrester. Among them, the neural network model uses two sets of convolution layers and pooling layers that are alternately set to perform double extraction and aggregation of the features of the SDP image, thereby improving the classification accuracy of the model.

[0164] Among them, the convolution layer extracts local features of the input image by using the convolution kernel until all pixels of the original input image are masked and the inner product is completed to generate a feature map. The pooling layer follows the convolution layer and is used to reduce the spatial resolution of the feature map while maintaining important feature information, which helps to achieve feature invariance. These two layers appear alternately in the neural network model, and together they build a network structure that can automatically and effectively learn useful features from raw data.

[0165] Specifically, the aggregation layer is a pooling layer.

[0166] The fully connected layer maps the learned features to the final output, corrects the error between the input and output through back propagation, and outputs the final classification result, that is, the working status of the lightning arrester.

[0167] Specifically, the fully connected layer includes a flat layer, a hidden layer and an output layer connected in sequence. The flat layer is used to convert the feature map output by the second pooling layer into one-dimensional data, the hidden layer is used to perform feature fusion on the one-dimensional data output by the flat layer and extract complex features, and the output layer is used to convert the output of the hidden layer into the working state of the arrester.

[0168] Specifically, the sizes of the convolution kernels of the first convolution layer and the second convolution layer are 3×3.

[0169] Specifically, the activation function of the neural network model is the ReLu function, and the formula of the ReLu function is:

[0170] ;

[0171] Where x is the input.

[0172] In order to enhance the nonlinear separability of the neural network model, an activation layer needs to be introduced. Since the ReLu function can significantly improve the training speed of the neural network model without affecting the accuracy of the model, the ReLu function is used as the activation function of the neural network model.

[0173] In addition, during implementation, the format of the input image of the neural network model (ie, the SDP image) is an RGB image with a size of 64×64 pixels.

[0174] In the embodiment of the present invention, before the SDP image and the resistive current waveform image are input into the arrester state detection model for classification, a step of training the arrester state detection model is also included.

[0175] It should be noted that before using the neural network model to classify the SDP image and the resistive current signal waveform to identify the working state of the obstacle avoider, the neural network model needs to be trained, and the trained neural network model is used to identify the working state of the arrester.

[0176] Specifically, training the arrester state detection model includes:

[0177] (1) Training a first state detection model based on an SDP image generated by a partial discharge signal of a lightning arrester includes the following steps:

[0178] Obtain the partial discharge signals of the arrester in normal state, internal valve damp state and structural damage state respectively;

[0179] Based on the partial discharge signal, SDP images of the arrester in a normal state, an internal valve damp state, and a structural damage state are generated respectively;

[0180] Annotate each SDP image with the working status of the corresponding arrester, and use the annotation as a label for the training model;

[0181] Divide the annotated SDP images into a first training set, a first validation set, and a first test set;

[0182] Using the first training set to train a first state detection model;

[0183] Using the first validation set to validate the trained first state detection model to adjust model parameters;

[0184] The verified first state detection model is tested using the first test set to evaluate the classification accuracy of the first state detection model.

[0185] (2) Training a second state detection model based on the waveform image of the resistive current obtained from the leakage current signal of the arrester, comprising the following steps:

[0186] Respectively obtain the leakage current signal of the arrester in the internal valve aging state and the insulation layer damage state;

[0187] extracting a resistive current signal from the leakage current signal, and generating a waveform image of the resistive current signal;

[0188] Annotating each waveform image of the resistive current signal with the corresponding working state of the arrester, and using the annotation as a label for the training model;

[0189] Dividing the labeled waveform image of the resistive current signal into a second training set, a second verification set, and a second test set;

[0190] Training a second state detection model using a second training set;

[0191] Using the second validation set to validate the trained second state detection model to adjust model parameters;

[0192] The verified second state detection model is tested using the second test set to evaluate the classification accuracy of the second state detection model.

