An on-line monitoring system for transformer health status

Abstract

The present application relates to a kind of transformer health state on-line monitoring system, it is related to transformer monitoring technical field, including data acquisition unit, modeling analysis unit, mechanical health assessment unit and defect positioning unit;The present application is by constructing three-dimensional finite element model and establishing multidimensional structure vibration transfer function matrix, using inverse matrix inversion algorithm, the low-frequency mechanical vibration signal of box surface is directly perspective mapping as the real-time attenuation of internal core clamping force and winding pre-tightening force, only using external conventional sensor can realize the quantitative perception to deep-level mechanical looseness and fatigue state, greatly improve the early active early warning capability of equipment mechanical defect, simultaneously by constructing three-dimensional non-uniform acoustic impedance model and introducing ray tracing algorithm, real inverse restores the refraction and diffraction physical path of ultrasonic wave in oil, paperboard, metal propagation, accurately locks the exact spatial coordinates of micro dendritic discharge or local insulation deterioration.

Description

Technical Field

[0001] This invention relates to the field of transformer monitoring technology, and in particular to an online monitoring system for the health status of transformers. Background Technology

[0002] As a core component of the power grid, the operational health of transformers directly impacts the grid's safety and stability. Current online transformer monitoring primarily relies on methods such as dissolved gas analysis in oil and high-frequency / ultra-high-frequency partial discharge monitoring. However, these mainstream monitoring methods are highly dependent on electrical and chemical characteristics, exhibiting the following significant drawbacks: Under long-term alternating thermal stress and electromagnetic forces, the core components of a transformer (such as the core clamping device and winding pads) are prone to microscopic deformation or loosening due to the attenuation of clamping force. Existing vibration monitoring methods based on the external tank wall are extremely difficult to accurately quantify the deep mechanical state of the transformer because the signal needs to pass through the complex attenuation of transformer oil and insulation structure.

[0003] Furthermore, when minute dendritic discharges or localized insulation degradation occur inside a transformer, the resulting ultrasonic or electromagnetic signals encounter severe obstruction from the windings and refraction and reflection from the complex three-dimensional metal structure such as the core during their outward propagation. This leads to significant multipath effect errors in the inversion calculations of traditional external sensor arrays based on time difference of arrival, making it difficult to accurately pinpoint the exact physical location of the defect. To address the aforementioned technical deficiencies, a solution is proposed. Summary of the Invention

[0004] The purpose of this invention is to: construct a three-dimensional finite element model and establish a multi-dimensional structural vibration transfer function matrix, and use an inverse matrix inversion algorithm to directly map the low-frequency mechanical vibration signal on the surface of the housing into the real-time attenuation of the internal iron core clamping force and winding preload. This allows for the quantitative perception of deep-seated mechanical loosening and fatigue states using only external conventional sensors, greatly improving the early proactive warning capability of equipment mechanical defects. At the same time, by constructing a three-dimensional non-uniform acoustic impedance model and introducing a ray tracing algorithm, the physical path of refraction and diffraction of ultrasonic waves propagating between oil, cardboard, and metal is realistically reversed, accurately pinpointing the precise spatial coordinates of micro-dendritic discharges or local insulation degradation.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: an online monitoring system for the health status of a transformer, comprising a data acquisition unit, a modeling and analysis unit, a mechanical health assessment unit, and a defect location unit, wherein: The data acquisition unit utilizes a UHF sensor array, an acoustic emission sensor array, and a low-frequency vibration acceleration sensor deployed outside the transformer to synchronously acquire electromagnetic wave signals, ultrasonic signals, and low-frequency mechanical vibration signals of partial discharge on the same time reference, and sends them to the mechanical health assessment unit and the defect location unit. The modeling and analysis unit is used to obtain the three-dimensional mechanical structure inside the transformer to construct a finite element analysis model. It divides the corresponding dielectric mesh region according to the different acoustic impedances and sound velocities of the transformer oil, winding paperboard, iron core metal and tank wall. Through frequency sweep excitation simulation, it establishes a multi-dimensional structural vibration transfer function matrix from each core component inside the transformer to each vibration measurement point on the outer wall of the tank. The mechanical health assessment unit is used to obtain low-frequency mechanical vibration signals and input them into the inverse matrix of the transfer function matrix. Combined with internal modal parameters, it inversely calculates the real-time attenuation of the internal core clamping force and winding preload, and obtains the quantitative level of the deep mechanical health status. The defect location unit is used to construct a three-dimensional non-uniform acoustic impedance model that includes transmission, refraction and diffraction attenuation parameters. In the three-dimensional non-uniform acoustic impedance model, the actual path of the sound wave propagating from each acoustic emission sensor position to the interior is simulated using a ray tracing algorithm to obtain the simulated acoustic path. The isochronous surface intersection region of each simulated acoustic path in three-dimensional space is calculated, and the three-dimensional coordinate point with the highest intersection probability density and conforming to the characteristics of transformer electric field concentration is locked as the defect location.

