Method for detecting partial discharge of unmanned aerial vehicle

Abstract

The invention provides a method for detecting partial discharge of an unmanned aerial vehicle, and the method comprises the steps: collecting a partial discharge signal and motion parameters of the unmanned aerial vehicle, and correcting the real motion state of the unmanned aerial vehicle in real time through a preset unmanned aerial vehicle kinetic model; beam forming array parameters of the unmanned aerial vehicle are adjusted through a preset dynamic phase compensation weight model, and signal phase distortion in a real motion state is compensated; a cross-modal combined likelihood function is determined by fusing the ultrasonic time difference characteristics and the radio frequency pulse energy information, dynamic particle filtering is driven through the cross-modal combined likelihood function, iterative convergence is carried out in a three-dimensional space until optimal discharge point coordinates are determined, and a discharge positioning result is obtained; and determining and controlling an unmanned aerial vehicle inspection mode to carry out partial discharge detection according to the credibility index value of the discharge positioning result. According to the method, the positioning accuracy is improved, more accurate positioning is realized, an inspection strategy is optimized, and the positioning accuracy and robustness of the mobile platform in a complex electromagnetic environment are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of detecting partial discharge, and particularly to a method for detecting partial discharge by using an unmanned aerial vehicle (UAV). Background Art

[0002] In the field of partial discharge location of power equipment, in the face of the movement of UAVs, the prior art is difficult to accurately compensate for the errors caused by the movement, resulting in limited positioning accuracy. The existing methods cannot effectively fuse multi-sensor data and have insufficient anti-interference ability in complex electromagnetic environments. In the inspection of power equipment, the existing partial discharge location technology is affected by the movement of UAVs in complex electromagnetic environments, and there are problems such as low positioning accuracy and poor reliability. Summary of the Invention

[0003] The purpose of the present invention is to propose a method for detecting partial discharge by using an unmanned aerial vehicle, so as to solve the technical problem of the interference of UAV movement on partial discharge location.

[0004] On the one hand, a method for detecting partial discharge by using an unmanned aerial vehicle is provided, including:

[0005] Collecting partial discharge signals and UAV movement parameters, and real-time correcting the true movement state of the UAV through a preset UAV dynamics model;

[0006] Adjusting the beamforming array parameters of the UAV through a preset dynamic phase compensation weight model to compensate for the signal phase distortion of the true movement state;

[0007] Determining a cross-modal joint likelihood function by fusing ultrasonic time difference features and radio frequency pulse energy information, and driving a dynamic particle filter through the cross-modal joint likelihood function to iteratively converge in three-dimensional space until the optimal discharge point coordinates are determined to obtain a discharge location result;

[0008] Determining to control the UAV inspection mode for detecting partial discharge through the credibility index value of the discharge location result.

[0009] Preferably, the UAV dynamics model includes,

[0010]

[0011] Wherein, is the UAV position vector [x, y, z] T ; v is the UAV velocity vector; w p , w v are the position and velocity measurement noises; a is the UAV acceleration vector; is the rotation matrix from the body coordinate system to the geographic coordinate system.

[0012] Preferably, the dynamic phase compensation weight model includes

[0013]

[0014] where is the variance of signal strength; v(t) is the instantaneous velocity vector of the UAV, and its magnitude ||v(t)|| represents the severity of motion; Δt is the time difference; θ drone (t) is the current yaw angle of the UAV; θ sig is the estimated azimuth angle of the signal source; is the variance of angle measurement error; t is the current time; W(t) is the beamforming weight vector at the current time.

[0015] Preferably, it further includes compensating for the signal phase distortion of the true motion state according to the following formula

[0016]

[0017] where φ comp (t) is the compensated element phase; φ raw (t) is the original measured phase, which is calculated from the signals received by each element of the ultrasonic array; λ is the signal wavelength; d i is the position vector of the i-th element in the body coordinate system.

[0018] Preferably, it further includes constraining the true motion state through the following spatial domain filter

[0019]

[0020] where w is the beamforming weight vector; R nn is the noise covariance matrix; μ is the regularization parameter; a(θ) is the array steering vector.

[0021] Preferably, the cross-modal joint likelihood function includes

[0022]

[0023] where p(z|x) is the conditional probability density function of the observed data z given the particle state x; ∏ sensor is the product operation of the multi-sensor joint likelihood probability; σ s is the standard deviation of the sensor measurement noise; t pred is the signal arrival time / energy value predicted based on the particle position x; t meas is the signal arrival time / energy value actually measured by the sensor.

[0024] Preferably, the dynamic particle filter driven by the cross-modal joint likelihood function includes uniformly initializing a plurality of dynamic particles in the surrounding space, and dynamically adjusting the number of particles when the number of effective particles is less than a preset threshold.

