Resistive current free PT signal measurement method for capacitive equipment of lightning arrester

Abstract

The application provides a capacitive equipment resistance current PT-free signal measurement method suitable for a lightning arrester, and relates to the technical field of capacitive equipment resistance current measurement. The method comprises the following steps: deploying a non-contact electric field coupling sensing array; acquiring a multi-channel main induction signal and a high-voltage side induction electric field signal of the equipment; performing cascade filtering processing to obtain a filtered reference signal; dynamically correcting the phase offset of the filtered reference signal to obtain a calibrated reference signal; collecting a leakage current signal and separating a resistance current component; constructing a dielectric loss factor multivariate dynamic compensation model based on a deep belief network, embedding a cumulative effect factor of the running time of the capacitive equipment, and updating the parameters in real time through an online learning algorithm to obtain a resistance current measurement result. The application realizes high-precision resistance current measurement through advanced technology, establishes a model in combination with multiple environmental parameters and the running time of the equipment, improves adaptability, and guarantees long-term accurate monitoring through online parameter updating.

Description

Technical Field

[0001] The present invention relates to the technical field of capacitive device resistive current measurement, and in particular to a capacitive device resistive current measurement method applicable to a lightning arrester without obtaining a PT signal. Background Art

[0002] In power systems, lightning arresters (SAs), critical capacitive devices that ensure the safe operation of electrical equipment, make accurate monitoring of their operating status crucial. Measuring resistive current is a key indicator for assessing the health of SLAs. Traditional methods often rely on voltage transformers (PTs) to obtain reference voltage signals. In practical applications, PT installation and maintenance present numerous challenges. Firstly, PT installation requires complex electrical connections, which is challenging in space-constrained and complex substations, increasing the complexity and cost of project implementation. Secondly, PTs inherently exhibit measurement errors, such as angular and ratio errors. These errors are transferred to the resistive current measurement results, reducing measurement accuracy. Furthermore, PTs can experience operational failures, such as insulation aging or secondary short circuits. Once a PT fails, the entire measurement system based on its signals ceases to function properly, severely impacting the monitoring of the SLA's operating status.

[0003] Furthermore, traditional measurement methods lack effective anti-interference measures when faced with complex electromagnetic interference environments. Substations contain a large number of electrical devices, and the electromagnetic noise they generate can severely interfere with measurement signals, leading to large fluctuations in measurement results and difficulty accurately reflecting the arrester's true resistive current. As power systems evolve toward intelligent and highly reliable systems, higher requirements are placed on the accuracy, real-time nature, and convenience of arrester resistive current measurement. A new measurement method is urgently needed that is independent of PT signals, exhibits strong anti-interference capabilities, and offers high measurement accuracy. Summary of the Invention

[0004] The present invention provides a PT signal-free measurement method for resistive current of capacitive equipment applicable to a lightning arrester, so as to solve the defects in the prior art.

[0005] The present invention provides a method for measuring the resistive current of a capacitive device applicable to a lightning arrester without obtaining a PT signal, comprising:

[0006] A non-contact electric field coupling sensing array is deployed in a preset sensing area of ​​the high-voltage conductor of the capacitive device. The sensing array includes a main sensing electrode group and a phase monitoring electrode group.

[0007] The main sensing electrode group captures the spatial distribution characteristics of the power-frequency electric field around the high-voltage conductor through the electric field coupling effect, generating a multi-channel main sensing signal. The phase monitoring electrode group senses the surface electric field of the high-voltage conductor through near-field coupling, generating an induced electric field signal on the high-voltage side of the equipment.

[0008] The power frequency characteristics are extracted from the multi-channel main induction signal to generate an initial reference signal with the same frequency and phase as the operating voltage of the capacitive device. The initial reference signal is then subjected to cascade filtering to obtain a filtered reference signal.

[0009] Based on the filtered reference signal and the induced electric field signal on the high-voltage side of the equipment, a phase closed-loop calibration mechanism is introduced. By comparing the phase difference between the two in real time, the phase offset of the filtered reference signal is dynamically corrected to obtain the calibration reference signal.

[0010] The leakage current signal of the capacitive device grounding wire is collected and subjected to analog-to-digital conversion and adaptive spectrum decomposition. Combined with the calibration reference signal, the resistive current component is separated.

[0011] Multi-dimensional environmental parameters of the preset sensing area are synchronously collected, and a multivariable dynamic compensation model of the dielectric loss factor based on the deep belief network is constructed. The cumulative effect factor of the operating time of the capacitive device is embedded, and the parameters are updated in real time through the online learning algorithm. The multi-dimensional environmental parameters are used as input, and the compensation value for the resistive current component is output to obtain the resistive current measurement result.