[0006] Furthermore, the specific process of constructing the finite element analysis model and dividing the medium mesh region is as follows: S11. Obtain the original three-dimensional computer-aided design model of the transformer. The original three-dimensional computer-aided design model includes the dimensional and distribution data of the core, windings, insulating pressure plate, transformer oil space, and external tank wall. The original three-dimensional computer-aided design model is simplified by fluid-structure interaction dynamics based on acoustic-mechanical conduction characteristics to obtain the finite element analysis model, as follows: Miniature fasteners, small chamfers in oil passages, and external non-load-bearing accessories whose impact on electromagnetic force transmission and sound wave refraction is less than a preset threshold are ignored. The multilayer laminated silicon-through-silicon core is equivalent to a solid anisotropic continuous medium topology with orthogonal anisotropic elastic modulus. The winding is equivalent to a homogenized viscoelastic composite entity with both mass effect and stiffness characteristics according to the volume average method. S12. In the finite element analysis model, based on the actual physical materials of each component inside the transformer, their density, elastic modulus, Poisson's ratio, and the unique acoustic impedance and sound velocity parameters of each medium are defined to construct a spatially non-uniform heterogeneous property field, as detailed below: Define the corresponding longitudinal wave velocity for the iron core metal region. transverse wave speed of sound and material density Its acoustic impedance is ; Define the corresponding equivalent composite velocity for the winding paperboard region. With anisotropic damping coefficient; For the transformer oil region, define the corresponding hydrostatic pressure, bulk elastic modulus, and velocity of sound. and fluid density Its fluid acoustic impedance ; For the tank wall region, the standard elastic modulus and isotropic sound velocity characteristics of the steel are defined; S13. A multi-scale, two-layer adaptive meshing strategy is adopted. A structural dynamics mesh layer is created for low-frequency mechanical vibration, and a high-frequency acoustic impedance mesh layer is created for high-frequency acoustic emission localization. Furthermore, fluid-structure interaction boundary conditions are defined at the interface between the inner wall of the oil tank and the transformer oil, and at the interface between the winding surface and the transformer oil. Specifically: In solid structural components, mesh generation is performed to obtain structural dynamics mesh layers, corresponding to the maximum linear size of the mesh elements. satisfy: ; in, This is the minimum shear wave velocity in the medium corresponding to the solid structure component; This is the highest mechanical vibration frequency of the sweep frequency excitation; In the fluid transformer oil region and the non-uniform dielectric region of the insulating paperboard directly wrapping the windings, a high-frequency acoustic impedance grid layer is obtained by dividing a fine acoustic grid to satisfy the simulation of high-frequency acoustic wave spatial refraction and diffraction. The maximum linear size of the corresponding grid cell is... Limited by the shortest wavelength of partial discharge ultrasound in oil, the following conditions must be met: in, The center response frequency of the acoustic emission sensor array is used.

[0007] Furthermore, the specific process for establishing the multidimensional structural vibration transfer function matrix is ​​as follows: S21. In the finite element analysis model with well-defined mesh, set discrete virtual unit excitation source nodes at the core center, core clamp fastening bolts, winding middle and pressure plate, and set response extraction nodes at the actual vibration acceleration sensor installation coordinates corresponding to the external tank wall. S22. Inject a preset frequency step-sweep harmonic response excitation into the finite element solver. By extracting the steady-state response ratio of the output node to the input node at each frequency point, construct a multidimensional structural vibration transfer function matrix in the complex domain. Each element of the vibration transfer function matrix quantitatively characterizes the amplitude attenuation and phase lag law when the unit force vibration of the j-th core mechanical component inside is transmitted to the i-th surface measurement point outside.

