Low-voltage electricity testing equipment and electricity testing method thereof

Abstract

The invention discloses low-voltage electricity testing equipment and an electricity testing method thereof, and belongs to the field of electricity testing equipment. S2, performing low-noise amplification and analog-to-digital conversion on the analog signal to generate an initial digital signal data stream; s3, inputting the initial digital signal data stream into a covariance matrix adaptive evolution module, updating a covariance matrix of the leakage potential signal online through an exponential weighted sliding window, and extracting a signal subspace basis vector and a noise subspace basis vector; s4, outputting an optimal processing parameter set; s5, obtaining an enhanced weak signal feature sequence; and S6, judging the electrification state through naive Bayes confidence inference, and generating an electrification judgment result. According to the invention, the system can still maintain more than 99.5% of judgment consistency in temperature and humidity alternating and wind power induction significant scenes.

Description

Technical Field

[0001] This invention relates to the field of conductor technology, and in particular to a low-voltage testing device and its testing method. Background Technology

[0002] With the continuous increase in the intelligence of power systems and the frequency of live-line work, AC conductor voltage testing technology has received increasing attention as an important means to ensure the personal safety of operation and maintenance personnel and controllable operational risks. In high-voltage AC transmission scenarios of 400V and above, traditional simple equipment such as manual voltage testers and fixed threshold triggers are still widely used. However, in complex operating environments such as high altitude and open air, the leakage potential of AC conductors is usually in the millivolt level weak signal range and is severely submerged in multi-source interference such as power frequency harmonics, lightning induction, welding sparks and radio waves, which poses a severe challenge to traditional equipment.

[0003] Current voltage detection technologies mostly rely on fixed voltage threshold triggering methods or static analog filter response mechanisms. These methods generally have the following limitations: First, fixed thresholds cannot adapt to the dynamic changes in leakage potential under different conditions such as altitude, humidity, and temperature, which can easily lead to "no alarm when energized" or "false alarm" phenomena. Second, static filters cannot be adjusted according to changes in the noise spectrum structure, and cannot effectively extract weak effective signals from background noise, especially in environments with strong electromagnetic interference, where the false alarm rate is extremely high. Third, existing methods generally lack self-learning and adaptive mechanisms, requiring repeated manual calibration or reliance on external compensation circuits, which is cumbersome and cannot be deployed quickly. In addition, some intelligent voltage detection methods attempt to tune parameters using a single optimization algorithm, but they are prone to getting trapped in local extrema and cannot achieve robust filtering and discrimination in non-stationary noise environments.

[0004] In conclusion, there is an urgent need for an innovative approach that features dynamic modeling, global optimization, and low resource overhead to achieve a breakthrough. Summary of the Invention

[0005] One objective of this invention is to provide a low-voltage testing device and its testing method. In scenarios with alternating temperature and humidity and significant wind power induction, this system can still maintain a decision consistency rate of over 99.5%.

[0006] To achieve the above objectives, the present invention first provides a method for testing voltage using a low-voltage testing device, comprising the following steps: S1. Using an AC conductor voltage testing device, collect the near-field leakage potential signal of the AC conductor without damaging the insulation distance, and output an analog signal; S2. Perform low-noise amplification and analog-to-digital conversion on the analog signal to generate an initial digital signal data stream; S3. Input the initial digital signal data stream into the covariance matrix adaptive evolution module, update the covariance matrix of the leakage potential signal online through an exponentially weighted sliding window, and extract the signal subspace basis vector and noise subspace basis vector; S4. Provide the signal subspace basis vector and noise subspace basis vector to the Grey Wolf optimization module. Based on the objective function of maximizing signal-to-noise ratio, use the Grey Wolf optimization algorithm to perform a global search on the filter coefficients, envelope detection gain and decision threshold parameters, and output the optimal set of processing parameters. S5. Apply the optimal set of processing parameters to perform bandpass filtering, envelope extraction, and threshold decision on the initial digital signal data stream to obtain an enhanced weak signal feature sequence; S6. Input the enhanced weak signal feature sequence into the voltage detection decision module, integrate the time domain amplitude, frequency domain harmonic components and phase stability features, and determine the charging state through Naive Bayes confidence inference to generate the charging decision result.

[0007] Optionally, S1 includes the following steps: S11. Without compromising the safe insulation distance, an AC conductor voltage detector with dual-channel sensing capability is used for signal sensing. The voltage detector includes an electric field coupling sensor and a magnetic field induction coil, which are used to collect near-field leakage potential signals and induced magnetic field interference signals of the AC conductor, respectively. S12. The electric field coupling sensor outputs a simulated leakage potential signal. The amplitude of the simulated leakage potential signal is less than 1mV, and it is superimposed with a 50 / 60Hz main power frequency component and multiple harmonic interference components. The magnetic field induction coil outputs a simulated induced magnetic field signal. Used to assist in noise source identification; S13. Adjust the gain of the output signals of the electric field coupling sensor and the magnetic field induction coil respectively. The gain adjustment process is based on the initial gain parameter set of the control unit in the AC conductor voltage detector. ,in For voltage channel gain, The gain of the magnetic induction channel makes the dynamic range of the two channel signals controllable and the amplitude within the linear region of the analog-to-digital converter. S14. The adjusted leakage potential analog signal As the main channel input, it is converted into an analog voltage signal stream; simultaneously, the induced magnetic field analog signal is converted into... As an auxiliary reference signal, the leakage potential analog signal The weak electric field coupling potential generated in space by the non-ideal insulation properties of the AC conductor; the induced magnetic field simulates the signal. This is the induced magnetic disturbance signal when current flows through a nearby conductor; S15. Use a sampling frequency of Analog-to-digital converters for analog voltage signal streams Discretization is performed, and the sampling frequency is... The value range is set according to the power frequency and harmonic content, not lower than 1kHz, covering the main frequency of the signal and the 3rd, 5th, and 7th harmonics, generating the original digital signal data sequence. ,in, , for the first One sampling point, For the corresponding sampling time, S16. Total number of sampling points; S16. Convert the original digital signal data sequence It is stored synchronously with the simulated signal of the induced magnetic field at the corresponding moment.