[0025] Preferably, a plurality of dynamic particles are initialized according to the following formula:

[0026]

[0027] where is the weight of the i-th particle at time k; is the observation likelihood function, which is the conditional probability of the sensor data z given the particle state ; k is the conditional probability of; is the state transition probability, which describes the motion propagation model of the particle from to ; q(·) is the proposal distribution.

[0028] Preferably, it further includes determining a credibility index value according to the following formula:

[0029]

[0030] where N eff is the number of effective particles; Entropy(p(x)) is the state distribution entropy; CRLB is the Cramer-Rao lower bound; λ1, λ2, λ3 are weighting coefficients.

[0031] Preferably, the determination of controlling the UAV inspection mode for partial discharge detection includes, when the credibility index value is less than a preset credibility threshold, controlling the UAV to enable the three-axis gimbal lock to improve the hovering stability and switch to the fine scanning mode, and requesting neighboring UAVs to cooperate in observation.

[0032] In summary, implementing the embodiments of the present invention has the following beneficial effects:

[0033] The method for UAV detection of partial discharge provided by the present invention solves the problem of interference of UAV motion on partial discharge positioning through the UAV motion compensation positioning algorithm based on dynamic adaptive beamforming and multi-modal data fusion, and improves the positioning accuracy by using multi-modal data synchronous acquisition and spatio-temporal registration; adopts dynamic weight particle filter positioning and credibility feedback control to achieve more accurate positioning and optimize the inspection strategy, effectively improving the positioning accuracy and robustness of the mobile platform in a complex electromagnetic environment. Description of the Drawings

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, obtaining other drawings based on these drawings still belongs to the scope of the present invention.

[0035] Figure 1 This is a schematic diagram of the main process of a method for a drone to detect partial discharge in an embodiment of the present invention. Specific embodiments

[0036] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will further describe the present invention in detail with reference to the drawings.

[0037] As Figure 1 shown, this is a schematic diagram of an embodiment of a method for a drone to detect partial discharge provided by the present invention. In this embodiment, the method includes the following steps:

[0038] Step S1, collect partial discharge signals and drone motion parameters, and real-time correct the true motion state of the drone through a preset drone dynamics model; synchronously collect partial discharge signals and drone motion parameters through multi-modal sensors (ultrasonic array, UHF radio frequency, IMU, and RTK-GNSS), and use the IEEE 1588v2 protocol to achieve microsecond-level spatio-temporal alignment; decouple platform vibration and true motion state in real-time based on the drone dynamics model and Kalman filter.

[0039] One embodiment, multi-modal data synchronous acquisition: Configure 4 × ultrasonic arrays (sampling at 200 kHz) to detect ultrasonic signals generated by partial discharge, a dual-antenna UHF radio frequency sensor to receive radio frequency signals, a six-axis IMU (including gyroscope / accelerometer) to obtain the motion attitude information of the drone, and an RTK-GNSS positioning module (accuracy of ±1 cm) to determine the position of the drone. Use the IEEE1588v2 precision clock protocol for data synchronization, with a timestamp alignment accuracy of up to 10 μs level, and establish a conversion matrix between the drone body coordinate system and the geographic coordinate system.

[0040] In a specific embodiment, the drone dynamics model includes

[0041]

[0042] Wherein, is the drone position vector (unit: m), and the three-dimensional coordinates [x, y, z] in the geographic coordinate system T ; v is the drone velocity vector (unit: m / s); w p , wv are the position and velocity measurement noises (unit: m, m / s), which follow a zero-mean Gaussian distribution, and the covariance matrix is determined by the GNSS positioning accuracy (horizontal ±1 cm, vertical ±2 cm); a is the acceleration vector of the UAV (unit: m / s 2 ), which is calculated from the motor thrust model; is the rotation matrix from the body coordinate system to the geographic coordinate system, which is calculated by the Rodriguez formula from the Euler angles (roll, pitch, yaw); perform Kalman prediction-correction. In the prediction stage, the motion state is estimated based on the motor speed and the aerodynamic model, and in the correction stage, the IMU, GNSS, and visual odometer data are fused.

[0043] Step S2, adjust the beamforming array parameters of the UAV through a preset dynamic phase compensation weight model to compensate for the signal phase distortion in the real motion state; perform adaptive beamforming processing, calculate the weights of each element according to the dynamic phase compensation weight model, and compensate for the phase deviation caused by the vibration of the UAV.