[0008] Furthermore, the specific process for obtaining the real-time attenuation is as follows: S31. Acquire the low-frequency mechanical vibration signal on the surface of the enclosure and convert it into a time-domain signal. Remove the DC bias, add a Hanning window to prevent energy leakage, and perform bandpass filtering on the time-domain signal to filter out high-frequency white noise from the environment. Then, use fast Fourier transform to convert the time-domain signal into a frequency-domain signal vector. S32. Obtain the multidimensional structural vibration transfer function matrix. According to the dynamics theory of linear systems, the external response Response of internal excitation source Satisfies the positive transitivity relationship: The generalized inverse matrix algorithm with regularization constraints is used for inversion calculation to obtain the true vibration amplitude and dynamic excitation force of core components such as internal iron core clamps and winding pressure plates at various characteristic frequencies. The inversion formula is as follows: ; in, It is the conjugate transpose of the transfer function matrix; It is the identity matrix; The regularization parameter is adaptively adjusted based on the signal-to-noise ratio; S33. Based on the actual vibration amplitude, retrieve the internal modal parameters of the finite element model. According to the dynamic relationship of the multi-degree-of-freedom system, substitute the inverted internal vibration state changes into the nonlinear stiffness mapping curve to calculate the axial equivalent stiffness K. The specific formula is as follows: ; in For each order of natural frequency, For modal mass; The axial equivalent stiffness K has a nonlinear positive correlation with the clamping force, which can be converted into the core clamping force and the winding preload. S34. The core clamping force and winding preload are compared with the reference clamping force stored during initial calibration. Comparative calculations were performed to obtain the real-time attenuation percentage. : ; in This refers to the core clamping force or winding preload. S35. Evaluate the percentage of real-time attenuation based on the preset mechanical health status assessment threshold to obtain the quantitative level of deep mechanical health status.

[0009] Furthermore, the specific process of constructing the three-dimensional non-uniform acoustic impedance model is as follows: Construct a three-dimensional non-uniform acoustic impedance model that includes transmission, refraction, and diffraction attenuation parameters. S41, Based on a three-dimensional spatial coordinate system A non-uniform acoustic impedance field is constructed, dividing the internal space of the transformer into a fluid region (transformer oil), a porous composite region (winding insulation paperboard), and a high-density solid region (core metal), and then applying the formula... This generates a three-dimensional acoustic impedance distribution matrix covering the entire domain. S42. At the physical interface of different media, a three-dimensional refraction vector equation system based on Snell's law is introduced, and empirical transmission attenuation coefficients and edge diffraction damping parameters are assigned to form a three-dimensional non-uniform acoustic impedance model.

[0010] Furthermore, the specific process for locating the defect is as follows: S51. Extract the arrival time of the electromagnetic wave signal as the initial absolute zero point t0 of the partial discharge. Extract the precise time t1, t2, ..., tN when the N acoustic emission sensors deployed on the outer wall of the oil tank receive the first wave of the ultrasonic wave. Obtain the actual flight time of the ultrasonic wave from the unknown defect source to each acoustic emission sensor. S52. In the three-dimensional non-uniform acoustic impedance model, the three-dimensional coordinate points of N acoustic emission sensors distributed on the outer boundary are used as virtual emission sources. Following the principle of Markovnikov, in each iteration time step, the simulated acoustic ray is pushed back into the inner three-dimensional space. When the pushed ray encounters the core and winding boundary in the three-dimensional non-uniform acoustic impedance model, its three-dimensional spatial refraction angle and diffraction path are calculated to generate multiple simulated acoustic paths that bend with the properties of the medium. S53. For the reverse ray emitted by the i-th acoustic emission sensor, when the cumulative acoustic flight time in the three-dimensional non-uniform acoustic impedance model is equal to the actual flight time, the set of all spatial endpoints reached by the reverse ray constitutes a highly distorted three-dimensional surface, i.e., the isochronous surface. S54. Calculate the topological intersection of N isochronous surfaces corresponding to N acoustic emission sensors in three-dimensional space to form an intersection and dispersion region, i.e., the isochronous surface intersection region. Introduce the Gaussian mixture probability density function to assign probability weights to the voxel mesh in the intersection and dispersion region and find the core coordinate region with the largest spatial intersection probability density. S55. Import the core coordinate region into the transformer electrostatic field twin model. Through spatial Boolean intersection operation, verify whether the core coordinate region is located in the characteristic space where the electric field strength is greater than the partial discharge threshold. If it is, confirm the core coordinate region as the final defect location.