[0008] Optionally, S2 includes the following steps: S21. Simulate the leakage potential signal The input is fed to a low-noise signal amplification module, which consists of a high-input-impedance buffer and a cascaded amplifier with adjustable voltage gain. The voltage gain factor of the amplifier with adjustable voltage gain is [value missing]. The voltage gain factor is dynamically adjusted by the control unit in the AC conductor voltage tester, so that the amplified leakage potential analog signal... It remains within the linear input dynamic range of the analog-to-digital converter; S22. Simulation signal of amplified leakage potential Anti-aliasing filtering is performed using a low-pass filter, with the cutoff frequency set to the sampling frequency. Half of it is used to filter out interference signals that are higher than the sampling bandwidth, preventing high-frequency components from aliasing into the subsequent digital signal; S23. The simulated leakage potential signal after anti-aliasing filtering is processed. The input is fed into the analog-to-digital converter (ADC), which uses a successive approximation ADC to perform discretization sampling to obtain the initial digital signal data stream. Each sampling point Indicates at time The voltage value at that point, at any time Defined as the first The time points corresponding to each sampling point; S24. Initial digital signal data stream Integrity verification is performed, including verification of the effective number of quantization bits and overload value judgment for each sampling point, and sampling point voltage value. The voltage must be greater than or equal to zero and less than or equal to the reference voltage of the analog-to-digital converter. If any sampling point exceeds this range, it is marked as an overload sample, and the control unit is triggered to execute the voltage gain factor. The readjustment mechanism.

[0009] Optionally, S3 includes the following steps: S31. Transfer the initial digital signal data stream Input covariance matrix adaptive evolution module to construct time series sliding window And introduce a weighted time-frequency response function. Adjust the sample weights and construct a weighted covariance matrix: in, The mean of the center sample of the sliding window; To match the power frequency center frequency The weighted time-frequency response function related to phase stability, through the weighted time-frequency response function Optimize the contribution of samples that match the power frequency steady-state signal to the covariance; S32. Perform eigenvalue decomposition on the weighted covariance matrix at each sliding window time step to obtain the eigenvalue set of all principal component directions at that time step. and its corresponding set of feature vectors For each eigenvalue in the eigenvalue set Calculate its proportion in the total energy to obtain the energy normalized ratio. The energy normalization ratio represents the current number of times. The degree of contribution of each principal component direction to the total signal energy of the sliding window; S33. For each eigenvalue Combined with the voltage amplitude threshold of the leakage potential signal and noise energy determination threshold It is determined whether the direction belongs to the principal component with weak amplitude and low energy. If the above conditions are met, a dynamic signal enhancement factor is applied to the principal component direction. The value of the dynamic signal enhancement factor is related to the relative difference between the eigenvalue and the noise threshold, and the enhancement degree is determined by the adjustment coefficient. control, The value of needs to be adjusted in combination with factors such as the noise level of the signal and the initial strength of the weak signal. The value range is generally 0.1-10. The role of the dynamic signal enhancement factor is to improve the covariance contribution of the weak signal in the low energy direction. S34. Extract each eigenvector from the eigenvector set. Its corresponding energy normalization ratio and signal enhancement factor A joint judgment is made, and if the product exceeds the set signal recognition threshold... Then the eigenvector is assigned to the signal subspace basis vector set. Otherwise, classify it into the set of basis vectors in the noise subspace. And the set of signal subspace basis vectors obtained at the current sliding window time. and the set of basis vectors of the noise subspace Output as the result of feature space constraints.

[0010] Optionally, S3 includes the following steps: S41. Set the basis vectors of the signal subspace. With the set of basis vectors in the noise subspace The input is fed into the Grey Wolf optimization module, which constructs an orthogonal projection structure of the signal direction based on the basis vector set of the signal subspace, thus obtaining the signal projection operator. An orthogonal projection structure of the noise direction is constructed based on the basis vector set of the noise subspace, resulting in the noise projection operator. The signal projection operator and the noise projection operator are used to extract the components of the filtering result in the signal direction and the noise direction, respectively. S42. Define an optimization objective function to maximize the signal-to-noise ratio in the Grey Wolf optimization module. The optimization objective function is to optimize the signal-to-noise ratio under the current parameter configuration. The resulting filtered signal sequence exhibits a large projection energy in the signal direction and a small projection energy in the noise direction. The projection energy in the signal direction is determined by the signal projection operator. The noise direction projection energy, obtained by applying the filtered signal sequence, is determined by the noise projection operator. The value of the objective function obtained by applying the same filtering result is equal to the ratio of the two projected energies. The larger the ratio, the better the current parameter configuration. in, The squared 2-norm of a vector is expressed in units of V. 2 ; This represents the projection component of the filtered signal in the signal subspace direction; This indicates the residual components of the filtered signal in the noise subspace direction; To avoid infinitesimal constants with a denominator of zero, the unit is V. 2 The larger the value of this function, the better the signal enhancement effect and the stronger the noise suppression capability. S43. Initialize the population for the gray wolf optimization algorithm. Each individual in the population corresponds to a set of parameter vectors to be optimized. The parameter vector to be optimized includes the set of filter coefficients. Envelope detection gain and decision threshold parameters For each individual, the signal-to-noise ratio objective function is evaluated in the current signal window. The evaluation result is used as the fitness value of the individual. The individual with the best fitness is selected as α wolf, the second best as β wolf, and the third best as δ wolf. S44. Based on the gray wolf optimization strategy, in each iteration, the average vector of the current parameter vectors of α wolf, β wolf, and δ wolf is calculated as the new center target position. The difference distance between the current individual and the target position is calculated, and a shrinkage control factor is used as the basis for this difference. Execute position update, shrink control factor As the algebra decreases, early updates tend to be more global, while later updates tend to be more local refinement. S45. During the iteration process, a mutation perturbation mechanism is introduced to prevent getting trapped in local optima. When the fitness improvement of the entire population is lower than a preset threshold over several consecutive generations, a Gaussian perturbation mechanism is triggered, adjusting the parameter vector with the current optimal fitness. Add small perturbations to readjust its search direction in order to escape local traps; S46. When the optimization iteration reaches the maximum number of rounds, or the rate of change of the optimal individual fitness falls below the preset stopping condition, the Gray Wolf Optimization Algorithm stops running and outputs the current globally optimal parameter vector. The current globally optimal parameter vector consists of three parts: the set of filter coefficients. Used to set the filter structure and frequency characteristics; envelope detection gain The amplification factor used to determine signal amplitude detection; the decision threshold parameter. .

[0011] Optionally, S5 includes the following steps: S51. Optimal processing parameter vector Applied to the initial digital signal data stream For the initial digital signal data stream When applying bandpass filters, the structure and characteristic parameters of the bandpass filter are determined by the optimal set of filter coefficients. The optimal filter coefficient set is determined by parameters including filter type, center frequency, passband bandwidth, and bandpass filter order. The bandpass filter, based on a set center frequency, preserves the target frequency band energy in the initial digital signal data stream while suppressing interference from other frequency bands. The filtering result forms the filter output signal sequence. Each data point Indicates the first The filtered amplitude corresponding to each sampling point under the action of the optimal filter parameters; S52. For the filter output signal sequence Envelope extraction is performed by constructing an instantaneous complex envelope using Hilbert transform, and then extracting its magnitude as the signal envelope amplitude. The envelope detection gain coefficient is used during the signal envelope extraction process. The envelope amplitude is uniformly gain-adjusted to ultimately form an envelope signal sequence. Each envelope amplitude Indicates the first The envelope amplitude value of the filtered signal at each sampling point, multiplied by the current optimal envelope detection gain coefficient.