[0012] According to the method for measuring resistive current of capacitive equipment of a lightning arrester without obtaining a PT signal provided by the present invention, the process of obtaining a multi-channel main induction signal includes:

[0013] Each sensing electrode in the main sensing electrode group generates induced charges in the power frequency electric field around the high-voltage conductor through the electric field coupling effect, forming a micro-current signal that changes with the electric field.

[0014] The microcurrent signal is collected by a multi-channel synchronous acquisition method, and the microcurrent signal is subjected to impedance conversion and multi-stage amplification processing through a signal conditioning circuit to obtain an amplified signal.

[0015] The amplified signal is preliminarily filtered to form a multi-channel main sensing signal.

[0016] According to the method for measuring the resistive current of a capacitive device of a lightning arrester without obtaining a PT signal provided by the present invention, the process of obtaining the induced electric field signal on the high-voltage side of the device includes:

[0017] The phase monitoring electrode group senses the electric field on the surface of the high-voltage conductor through near-field coupling to obtain micro-electric signals.

[0018] A low-noise amplifier circuit is used to amplify the micro-electric signal, and the amplified micro-electric signal is subjected to band-pass filtering.

[0019] The phase of the filtered micro-electric signal is corrected through the phase holding circuit to form the induced electric field signal on the high-voltage side of the equipment.

[0020] According to the method for measuring the resistive current of a capacitive device of a lightning arrester without obtaining a PT signal provided by the present invention, the process of generating an initial reference signal having the same frequency and phase as the operating voltage of the capacitive device includes:

[0021] The blind source separation algorithm based on the kurtosis maximization criterion is used to process the multi-channel main induction signals. The power frequency signal and the interference signal are separated by iteratively optimizing the separation matrix to obtain the target signal containing the power frequency characteristics.

[0022] Perform spectrum analysis on the target signal to extract the frequency and phase information of the power frequency fundamental component.

[0023] Based on the extracted frequency and phase information, the cubic spline interpolation method is used to reconstruct and generate a continuous initial reference signal waveform.

[0024] Perform phase smoothing processing on the initial reference signal waveform to obtain an initial reference signal.

[0025] According to the method for measuring resistive current of capacitive equipment of a lightning arrester without obtaining a PT signal provided by the present invention, the process of obtaining a filtered reference signal includes:

[0026] The initial reference signal is input into the adaptive notch filter to eliminate the periodic interference in a specific frequency band and obtain the initial filtered signal.

[0027] Perform wavelet threshold denoising on the initial filtered signal to obtain a denoised filtered signal.

[0028] The denoising filter signal is smoothed to obtain a filter reference signal, and the filter reference signal is amplitude normalized.

[0029] According to the method for measuring resistive current of capacitive equipment of a lightning arrester without obtaining a PT signal provided by the present invention, the process of obtaining a calibration reference signal includes:

[0030] The filtered reference signal and the induced electric field signal on the high-voltage side of the equipment are input into the phase comparator, the phase difference between the two is calculated, and a phase error signal is generated.

[0031] The phase error signal is input into the Kalman filter for filtering to eliminate noise interference in the error signal.

[0032] The filtered phase error signal is input into the proportional-integral-derivative controller to generate the phase compensation amount.

[0033] The phase compensation amount is applied to the filtered reference signal through an orthogonal modulator, and the phase of the signal is dynamically corrected until the phase error is less than a preset threshold to obtain a calibration reference signal.

[0034] According to the method for measuring resistive current of capacitive equipment of a lightning arrester without obtaining a PT signal provided by the present invention, the process of separating and obtaining the resistive current component includes:

[0035] A magneto-optical current sensor is used to collect leakage current signals from the grounding wire of capacitive equipment.

[0036] The leakage current signal is converted into a digital signal through an analog-to-digital converter.

[0037] Adaptive spectrum decomposition based on Hilbert-Huang transform is used for digital signals. The process includes:

[0038] The mirror extension boundary processing technology is used to pre-process the digital signal to suppress the endpoint effect.

[0039] The digital signal is decomposed into multiple intrinsic mode functions through empirical mode decomposition.

[0040] Perform Hilbert transform on each intrinsic mode function to obtain the corresponding Hilbert spectrum.

[0041] According to the phase information of the calibration reference signal, the in-phase component is extracted from the decomposed Hilbert spectrum, and the in-phase components are superimposed to obtain the resistive current component.

[0042] According to the method for measuring resistive current of capacitive equipment applicable to lightning arresters without obtaining PT signals provided by the present invention, the process of constructing a multivariable dynamic compensation model for dielectric loss factor based on a deep belief network includes:

[0043] The structure of the deep belief network is designed, including an input layer, multiple restricted Boltzmann machine hidden layers and an output layer. The number of nodes in the input layer is consistent with the dimension of the multi-dimensional environmental parameters. The output layer is a single node, corresponding to the compensation value of the resistive current component. The multi-dimensional environmental parameters include temperature, humidity, air pressure and air medium parameters.