[0011] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: The online monitoring system for the health status of this transformer constructs a three-dimensional finite element model and establishes a multi-dimensional structural vibration transfer function matrix. Using an inverse matrix inversion algorithm, it directly maps the low-frequency mechanical vibration signal on the surface of the tank into the real-time attenuation of the internal core clamping force and winding preload. It can achieve quantitative perception of deep-seated mechanical loosening and fatigue conditions using only external conventional sensors, greatly improving the early proactive warning capability of equipment mechanical defects. At the same time, by constructing a three-dimensional non-uniform acoustic impedance model and introducing a ray tracing algorithm, it realistically reverse-engineers the physical path of refraction and diffraction of ultrasonic waves propagating between oil, cardboard, and metal, accurately pinpointing the exact spatial coordinates of micro-dendritic discharges or local insulation degradation. Attached Figure Description

[0012] Figure 1 A schematic diagram of the overall structure of the present invention is shown; Figure 2 A schematic diagram of the overall method flow of the present invention is shown. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Example: like Figure 1-2 As shown, an online monitoring system for the health status of a transformer includes a data acquisition unit, a modeling and analysis unit, a mechanical health assessment unit, and a defect location unit, wherein: The data acquisition unit utilizes a UHF sensor array, an acoustic emission sensor array, and a low-frequency vibration acceleration sensor deployed outside the transformer to synchronously acquire electromagnetic wave signals, ultrasonic signals, and low-frequency mechanical vibration signals of partial discharge on the same time reference, and sends them to the mechanical health assessment unit and the defect location unit. The modeling and analysis unit is used to obtain the three-dimensional mechanical structure inside the transformer to construct a finite element analysis model. It divides the corresponding dielectric mesh region according to the different acoustic impedances and sound velocities of the transformer oil, winding paperboard, iron core metal and tank wall. Through frequency sweep excitation simulation, it establishes a multi-dimensional structural vibration transfer function matrix from each core component inside the transformer to each vibration measurement point on the outer wall of the tank. The specific process of constructing the finite element analysis model and dividing the medium mesh region is as follows: S11. Obtain the original three-dimensional computer-aided design model of the transformer. The original three-dimensional computer-aided design model includes the dimensional and distribution data of the core (yoke, core column, tension strip and clamps), windings (high voltage winding, low voltage winding, interlayer insulation paperboard), insulation pressure plate, transformer oil space, and external tank wall. The original three-dimensional computer-aided design model is simplified using fluid-structure interaction dynamics based on acoustic-mechanical conduction characteristics to obtain a finite element analysis model, as follows: Miniature fasteners, small chamfers in oil passages, and external non-load-bearing accessories whose impact on electromagnetic force transmission and sound wave refraction is less than a preset threshold are ignored. The multilayer laminated silicon-through-silicon core is equivalent to a solid anisotropic continuous medium topology with orthogonal anisotropic elastic modulus. The winding (a composite of copper coil and inter-plate paper pad) is equivalent to a homogenized viscoelastic composite entity with both mass effect and stiffness characteristics according to the volume average method. S12. In the finite element analysis model, based on the actual physical materials of each component inside the transformer, their density, elastic modulus, Poisson's ratio, and the unique acoustic impedance and sound velocity parameters of each medium are defined to construct a spatially non-uniform heterogeneous property field, as detailed below: Define the corresponding longitudinal wave velocity for the iron core metal region. transverse wave speed of sound and material density Its acoustic impedance is ; Define the corresponding equivalent composite velocity for the winding paperboard region. With anisotropic damping coefficient; For the transformer oil region, define the corresponding hydrostatic pressure, bulk elastic modulus, and velocity of sound. and fluid density Its fluid acoustic impedance ; For the tank wall region, the standard elastic modulus and isotropic sound velocity characteristics of the steel are defined; S13. A multi-scale, two-layer adaptive meshing strategy is adopted. A structural dynamics mesh layer is created for low-frequency mechanical vibration, and a high-frequency acoustic impedance mesh layer is created for high-frequency acoustic emission localization. Furthermore, fluid-structure interaction boundary conditions are defined at the interface between the inner wall of the oil tank and the transformer oil, and at the interface between the winding surface and the transformer oil. Specifically: In solid structural components, mesh generation is performed to obtain structural dynamics mesh layers, corresponding to the maximum linear size of the mesh elements. satisfy: ; in, This is the minimum shear wave velocity in the medium corresponding to the solid structure component; This is the highest mechanical vibration frequency of the sweep frequency excitation; In the fluid transformer oil region and the non-uniform dielectric region of the insulating paperboard directly wrapping the windings, a high-frequency acoustic impedance grid layer is obtained by dividing a fine acoustic grid to satisfy the simulation of high-frequency acoustic wave spatial refraction and diffraction. The maximum linear size of the corresponding grid cell is... Limited by the shortest wavelength of partial discharge ultrasound in oil, the following conditions must be met: in, The center response frequency of the acoustic emission sensor array is used.