[0012] S53. Envelope signal sequence With the current optimal decision threshold parameter Comparison was performed to construct a weak signal feature sequence. The weak signal feature sequence is used to represent the judgment result of whether each sampling point meets the characteristics of the leakage potential signal; when a certain envelope amplitude Greater than or equal to the current optimal decision threshold parameter At that time, it is assumed that there is a leakage potential signal at the sampling point, and the characteristic value of the corresponding location is marked. Set to 1; when the envelope amplitude Less than the current optimal decision threshold parameter When this occurs, it is considered that no valid signal was detected at the sampling point, and the feature label value is set. Setting it to 0, thus forming a weak signal feature sequence As a dominant discriminant label for electrical signals in the time domain.

[0013] Optionally, S6 includes the following steps: S61. Input the enhanced weak signal feature sequence and the corresponding envelope signal sequence into the voltage detection decision module. In the voltage detection decision module, extract the following three types of features in sequence to construct the basis for judging the energized state: The first category is time-domain amplitude characteristics. , which represents the average value of the envelope amplitude of all sampling points within the current sliding time window, and is used to reflect the overall energy level of the leakage potential signal in the time domain; The second category is frequency domain harmonic energy characteristics. By performing a fast Fourier transform on the filter output signal, the amplitude of the main frequency component, which includes the 50 or 60 Hz fundamental wave and the 3rd, 5th, and 7th harmonics, is extracted to characterize the energy distribution of the current signal in the typical power frequency electrical disturbance frequency band. The third category is phase stability characteristics. The calculation is based on the phase angle change of the main frequency signal within several consecutive sampling windows, which is expressed as the standard deviation of the main frequency phase angle and is used to measure the temporal consistency of the leakage potential signal in multiple time periods. S62. Extract the time-domain amplitude features Frequency domain harmonic energy characteristics Phase stability characteristics Both are input into the Naive Bayes confidence inference model, which infers based on the conditional probability table constructed from historical labeled samples and outputs the confidence probability value that the currently observed sample is in a charged state. confidence probability value of being in an uncharged state ; S63. Based on the confidence probability value of the charged state The magnitude is used to make a three-class classification decision, resulting in a charge determination result.

[0014] Optionally, the charged decision result includes: When the confidence probability value When the value is greater than or equal to 0.995, it is determined that the AC conductor is energized at that moment, and the tag is output as energized; When the confidence probability value When the value is less than 0.200, it is determined that the AC conductor is in an unenergized state at that moment, and the tag is output as unenergized; When the confidence probability value When the value is greater than or equal to 0.200 and less than 0.995, the system determines that the signal is in a suspected energized state and outputs a label indicating suspected energization. The system will then automatically extend the duration of the current signal acquisition window.

[0015] To achieve the above objectives, the present invention also provides a low-voltage testing device applied to the above-described voltage testing method. This device includes a voltage detector, a telescopic rod, a remote control wrench, and a puncture structure. The voltage detector is used to check whether a conductor is live. The telescopic rod is used to move the voltage detector and the puncture structure. The remote control wrench is used to drive the puncture structure. The puncture structure includes an elbow with a wire groove and a raised strip. A puncture head is slidably connected to the elbow along the raised strip. A puncture needle is provided at the top of the puncture head, and a puncture point is provided at the bottom of the puncture head. The device has a cavity, inside which a circular plate is placed. A threaded rod is fixed to the bottom of the circular plate, and the threaded rod passes through and is rotatably connected to the puncture head. The puncture head is connected to the probe of the electroscope via a copper wire. A hexagonal nut is fixed to the bottom of the puncture head. A hexagonal groove is provided on the top of the remote control wrench, and a bolt is fixed to the bottom of the remote control wrench. A threaded groove is provided on the top of the telescopic rod. Three elastic retaining rings are fixed to one side of the remote control wrench. The puncture head is arc-shaped, and the puncture needle is made of metal.

[0016] The beneficial effects of this invention are: This invention introduces a time-frequency similarity weighting function during the construction of the covariance matrix. The sample weights are dynamically adjusted based on the phase stability of the electrical signal on the main power frequency component, thereby increasing the covariance contribution of samples that match the AC leakage potential. At the same time, by combining the energy normalization ratio and the dynamic signal enhancement factor, a fine division mechanism between the signal subspace and the noise subspace is established. Compared with traditional PCA or static filtering separation methods, the main direction can still be stably extracted under signal conditions below 0.2mV, effectively improving the discrimination capability of the signal-to-noise subspace. The measured signal-to-noise ratio is improved by more than 12dB.

[0017] This invention constructs an orthogonal projection operator based on the basis vector set of the signal subspace and the basis vector set of the noise subspace. The ratio of maximizing the projected energy in the signal direction to minimizing the residual energy in the noise direction is used as the fitness objective function of the gray wolf optimization. This ensures that the parameter optimization direction is completely aligned with the physical principal components of the weak electrical signal, effectively suppressing the misleading interference of welding sparks and unstructured noise from wireless interference on the filter structure. This enables the filter parameters to have cross-scenario generalization ability, reducing the system false alarm rate to less than 0.7%.

