Partial discharge detection method and system for unmanned aerial vehicle

Abstract

The invention provides a partial discharge detection method and system for an unmanned aerial vehicle, and the method comprises the steps: collecting infrared thermal image data through a preset sensor; background interference in thermal imaging is corrected by fusing the environment temperature, the space smooth temperature field and the surface curvature information, and the temperature contrast of the weak discharge signal is improved according to the corrected background interference; selective filtering is carried out by repairing the direction of the wavelet, multi-scale temperature gradient characteristics are extracted, and edge and texture information of the partial discharge area is enhanced according to the multi-scale temperature gradient characteristics; the energy ratio of the discharge characteristic is enhanced through nonlinear mapping, and a false signal introduced by hovering vibration of the unmanned aerial vehicle is suppressed; and based on the physical correlation between the infrared radiation energy and the discharge capacity, non-contact discharge capacity quantitative evaluation is realized through an integral temperature rise area, and a partial discharge detection result is determined. According to the method, the problem of accurate identification of weak partial discharge in a complex electromagnetic environment is solved, the problem of low positioning accuracy is solved, and the detection efficiency is improved.

Description

Technical Field

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

[0002] In the field of partial discharge detection, the existing technologies have problems such as low accuracy and difficulty in recognition when detecting weak partial discharges in complex electromagnetic environments. Currently, some detection methods may not be able to effectively remove background noise interference, resulting in inaccurate detection results; some methods have low positioning accuracy and are difficult to meet actual requirements. Summary of the Invention

[0003] The object of the present invention is to provide a method and system for partial discharge detection of an unmanned aerial vehicle, so as to solve the technical problem of improving the accuracy and precision of partial discharge detection.

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

[0005] Collecting infrared thermal image data of different frequency bands through a preset sensor; wherein, the sensor at least includes an ultraviolet pulse sensor and a visible light camera;

[0006] Correcting background interference in the thermal imaging by fusing the environmental temperature, spatial smoothed temperature field and surface curvature information in the infrared thermal image data, and improving the temperature contrast of weak discharge signals according to the corrected background interference;

[0007] Performing selective filtering through the direction of the wavelet transform to extract multi-scale temperature gradient features, and enhancing the edge and texture information of the partial discharge region according to the multi-scale temperature gradient features;

[0008] Enhancing the energy ratio of discharge features through non-linear mapping, and suppressing false signals introduced by the hovering vibration of the unmanned aerial vehicle;

[0009] Based on the physical relationship between infrared radiation energy and discharge quantity, realizing non-contact discharge quantity quantization evaluation through integrating the temperature rise region, and determining the partial discharge detection result according to the evaluation result.

[0010] Preferably, it further includes correcting the background interference in the thermal imaging according to the following formula:

[0011]

[0012] wherein, T c (x, y, t) is the compensated target temperature field; α, β, γ are adaptive weight coefficients, dynamically adjusted according to the environment; T amb (t) is the environmental reference temperature; G σis a Gaussian convolution kernel; * is the convolution operation; I(x, y, t) is the temperature field collected by the original infrared thermal imager; dA is the spatial integration microelement; is the Laplacian operator; T surf is the temperature field on the surface of the device.

[0013] Preferably, it further includes selectively filtering according to the direction of repairing the wavelet by the following formula:

[0014]

[0015] where W s,θ is the wavelet coefficient; is the complex wavelet basis function; s is the scale parameter; θ is the direction parameter; w0 is the center frequency; m, n are the spatial translation parameters.

[0016] Preferably, it further includes enhancing the energy proportion of the discharge characteristics by the following formula:

[0017] Enh(F) = tanh(λ × |F| k ) × sign(F) + μ × F ⊙ M

[0018] where F is the input feature tensor; λ is the dynamic gain factor; k is the nonlinear exponent; tanh(·) is the hyperbolic tangent function; μ is the motion artifact suppression coefficient; ⊙ is the Hadamard product; M is the binary mask matrix.

[0019] Preferably, the non-contact discharge quantity quantization evaluation by integrating the temperature rise region includes,

[0020] Construct a deep residual spatio-temporal network to jointly optimize the discharge recognition, positioning accuracy and feature robustness, and constrain the overfitting of a single task;

[0021] Establish a discharge quantity inversion model and calculate the discharge quantity according to the discharge quantity inversion model.