[0015] The specific process of establishing the multidimensional structural vibration transfer function matrix is ​​as follows: S21. In the finite element analysis model with well-defined mesh, set discrete virtual unit excitation source nodes at the core center, core clamp fastening bolts, winding middle and pressure plate, and set response extraction nodes at the actual vibration acceleration sensor installation coordinates corresponding to the external tank wall. S22. Inject a preset frequency step-sweep harmonic response excitation into the finite element solver. By extracting the steady-state response ratio of the output node to the input node at each frequency point, construct a multidimensional structural vibration transfer function matrix in the complex domain. Each element of the vibration transfer function matrix quantitatively characterizes the amplitude attenuation and phase lag law when the unit force vibration of the j-th core mechanical component inside is transmitted to the i-th surface measurement point outside.

[0016] The mechanical health assessment unit is used to obtain low-frequency mechanical vibration signals and input them into the inverse matrix of the transfer function matrix. Combined with internal modal parameters, it inversely calculates the real-time attenuation of the internal core clamping force and winding preload, and obtains the quantitative level of the deep mechanical health status. The specific process for obtaining the real-time attenuation is as follows: S31. Acquire the low-frequency mechanical vibration signal on the surface of the enclosure and convert it into a time-domain signal. Remove the DC bias, add a Hanning window to prevent energy leakage, and perform bandpass filtering on the time-domain signal to filter out high-frequency white noise from the environment. Then, use fast Fourier transform to convert the time-domain signal into a frequency-domain signal vector. S32. Obtain the multidimensional structural vibration transfer function matrix. According to the dynamics theory of linear systems, the external response Response of internal excitation source Satisfies the positive transitivity relationship: The generalized inverse matrix algorithm with regularization constraints is used for inversion calculation to obtain the true vibration amplitude and dynamic excitation force of core components such as internal iron core clamps and winding pressure plates at various characteristic frequencies. The inversion formula is as follows: ; in, It is the conjugate transpose of the transfer function matrix; It is the identity matrix; The regularization parameter is adaptively adjusted based on the signal-to-noise ratio; S33. Based on the actual vibration amplitude, retrieve the internal modal parameters of the finite element model. According to the dynamic relationship of the multi-degree-of-freedom system, substitute the inverted internal vibration state changes into the nonlinear stiffness mapping curve to calculate the axial equivalent stiffness K. The specific formula is as follows: ; in For each order of natural frequency, For modal mass; Since both the iron core silicon steel sheet laminations and the winding insulation paperboard are nonlinear elastic bodies under compression, the axial equivalent stiffness K has a nonlinear positive correlation function with the clamping force, thus converting the axial equivalent stiffness K into the iron core clamping force and the winding preload. S34. The core clamping force and winding preload are compared with the reference clamping force stored during initial calibration. Comparative calculations were performed to obtain the real-time attenuation percentage. : ; in This refers to the core clamping force or winding preload. S35. Evaluate the percentage of real-time attenuation based on the preset mechanical health status assessment threshold to obtain the quantitative level of deep mechanical health status.

[0017] For example: a decay rate of < 5% indicates "excellent health status". A decay rate of 15% to 5% indicates "slight loosening / requires attention". A decay rate greater than 15% is considered a "serious loosening defect / warning".

[0018] The defect location unit is used to construct a three-dimensional non-uniform acoustic impedance model that includes transmission, refraction and diffraction attenuation parameters. In the three-dimensional non-uniform acoustic impedance model, the actual path of the sound wave propagating from each acoustic emission sensor position to the interior is simulated using a ray tracing algorithm to obtain the simulated acoustic path. The isochronous surface intersection region of each simulated acoustic path in three-dimensional space is calculated, and the three-dimensional coordinate point with the highest intersection probability density and conforming to the characteristics of transformer electric field concentration is locked as the defect location.

