Method and device for abnormal emergency protection of oil-immersed transformer

Abstract

This invention discloses an emergency protection method and device for oil-immersed transformers, relating to the field of transformer-related technology. The method includes: synchronously acquiring two types of signals—high-frequency transient and multi-physics field evolution—through a built-in sensor network; inputting these signals into a multi-physics field coupled simulation model; dynamically analyzing and outputting the inferred fault type and two remaining time parameters; matching graded active intervention commands from a multi-level intervention strategy library; fine-tuning the commands based on fault confidence and device health; generating targeted intervention commands; and controlling the physical intervention device to implement directional energy cancellation or blocking at the presumed fault location. This invention solves the technical problems of delayed protection action response and inability to accurately locate and actively intervene in the nascent stage of fault energy, leading to structural damage to the transformer during internal arc faults. It achieves the technical effect of improving the accuracy and timeliness of graded active intervention and the transformer's ability to autonomously recover its health after fault disturbances.

Description

Technical Field

[0001] This invention relates to the field of transformer-related technologies, specifically to an emergency protection method and device for oil-immersed transformers. Background Technology

[0002] As a core hub in the power grid system, the operational reliability of oil-immersed transformers directly affects the safety and stability of the power network. However, when a sudden insulation breakdown occurs inside the transformer, inducing a rapidly developing arc fault, traditional protection systems suffer from numerous problems, including response lag, insufficient accuracy, and irreversible destructive consequences. First, existing transformer relay protection relies on changes in electrical quantities for fault identification. Limited by the weak electrical characteristics in the early stages of a fault, the complexity of electromagnetic transient processes, and the inherent algorithm delays of protection devices, it can typically only reliably operate when the fault has developed to a severe stage (such as phase-to-phase short circuits or ground faults), failing to provide effective intervention when the fault energy is still in a controllable nascent stage. Second, non-electrical quantity protection methods, including gas relays, pressure relief valves, and oil flow relays, have response speeds that are... The action threshold determines that it cannot achieve early warning and active blocking, and can only reduce the severity of the fault consequences, but cannot prevent the transformer body from suffering irreversible mechanical, electrical and thermal damage. Furthermore, the automatic fire extinguishing system, such as oil draining and nitrogen injection or water spray device, is limited to passive suppression after the fire, and is completely unable to deal with instantaneous disasters such as oil tank explosion, and has an essential protection blind spot in the time dimension. Moreover, the current protection strategy lacks the ability to dynamically perceive the multi-physics field of the fault evolution process, cannot accurately locate the energy release source, assess its energy level and expansion path in the early stage of the fault, and does not have the ability to carry out "surgical" energy cancellation or arc blocking for the core area of ​​the fault.

[0003] Therefore, current technologies suffer from technical problems such as delayed protection response and inability to accurately locate and actively intervene in the early stages of fault energy, leading to structural damage to transformers during internal arcing faults. Summary of the Invention

[0004] This application provides an emergency protection method and device for oil-immersed transformers, which solves the technical problems in the prior art where the protection action response is delayed and the inability to accurately locate and actively intervene in the nascent stage of fault energy leads to structural damage to the transformer in the internal arc fault. It achieves the technical effect of improving the accuracy and timeliness of graded active intervention and the ability of the transformer to autonomously recover to a healthy state after fault disturbance.

[0005] This application provides an emergency protection method for oil-immersed transformers, the method comprising: simultaneously acquiring a first type of high-frequency transient signal and a second type of multi-physics field evolution signal characterizing the energy release process of an insulation fault through a sensor network embedded in the transformer body; simultaneously inputting the first type of high-frequency transient signal and the second type of multi-physics field evolution signal into a built-in multi-physics field coupled simulation model, dynamically analyzing and outputting a predicted fault type, a first remaining time parameter, and a second remaining time parameter characterizing the current fault state and future evolution risk; based on the first remaining time parameter, the second remaining time parameter, and the predicted fault type, dynamically matching graded active intervention commands from a pre-stored multi-level intervention strategy library, wherein the multi-level intervention strategy library includes at least an energy cancellation strategy and an arc surgery strategy; adaptively fine-tuning the graded active intervention commands according to the confidence parameter of the predicted fault type and the health status parameter of the physical intervention device to generate targeted intervention commands; and controlling the physical intervention device deployed at the presumed location of the fault point according to the targeted intervention commands to directionally cancel or block the development process of the fault energy.