[0022] Preferably, the deep residual spatio-temporal network includes,

[0023] y = Softmax(ResBlock5(Conv3D(LSTM(F))))

[0024] where F is the input feature tensor; LSTM is the long short-term memory network; Conv3D is the three-dimensional convolution operation; ResBlock5 is the fifth residual block, and the residual block is the basic unit in the deep residual network; Softmax is an activation function that can be used to judge the possibility that the input data belongs to different categories.

[0025] Preferably, the deep residual spatio-temporal network further includes a multi-task loss function:

[0026] L = αL cls + βL reg + γL recon

[0027] where L cls is the classification loss; L reg is the regression loss; L recon is the reconstruction loss; α, β, and γ are loss weight coefficients.

[0028] Preferably, the discharge amount inversion model includes

[0029]

[0030] where Q is the equivalent discharge amount; σ is the Stefan-Boltzmann constant; T(x, y) is the local temperature field; T0 is the environmental reference temperature; ε is the material emissivity; η is the heat conduction efficiency coefficient; τ is the discharge duration.

[0031] On the other hand, a system for partial discharge detection of an unmanned aerial vehicle is also provided, which is used to implement the method for partial discharge detection of an unmanned aerial vehicle, and includes

[0032] a data acquisition module, which is used to collect infrared thermal image data of different frequency bands through a preset sensor; where the sensor at least includes an ultraviolet pulse sensor and a visible light camera;

[0033] a thermal imaging correction module, which is used to correct the background interference in the thermal imaging by fusing the environmental temperature, spatial smooth temperature field, and surface curvature information in the infrared thermal image data, and improve the temperature contrast of weak discharge signals according to the corrected background interference;

[0034] a filtering module, which is used to perform selective filtering through the direction of the repaired wavelet and extract multi-scale temperature gradient features, and enhance the edge and texture information of the partial discharge area according to the multi-scale temperature gradient features;

[0035] an enhancement module, which is used to enhance the energy ratio of discharge features through non-linear mapping and suppress the false signals introduced by the hovering vibration of the unmanned aerial vehicle;

[0036] an evaluation module, which is used to perform non-contact discharge amount quantization evaluation through integrating the temperature rise area based on the physical association between infrared radiation energy and discharge amount, and determine the partial discharge detection result according to the evaluation result.

[0037] In summary, implementing the embodiments of the present invention has the following beneficial effects:

[0038] The method and system for partial discharge detection of drones provided by the present invention introduce a spatio-temporal domain joint analysis framework, adaptive wavelet threshold denoising, and a dynamic background temperature compensation mechanism, overcoming the problem of difficult accurate identification of weak partial discharges in complex electromagnetic environments; by constructing a deep residual spatio-temporal network and designing a multi-task loss function, sub-pixel level positioning is achieved, overcoming the problem of low positioning accuracy; a discharge quantity inversion model is established to more accurately obtain discharge quantity information; an edge computing module is developed to achieve real-time mapping, improving the detection efficiency. Description of the Drawings

[0039] In order 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 use in 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.

[0040] Figure 1 It is a schematic diagram of the main process of a method for partial discharge detection of drones in an embodiment of the present invention. Detailed Embodiments

[0041] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the drawings.

[0042] As Figure 1 shown, it is a schematic diagram of an embodiment of a method for partial discharge detection of drones provided by the present invention. In this embodiment, the method includes the following steps:

[0043] Step S1, collecting infrared thermal image data of different frequency bands through a preset sensor; wherein, the sensor at least includes an ultraviolet pulse sensor and a visible light camera; it can be understood that in this example, the drone is equipped with a dual-band infrared thermal imager to obtain infrared thermal image data of different frequency bands, reflecting the thermal characteristics generated by partial discharges from different angles. Integrating an ultraviolet pulse sensor and a visible light camera, the ultraviolet pulse sensor can detect the ultraviolet pulse signal generated by partial discharges, and the visible light camera is used to provide visual information of the detection site to assist in positioning and analysis. Designing a four-axis drone adaptive hovering system to ensure that the drone hovers stably during the detection process and guarantee the accuracy of the detection data. Developing a multi-source data synchronous acquisition card to achieve precise synchronous acquisition of data from multiple sensors, making subsequent data analysis more reliable.