[0044] In one embodiment, the dynamic phase compensation weight model includes

[0045]

[0046] where is the variance of the signal strength (unit: V 2 ), which reflects the signal-to-noise ratio of the ultrasonic / RF sensor at the current moment, and is calculated as the square of the standard deviation of the signal energy in the previous 1 ms through a sliding window; v(t) is the instantaneous velocity vector of the UAV (unit: m / s), which is the three-dimensional velocity component [v x , v y , v z obtained by fusing the IMU and GNSS; T , and its modulus ||v(t)|| represents the degree of motion intensity; Δt is the time difference (unit: s), which is the interval time between two adjacent beamforming calculations (default 5 ms); θ drone (t) is the current yaw angle of the UAV (unit: radian), which is obtained by integrating the gyroscope, and the defined range is [-π, π); θ sig is the estimated azimuth angle of the signal source (unit: radian), which is the azimuth of the discharge point predicted based on the particle filter result at the previous moment; is the variance of the angle measurement error (unit: radian²), which is determined by the array aperture and the signal wavelength.

[0047] Furthermore, compensate for the signal phase distortion in the real motion state according to the following formula

[0048]

[0049] where φcomp (t) is the compensated element phase (unit: radian), used to eliminate the phase shift caused by movement; φ raw (t) is the original measured phase (unit: radian), calculated from the signals received by each element of the ultrasonic array; λ is the signal wavelength (unit: m); d i is the position vector of the i-th element in the body coordinate system (unit: m), and its value is determined by the array geometry layout.

[0050] It should be noted that it also includes establishing a spatial filter, and the formula is:

[0051]

[0052] In the formula, w is the beamforming weight vector (complex form), and its dimension is equal to the number of array elements (for example, a 4× ultrasonic array is 4-dimensional); R nn is the noise covariance matrix (unit: V 2 ), estimated from the received data during the no-signal period; μ is the regularization parameter (dimensionless), used to balance noise suppression and main lobe gain, and the empirical value is taken as 0.1 - 1.0; a(θ) is the array steering vector (complex form); the constraint condition includes the motion state covariance matrix.

[0053] Step S3, determine the cross-modal joint likelihood function by fusing ultrasonic time difference features and RF pulse energy information, and drive the dynamic particle filter through the cross-modal joint likelihood function to iteratively converge in three-dimensional space until the optimal discharge point coordinates are determined to obtain the discharge location result; construct a joint feature space, including ultrasonic time difference features RF pulse envelope feature E RF(t) and environmental parameters (temperature, humidity, wind speed); perform hierarchical association matching, calculate the cross-modal similarity matrix, and use the Hungarian algorithm to solve the optimal association; dynamic weight particle filter positioning.

[0054] In one embodiment, 5000 particles are uniformly initialized in the 5m space around the device, and each particle carries the state quantity (x, y, z, σ, q); perform importance sampling: 3

[0055]

[0056] In the formula, is the weight of the i-th particle at time k (dimensionless), which represents the probability density of the particle after normalization; is the observation likelihood function, which is the conditional probability of the sensor data z given the particle state ; k is the state transition probability, which describes the particle from to​​ The motion propagation model (such as Gaussian random walk); q(·) is the proposal distribution. Here, the prior distribution is selected, that is, q = p(x k |x k-1 );

[0057] Among them, the observation likelihood function is:

[0058]

[0059] In the formula, p(z|x) is the conditional probability density function of the observed data z when the particle state is x; ∏ sensor is the multiplication operation of the joint likelihood probability of multiple sensors; σ s is the standard deviation of the sensor measurement noise (unit: seconds / meter / volt, etc., depending on the sensor type); t pred is the signal arrival time / energy value predicted based on the particle position x (theoretical value); t meas is the signal arrival time / energy value actually measured by the sensor (measured value)

[0060] The dynamic particle filtering driven by the cross-modal joint likelihood function includes uniformly initializing multiple dynamic particles in the surrounding space, and dynamically adjusting the number of particles when the number of effective particles is less than a preset threshold; using the residual resampling method, when the number of effective particles N eff < 0.3N, the number of particles is dynamically adjusted.

[0061] Step S4, determine to control the UAV inspection mode for partial discharge detection through the credibility index value of the discharge positioning result. The UAV inspection mode is controlled in real time through the positioning credibility index Q value to achieve the closed-loop optimization of motion compensation-signal processing-positioning decision-making, effectively improving the positioning accuracy and robustness of the mobile platform in a complex electromagnetic environment.

[0062] In one embodiment, the credibility index value is determined according to the following formula

[0063]

[0064] Among them, N eff is the number of effective particles (dimensionless), reflecting the degree of particle degradation; Entropy(p(x)) is the state distribution entropy (unit: nat), measuring the uncertainty of the positioning result; CRLB is the Cramér-Rao lower bound (unit: m 2 ), the theoretical minimum positioning variance, calculated from the sensor array geometry and the signal-to-noise ratio; λ1, λ2, λ3 are weighting coefficients (dimensionless).

[0065] When the credibility index value is less than the preset credibility threshold, the drone is controlled to enable the three-axis gimbal lock to improve the hovering stability and switch to the fine scanning mode, and request neighboring drones to cooperate in observation. When Q < Q th , the drone improves the hovering stability (enables the three-axis gimbal lock), switches to the fine scanning mode (spiral progressive trajectory), and requests neighboring drones to cooperate in observation.