[0044] Multi-dimensional environmental parameters and corresponding measured values ​​of dielectric loss factor under different environmental conditions are collected to construct training data sets and validation data sets.

[0045] The deep belief network is pre-trained using a layer-by-layer greedy training algorithm, and the parameters of the restricted Boltzmann machine in each layer are optimized using the contrastive divergence algorithm.

[0046] The pre-trained deep belief network is used as the initial model, and the back propagation algorithm is used to adjust the model parameters to minimize the error between the predicted compensation value and the actual compensation value.

[0047] The model is verified through the validation data set, and the network structure and parameters are adjusted until the prediction accuracy of the model meets the preset requirements.

[0048] According to the method for measuring the resistive current of a capacitive device of a lightning arrester without obtaining a PT signal provided by the present invention, the process of embedding the cumulative effect factor of the device operation time includes:

[0049] Define the cumulative effect factor of equipment operation time.

[0050] Add an input node to the deep belief network to input the cumulative effect factor of the device running time.

[0051] An attention mechanism is introduced to dynamically adjust the weight of the cumulative effect factor of device operating time in the model.

[0052] The cumulative effect factor of equipment operating time is fused with multi-dimensional environmental parameters as the input of the deep belief network.

[0053] According to the method for measuring resistive current of capacitive equipment applicable to a lightning arrester without obtaining a PT signal, a process of updating parameters in real time through an online learning algorithm includes:

[0054] The online learning algorithm adopts the online sequence extreme learning machine as the model, and the output of the deep belief network is used as the input of the online sequence extreme learning machine.

[0055] Set the forgetting factor, which is used to control the impact of historical data on model parameter updates.

[0056] New measurement data is collected in real time, including multi-dimensional environmental parameters, cumulative effect factors of equipment operation time, and the error between the actual measurement value of the resistive current component and the model prediction value.

[0057] Based on the newly collected data, the model parameters are updated through an online sequential extreme learning machine.

[0058] Set the parameter update threshold. When the error between the new data and the model prediction result exceeds the parameter update threshold, the model parameter update is triggered.

[0059] The present invention provides a method for measuring the resistive current of a capacitive device suitable for a lightning arrester without obtaining a PT signal. By deploying a non-contact electric field coupling sensor array in a preset sensing area of ​​the high-voltage conductor of the capacitive device, the signal is directly obtained from the electric field around the high-voltage conductor, completely avoiding dependence on the PT signal. The structure of the measurement system is simplified, the installation and maintenance costs and failure risks caused by the PT equipment are reduced, and the reliability and stability of the measurement system are improved. Through advanced signal processing technology, the accuracy of the resistive current measurement is effectively improved, and the actual operating status of the lightning arrester can be more accurately reflected, providing a reliable basis for the status assessment and fault warning of the equipment. A multivariate dynamic compensation model for the dielectric loss factor based on a deep belief network is constructed. It can not only take into account the impact of environmental factors on the measurement results in real time, but also comprehensively analyze the operating status of the equipment throughout its life cycle by embedding the cumulative effect factor of the equipment's operating time, so that the measurement results are more in line with the actual situation and the adaptability of the measurement method under different environmental conditions is improved. With the help of an online sequence extreme learning machine as an online learning algorithm, new measurement data is collected in real time to update the model parameters, ensuring that the model always keeps pace with the actual situation. This ensures that the model can maintain high accuracy as the equipment operating status and environment change, providing a strong guarantee for long-term, stable and accurate monitoring of the arrester resistive current. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0061] Figure 1 1. It is a flow chart of a method for measuring resistive current of a capacitive device of a lightning arrester without obtaining a PT signal, provided in an embodiment of the present invention;

[0062] Figure 2 This is a schematic diagram of a process for generating an initial reference signal having the same frequency and phase as the operating voltage of a capacitive device in an embodiment of the present invention;

[0063] Figure 3 is a schematic diagram of a process for obtaining a calibration reference signal in an embodiment of the present invention;

[0064] Figure 4 It is a schematic diagram of a process for separating and obtaining a resistive current component in an embodiment of the present invention. DETAILED DESCRIPTION

[0065] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0066] The following combination Figures 1-4 The present invention describes a method for measuring resistive current of capacitive equipment applicable to lightning arresters without obtaining PT signals.

[0067] Figure 1 The present invention provides a structural diagram of a method for measuring resistive current of a capacitive device of a lightning arrester without obtaining a PT signal, provided in an embodiment of the present invention.

[0068] like Figure 1 As shown, the embodiment of the present invention provides a method for measuring the resistive current of a capacitive device of a lightning arrester without obtaining a PT signal, including:

[0069] A non-contact electric field coupling sensing array is deployed in a preset sensing area of ​​the high-voltage conductor of the capacitive device. The sensing array includes a main sensing electrode group and a phase monitoring electrode group.