[0006] This application also provides an emergency protection device for an oil-immersed transformer, the device comprising: a signal acquisition module, used to simultaneously acquire a first type of high-frequency transient signal and a second type of multi-physics field evolution signal characterizing the energy release process of an insulation fault through a sensor network embedded in the transformer body; a signal analysis module, used to simultaneously input the first type of high-frequency transient signal and the second type of multi-physics field evolution signal into a built-in multi-physics field coupled simulation model, dynamically analyze and output the deduced fault type, a first remaining time parameter, and a second remaining time parameter characterizing the current fault state and future evolution risk; and a command matching module, used to match the first type of high-frequency transient signal and the second type of multi-physics field evolution signal based on the first type of high-frequency transient signal and the second type of multi-physics field evolution signal. The remaining time parameter, the second remaining time parameter, and the deduced fault type are used to dynamically match graded active intervention instructions from a pre-stored multi-level intervention strategy library, wherein the multi-level intervention strategy library includes at least energy cancellation strategies and arc surgery strategies; the instruction fine-tuning module is used to adaptively fine-tune the graded active intervention instructions based on the confidence parameter of the deduced fault type and the health status parameter of the physical intervention device to generate targeted intervention instructions; the fault handling module is used to control the physical intervention device deployed at the estimated location of the fault point to perform targeted cancellation or blocking of the fault energy development process according to the targeted intervention instructions.

[0007] The proposed emergency protection method and device for oil-immersed transformers utilizes a built-in sensor network to synchronously acquire two types of signals: high-frequency transient signals and multi-physics field evolution signals. These signals are input into a multi-physics field coupled simulation model, which dynamically analyzes and outputs the predicted fault type and two remaining time parameters. Graded active intervention commands are matched from a multi-level intervention strategy library. The commands are fine-tuned based on fault confidence and device health, generating targeted intervention commands to control the physical intervention device to perform directional energy cancellation or blocking at the presumed fault location. This solves the technical problems of delayed protection action response and inability to accurately locate and actively intervene in the nascent stage of fault energy, leading to structural damage to the transformer during internal arcing faults. It achieves the technical effect of improving the accuracy and timeliness of graded active intervention and the transformer's ability to autonomously recover its health after fault disturbances. Attached Figure Description

[0008] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the apparatus according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0009] Figure 1 This is a schematic diagram of the emergency protection method for an oil-immersed transformer provided in an embodiment of this application.

[0010] Figure 2 This is a schematic diagram of the abnormal emergency protection device for an oil-immersed transformer provided in an embodiment of this application.

[0011] Explanation of reference numerals in the attached diagram: Signal acquisition module 10, signal analysis module 20, command matching module 30, command fine-tuning module 40, fault handling module 50. Detailed Implementation

[0012] To further illustrate the technical means and effects adopted by the present invention in order to achieve the intended purpose, the following detailed description is provided in conjunction with the accompanying drawings and preferred embodiments, based on the specific implementation methods, structures, features and effects of the present invention.

[0013] This application provides an emergency protection method for oil-immersed transformers in case of abnormalities, such as... Figure 1 As shown, the method includes:

[0014] Step S100: Through the sensing network embedded in the transformer body, the first type of high-frequency transient signal and the second type of multi-physics field evolution signal characterizing the energy release process of insulation fault are simultaneously acquired.

[0015] Step S100 further includes: using a piezoelectric thin film sensor array embedded in the winding insulation with a non-uniform topology to collect a first type of high-frequency transient signal reflecting the stress wave characteristics of partial discharge and arc initiation stage; and using an optical fiber sensor network and an immersion micro differential pressure sensor embedded in key locations of the transformer body to collect a second type of multi-physics field evolution signal reflecting temperature field gradient changes, bubble dynamics, and local pressure disturbances.

[0016] Preferably, when piezoelectric materials are subjected to mechanical stress (such as pressure or vibration), they generate charge or voltage signals. When partial discharge or arcing occurs, tiny, impactful stress waves are generated instantaneously. The piezoelectric film is very thin and flexible, allowing it to be embedded in narrow spaces. It has an extremely fast response speed and can capture transient stress waves at the microsecond or even nanosecond level. The sensors are not uniformly arranged, but strategically deployed with varying density according to the transformer's internal electric field, temperature, and structural weaknesses, such as high-fault areas like winding ends, corners, and lead connections. This achieves the highest precision monitoring of critical areas with the fewest sensors, embedded within the winding insulation. The piezoelectric film sensor array collects the impactful stress wave pulses generated when the material fractures instantaneously and the charge is violently neutralized during a fault, reflecting the stress wave characteristics of the initial stage of partial discharge and arcing. The signal is very short and has a high frequency (high-frequency transient), thus identifying the first type of high-frequency transient signal.