[0044] Step S2, correct the background interference in the thermal imaging by fusing the ambient temperature, the spatial smoothed temperature field, and the surface curvature information in the infrared thermal image data, and improve the temperature contrast of the weak discharge signal according to the corrected background interference; that is, perform dynamic background temperature field modeling, and dynamically correct the background interference in the thermal imaging by establishing an ambient temperature compensation equation to improve the temperature contrast of the weak discharge signal.

[0045] In one embodiment, correct the background interference in the thermal imaging according to the following formula:

[0046]

[0047] where, T c (x, y, t) is the compensated target temperature field (unit: °C); α, β, γ are adaptive weight coefficients, dynamically adjusted according to the environment; T amb (t) is the ambient reference temperature (measured value varying with time, unit: °C); G σ is the Gaussian convolution kernel (with a standard deviation of σ, for spatial smoothing); * is the convolution operation; I(x, y, t) is the temperature field collected by the original infrared thermal imager (unit: °C); dA is the spatial integration microelement (unit: m2); is the Laplace operator (characterizing the curvature change of the temperature field); T surf is the device surface temperature field (unit: °C).

[0048] Step S3, perform selective filtering through the direction of the wavelet transform and extract multi-scale temperature gradient features, and enhance the edge and texture information of the partial discharge region according to the multi-scale temperature gradient features; that is, through the direction selective filtering of the complex wavelet transform, extract multi-scale temperature gradient features in the time-frequency domain joint analysis to enhance the edge and texture information of the partial discharge region.

[0049] In one embodiment, it further includes performing selective filtering through the direction of the wavelet transform according to the following formula:

[0050]

[0051] where, W s,θ is the wavelet coefficient (the decomposition result at scale s and direction θ); is the complex wavelet basis function (with direction selectivity); s is the scale parameter (controlling the frequency resolution); θ is the direction parameter; w0 is the center frequency (inversely proportional to the scale s); m, n are the spatial translation parameters (unit: pixel).

[0052] Step S4, enhance the energy proportion of the discharge feature through non-linear mapping and suppress the false signals introduced by the hovering vibration of the drone; that is, enhance the energy proportion of the discharge feature through non-linear mapping, and at the same time use the masking mechanism to suppress the false signals introduced by the hovering vibration of the drone.

[0053] In one embodiment, enhance the energy proportion of the discharge feature according to the following formula:

[0054] Enh(F) = tanh(λ × |F| k ) × sigη(F) + μ × F ⊙ M

[0055] where F is the input feature tensor (including three-dimensional time-space-frequency information); λ is the dynamic gain factor (adaptively adjusted according to the signal-to-noise ratio); k is the non-linear exponent (the value range is 1.5 - 2.5, controlling the enhancement intensity); tanh(·) is the hyperbolic tangent function (compressing large values and retaining the sign); μ is the motion artifact suppression coefficient (0 - 1, suppressing the vibration noise of the drone); ⊙ is the Hadamard product (element-wise multiplication); M is the binary mask matrix (marking the motion artifact area, enhancing the discharge feature, and improving the detection accuracy).

[0056] Step S5, based on the physical correlation between the infrared radiation energy and the discharge amount, realize non-contact discharge amount quantization evaluation through integrating the temperature rise area, and determine the partial discharge detection result according to the evaluation result. That is, verify through the three-dimensional test field and output the GIS real-time mapping by the edge computing module to form a closed-loop detection system from data acquisition, feature fusion to intelligent diagnosis.

[0057] In one embodiment, the realization of non-contact discharge amount quantization evaluation through integrating the temperature rise area includes constructing a deep residual spatio-temporal network, jointly optimizing the discharge recognition, positioning accuracy and feature robustness, and constraining the overfitting of a single task; establishing a discharge amount inversion model, and calculating the discharge amount according to the discharge amount inversion model. Specifically, the deep residual spatio-temporal network includes,

[0058] y = Softmax(ResBlock5(Conv3D(LSTM(F))))

[0059] where F is the input feature tensor; LSTM is the long short-term memory network; Conv3D is the three-dimensional convolution operation; ResBlock5 is the fifth residual block, and the residual block is the basic unit in the deep residual network; Softmax is an activation function, which can be used to judge the possibility that the input data belongs to different categories.

[0060] The deep residual spatio-temporal network also includes designing a multi-task loss function to achieve sub-pixel level positioning, and the multi-task loss function:

[0061] L = αLcls +βL reg +γL recon

[0062] where L cls is the classification loss (cross entropy, identifying the discharge type); L reg is the regression loss (mean square error, predicting the positioning coordinates); L recon is the reconstruction loss (autoencoder, ensuring feature consistency); α, β, and γ are loss weight coefficients (optimized through grid search).