[0018] This invention employs a Naive Bayes confidence inference algorithm at the decision logic layer, while integrating three physically interpretable multidimensional features—time domain amplitude, frequency domain harmonic energy, and dominant frequency phase stability—as input variables. By constructing a conditional probability distribution model trained on historical data, it outputs the confidence level of the charged state, achieving a system-level closed loop from sampled signals to cognitive decision-making. Furthermore, it introduces a suspected charged intermediate state and a self-expanding sampling window mechanism to further reduce the risk of misjudgment in ambiguous boundary sections. In scenarios with alternating temperature and humidity and significant wind power induction, this system can still maintain decision consistency of over 99.5%. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a low-voltage testing device and its testing method proposed in this invention; Figure 2 The low-voltage testing equipment and testing method proposed in this invention are based on the gray wolf optimization algorithm structure constructed from the signal subspace and noise subspace. Figure 3 The present invention provides a low-voltage voltage testing device and its testing method, which is a Naive Bayes voltage testing decision module based on the fusion of time amplitude, frequency domain harmonics and phase stability characteristics. Figure 4 This is an external view of a low-voltage testing device and its testing method proposed in this invention. Figure 5This invention provides a low-voltage testing device and its testing method. Figure 4 Structural separation diagram; Figure 6 This is a schematic diagram of the puncture structure of a low-voltage electrical testing device and its testing method proposed in this invention. Figure 7 This is a schematic diagram of point A of a low-voltage testing device and its testing method proposed in this invention; Figure 8 This is a schematic diagram of the detector and elastic retaining ring structure of a low-voltage voltage testing device and its voltage testing method proposed in this invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0021] refer to Figures 1-8 A low-voltage voltage testing device and its testing method, comprising the following steps: S1. Using an AC conductor voltage testing device, collect the near-field leakage potential signal of the AC conductor without damaging the insulation distance, and output an analog signal; S2. Perform low-noise amplification and analog-to-digital conversion on the analog signal to generate an initial digital signal data stream; S3. Input the initial digital signal data stream into the covariance matrix adaptive evolution module, update the covariance matrix of the leakage potential signal online through an exponentially weighted sliding window, and extract the signal subspace basis vector and noise subspace basis vector; S4. Provide the signal subspace basis vector and noise subspace basis vector to the Grey Wolf optimization module. Based on the objective function of maximizing signal-to-noise ratio, use the Grey Wolf optimization algorithm to perform a global search on the filter coefficients, envelope detection gain and decision threshold parameters, and output the optimal set of processing parameters. S5. Apply the optimal set of processing parameters to perform bandpass filtering, envelope extraction, and threshold decision on the initial digital signal data stream to obtain the enhanced weak signal feature sequence; S6. Input the enhanced weak signal feature sequence into the voltage detection decision module, integrate the time domain amplitude, frequency domain harmonic components and phase stability features, and determine the charging state through Naive Bayes confidence inference to generate the charging decision result.

[0022] In this embodiment, S1 includes the following steps: S11. Without compromising the safe insulation distance, an AC conductor voltage detector with dual-channel sensing capability is used for signal sensing. The voltage detector includes an electric field coupling sensor and a magnetic field induction coil, which are used to collect near-field leakage potential signals and induced magnetic field interference signals of the AC conductor, respectively. S12. The electric field coupling sensor outputs a simulated leakage potential signal. The amplitude of the simulated leakage potential signal is less than 1mV, and it is superimposed with a 50 / 60Hz main power frequency component and multiple harmonic interference components. The magnetic field induction coil outputs a simulated induced magnetic field signal. Used to assist in noise source identification; S13. Adjust the gain of the output signals of the electric field coupling sensor and the magnetic field induction coil respectively. The gain adjustment process is based on the initial gain parameter set of the control unit in the AC conductor voltage detector. ,in For voltage channel gain, The gain of the magnetic induction channel makes the dynamic range of the two channel signals controllable and the amplitude within the linear region of the analog-to-digital converter. S14. The adjusted leakage potential analog signal As the main channel input, it is converted into an analog voltage signal stream; simultaneously, the induced magnetic field analog signal is converted into... As an auxiliary reference signal, the leakage potential analog signal The weak electric field coupling potential generated in space by the non-ideal insulation properties of the AC conductor; the induced magnetic field simulates the signal. This is the induced magnetic disturbance signal when current flows through a nearby conductor; S15. Use a sampling frequency of Analog-to-digital converters for analog voltage signal streams Discretization is performed, and the sampling frequency is... The value range is set according to the power frequency and harmonic content, not lower than 1kHz, covering the main frequency of the signal and the 3rd, 5th, and 7th harmonics, generating the initial digital signal data stream. ,in, , for the first One sampling point, , where is the corresponding sampling time. S16. Total number of sampling points; S16. Initial digital signal data stream It is stored synchronously with the simulated signal of the induced magnetic field at the corresponding moment.

[0023] In this embodiment, S2 includes the following steps: S21. Simulate the leakage potential signal The input is fed to a low-noise signal amplification module, which consists of a high-input-impedance buffer and a cascaded amplifier with adjustable voltage gain. The voltage gain factor of the amplifier with adjustable voltage gain is [value missing]. The voltage gain factor is dynamically adjusted by the control unit in the AC conductor voltage tester, so that the amplified leakage potential analog signal... It remains within the linear input dynamic range of the analog-to-digital converter; S22. Simulation signal of amplified leakage potential Anti-aliasing filtering is performed using a low-pass filter, with the cutoff frequency set to the sampling frequency. Half of it is used to filter out interference signals that are higher than the sampling bandwidth, preventing high-frequency components from aliasing into the subsequent digital signal; S23. The simulated leakage potential signal after anti-aliasing filtering is processed. The input is fed into the analog-to-digital converter (ADC), which uses a successive approximation ADC to perform discretization sampling to obtain the initial digital signal data stream. Each sampling point Indicates at time The voltage value at that point, at any time Defined as the first The time points corresponding to each sampling point; S24. Initial digital signal data stream Integrity verification is performed, including verification of the effective number of quantization bits and overload value judgment for each sampling point, and sampling point voltage value. The voltage must be greater than or equal to zero and less than or equal to the reference voltage of the analog-to-digital converter. If any sampling point exceeds this range, it is marked as an overload sample, and the control unit is triggered to execute the voltage gain factor. The readjustment mechanism.

[0024] In this embodiment, S3 includes the following steps: S31. Transfer the initial digital signal data stream Input covariance matrix adaptive evolution module to construct time series sliding window And introduce a weighted time-frequency response function. Adjust the sample weights and construct a weighted covariance matrix: in, The mean of the center sample of the sliding window; To match the power frequency center frequency The weighted time-frequency response function related to phase stability, through the weighted time-frequency response function Optimize the contribution of samples that match the power frequency steady-state signal to the covariance; S32. Perform eigenvalue decomposition on the weighted covariance matrix at each sliding window time step to obtain the eigenvalue set of all principal component directions at that time step. and its corresponding set of feature vectors For each eigenvalue in the eigenvalue set Calculate its proportion in the total energy to obtain the energy normalized ratio. The energy normalization ratio represents the current number of times. The degree of contribution of each principal component direction to the total signal energy of the sliding window; S33. For each eigenvalue Combined with the voltage amplitude threshold of the leakage potential signal and noise energy determination threshold It is determined whether the direction belongs to the principal component with weak amplitude and low energy. If the above conditions are met, a dynamic signal enhancement factor is applied to the principal component direction. The value of the dynamic signal enhancement factor is related to the relative difference between the eigenvalue and the noise threshold, and the enhancement degree is determined by the adjustment coefficient. control, The value of needs to be adjusted in combination with factors such as the noise level of the signal and the initial strength of the weak signal. The value range is generally 0.1-10. The role of the dynamic signal enhancement factor is to improve the covariance contribution of the weak signal in the low energy direction. S34. Extract each eigenvector from the eigenvector set. Its corresponding energy normalization ratio and signal enhancement factor A joint judgment is made, and if the product exceeds the set signal recognition threshold... Then the eigenvector is assigned to the signal subspace basis vector set. Otherwise, classify it into the set of basis vectors in the noise subspace. And the set of signal subspace basis vectors obtained at the current sliding window time. and the set of basis vectors of the noise subspace Output as the result of feature space constraints.