[0019] The specific process of constructing a three-dimensional non-uniform acoustic impedance model is as follows: Construct a three-dimensional non-uniform acoustic impedance model that includes transmission, refraction, and diffraction attenuation parameters. S41, Based on a three-dimensional spatial coordinate system A non-uniform acoustic impedance field is constructed, dividing the internal space of the transformer into a fluid region (transformer oil), a porous composite region (winding insulation paperboard), and a high-density solid region (core metal), and then applying the formula... This generates a three-dimensional acoustic impedance distribution matrix covering the entire domain. S42. At the physical interfaces of different media (such as oil-paper boundary, oil-iron boundary), a three-dimensional refraction vector equation system based on Snell's law is introduced, and empirical transmission attenuation coefficients and edge diffraction damping parameters are assigned to form a three-dimensional non-uniform acoustic impedance model.

[0020] The specific process for locating the defect is as follows: S51. Extract the arrival time of the electromagnetic wave signal as the initial absolute zero point t0 of the partial discharge. Extract the precise time t1, t2, ..., tN when the N acoustic emission sensors deployed on the outer wall of the oil tank receive the first wave of the ultrasonic wave. Obtain the actual flight time of the ultrasonic wave from the unknown defect source to each acoustic emission sensor. S52. In the three-dimensional non-uniform acoustic impedance model, the three-dimensional coordinate points of N acoustic emission sensors distributed on the outer boundary are used as virtual emission sources. Following the principle of Markovnikov, in each iteration time step, the simulated acoustic ray is pushed back into the inner three-dimensional space. When the pushed ray encounters the core and winding boundary in the three-dimensional non-uniform acoustic impedance model, its three-dimensional spatial refraction angle and diffraction path are calculated to generate multiple simulated acoustic paths that bend with the properties of the medium. S53. For the reverse ray emitted by the i-th acoustic emission sensor, when the cumulative acoustic flight time in the three-dimensional non-uniform acoustic impedance model is equal to the actual flight time, the set of all spatial endpoints reached by the reverse ray constitutes a highly distorted three-dimensional surface, i.e., the isochronous surface. S54. Calculate the topological intersection of N isochronous surfaces corresponding to N acoustic emission sensors in three-dimensional space to form an intersection and dispersion region, i.e., the isochronous surface intersection region. Introduce the Gaussian mixture probability density function to assign probability weights to the voxel mesh in the intersection and dispersion region and find the core coordinate region with the largest spatial intersection probability density. S55. Import the core coordinate region into the transformer electrostatic field twin model. Through spatial Boolean intersection operation, verify whether the core coordinate region is located in the characteristic space where the electric field strength is greater than the partial discharge threshold. If it is, confirm the core coordinate region as the final defect location.

[0021] This invention constructs a three-dimensional finite element model and establishes a multi-dimensional structural vibration transfer function matrix. Using an inverse matrix inversion algorithm, it directly maps the low-frequency mechanical vibration signal on the surface of the housing into the real-time attenuation of the internal iron core clamping force and winding preload. It can achieve quantitative perception of deep-seated mechanical loosening and fatigue state using only external conventional sensors, greatly improving the early proactive warning capability of equipment mechanical defects. At the same time, by constructing a three-dimensional non-uniform acoustic impedance model and introducing a ray tracing algorithm, it realistically reverse-engineers the physical path of refraction and diffraction of ultrasonic waves propagating between oil, cardboard, and metal, accurately locking the precise spatial coordinates of micro-dendritic discharges or local insulation degradation.

[0022] The size of the interval and threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value.

[0023] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation. 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. An online monitoring system for the health status of a transformer, characterized in that, It includes a data acquisition unit, a modeling and analysis unit, a mechanical health assessment unit, and a defect location unit, wherein: The data acquisition unit utilizes a UHF sensor array, an acoustic emission sensor array, and a low-frequency vibration acceleration sensor deployed outside the transformer to synchronously acquire electromagnetic wave signals, ultrasonic signals, and low-frequency mechanical vibration signals of partial discharge on the same time reference, and sends them to the mechanical health assessment unit and the defect location unit. The modeling and analysis unit is used to obtain the three-dimensional mechanical structure inside the transformer to construct a finite element analysis model. It divides the corresponding dielectric mesh region according to the different acoustic impedances and sound velocities of the transformer oil, winding paperboard, iron core metal and tank wall. Through frequency sweep excitation simulation, it establishes a multi-dimensional structural vibration transfer function matrix from each core component inside the transformer to each vibration measurement point on the outer wall of the tank. The mechanical health assessment unit is used to obtain low-frequency mechanical vibration signals and input them into the inverse matrix of the transfer function matrix. Combined with internal modal parameters, it inversely calculates the real-time attenuation of the internal core clamping force and winding preload, and obtains the quantitative level of the deep mechanical health status. The defect location unit is used to construct a three-dimensional non-uniform acoustic impedance model that includes transmission, refraction and diffraction attenuation parameters. In the three-dimensional non-uniform acoustic impedance model, the actual path of the sound wave propagating from each acoustic emission sensor position to the interior is simulated using a ray tracing algorithm to obtain the simulated acoustic path. The isochronous surface intersection region of each simulated acoustic path in three-dimensional space is calculated, and the three-dimensional coordinate point with the highest intersection probability density and conforming to the characteristics of transformer electric field concentration is locked as the defect location.