[0070] When deploying a contactless electric field coupling sensor array within the pre-set sensing area of ​​a capacitive device's high-voltage conductor, a circular sensing area 0.5-1.5 meters from the high-voltage conductor is first determined through 3D modeling to ensure there is no metal obstruction. The main sensing electrode group utilizes 6-8 honeycomb-shaped porous electrodes evenly distributed in a 360-degree circular pattern. The electrode edges are serrated and coated with a graphene-silicon dioxide composite nanocoating, which forms a hydrophobic insulating layer through a molecular self-assembly process.

[0071] The phase monitoring electrode group consists of four micro needle electrodes embedded in the pre-set monitoring points of the high-voltage conductor insulation support structure, maintaining a distance of 5-10mm from the conductor surface. The two sets of electrodes are fixed to the stepped insulation platform through a three-dimensional fine-tuning bracket. The platform is made of epoxy glass cloth laminate, and the surface is anodized to achieve multi-dimensional electric field shielding. Finally, the insulation performance test is carried out to ensure that the insulation resistance between the electrodes is greater than 10 12 Ω.

[0072] The main sensing electrode group captures the spatial distribution characteristics of the power-frequency electric field around the high-voltage conductor through the electric field coupling effect, generating a multi-channel main sensing signal. The phase monitoring electrode group senses the surface electric field of the high-voltage conductor through near-field coupling, generating an induced electric field signal on the high-voltage side of the equipment.

[0073] The process of obtaining the multi-channel main sensing signal includes:

[0074] Each sensing electrode in the main sensing electrode group generates induced charges in the power frequency electric field around the high-voltage conductor through the electric field coupling effect, forming a micro-current signal that changes with the electric field.

[0075] The microcurrent signal is collected by multi-channel synchronous acquisition method, and the sampling frequency is set to 20 times the industrial frequency of 50Hz, that is, 1kHz. The microcurrent signal is subjected to impedance conversion and multi-stage amplification processing through the signal conditioning circuit to obtain the amplified signal.

[0076] The signal conditioning circuit first converts the micro-current signal into a voltage signal (conversion coefficient 1V / 100pA) through an I / V conversion circuit composed of a high-precision operational amplifier, and then passes through a three-stage amplification circuit (total gain 10 4 -10 5 The signal amplitude is adjusted to the standard processing range of 0-5V, and an RC low-pass filter is set between each stage of the amplifier circuit to suppress high-frequency noise.

[0077] The initial filtering uses a Butterworth second-order high-pass filter (cut-off frequency 0.5Hz) to remove the DC component, and then uses a second-order low-pass filter (cut-off frequency 150Hz) to filter out high-frequency interference, ultimately forming a multi-channel main induction signal containing the power frequency fundamental wave and a small amount of harmonic components. The signal-to-noise ratio of each channel signal is not less than 60dB.

[0078] The process of obtaining the induced electric field signal on the high-voltage side of the equipment includes:

[0079] The phase monitoring electrode group senses the electric field on the surface of the high-voltage conductor through near-field coupling, generating a micro-electrical signal. The micro-needle electrodes of the phase monitoring electrode group are placed 5-10mm close to the surface of the high-voltage conductor. Through near-field coupling, they directly sense the electric field on the conductor surface, generating a weak electrical signal.

[0080] A low-noise amplifier circuit is used to amplify the micro-electric signal, and the amplified micro-electric signal is subjected to band-pass filtering.

[0081] A low-noise amplifier circuit (input offset voltage ≤ 10μV, common mode rejection ratio ≥ 120dB) composed of an instrument amplifier is used to amplify the micro-electric signal, and the amplification factor is set to 10 3 The bandpass filter uses a Chebyshev I-type second-order filter with a passband range of 45-55Hz and an attenuation characteristic of ≥80dB / dec, strictly limiting the power frequency signal band.

[0082] The phase of the filtered micro-electric signal is corrected through the phase holding circuit to form the induced electric field signal on the high-voltage side of the equipment.

[0083] The phase keeping circuit adopts a all-pass filter structure, and through accurate matching of resistance and capacitance parameters (error ≤ 1%), phase error compensation of ±0.1° is realized in a frequency band of 45-55 Hz, so that the phase of the processed signal is consistent with the phase of the surface electric field of the high-voltage conductor, and finally the high-voltage side induced electric field signal of the equipment is formed, and the phase stability is ≤0.5° / h.

[0084] The power frequency feature is extracted from the multi-channel main induction signal, an initial reference signal with the same frequency and phase as the operating voltage of the capacitive equipment is generated, and the initial reference signal is subjected to cascade filtering processing to obtain a filtered reference signal.

[0085] Figure 2 It is a flowchart of generating an initial reference signal with the same frequency and phase as the operating voltage of the capacitive equipment in the embodiment of the application.