[0017] Preferably, optical fibers can transmit optical signals, and their internal structures, such as Bragg gratings, are extremely sensitive to temperature and strain. By measuring the wavelength changes of scattered light, distributed temperature and strain measurements can be achieved. A single optical fiber can replace hundreds or thousands of point sensors. Immersion micro-differential pressure sensors are miniature pressure sensors directly immersed in transformer oil, used to measure extremely small pressure differences between different locations. They collect second-type multiphysics field evolution signals reflecting changes in temperature field gradients, bubble dynamics, and local pressure disturbances. Among these, partial discharge or overheating will first cause local oil temperature anomalies. The optical fiber network can map the entire transformer in real time. The internal temperature contour map of the device indicates the location of hot spots and the rate of heat diffusion, and determines the temperature field gradient change. The electric arc in the oil will instantly vaporize the insulating oil, producing a large amount of characteristic gases such as hydrogen and acetylene to form bubbles. Fiber optics and micro differential pressure sensors can detect the nucleation, growth, movement, merging and even collapse of the bubbles. The behavior of the bubbles serves as a key indicator for judging the energy and development trend of the electric arc. When bubbles are generated, expand or move, even if the overall oil tank pressure has not yet risen, a small pressure fluctuation will occur in the limited space area near the fault point. The immersion micro differential pressure sensor can capture the initial pressure signal and determine the second type of multiphysics field evolution signal.

[0018] Step S200: The first type of high-frequency transient signal and the second type of multiphysics field evolution signal are synchronously input into a built-in multiphysics field coupling simulation model, and the model is dynamically analyzed and output to represent the current fault state and future evolution risk, the deduced fault type, the first remaining time parameter and the second remaining time parameter.

[0019] Step S200 further includes: utilizing the spatially directional stress wavefront information in the first type of high-frequency transient signal, performing dual-constraint localization of potential fault sources based on time-of-arrival difference and waveform attenuation characteristics in the multiphysics coupled simulation model to obtain an initial set of fault sources; using the spatiotemporal sequence in the second type of multiphysics evolution signal as a dynamic observation vector to recursively update the preset fluid-thermal-chemical coupling equations in the multiphysics coupled simulation model; after each state update, using the simulation state of the previous moment as the initial value, running the updated fluid-thermal-chemical coupling equations for rapid finite-time domain analysis. Forward simulation is performed to model the energy release and propagation process of potential faults in the initial set of fault sources, generating forward simulation results. The proportional relationship of energy conversion to preset gas-generating components during the rapid forward simulation is analyzed, and compared with preset fault characteristic spectra to determine the type of fault being simulated. Based on the spatiotemporal extrema of the mechanical stress field of the tank and the thermal aging rate field of the insulating material in the forward simulation results, a first remaining time parameter and a second remaining time parameter are calculated. The first remaining time parameter is the time when the simulated pressure reaches the tank structure failure threshold, and the second remaining time parameter is the time when the simulated thermal damage to the insulating material accumulates to the irreversible threshold.

[0020] Preferably, the first type of high-frequency transient signal and the second type of multiphysics field evolution signal are simultaneously input into the multiphysics field coupled simulation model for analysis. The timing of the reception of the same stress wave by sensors at different locations in the piezoelectric thin film sensor array varies. Spatially directional stress wavefront information is extracted from the first type of high-frequency transient signal, including the time when the stress wave first arrives at each sensor, the steepness of the rising edge of the wavefront, and the amplitude attenuation rate of the wavefront. Furthermore, the velocity differences and amplitude attenuation rate differences of the stress wave in different propagation directions within the transformer are related to the anisotropic characteristics of the insulating medium (oil, paper, pressure plate, etc.). In the multiphysics field coupled simulation model, potential fault sources are located based on the dual constraints of time-of-arrival (TOA) and waveform attenuation characteristics. The TOA constraint calculates the time difference between the reception of the same stress wave front by any two sensors. Based on the propagation velocity model of the stress wave in the transformer's internal medium, the time difference is... The values ​​are converted into distance differences. With each pair of sensors as the focus and the distance difference as a constant, a set of hyperboloid equations is constructed. The intersection region of multiple hyperboloid equations is the possible location range of the fault source. The waveform attenuation characteristic constraint measures the peak amplitude of the stress wave received by different sensors. Based on the attenuation coefficient of the stress wave in each medium of the transformer, a functional relationship between amplitude attenuation and propagation distance is established. With each sensor as the center and the theoretical propagation distance corresponding to its measured amplitude as the radius, a set of spherical or ellipsoidal surfaces is constructed. The intersection region of multiple spherical surfaces further narrows the possible location range of the fault source. Then, the intersection of the hyperboloids obtained from the time of arrival constraint and the intersection of the spherical surfaces obtained from the waveform attenuation constraint are spatially overlapped. The spatial region pointed to by both constraints is retained as the candidate location of the fault source, forming the initial fault source set. Each element in the set corresponds to a possible fault initiation location, and its magnitude reflects the degree of uncertainty in the location.