[0025] In this embodiment, S4 includes the following steps: S41. Set the basis vectors of the signal subspace. With the set of basis vectors in the noise subspace The input is fed into the Grey Wolf optimization module, which constructs an orthogonal projection structure of the signal direction based on the basis vector set of the signal subspace, thus obtaining the signal projection operator. An orthogonal projection structure of the noise direction is constructed based on the basis vector set of the noise subspace, resulting in the noise projection operator. The signal projection operator and the noise projection operator are used to extract the components of the filtering result in the signal direction and the noise direction, respectively. S42. Define an optimization objective function to maximize the signal-to-noise ratio in the Grey Wolf optimization module. The optimization objective function is to optimize the signal-to-noise ratio under the current parameter configuration. The resulting filtered signal sequence exhibits a large projection energy in the signal direction and a small projection energy in the noise direction. The projection energy in the signal direction is determined by the signal projection operator. The noise direction projection energy, obtained by applying the filtered signal sequence, is determined by the noise projection operator. The value of the objective function obtained by applying the same filtering result is equal to the ratio of the two projected energies. The larger the ratio, the better the current parameter configuration. in, The squared 2-norm of a vector is expressed in units of V. 2 ; This represents the projection component of the filtered signal in the signal subspace direction; This indicates the residual components of the filtered signal in the noise subspace direction; To avoid infinitesimal constants with a denominator of zero, the unit is V. 2 The larger the value of this function, the better the signal enhancement effect and the stronger the noise suppression capability. S43. Initialize the population for the gray wolf optimization algorithm. Each individual in the population corresponds to a set of parameter vectors to be optimized. The parameter vector to be optimized includes the set of filter coefficients. Envelope detection gain and decision threshold parameters For each individual, the signal-to-noise ratio objective function is evaluated in the current signal window. The evaluation result is used as the fitness value of the individual. The individual with the best fitness is selected as α wolf, the second best as β wolf, and the third best as δ wolf. S44. Based on the gray wolf optimization strategy, in each iteration, the average vector of the current parameter vectors of α wolf, β wolf, and δ wolf is calculated as the new center target position. The difference distance between the current individual and the target position is calculated, and a shrinkage control factor is used as the basis for this difference. Execute position update, shrink control factor As the algebra decreases, early updates tend to be more global, while later updates tend to be more local refinement. S45. During the iteration process, a mutation perturbation mechanism is introduced to prevent getting trapped in local optima. When the fitness improvement of the entire population is lower than a preset threshold over several consecutive generations, a Gaussian perturbation mechanism is triggered, adjusting the parameter vector with the current optimal fitness. Add small perturbations to readjust its search direction in order to escape local traps; S46. When the optimization iteration reaches the maximum number of rounds, or the rate of change of the optimal individual fitness falls below the preset stopping condition, the Gray Wolf Optimization Algorithm stops running and outputs the current globally optimal parameter vector. The current globally optimal parameter vector consists of three parts: the set of filter coefficients. Used to set the filter structure and frequency characteristics; envelope detection gain The amplification factor used to determine signal amplitude detection; the decision threshold parameter. .

[0026] In this embodiment, S5 includes the following steps: S51. Optimal processing parameter vector Applied to the initial digital signal data stream For the initial digital signal data stream When applying bandpass filters, the structure and characteristic parameters of the bandpass filter are determined by the optimal set of filter coefficients. The optimal filter coefficient set is determined by parameters including filter type, center frequency, passband bandwidth, and bandpass filter order. The bandpass filter, based on a set center frequency, preserves the target frequency band energy in the initial digital signal data stream while suppressing interference from other frequency bands. The filtering result forms the filter output signal sequence. Each data point Indicates the first The filtered amplitude corresponding to each sampling point under the action of the optimal filter parameters; S52. For the filter output signal sequence Envelope extraction is performed by constructing an instantaneous complex envelope using Hilbert transform, and then extracting its magnitude as the signal envelope amplitude. The envelope detection gain coefficient is used during the signal envelope extraction process. The envelope amplitude is uniformly gain-adjusted to ultimately form an envelope signal sequence. Each envelope amplitude Indicates the first The envelope amplitude value of the filtered signal at each sampling point, multiplied by the current optimal envelope detection gain coefficient.

[0027] S53. Envelope signal sequence With the current optimal decision threshold parameter Comparison was performed to construct a weak signal feature sequence. The weak signal feature sequence is used to represent the judgment result of whether each sampling point meets the characteristics of the leakage potential signal; when a certain envelope amplitude Greater than or equal to the current optimal decision threshold parameter At that time, it is assumed that there is a leakage potential signal at the sampling point, and the characteristic value of the corresponding location is marked. Set to 1; when the envelope amplitude Less than the current optimal decision threshold parameter When this occurs, it is considered that no valid signal was detected at the sampling point, and the feature label value is set. Setting it to 0, thus forming a weak signal feature sequence As a dominant discriminant label for electrical signals in the time domain.

[0028] In this embodiment, S6 includes the following steps: S61. Input the enhanced weak signal feature sequence and the corresponding envelope signal sequence into the voltage detection decision module. In the voltage detection decision module, extract the following three types of features in sequence to construct the basis for judging the energized state: The first category is time-domain amplitude characteristics. , which represents the average value of the envelope amplitude of all sampling points within the current sliding time window, and is used to reflect the overall energy level of the leakage potential signal in the time domain; The second category is frequency domain harmonic energy characteristics. By performing a fast Fourier transform on the filter output signal, the amplitude of the main frequency component, which includes the 50 or 60 Hz fundamental wave and the 3rd, 5th, and 7th harmonics, is extracted to characterize the energy distribution of the current signal in the typical power frequency electrical disturbance frequency band. The third category is phase stability characteristics. The calculation is based on the phase angle change of the main frequency signal within several consecutive sampling windows, which is expressed as the standard deviation of the main frequency phase angle and is used to measure the temporal consistency of the leakage potential signal in multiple time periods. S62. Extract the time-domain amplitude features Frequency domain harmonic energy characteristics Phase stability characteristics Both are input into the Naive Bayes confidence inference model, which infers based on the conditional probability table constructed from historical labeled samples and outputs the confidence probability value that the currently observed sample is in a charged state. confidence probability value of being in an uncharged state ; S63. Based on the confidence probability value of the charged state The magnitude is used to make a three-class classification decision, resulting in a charge determination result.