2. The online monitoring system for transformer health status according to claim 1, characterized in that, The specific process of constructing the finite element analysis model and dividing the medium mesh region is as follows: S11. Obtain the original three-dimensional computer-aided design model of the transformer. The original three-dimensional computer-aided design model includes the dimensional and distribution data of the core, windings, insulating pressure plate, transformer oil space, and external tank wall. The original three-dimensional computer-aided design model is simplified by fluid-structure interaction dynamics based on acoustic-mechanical conduction characteristics to obtain the finite element analysis model, as follows: Miniature fasteners, small chamfers in oil passages, and external non-load-bearing accessories whose impact on electromagnetic force transmission and sound wave refraction is less than a preset threshold are ignored. The multilayer laminated silicon-through-silicon core is equivalent to a solid anisotropic continuous medium topology with orthogonal anisotropic elastic modulus. The winding is equivalent to a homogenized viscoelastic composite entity with both mass effect and stiffness characteristics according to the volume average method. S12. In the finite element analysis model, based on the actual physical materials of each component inside the transformer, their density, elastic modulus, Poisson's ratio, and the unique acoustic impedance and sound velocity parameters of each medium are defined to construct a spatially non-uniform heterogeneous property field, as detailed below: Define the corresponding longitudinal wave velocity for the iron core metal region. transverse wave speed of sound and material density Its acoustic impedance is ; Define the corresponding equivalent composite velocity for the winding paperboard region. With anisotropic damping coefficient; For the transformer oil region, define the corresponding hydrostatic pressure, bulk elastic modulus, and velocity of sound. and fluid density Its fluid acoustic impedance ; For the tank wall region, the standard elastic modulus and isotropic sound velocity characteristics of the steel are defined; S13. A multi-scale, two-layer adaptive meshing strategy is adopted. A structural dynamics mesh layer is created for low-frequency mechanical vibration, and a high-frequency acoustic impedance mesh layer is created for high-frequency acoustic emission localization. Furthermore, fluid-structure interaction boundary conditions are defined at the interface between the inner wall of the oil tank and the transformer oil, and at the interface between the winding surface and the transformer oil. Specifically: In solid structural components, mesh generation is performed to obtain structural dynamics mesh layers, corresponding to the maximum linear size of the mesh elements. satisfy: ； in, This is the minimum shear wave velocity in the medium corresponding to the solid structure component; This is the highest mechanical vibration frequency of the sweep frequency excitation; In the fluid transformer oil region and the non-uniform dielectric region of the insulating paperboard directly wrapping the windings, a high-frequency acoustic impedance grid layer is obtained by dividing a fine acoustic grid to satisfy the simulation of high-frequency acoustic wave spatial refraction and diffraction. The maximum linear size of the corresponding grid cell is... Limited by the shortest wavelength of partial discharge ultrasound in oil, the following conditions must be met: in, The center response frequency of the acoustic emission sensor array is used.

3. The online monitoring system for transformer health status according to claim 1, characterized in that, The specific process of establishing the multidimensional structural vibration transfer function matrix is ​​as follows: S21. In the finite element analysis model with well-defined mesh, set discrete virtual unit excitation source nodes at the core center, core clamp fastening bolts, winding middle and pressure plate, and set response extraction nodes at the actual vibration acceleration sensor installation coordinates corresponding to the external tank wall. S22. Inject a preset frequency step-sweep harmonic response excitation into the finite element solver. By extracting the steady-state response ratio of the output node to the input node at each frequency point, construct a multidimensional structural vibration transfer function matrix in the complex domain. Each element of the vibration transfer function matrix quantitatively characterizes the amplitude attenuation and phase lag law when the unit force vibration of the j-th core mechanical component inside is transmitted to the i-th surface measurement point outside.