[0193] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0194] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A lightning arrester status monitoring device based on partial discharge and leakage current, characterized in that: The device comprises: Partial discharge collection circuit, collecting partial discharge signals of lightning arresters; An SDP image generation module, generating an SDP image based on the partial discharge signal; A resistive current extraction circuit collects the leakage current of the arrester and extracts a resistive current signal from the leakage current; A resistive current waveform generating module, generating a waveform image of the resistive current signal; A state detection model, which determines, based on the SDP image, that the arrester is in a normal state, an internal valve is damp, or a structurally damaged state; when the arrester is in a structurally damaged state, determines, based on the waveform image of the resistive current signal, that the arrester is in an internal valve aging state or an insulation layer damaged state; The state detection model includes a first state detection model; Inputting the SDP image into a first state detection model for classification and identification, and determining whether the arrester is in a normal state, an internal valve damp state, or a structural damage state according to the classification result output by the first state detection model; The state detection model also includes a second state detection model; When it is determined that the arrester is in a structural damage state according to the classification result output by the first state detection model, the waveform image of the resistive current signal is input into the second state detection model for classification and identification, and it is determined that the arrester is in an internal valve aging state or an insulation layer damage state according to the classification result output by the second state detection model; The first state detection model and the second state detection model are both neural network models.

2. The device according to claim 1, characterized in that The partial discharge acquisition circuit comprises: a first current detection device, an autotransformer, a first resistor, an inductor, a first capacitor and a second resistor, Among them, the input end of the first current detection device is connected to the grounding wire of the lightning arrester, the output end of the first current detection device is connected to the input end of the autotransformer, one output end of the autotransformer is connected to one end of the first resistor, the other end of the first resistor is connected to one end of the inductor, the other end of the inductor is connected to one end of the first capacitor, the other end of the first capacitor is connected to the other output end of the autotransformer, the two ends of the second resistor are respectively connected to the two ends of the first capacitor, and the two ends of the second resistor output the partial discharge signal.

3. The device according to claim 2, characterized in that The partial discharge acquisition circuit further includes: a signal amplifier, the input end of the signal amplifier is connected to the two ends of the second resistor, and the signal amplifier amplifies the partial discharge signal and then outputs it.

4. The device according to claim 2, characterized in that The first current detection device is a current transformer or a Rogowski coil.

5. The device according to any one of claims 1 to 4, characterized in that: The resistive current extraction circuit includes: a second current detection device, a capacitive current acquisition module and a differential amplifier. The input end of the second current detection device is connected to the grounding wire of the arrester, and is used to collect the leakage current of the arrester, and the output end of the second current detection device is connected to the positive input end of the differential amplifier; The capacitive current acquisition module is used to collect the capacitive current in the leakage current, the input end of which is connected between the high-voltage transmission line to which the arrester is connected and the grounding terminal, and the output end is connected to the negative input end of the differential amplifier; The output end of the differential amplifier outputs a resistive current signal in the leakage current.

6. The device according to claim 5, characterized in that The capacitive current acquisition module includes: a voltage transformer, a differential phase shift circuit and a controllable gain amplifier. The input end of the voltage transformer is connected between the high-voltage transmission line to which the lightning arrester is connected and the grounding end, the output end of the voltage transformer is connected to the input end of the differential phase shift circuit, the output end of the differential phase shift circuit is connected to the input end of the controllable gain amplifier, and the output end of the controllable gain amplifier is connected to the negative input end of the differential amplifier.

7. The device according to claim 6, characterized in that The capacitive current acquisition module further includes: a photoelectric isolation module, and the photoelectric isolation module is connected between the voltage transformer and the differential phase shift circuit.

8. The device according to claim 5, characterized in that The second current detection device is a current transformer or a Rogowski coil.

9. The device according to any one of claims 1 to 4, characterized in that The resistive current waveform generation module includes: A first analog-to-digital conversion module converts the extracted resistive current signal from an analog signal to a digital signal; An image drawing module draws a waveform diagram of the resistive current signal according to the resistive current signal converted into a digital signal.