[0029] In this embodiment, the liveness determination result includes: When the confidence probability value When the value is greater than or equal to 0.995, it is determined that the AC conductor is energized at that moment, and the tag is output as energized; When the confidence probability value When the value is less than 0.200, it is determined that the AC conductor is in an unenergized state at that moment, and the tag is output as unenergized; When the confidence probability value When the value is greater than or equal to 0.200 and less than 0.995, the system determines that the signal is in a suspected energized state and outputs a label indicating suspected energization. The system will then automatically extend the duration of the current signal acquisition window.

[0030] In this embodiment, the voltage testing equipment includes a voltage detector 1, a telescopic rod 2, a remote control wrench 3, and a puncture structure. The voltage detector 1 is used to test whether the wire is live. The telescopic rod 2 is used to move the voltage detector 1 and the puncture structure. The remote control wrench 3 is used to drive the puncture structure. The puncture structure includes a bend 4, which has a wire groove 5 and a raised strip 6. A puncture head 7 is slidably connected to the bend 4 along the raised strip 6. A puncture needle 8 is provided at the top of the puncture head 7, and a cavity 9 is provided at the bottom of the puncture head 7. A circular plate 1 is placed inside the cavity 9. 0. A threaded rod 11 is fixed to the bottom of the circular plate 10. The threaded rod 11 passes through the puncture head 7 and is rotatably connected to the puncture head 7. The puncture head 7 is connected to the probe of the electroscope 1 through a copper wire 17. A hexagonal nut 12 is fixed to the bottom of the puncture head 7. A hexagonal groove 13 is provided on the top of the remote control wrench 3. A bolt 14 is fixed to the bottom of the remote control wrench 3. A threaded groove 15 is provided on the top of the telescopic rod 2. An elastic retaining ring 16 is fixed on one side of the remote control wrench 3. The elastic retaining ring 16 is vertically distributed in three places. The puncture head 7 is arc-shaped. The puncture needle 8 is made of metal.

[0031] Working principle: Open the box and take out the various components of the electrical testing equipment for assembly. Insert the voltage detector 1 into the elastic retaining ring 16 to complete the installation of the voltage detector 1 and the remote control wrench 3. Then, insert the bolt 14 at the bottom of the remote control wrench 3 into the threaded groove 15. Next, rotate the telescopic rod 2 to connect it to the remote control wrench 3. Finally, place the hexagonal nut 12 of the piercing structure into the hexagonal groove 13 to complete the installation of the piercing structure, thus assembling the entire electrical testing equipment. Figure 1As shown, when it is necessary to test the voltage of an AC conductor with an insulated layer, the telescopic rod 2 extends, raising the piercing structure and the voltage detector 1. Then, the bend 4 of the piercing structure wraps around the AC conductor. Afterwards, the technician controls the remote control wrench 3 to rotate the threaded rod 11. The threaded rod 11 and the circular plate 10 rotate, and then the threaded rod 11 rotates and rises, sliding the piercing head 7 along the raised strip 6 through the circular plate 10. Then, the piercing needle 8 penetrates the insulation layer of the conductor, and the metal piercing needle 8 directly contacts the conductor. At this time, the AC voltage signal of the conductor is conducted through the piercing needle 8. The copper wire 17, as a conductive medium, directly connects the piercing needle 8 and the probe of the voltage detector 1 to form a conductive path, transmitting the AC voltage signal on the conductor to the input end of the probe of the voltage detector 1. After the probe of the voltage detector 1 receives the AC voltage signal, the internal detection circuit (such as capacitive coupling, electric field induction, or direct contact detection) will analyze the signal characteristics. If the conductor is energized, the voltage detector will indicate the energized state through audible and visual alarms, indicator lights, or displays, thereby realizing the voltage testing of the AC insulated conductor.

[0032] Application Example 1: A power grid company plans to conduct live-line testing on a section of 400V overhead distribution line located in the suburbs of Mianning County, Xichang City, Liangshan Prefecture, Sichuan Province. This section of the line is situated in a transitional area between mountains and hills, at an altitude of approximately 2400 meters. The area is characterized by frequent cloud cover and low pressure throughout the year, large diurnal temperature variations, and is a typical high-altitude environment with strong interference. Furthermore, it is frequently accompanied by mountain thunderstorms, wind induced current, and frequent wireless communication. As a result, under non-contact monitoring conditions, the leakage potential signal strength of the AC conductors in this area is generally below 0.4mV, with the effective signal completely submerged in power frequency noise centered at 50 / 60Hz, harmonic crosstalk, and high-frequency electromagnetic interference.

[0033] In this voltage testing, the maintenance personnel used the novel portable intelligent voltage testing equipment described in this invention. This equipment has an embedded signal processing and decision system based on covariance matrix adaptive evolution and gray wolf optimization algorithm, and is equipped with an STM32F4 series low-power microcontroller. Together with a bipolar electric field probe and an iron core induction coil, it can detect the weak signal of AC wire leakage potential without contacting the wire and maintaining a safe distance of 1.5 meters. The adjustment coefficient was set to 3.0.

[0034] The testing was conducted under cloudy and windy conditions, with an ambient temperature of 8°C and a relative humidity of 78%. Three independent monitoring points were set up in the testing area: tower base A (near the welding area), tower base B (near the communication transmission point), and tower base C (on the wind turbine transformer outlet side). Each point underwent continuous dynamic monitoring sampling for 10 minutes, with 1000 samples per second, for a total of 60,000 sampling points.

[0035] At tower base A, a traditional voltage tester did not issue any liveness warning. After manual verification with a contact voltage tester, the wire was confirmed to be live. The device of this invention acquired a valid leakage signal within 10 seconds, with an envelope peak value of 0.37mV. The Gray Wolf optimized output bandpass filter has a center frequency of 50.6Hz, a bandwidth of 12Hz, a third-order structure, an envelope detection gain of 7.8 times, and a Bayesian confidence output result of 0.998 for liveness confidence, automatically triggering a "live" alarm.

[0036] At tower base B, the traditional analog filter method generated false alarms, continuously reporting "energized" signals, but subsequent contact voltage testing confirmed a power-off state. Analysis revealed that the area experienced strong interference from communication base stations, with the dominant interference frequency at 59.8Hz, which traditional bandpass filters could not effectively distinguish. The device of this invention, utilizing a covariance evolution structure, successfully classified this interference into the noise subspace. After filtering, the effective signal amplitude was below 0.09mV, and the Bayesian confidence output showed a "not energized" confidence level of 0.985, thus avoiding false alarms.