4. The online monitoring system for transformer health status according to claim 1, characterized in that, The specific process for obtaining the real-time attenuation is as follows: S31. Acquire the low-frequency mechanical vibration signal on the surface of the enclosure and convert it into a time-domain signal. Remove the DC bias, add a Hanning window to prevent energy leakage, and perform bandpass filtering on the time-domain signal to filter out high-frequency white noise from the environment. Then, use fast Fourier transform to convert the time-domain signal into a frequency-domain signal vector. S32. Obtain the multidimensional structural vibration transfer function matrix. According to the dynamics theory of linear systems, the external response Response of internal excitation source Satisfies the positive transitivity relationship: The generalized inverse matrix algorithm with regularization constraints is used for inversion calculation to obtain the true vibration amplitude and dynamic excitation force of core components such as internal iron core clamps and winding pressure plates at various characteristic frequencies. The inversion formula is as follows: ； in, It is the conjugate transpose of the transfer function matrix; It is the identity matrix; The regularization parameter is adaptively adjusted based on the signal-to-noise ratio; S33. Based on the actual vibration amplitude, retrieve the internal modal parameters of the finite element model. According to the dynamic relationship of the multi-degree-of-freedom system, substitute the inverted internal vibration state changes into the nonlinear stiffness mapping curve to calculate the axial equivalent stiffness K. The specific formula is as follows: ； in For each order of natural frequency, For modal mass; The axial equivalent stiffness K has a nonlinear positive correlation with the clamping force, which can be converted into the core clamping force and the winding preload. S34. The core clamping force and winding preload are compared with the reference clamping force stored during initial calibration. Comparative calculations were performed to obtain the real-time attenuation percentage. : ； in This refers to the core clamping force or winding preload. S35. Evaluate the percentage of real-time attenuation based on the preset mechanical health status assessment threshold to obtain the quantitative level of deep mechanical health status.

5. The online monitoring system for transformer health status according to claim 1, characterized in that, The specific process of constructing a three-dimensional non-uniform acoustic impedance model is as follows: Construct a three-dimensional non-uniform acoustic impedance model that includes transmission, refraction, and diffraction attenuation parameters. S41, Based on a three-dimensional spatial coordinate system A non-uniform acoustic impedance field is constructed, dividing the internal space of the transformer into a fluid region (transformer oil), a porous composite region (winding insulation paperboard), and a high-density solid region (core metal), and then applying the formula... This generates a three-dimensional acoustic impedance distribution matrix covering the entire domain. S42. At the physical interface of different media, a three-dimensional refraction vector equation system based on Snell's law is introduced, and empirical transmission attenuation coefficients and edge diffraction damping parameters are assigned to form a three-dimensional non-uniform acoustic impedance model.

6. The online monitoring system for transformer health status according to claim 1, characterized in that, The specific process for locating the defect is as follows: S51. Extract the arrival time of the electromagnetic wave signal as the initial absolute zero point t0 of the partial discharge. Extract the precise time t1, t2, ..., tN when the N acoustic emission sensors deployed on the outer wall of the oil tank receive the first wave of the ultrasonic wave. Obtain the actual flight time of the ultrasonic wave from the unknown defect source to each acoustic emission sensor. S52. In the three-dimensional non-uniform acoustic impedance model, the three-dimensional coordinate points of N acoustic emission sensors distributed on the outer boundary are used as virtual emission sources. Following the principle of Markovnikov, in each iteration time step, the simulated acoustic ray is pushed back into the inner three-dimensional space. When the pushed ray encounters the core and winding boundary in the three-dimensional non-uniform acoustic impedance model, its three-dimensional spatial refraction angle and diffraction path are calculated to generate multiple simulated acoustic paths that bend with the properties of the medium. S53. For the reverse ray emitted by the i-th acoustic emission sensor, when the cumulative acoustic flight time in the three-dimensional non-uniform acoustic impedance model is equal to the actual flight time, the set of all spatial endpoints reached by the reverse ray constitutes a highly distorted three-dimensional surface, i.e., the isochronous surface. S54. Calculate the topological intersection of N isochronous surfaces corresponding to N acoustic emission sensors in three-dimensional space to form an intersection and dispersion region, i.e., the isochronous surface intersection region. Introduce the Gaussian mixture probability density function to assign probability weights to the voxel mesh in the intersection and dispersion region and find the core coordinate region with the largest spatial intersection probability density. S55. Import the core coordinate region into the transformer electrostatic field twin model. Through spatial Boolean intersection operation, verify whether the core coordinate region is located in the characteristic space where the electric field strength is greater than the partial discharge threshold. If it is, confirm the core coordinate region as the final defect location.

Images