10. The device according to any one of claims 1 to 4, characterized in that The SDP image generation module includes: A second analog-to-digital conversion module converts the collected partial discharge signal from an analog signal to a digital signal; A signal processing module, performing signal processing on the partial discharge signal converted into a digital signal, and using the processed signal as a state identification signal; The image conversion module converts the state recognition signal into an SDP image.

11. The device according to claim 10, characterized in that The signal processing module uses Hilbert-Huang transform to perform signal processing to obtain a state recognition signal, including the following steps: Decomposing the partial discharge signal into a plurality of intrinsic mode components and corresponding residual terms by using an empirical mode decomposition method; Performing Hilbert transform on each of the intrinsic modal components to obtain a corresponding analytical signal; The analytical signals of all the intrinsic modal components and the residual terms are combined to obtain a state identification signal corresponding to the partial discharge signal.

12. The device according to claim 11, characterized in that The formula for performing Hilbert transform on each of the intrinsic modal components to obtain the corresponding analytical signal is: ; ; In the formula, is the ith natural mode component, for Hilbert transform, t is the signal time point, is a variable in the integral, representing any time point from the upper limit to the lower limit of the integral. is the Hilport nucleus; yes The analytical signal, Is an imaginary unit.

13. The device according to claim 10, characterized in that The image conversion module converts the state recognition signal into an SDP image, including the following steps: Mapping each discrete point in the state identification signal into a polar coordinate space to obtain the polar coordinates of each discrete point; In the polar coordinate space, symmetrizing each discrete point to obtain a corresponding polar coordinate point; Each discrete point and the polar coordinate point after symmetry processing are graphically displayed in the polar coordinate space to obtain the SDP image.

14. The device according to claim 13, characterized in that Mapping each discrete point in the state recognition signal to a polar coordinate space to obtain the polar coordinates of each discrete point includes: The polar coordinate radius, the clockwise rotation angle of the initial line, and the counterclockwise rotation angle of the initial line corresponding to each discrete point in the polar coordinate space are calculated according to the state recognition signal. The formula is: ； ； ； In the formula, for The polar coordinate radius of the discrete point at the time, is the clockwise rotation angle of the initial line, is the counterclockwise rotation angle of the initial line, for The signal amplitude at discrete points in time, is the minimum signal amplitude of the state identification signal, is the maximum signal amplitude of the state identification signal, is the lag time, is the initial rotation angle, k is the magnification; According to the polar coordinate radius, the clockwise rotation angle of the initial line, and the counterclockwise rotation angle of the initial line, the polar coordinates of the discrete point are determined as .

15. The device according to claim 1, characterized in that The neural network model includes: The first convolutional layer is used to extract features from the input image and output a feature map; The first pooling layer is used to aggregate the feature maps output by the first convolutional layer and output the feature maps; The second convolutional layer is used to extract features from the feature map output by the first pooling layer and output the feature map; A second pooling layer, used to aggregate the feature maps output by the second convolutional layer and output the feature maps; and The fully connected layer is used to fuse the feature maps output by the second pooling layer and output the working status of the arrester.

16. The device according to claim 15, characterized in that The fully connected layer includes a flat layer, a hidden layer and an output layer connected in sequence. Among them, the flat layer is used to convert the feature map output by the second pooling layer into one-dimensional data, the hidden layer is used to perform feature fusion on the one-dimensional data output by the flat layer and extract complex features, and the output layer is used to convert the output of the hidden layer into the working state of the arrester.

17. The device according to claim 15, characterized in that The sizes of the convolution kernels of the first convolution layer and the second convolution layer are 3×3.

18. The device according to claim 1, characterized in that The activation function of the neural network model is the ReLu function, and the formula of the ReLu function is: ； Where x is the input.

Images