[0037] At tower base C, traditional methods cannot detect intermittent leakage signals as low as 0.22mV, and the equipment judges it as "not energized." However, the device of this invention extracts phase stability characteristics after continuously capturing 7 windows. If the amplitude fluctuation of the main harmonic of the power frequency is less than 1.2%, it is ultimately determined to be "suspected to be live". The system automatically extends the sampling window to 15 seconds, and the Bayesian confidence level reaches 0.996, which finally triggers the "live" label. On-site verification confirms that the wire has residual induced voltage due to incomplete power disconnection. This judgment avoids serious operational risks.

[0038] To verify the robustness and accuracy of the method of this invention, the research team selected 360 manual "energized / non-energized" judgment samples from the Liangshan power grid operation and maintenance site over the past three years, constructed a sample set, and compared the voltage detection accuracy of the method of this invention with the traditional static filtering + fixed threshold method. The results are as follows: Table 1. Voltage detection accuracy of the method in this embodiment compared to the traditional static filtering + fixed threshold method. index Traditional methods Method of the present invention Minimum detectable signal amplitude 0.8mV 0.18mV Average false alarm rate 6.7% 0.7% Average false negative rate 12.1% 0.4% Signal-to-noise ratio improvement 5.6dB 12.3dB Response time (mean) 13.8 seconds 3.7 seconds Consistency assessment (comparison between live and manual methods) 85.2% 99.5% Algorithm computational resource overhead 112kB / 48kB 65kB / 28kB Furthermore, during the algorithm iteration, we tested 20 combinations of filters and envelope gains for the gray wolf optimized output. By comparing the fitness convergence curves of the "projection SNR objective function" and the "traditional frequency domain amplitude maximization", we found that the traditional optimization function gets stuck in a local minimum after the 13th generation. However, the method of this invention converges rapidly from the 6th generation due to the dynamic projection direction of the fused signal subspace and noise subspace, and reaches the optimal solution in the 18th generation. The final SNR is improved by about 2.5 times, which verifies the global optimal performance of the fusion structure.

[0039] In summary, through simulation of real power field scenarios and verification using large-scale historical samples, this invention not only significantly improves the detection capability of weak electrical signals, the false alarm control capability, and the adaptive optimization capability, but also realizes a highly reliable and practical portable voltage detection system on a hardware-restricted MCU platform, demonstrating extremely high engineering value and promotion potential.

[0040] This invention introduces a time-frequency similarity weighting function during the construction of the covariance matrix. The sample weights are dynamically adjusted based on the phase stability of the electrical signal on the main power frequency component, thereby increasing the covariance contribution of samples that match the AC leakage potential. At the same time, by combining the energy normalization ratio and the dynamic signal enhancement factor, a fine division mechanism between the signal subspace and the noise subspace is established. Compared with traditional PCA or static filtering separation methods, the main direction can still be stably extracted under signal conditions below 0.2mV, effectively improving the discrimination capability of the signal-to-noise subspace. The measured signal-to-noise ratio is improved by more than 12dB.

[0041] This invention constructs an orthogonal projection operator based on the basis vector set of the signal subspace and the basis vector set of the noise subspace. The ratio of maximizing the projected energy in the signal direction to minimizing the residual energy in the noise direction is used as the fitness objective function of the gray wolf optimization. This ensures that the parameter optimization direction is completely aligned with the physical principal components of the weak electrical signal, effectively suppressing the misleading interference of welding sparks and unstructured noise from wireless interference on the filter structure. This enables the filter parameters to have cross-scenario generalization ability, reducing the system false alarm rate to less than 0.7%.

[0042] This invention employs a Naive Bayes confidence inference algorithm at the decision logic layer, while integrating three physically interpretable multidimensional features—time domain amplitude, frequency domain harmonic energy, and dominant frequency phase stability—as input variables. By constructing a conditional probability distribution model trained on historical data, it outputs the confidence level of the charged state, achieving a system-level closed loop from sampled signals to cognitive decision-making. Furthermore, it introduces a suspected charged intermediate state and a self-expanding sampling window mechanism to further reduce the risk of misjudgment in ambiguous boundary sections. In scenarios with alternating temperature and humidity and significant wind power induction, this system can still maintain decision consistency of over 99.5%.

[0043] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for testing voltage using a low-voltage testing device, characterized in that, Includes the following steps: S1. Acquire electromagnetic field signals without disrupting the insulation distance and output analog signals; S2. Preprocess the analog signal to generate an initial digital signal data stream; S3. Input the initial digital signal data stream into the covariance matrix adaptive evolution module, update the covariance matrix of the electromagnetic field signal online through an exponentially weighted sliding window, and extract the signal subspace basis vector and noise subspace basis vector; S4. Provide the signal subspace basis vector and noise subspace basis vector to the Grey Wolf optimization module, and output the optimal set of processing parameters based on the objective function of maximizing the signal-to-noise ratio; S5. Apply the optimal set of processing parameters to perform bandpass filtering, envelope extraction, and threshold decision on the initial digital signal data stream to obtain an enhanced weak signal feature sequence; S6. Input the enhanced weak signal feature sequence into the voltage detection decision module to generate a live decision result.

2. The voltage testing method for low-voltage testing equipment according to claim 1, characterized in that, S1 includes the following steps: S11. An AC conductor voltage detector with dual-channel sensing capability is used to sense signals and collect electromagnetic field signals, wherein the electromagnetic field signals include AC conductor near-field leakage potential signals and induced magnetic field interference signals. S12. The electric field coupling sensor outputs a simulated signal of near-field leakage potential of the AC conductor, which is superimposed with the main power frequency component and multi-order harmonic interference components. The magnetic field induction coil outputs a simulated signal of induced magnetic field. S13. Adjust the gain of the output signals of the electric field coupling sensor and the magnetic field induction coil respectively, so that the dynamic range of the two channels is controllable and the amplitude is in the linear region of the analog-to-digital converter. S14. The analog signal of near-field leakage potential of AC conductor after gain adjustment is used as the main channel input and converted into an analog voltage signal stream; at the same time, the analog signal of induced magnetic field after gain adjustment is used as an auxiliary reference signal. S15. Use an analog-to-digital converter to discretize the analog voltage signal stream and generate the original digital signal data sequence.