[0066] In summary, implementing the embodiments of the present invention has the following beneficial effects:

[0067] The method for detecting partial discharge by drones provided by the present invention solves the problem of interference of drone movement on partial discharge positioning through the drone motion compensation and positioning algorithm based on dynamic adaptive beamforming and multi-modal data fusion (DA-MDF), and improves the positioning accuracy by using multi-modal data synchronous acquisition and spatio-temporal registration; adopts dynamic weight particle filter positioning and credibility feedback control to achieve more accurate positioning and optimize the inspection strategy, effectively improving the positioning accuracy and robustness of the mobile platform in complex electromagnetic environments.

[0068] The above-disclosed are only the preferred embodiments of the present invention. Of course, the scope of the rights of the present invention cannot be limited thereby. Therefore, equivalent changes made according to the claims of the present invention still fall within the scope covered by the present invention.

Claims

1. A method for detecting partial discharge in an unmanned aerial vehicle, characterized in that, Including: Collecting partial discharge signals and UAV motion parameters, and real-time correcting the real motion state of the UAV through a preset UAV dynamics model; Adjusting the beamforming array parameters of the UAV through a preset dynamic phase compensation weight model to compensate for the signal phase distortion of the real motion state; Determining a cross-modal joint likelihood function by fusing ultrasonic time difference features and RF pulse energy information, and driving a dynamic particle filter through the cross-modal joint likelihood function to iteratively converge in three-dimensional space until the optimal discharge point coordinates are determined to obtain a discharge location result; Determining to control the UAV inspection mode to detect partial discharge through the credibility index value of the discharge location result.

2. The method according to claim 1, wherein The UAV dynamics model includes wherein, is the UAV position vector [x, y, z] T ; v is the UAV velocity vector; w p , w v are the position and velocity measurement noises; a is the UAV acceleration vector; is the rotation matrix from the body coordinate system to the geographic coordinate system.

3. The method according to claim 2, wherein The dynamic phase compensation weight model includes Among them, is the variance of signal strength; v(t) is the instantaneous velocity vector of the UAV, and its magnitude ||v(t)|| represents the degree of motion intensity; Δt is the time difference; θ drone (t) is the current yaw angle of the UAV; θ sig is the estimated azimuth angle of the signal source; is the variance of angle measurement error; t is the current time; W(t) is the beamforming weight vector at the current time.

4. The method according to claim 3, characterized in that, It also includes compensating for the signal phase distortion of the real motion state according to the following formula Among them, φ comp (t) is the compensated element phase; φ raw (t) is the original measured phase, which is calculated from the signals received by each element of the ultrasonic array; λ is the signal wavelength; d i is the position vector of the i-th element in the body coordinate system.

5. The method according to claim 4, characterized in that, It also includes constraining the real motion state through the following spatial domain filter where \(w\) is the beamforming weight vector; \(R\) nn is the noise covariance matrix; \(\mu\) is the regularization parameter; \(a(\theta)\) is the array steering vector.

6. The method according to claim 5, characterized in that, The cross-modal joint likelihood function includes Among them, p(z|x) is the conditional probability density function of the observed data z given the particle state x; ∏ sensor is the multiplication operation of the joint likelihood probability of multiple sensors; σ s is the standard deviation of the sensor measurement noise; t pred is the time-of-arrival / energy value of the signal predicted based on the particle position x; t meas is the time-of-arrival / energy value of the signal actually measured by the sensor.

7. The method according to claim 6, wherein The driving of the dynamic particle filter through the cross-modal joint likelihood function includes uniformly initializing a plurality of dynamic particles in the surrounding space, and dynamically adjusting the number of particles when the number of effective particles is less than a preset threshold.

8. The method according to claim 7, wherein Initializing a plurality of dynamic particles according to the following formula wherein, is the weight of the i-th particle at time k; is the observation likelihood function, which is the conditional probability of the sensor data z given the particle state ; k is; is the state transition probability, which describes the motion propagation model of the particle from to ; q(·) is the proposal distribution.

9. The method according to claim 8, wherein It also includes determining the credibility index value according to the following formula Among them, N eff is the number of effective particles; Entropy(p(x)) is the state distribution entropy; CRLB is the Cramér-Rao lower bound; λ1, λ2, λ3 are the weighting coefficients.

10. The method according to claim 9, wherein The determination of controlling the UAV inspection mode to detect partial discharge includes, when the credibility index value is less than a preset credibility threshold, controlling the UAV to enable the three-axis gimbal lock to improve the hovering stability and switch to the fine scanning mode, and requesting neighboring UAVs to cooperate in observation.

Images