[0063] Jointly optimize the discharge identification, positioning accuracy, and feature robustness to avoid overfitting in a single task. The discharge amount inversion model includes

[0064]

[0065] where Q is the equivalent discharge amount (unit: pC); σ is the Stefan–Boltzmann constant (5.67×10 -8 W·m -2 ·K -4 ⁻⁸); T(x, y) is the local temperature field (unit: K); T0 is the ambient reference temperature (unit: K); ε is the material emissivity (0.85–0.95, related to the surface material of the device); η is the heat conduction efficiency coefficient (0.6–0.9, determined by the device structure); τ is the discharge duration (unit: s).

[0066] Specific embodiments include setting up a three-dimensional test field (including 10 typical discharge models), optimizing the network parameters using a transfer learning strategy, developing an edge computing module (with a processing delay <200 ms) to achieve real-time mapping of the GIS system (refresh rate 30 Hz), and continuously optimizing the performance of the detection system. First, multi-source data synchronization is collected through a drone platform equipped with a dual-band infrared thermal imager, an ultraviolet sensor, and an adaptive hovering system. A dynamic background temperature compensation model is constructed using a Gaussian convolution kernel and a Laplace operator. Subsequently, a direction-selective dual-tree complex wavelet transform is used to extract the spatio-temporal-frequency three-dimensional feature tensor. A nonlinear enhancement function is used to suppress motion artifacts and enhance the discharge features. Then, a deep residual spatio-temporal network is used for multi-task learning to achieve sub-pixel-level positioning and discharge amount inversion. Finally, the GIS real-time mapping is verified through a three-dimensional test field and output by the edge computing module, forming a closed-loop detection system from data collection, feature fusion to intelligent diagnosis.

[0067] An embodiment of the present invention further provides a system for partial discharge detection of a drone, for implementing the method for partial discharge detection of a drone, including

[0068] A data acquisition module for collecting infrared thermal image data of different frequency bands through a preset sensor; wherein, the sensor at least includes an ultraviolet pulse sensor and a visible light camera;

[0069] A thermal imaging correction module for correcting background interference in thermal imaging by fusing ambient temperature, spatial smooth temperature field, and surface curvature information in the infrared thermal image data, and improving the temperature contrast of weak discharge signals according to the corrected background interference;

[0070] A filtering module for selectively filtering through the direction of wavelet repair and extracting multi-scale temperature gradient features, and enhancing the edge and texture information of the partial discharge area according to the multi-scale temperature gradient features;

[0071] An enhancement module for enhancing the energy proportion of discharge features through non-linear mapping and suppressing false signals introduced by the hovering vibration of the drone;

[0072] An evaluation module for realizing non-contact discharge quantity quantization evaluation through integrating the temperature rise area based on the physical association between infrared radiation energy and discharge quantity, and determining the partial discharge detection result according to the evaluation result.

[0073] It should be noted that the system described in the above embodiment corresponds to the method described in the above embodiment. Therefore, the parts not detailed in the system described in the above embodiment can be obtained by referring to the content of the method described in the above embodiment, and will not be elaborated here.

[0074] In summary, implementing the embodiments of the present invention has the following beneficial effects:

[0075] The method and system for partial discharge detection for drones provided by the present invention introduce a spatio-temporal domain joint analysis framework, adaptive wavelet threshold denoising, and a dynamic background temperature compensation mechanism, overcoming the problem of difficult accurate identification of weak partial discharges in complex electromagnetic environments; by constructing a deep residual spatio-temporal network and designing a multi-task loss function, sub-pixel level positioning is achieved, overcoming the problem of low positioning accuracy; a discharge quantity inversion model is established to more accurately obtain discharge quantity information; an edge computing module is developed to achieve real-time mapping, improving the detection efficiency.