[0086] As shown in Figure 2 , the process of generating an initial reference signal with the same frequency and phase as the operating voltage of the capacitive equipment includes:

[0087] First, a blind source separation algorithm based on the maximum kurtosis criterion is applied to the multi-channel main induction signal, the separation matrix is initialized as a unit matrix, the kurtosis value of the signal is calculated through iteration, and the matrix parameters are optimized through gradient descent method in each round of iteration until the kurtosis value converges, so that the target signal containing the power frequency feature is separated.

[0088] The target signal is subjected to fast Fourier transform, the spectrum of the 0-200Hz frequency band is intercepted, the fundamental wave peak value around 50Hz is identified, and the frequency and phase thereof are extracted. Based on the extracted frequency and phase, a cubic spline interpolation method is used to reconstruct the waveform: a time sequence t is generated at a sampling rate of 1kHz, the discrete sampling points are fitted through an interpolation function, and a continuous sinusoidal waveform is reconstructed. The interpolation function formula is:

[0089]

[0090] In the formula, is a cubic B-spline basis function, is an interpolation coefficient.

[0091] Finally, phase smoothing processing is performed, the phase mean value is calculated by using a sliding window, the points with phase jump exceeding 5° in the window are corrected, and the initial reference signal with the same frequency and phase as the operating voltage of the capacitive equipment is generated, and the phase error is controlled within ±1°.

[0092] The process of obtaining the filtered reference signal includes:

[0093] The initial reference signal is input into the adaptive notch filter. The adaptive notch filter adopts a second-order infinite impulse response structure. It uses a recursive least squares algorithm to track the interference frequency drift in real time, dynamically adjust the notch bandwidth, eliminate the periodic interference in a specific frequency band, and obtain the initial filtered signal.

[0094] The initial filtered signal is subjected to wavelet threshold denoising. The improved Birgé-Massart threshold strategy is adopted to achieve differentiated suppression of noise in different frequency bands through hierarchical threshold calculation to obtain the denoised filtered signal.

[0095] The denoising filtered signal is smoothed and the moving average filtering algorithm is used to eliminate the burrs in the signal to obtain the filtered reference signal.

[0096] The filtered reference signal is amplitude normalized to ensure that the signal amplitude is within a preset range.

[0097] Based on the filtered reference signal and the induced electric field signal on the high-voltage side of the equipment, a phase closed-loop calibration mechanism is introduced. By comparing the phase difference between the two in real time, the phase offset of the filtered reference signal is dynamically corrected to obtain the calibration reference signal.

[0098] Figure 3 4 is a flow chart of obtaining a calibration reference signal in an embodiment of the present invention.

[0099] like Figure 3 As shown, the process of obtaining the calibration reference signal includes:

[0100] The filtered reference signal and the induced electric field signal on the high-voltage side of the equipment are input into the phase comparator, the phase difference between the two is calculated, and a phase error signal is generated.

[0101] The phase error signal is input into the Kalman filter for filtering to eliminate noise interference in the error signal.

[0102] The filtered phase error signal is input into the proportional-integral-derivative controller to generate the phase compensation amount.

[0103] The phase compensation amount is applied to the filtered reference signal through an orthogonal modulator, and the phase of the signal is dynamically corrected until the phase error is less than a preset threshold to obtain a calibration reference signal.

[0104] The leakage current signal of the capacitive device grounding wire is collected and subjected to analog-to-digital conversion and adaptive spectrum decomposition. Combined with the calibration reference signal, the resistive current component is separated.

[0105] Figure 4 It is a schematic diagram of a process for separating and obtaining a resistive current component in an embodiment of the present invention.

[0106] like Figure 4As shown, the process of separating the resistive current component includes:

[0107] The leakage current signal of the grounding line of the capacitive equipment is collected by using a magneto-optical current sensor, which realizes non-contact measurement of the current based on the Faraday magneto-optical effect.

[0108] The leakage current signal is converted into a digital signal by an analog-to-digital converter.

[0109] The adaptive spectral decomposition based on Hilbert-Huang transform is used for the digital signal, and the process includes:

[0110] The mirror extension boundary processing technology is used for pre-processing the digital signal to suppress the end effect.

[0111] The digital signal is decomposed into a plurality of intrinsic modal functions by empirical mode decomposition.

[0112] The Hilbert transform is performed on each intrinsic modal function to obtain the corresponding Hilbert spectrum.

[0113] According to the phase information of the calibration reference signal, the in-phase component is extracted from the Hilbert spectrum obtained by decomposition, and the in-phase components are superimposed to obtain the resistive current component.

[0114] Synchronously collect multi-dimensional environmental parameters of a preset sensing area, construct a dielectric loss factor multivariate dynamic compensation model based on a deep belief network, embed a cumulative effect factor of the running time of the capacitive equipment, and update the parameters in real time through an online learning algorithm. The multi-dimensional environmental parameters are used as inputs, and the compensation value for the resistive current component is output, and the resistive current measurement result is obtained.