3. The voltage testing method for low-voltage testing equipment according to claim 1, characterized in that, S2 includes the following steps: S21. Input the analog signal of AC wire near-field leakage potential after gain adjustment to the low-noise signal amplification module so that the amplified analog signal of AC wire near-field leakage potential is kept within the linear input dynamic range of the analog-to-digital converter. S22. Perform anti-aliasing filtering on the amplified AC conductor near-field leakage potential analog signal; S23. Input the analog signal of the near-field leakage potential of the AC conductor after anti-aliasing filtering to the analog-to-digital conversion module to obtain the initial digital signal data stream; S24. Perform integrity verification on the initial digital signal data stream, verify the effective number of quantization bits and judge the overload value for each sampling point. If there is a sampling point that exceeds the range, it is marked as an overload sample and the control unit is triggered to execute the voltage gain factor readjustment mechanism.

4. The voltage testing method for low-voltage testing equipment according to claim 1, characterized in that, S3 includes the following steps: S31. Input the initial digital signal data stream into the covariance matrix adaptive evolution module, construct a time series sliding window, and introduce a weighted time-frequency response function to adjust the sample weights and construct a weighted covariance matrix; S32. Perform eigenvalue decomposition on the weighted covariance matrix at each sliding window time to obtain the set of eigenvalues ​​and the set of eigenvectors of all principal component directions at that time. Calculate the proportion of each eigenvalue in the total energy for each eigenvalue in the eigenvalue set to obtain the energy normalization ratio. S33. For each feature value, combine the voltage amplitude threshold and noise energy determination threshold of the AC conductor near-field leakage potential signal to determine whether it belongs to the principal component direction with weak amplitude and low energy. If the above conditions are met, apply a dynamic signal enhancement factor to the principal component direction. S34. Jointly judge each eigenvector in the eigenvector set with its corresponding energy normalization ratio and dynamic signal enhancement factor, and output the signal subspace basis vector set and noise subspace basis vector set obtained at the current sliding window time as the eigenspace constraint result.

5. The voltage testing method for low-voltage testing equipment according to claim 1, characterized in that, S4 includes the following steps: S41. Input the signal subspace basis vector set and the noise subspace basis vector set into the Grey Wolf optimization module to obtain the signal projection operator and the noise projection operator; S42. Define the objective function for maximizing the signal-to-noise ratio in the Grey Wolf optimization module; S43. Initialize the population for the gray wolf optimization algorithm, evaluate the signal-to-noise ratio objective function for each individual in the current signal window, use the evaluation result as the fitness value of the individual, and select the individual with the best fitness as α wolf, the second best as β wolf, and the third best as δ wolf. S44. Based on the gray wolf optimization strategy, in each iteration, the average vector of the three alpha wolves, beta wolves, and delta wolves is calculated as the new center target position. The difference distance between the current individual and the target position is calculated, and the position is updated based on the difference distance and a shrinkage control factor. The shrinkage control factor decreases with each generation. S45. During the iteration process, a mutation perturbation mechanism is introduced to prevent getting trapped in local optima. When the fitness improvement of the entire population is lower than the preset threshold over several consecutive generations, the Gaussian perturbation mechanism is triggered to add a small perturbation to the parameter vector with the current fitness optimum and readjust its search direction to escape the local trap. S46. When the optimization iteration reaches the maximum number of rounds, or the rate of change of the fitness of the best individual is lower than the preset stopping condition, the Grey Wolf Optimization Algorithm stops running and outputs the current global optimal parameter vector.

6. The voltage testing method for low-voltage testing equipment according to claim 1, characterized in that, S5 includes the following steps: S51. Apply the optimal processing parameter vector to the initial digital signal data stream, apply a bandpass filter to the initial digital signal data stream, and the filtering result forms the filter output signal sequence. S52. Perform envelope extraction processing on the filter output signal sequence. During the signal envelope extraction process, the envelope detection gain coefficient is used to uniformly adjust the gain of the envelope amplitude to finally form the envelope signal sequence. S53. Compare the envelope signal sequence with the current optimal decision threshold parameter to construct a weak signal feature sequence that represents the judgment result of whether each sampling point meets the characteristics of the AC conductor near-field leakage potential signal.

7. The voltage testing method for low-voltage testing equipment according to claim 1, characterized in that, S6 includes the following steps: S61. Input the enhanced weak signal feature sequence and the corresponding envelope signal sequence into the voltage detection decision module. In the voltage detection decision module, extract the time domain amplitude feature, frequency domain harmonic energy feature, and phase stability feature in sequence to construct the basis for judging the charged state: S62. Input the extracted time-domain amplitude features, frequency-domain harmonic energy features, and phase stability features into the Naive Bayes confidence inference model. The Naive Bayes confidence inference model outputs the confidence probability value of the current observed sample being in a charged state and the confidence probability value of it being in an uncharged state. S63. Based on the magnitude of the confidence probability value of the charged state, a three-classification decision is made to form the charged state decision result.

8. The voltage testing method for low-voltage testing equipment according to claim 1, characterized in that, The charged-up decision results include: When the confidence probability value is greater than or equal to the maximum judgment value, it is determined that the AC conductor is energized at that moment, and the tag "energized" is output. When the confidence probability value is less than the minimum judgment value, it is determined that the AC conductor is in an unenergized state at that moment, and the tag "unenergized" is output. When the confidence probability value is greater than or equal to the minimum judgment value and less than the maximum judgment value, the system determines that the current moment is in a suspected charged state and outputs a label indicating suspected charge. The system will then automatically extend the duration of the current signal acquisition window.

9. A low-voltage testing device for use in the steps of claims 1-8, characterized in that, The electrical testing equipment includes an electroscope (1), a telescopic rod (2), a remote control wrench (3), and a puncture structure; the electroscope (1) is used to test whether the wire is live, and its probe is connected to the puncture head (7) of the puncture structure through a copper wire (17); the telescopic rod (2) has a screw groove (15) at the top, which is used to connect and move the electroscope (1) and the remote control wrench (3); the remote control wrench (3) has a bolt (14) fixed at the bottom, a hexagonal groove (13) at the top, and three elastic retaining rings (16) vertically distributed on one side, which are used to drive the puncture structure; the puncture structure includes a... Includes elbow (4), puncture head (7) and matching components; elbow (4): has a wire groove (5) and a raised strip (6), and the puncture head (7) is slidably connected along the raised strip (6); puncture head (7): is arc-shaped, with a metal puncture needle (8) at the top and a cavity (9) at the bottom; a round plate (10) is placed in the cavity (9), and a threaded rod (11) is fixed at the bottom of the round plate (10), and the threaded rod (11) passes through the puncture head (7) and is rotatably connected to it; a hexagonal nut (12) is also fixed at the bottom of the puncture head (7), which is compatible with the hexagonal groove (13) of the remote control wrench (3).

Images