[0076] 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 partial discharge detection of an unmanned aerial vehicle, characterized in that: include: Collect infrared thermal imaging data of different frequency bands through preset sensors; wherein the sensors at least include ultraviolet pulse sensors and visible light cameras; Correcting background interference in thermal imaging by fusing ambient temperature, spatially smoothed temperature field and surface curvature information in the infrared thermal imaging data, and improving temperature contrast of weak discharge signals according to the corrected background interference; Selective filtering is performed by repairing the direction of the wavelet and extracting multi-scale temperature gradient features, and the edge and texture information of the partial discharge area is enhanced according to the multi-scale temperature gradient features; The energy proportion of discharge features is enhanced through nonlinear mapping, and false signals introduced by the hovering vibration of the UAV are suppressed; Based on the physical correlation between infrared radiation energy and discharge amount, non-contact discharge amount quantitative evaluation is achieved by integrating the temperature rise area, and the partial discharge detection result is determined according to the evaluation result.

2. The method according to claim 1, characterized in that It also includes the correction of background interference in thermal imaging according to the following formula: Among them, T c (x, y, t) is the target temperature field after compensation; α, β, γ are adaptive weight coefficients, which are dynamically adjusted according to the environment; T amb (t) is the ambient reference temperature; G σ is the Gaussian convolution kernel; * is the convolution operation; I(x, y, t) is the temperature field collected by the original infrared thermal imager; dA is the spatial integral differential element; is the Laplace operator; T surf is the surface temperature field of the equipment.

3. The method according to claim 2, characterized in that It also includes selective filtering by repairing the direction of the wavelet according to the following formula: Among them, W s,θ is the wavelet coefficient; is the complex wavelet basis function; s is the scale parameter; θ is the direction parameter; w0 is the center frequency; m, n are the spatial translation parameters.

4. The method according to claim 3, characterized in that It also includes the energy proportion of enhancing the discharge feature according to the following formula: Enh(F)=tanh(λ×|F| k )×sign(F)+μ×F⊙M Where F is the input feature tensor; λ is the dynamic gain factor; k is the nonlinear exponent; tanh(·) is the hyperbolic tangent function; μ is the motion artifact suppression coefficient; ⊙ is the Hadamard product; and M is the binary mask matrix.

5. The method according to claim 4, characterized in that The quantitative evaluation of non-contact discharge quantity by integrating the temperature rise area includes: Construct a deep residual spatiotemporal network to jointly optimize discharge recognition, positioning accuracy, and feature robustness, and constrain single task overfitting; A discharge capacity inversion model is established, and the discharge capacity is calculated based on the discharge capacity inversion model.

6. The method according to claim 5, characterized in that The deep residual spatiotemporal network includes, y=Softmax(ResBlock5(Conv3D(LSTM(F)))) Among them, F is the input feature tensor; LSTM is the long short-term memory network; Conv3D is the three-dimensional convolution operation; ResBlock5 is the fifth residual block, which is the basic building block in the deep residual network; Softmax is the activation function, which can be used to judge the possibility that the input data belongs to different categories.

7. The method according to claim 6, characterized in that The deep residual spatiotemporal network also includes a multi-task loss function: L=αL cls +βL reg +γL recon Among them, L cls is the classification loss; L reg is the regression loss; L recon is the reconstruction loss; α, β, γ are the loss weight coefficients.

8. The method according to claim 7, characterized in that The discharge inversion model includes: Among them, Q is the equivalent discharge amount; σ is the Stefan-Boltzmann constant; T(x, y) is the local temperature field; T0 is the ambient reference temperature; ε is the material emissivity; η is the thermal conductivity efficiency coefficient; τ is the discharge duration.

9. A system for partial discharge detection of unmanned aerial vehicles, used to implement the method according to any one of claims 1 to 8, characterized in that: include, A data acquisition module, used to collect infrared thermal imaging data of different frequency bands through a preset sensor; wherein the sensor includes at least an ultraviolet pulse sensor and a visible light camera; A thermal imaging correction module, used to correct background interference in thermal imaging by fusing ambient temperature, spatially smoothed temperature field and surface curvature information in the infrared thermal imaging data, and to improve the temperature contrast of weak discharge signals according to the corrected background interference; A filtering module is used to perform selective filtering and extract multi-scale temperature gradient features by repairing the direction of the wavelet, and enhance the edge and texture information of the local discharge area according to the multi-scale temperature gradient features; The enhancement module is used to enhance the energy proportion of the discharge characteristics through nonlinear mapping and suppress the false signals introduced by the hovering vibration of the drone; The evaluation module is used to realize non-contact quantitative evaluation of discharge amount by integrating temperature rise area based on the physical correlation between infrared radiation energy and discharge amount, and determine the partial discharge detection result according to the evaluation result.

Images