[0115] The process of constructing the dielectric loss factor multivariate dynamic compensation model based on the deep belief network includes:

[0116] The structure of the deep belief network is designed, including an input layer, a plurality of restricted Boltzmann machine hidden layers and an output layer, wherein the number of input layer nodes is consistent with the dimension of the multi-dimensional environmental parameters, the output layer is a single node corresponding to the compensation value of the resistive current component, and the multi-dimensional environmental parameters include temperature, humidity, air pressure and air medium parameters.

[0117] Collect multi-dimensional environmental parameters and corresponding dielectric loss factor measured values under different environmental conditions to construct a training data set and a verification data set.

[0118] The multi-dimensional environmental parameters in the training data set are normalized by using the min-max standardization method to map the data to the [0, 1] interval.

[0119] The deep belief network is pre-trained by using a layer-by-layer greedy training algorithm, and the parameters of each layer of the restricted Boltzmann machine are optimized by the contrast divergence algorithm.

[0120] The pre-trained deep belief network is used as the initial model, and the back propagation algorithm is used to adjust the model parameters to minimize the error between the predicted compensation value and the actual compensation value.

[0121] The loss function of the model adopts the mean square error function, which is expressed as follows:

[0122]

[0123] Where n is the number of samples, is the true value, is the predicted value.

[0124] The model is verified through the validation data set, and the network structure and parameters are adjusted until the prediction accuracy of the model meets the preset requirements.

[0125] The process of embedding the cumulative effect factor of device operation time includes:

[0126] Define the cumulative effect factor of equipment operation time, the formula is expressed as:

[0127]

[0128] Where t is the cumulative running time of the equipment, for The effective value of the leakage current of the equipment at the moment, and is the weight coefficient, which is determined according to the device type and operating environment.

[0129] Add a dedicated input node in the deep belief network to input the cumulative effect factor of device operation time .

[0130] By introducing the attention mechanism, the weight of the cumulative effect factor of the device running time in the model is dynamically adjusted, and its weight coefficient The calculation formula is expressed as:

[0131]

[0132] in, is the kth environmental parameter, m is the number of environmental parameters, is the adjustment parameter.

[0133] The cumulative effect factor of equipment operating time is fused with multi-dimensional environmental parameters as the input of the deep belief network to achieve dynamic compensation for equipment aging effects.

[0134] The process of updating parameters in real time through online learning algorithms includes:

[0135] An online sequence extreme learning machine is used as an online learning algorithm of the model, and an output of a deep belief network is used as an input of the online sequence extreme learning machine.

[0136] A forgetting factor is set to control the influence of historical data on the update of the model parameters, and the weight of recent data is higher than that of long-term data.

[0137] New measurement data are collected in real time, including multi-dimensional environmental parameters, cumulative effect factors of equipment running time, and errors between actual measurement values and model prediction values of the resistive current component.

[0138] According to the newly collected data, the model parameters are updated by the online sequence extreme learning machine, and the update formula is represented as:

[0139]

[0140]

[0141] wherein, is an estimated value of the model parameters at time t, is an estimated value of the model parameters at time t-1, is a covariance matrix at time t, is a covariance matrix at time t-1, is an input feature vector at time t, is a transpose matrix of is an expected output at time t, is a forgetting factor.

[0142] A parameter update threshold is set, and when the error between the new data and the model prediction result exceeds the parameter update threshold, the model parameter update is triggered.

[0143] The updated model is evaluated at a preset period, and if the model accuracy decreases by more than a preset range, the model training is performed again to ensure that the model always maintains a high compensation accuracy.

[0144] ​In summary, the embodiment provides a capacitive device resistance current PT signal measurement method suitable for lightning arresters. By deploying a non-contact electric field coupling sensor array in the pre-set induction area of the high-voltage conductor of the capacitive device, the signal is directly obtained from the electric field around the high-voltage conductor, completely avoiding the dependence on the PT signal. The structure of the measurement system is simplified, the installation and maintenance costs and the risk of failure caused by the PT device are reduced, and the reliability and stability of the measurement system are improved. Through advanced signal processing technology, the accuracy of the resistance current measurement is effectively improved, and the actual operating state of the lightning arrester can be more accurately reflected, providing a reliable basis for the state evaluation and fault warning of the device. By constructing a dielectric loss factor multivariate dynamic compensation model based on a deep belief network, not only can the influence of environmental factors on the measurement results be considered in real time, but also by embedding the device running time cumulative effect factor, the operating conditions of the device throughout its life cycle are comprehensively analyzed, making the measurement results more in line with the actual situation and improving the adaptability of the measurement method under different environmental conditions.

[0145] In a high-temperature environment, the dielectric properties of the lightning arrester will change, and if not compensated, the measurement results will be biased. The dynamic compensation model based on the deep belief network can accurately capture the complex relationship between these environmental parameters and resistance current measurement errors after a large number of sample training, and output accurate compensation values.

[0146] In terms of considering the characteristics of the device itself, as the running time of the device increases, the insulation materials inside the lightning arrester will gradually age, and their dielectric loss factor and other characteristics will change slowly, affecting the true value of the resistance current. By introducing the running time cumulative effect factor and dynamically adjusting its weight in the model through the attention mechanism, the model can fully consider the impact of device aging.

[0147] With the online sequence extreme learning machine as an online learning algorithm, new measurement data is collected in real time to update the model parameters. The operating environment and device state of the power system are constantly changing, and a model with fixed parameters is difficult to maintain high accuracy for a long time. The online learning algorithm can collect new measurement data in real time, including multi-dimensional environmental parameters, device running time cumulative effect factors, and the error between the actual measurement value and the model prediction value of the resistance current component, and update the model parameters in a timely manner. When the error between the new data and the model prediction result exceeds the set parameter update threshold, the model will immediately trigger an update. In practical applications, the parameter update response speed of this online learning algorithm is fast, usually completing within a few seconds, ensuring that the model always keeps pace with the actual situation. Ensuring that the model can continuously maintain high accuracy as the device operating state and environment change, providing a strong guarantee for long-term, stable, and accurate monitoring of the resistance current of the lightning arrester.

[0148] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0149] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for measuring resistive current of capacitive equipment of a lightning arrester without taking a PT signal, characterized in that: include: Deploying a non-contact electric field coupling sensing array in a preset sensing area of ​​a high-voltage conductor of a capacitive device, wherein the sensing array includes a main sensing electrode group and a phase monitoring electrode group; According to the main sensing electrode group, the spatial distribution characteristics of the power frequency electric field around the high-voltage conductor are captured through the electric field coupling effect to obtain a multi-channel main sensing signal; according to the phase monitoring electrode group, the surface electric field of the high-voltage conductor is sensed through near-field coupling to obtain the induced electric field signal on the high-voltage side of the equipment; Extracting power frequency characteristics from the multi-channel main induction signal to generate an initial reference signal having the same frequency and phase as the operating voltage of the capacitive device, and performing cascade filtering on the initial reference signal to obtain a filtered reference signal; Based on the filtered reference signal and the induced electric field signal on the high-voltage side of the equipment, a phase closed-loop calibration mechanism is introduced. By comparing the phase difference between the two in real time, the phase offset of the filtered reference signal is dynamically corrected to obtain the calibration reference signal. Collecting the leakage current signal of the capacitive device grounding wire, performing analog-to-digital conversion and adaptive spectrum decomposition, and combining it with the calibration reference signal to separate and obtain the resistive current component; Multi-dimensional environmental parameters of the preset sensing area are synchronously collected, and a multivariable dynamic compensation model of the dielectric loss factor based on the deep belief network is constructed. The cumulative effect factor of the capacitive device operating time is embedded, and the parameters are updated in real time through an online learning algorithm. The multi-dimensional environmental parameters are used as input, and the compensation value for the resistive current component is output to obtain the resistive current measurement result.

2. The method for measuring resistive current of capacitive equipment applicable to lightning arresters without taking PT signals according to claim 1, characterized in that: The process of obtaining the multi-channel main sensing signal includes: Each sensing electrode in the main sensing electrode group generates induced charges in the power frequency electric field around the high-voltage conductor through the electric field coupling effect, forming a micro-current signal that changes with the electric field; The microcurrent signal is collected by adopting a multi-channel synchronous acquisition method, and the microcurrent signal is subjected to impedance conversion and multi-stage amplification processing by a signal conditioning circuit to obtain an amplified signal; The amplified signal is preliminarily filtered to form a multi-channel main sensing signal.

3. The method for measuring resistive current of capacitive equipment applicable to lightning arresters without taking PT signals according to claim 1, characterized in that: The process of obtaining the induced electric field signal on the high-voltage side of the equipment includes: The phase monitoring electrode group senses the electric field on the surface of the high-voltage conductor through near-field coupling to obtain a micro-electric signal; A low-noise amplifier circuit is used to amplify the micro-electrical signal, and a band-pass filter is performed on the amplified micro-electrical signal; The phase of the filtered micro-electric signal is corrected through the phase holding circuit to form the induced electric field signal on the high-voltage side of the equipment.

4. The method for measuring resistive current of capacitive equipment applicable to lightning arresters without taking PT signals according to claim 1, characterized in that: The process of generating an initial reference signal that is in phase with the operating voltage of the capacitive device includes: The multi-channel main sensing signal is processed using a blind source separation algorithm based on the kurtosis maximization criterion, and the power frequency signal and the interference signal are separated by iteratively optimizing the separation matrix to obtain a target signal containing power frequency characteristics; Performing spectrum analysis on the target signal to extract frequency and phase information of the power frequency fundamental component; Based on the extracted frequency and phase information, a continuous initial reference signal waveform is reconstructed using the cubic spline interpolation method; Phase smoothing is performed on the initial reference signal waveform to obtain an initial reference signal.

5. The method for measuring resistive current of capacitive equipment applicable to lightning arresters without taking PT signals according to claim 1, characterized in that: The process of obtaining the filtered reference signal includes: Inputting the initial reference signal into an adaptive notch filter to eliminate periodic interference in a specific frequency band to obtain an initial filtered signal; Performing wavelet threshold denoising on the initial filtered signal to obtain a denoised filtered signal; The denoising filter signal is smoothed to obtain a filter reference signal, and the filter reference signal is amplitude normalized.

6. The method for measuring resistive current of capacitive equipment applicable to lightning arresters without obtaining PT signals according to claim 1, characterized in that: The process of obtaining the calibration reference signal includes: Inputting the filtered reference signal and the induced electric field signal on the high-voltage side of the device into a phase comparator, calculating the phase difference between the two, and generating a phase error signal; Inputting the phase error signal into a Kalman filter for filtering to eliminate noise interference in the error signal; The filtered phase error signal is input into the proportional-integral-derivative controller to generate a phase compensation amount; The phase compensation amount is applied to the filtered reference signal through an orthogonal modulator, and the phase of the signal is dynamically corrected until the phase error is less than a preset threshold, thereby obtaining a calibration reference signal.

7. The method for measuring resistive current of capacitive equipment applicable to lightning arresters without taking PT signals according to claim 1, characterized in that: The process of separating the resistive current components includes: A magneto-optical current sensor is used to collect leakage current signals from the grounding wire of capacitive equipment. Converting the leakage current signal into a digital signal through an analog-to-digital converter; Adaptive spectrum decomposition based on Hilbert-Huang transform is applied to the digital signal, and the process includes: Preprocessing the digital signal using a mirror extension boundary processing technology to suppress endpoint effects; Decomposing the digital signal into a plurality of intrinsic mode functions by empirical mode decomposition; Performing a Hilbert transform on each of the intrinsic mode functions to obtain a corresponding Hilbert spectrum; According to the phase information of the calibration reference signal, an in-phase component is extracted from the decomposed Hilbert spectrum, and the in-phase components are superimposed to obtain a resistive current component.

8. The method for measuring resistive current of capacitive equipment applicable to lightning arresters without taking PT signals according to claim 1, characterized in that: The process of constructing a multivariable dynamic compensation model for dielectric loss factor based on a deep belief network includes: Design a deep belief network structure, including an input layer, multiple restricted Boltzmann machine hidden layers, and an output layer. The number of nodes in the input layer is consistent with the dimensionality of the multidimensional environmental parameters, and the output layer is a single node corresponding to the compensation value of the resistive current component. The multidimensional environmental parameters include temperature, humidity, air pressure, and air medium parameters. Collect multi-dimensional environmental parameters and corresponding dielectric loss factor measured values ​​under different environmental conditions to construct training and validation datasets; The deep belief network is pre-trained using a layer-by-layer greedy training algorithm, and the parameters of the restricted Boltzmann machine in each layer are optimized using the contrastive divergence algorithm; The pre-trained deep belief network is used as the initial model, and the back propagation algorithm is used to adjust the model parameters to minimize the error between the predicted compensation value and the actual compensation value; The model is verified through the validation data set, and the network structure and parameters are adjusted until the prediction accuracy of the model meets the preset requirements.

9. The method for measuring resistive current of capacitive equipment applicable to lightning arresters without obtaining PT signals according to claim 1, characterized in that: The process of embedding the cumulative effect factor of device operation time includes: Define the cumulative effect factor of equipment operation time; Add an input node to the deep belief network to input the cumulative effect factor of the device's operating time; Introducing the attention mechanism to dynamically adjust the weight of the cumulative effect factor of device operating time in the model; The cumulative effect factor of equipment operating time is fused with multi-dimensional environmental parameters as the input of the deep belief network.

10. The method for measuring resistive current of capacitive equipment applicable to lightning arresters without obtaining PT signals according to claim 1, characterized in that: The process of updating parameters in real time through online learning algorithms includes: An online learning algorithm using an online sequence extreme learning machine as the model is used, and the output of the deep belief network is used as the input of the online sequence extreme learning machine; Setting a forgetting factor, which is used to control the impact of historical data on model parameter updates; Real-time collection of new measurement data, including multi-dimensional environmental parameters, cumulative effect factors of equipment operation time, and the error between the actual measured value of the resistive current component and the model prediction value; Based on the newly collected data, the model parameters are updated through the online sequential extreme learning machine; Set a parameter update threshold. When the error between the new data and the model prediction result exceeds the parameter update threshold, the model parameter update is triggered.

Images