[0021] Preferably, the temperature values ​​measured by the fiber optic sensor network over a continuous time series, each with a spatial coordinate label and a timestamp, and the pressure difference values ​​measured by the micro-differential pressure sensor over a continuous time series, each with a spatial coordinate label and a timestamp, are arranged in chronological order to form a spatially distributed data sequence that evolves over time. This is the spatiotemporal sequence in the second type of multiphysics evolution signal, which is then used as a dynamic observation vector to reflect the real-time changes in the physical field during fault evolution. Specifically, the temperature and pressure difference values ​​measured by all sensors at the current moment are arranged in a fixed order, with each dimension corresponding to a specific physical quantity at a specific spatial location, such as the temperature of a fiber optic grating point or the pressure difference value of a micro-differential pressure sensor. Then, a recursive state update is performed on the preset fluid-thermal-chemical coupling equations in the multiphysics coupling simulation model. The fluid equations describe the velocity and pressure field distribution of the insulating oil inside the transformer, including mass conservation equations and momentum conservation equations. The thermal equations... The system describes the internal temperature field distribution and heat conduction process of the transformer, including heat diffusion or energy conservation, encompassing heat source terms (heat generated by fault energy release), heat conduction terms, and heat convection terms (heat carried away by oil flow). A set of chemical equations describes the chemical reactions of insulating oil and insulating paper decomposing at high temperatures, including generation rate equations for gas-producing components such as hydrogen, acetylene, methane, ethylene, carbon monoxide, and carbon dioxide, and the functional relationships between the gas generation rate and temperature, pressure, arc energy, and material type. Furthermore, the density and viscosity terms in the fluid equations depend on temperature, and the heat source terms in the thermal equations depend on the fault energy release rate. Next, the difference between the current dynamic observation vector and the predicted state vector of the simulation model at the same time is calculated. Based on the magnitude and direction of the difference, the state variables of each physical field in the simulation model are adjusted, such as the temperature field distribution, pressure field distribution, and gas concentration field distribution. The adjusted state is used as the initial condition for the simulation calculation at the next time step. This recursive state update is repeated every time a new observation vector is received.

[0022] Preferably, after each state update, the updated fluid-thermal-chemical coupling equations are run for rapid forward derivation in the finite-time domain. Specifically, the updated physical field distribution at the current moment is used as the initial condition, and the fluid-thermal-chemical coupling equations are numerically solved using the finite element method, finite volume method, or finite difference method at a set time step. A preset finite time window (e.g., 5 milliseconds, 10 milliseconds, or 20 milliseconds) is used to deduce the equations. The length of the derivation time window is less than the time it takes for the fault to develop from the current state to irreversible damage. Adaptive time step and reduced-order numerical methods are used to make the calculation time of a single derivation much smaller than the length of the derivation time window, achieving completion within millisecond-level physical time. The calculation is performed, and then the energy release and propagation process of potential faults in the initial fault source set is simulated. This includes assuming that there is an energy release source at each candidate fault location in the initial fault source set, setting the power-time curve of energy release for each candidate source, such as constant power release and exponential decay release, adding the energy source term to the thermal equation and chemical equation, driving the simulation model to calculate the propagation path of energy inside the transformer, the energy distribution ratio, and converting it into the proportion of heat, pressure wave, and material decomposition chemical energy. The forward deduction results corresponding to each candidate source are output, including the temperature field, pressure field, oil flow velocity field, concentration field of each gas-producing component, mechanical stress field of the tank wall, and cumulative thermal damage field of the insulation material.

[0023] Preferably, during the rapid forward simulation, the proportions of energy conversion to preset gas-producing components are analyzed, such as the hydrogen / acetylene ratio, acetylene / ethylene ratio, and total hydrocarbon / hydrogen ratio. The preset fault feature spectrum is a feature vector of gas-producing component proportions corresponding to a specific fault type, which is calibrated in advance through a large number of experiments or simulations. For example, the feature spectrum corresponding to a local overheating fault is high proportions of methane and ethylene, medium proportions of hydrogen, and extremely low proportions of acetylene. The feature spectrum corresponding to a high-energy arc fault is high proportions of hydrogen and acetylene and low proportions of methane. Then, the similarity calculation is performed between the gas-producing component proportion vector output by the current simulation and each feature vector in the preset fault feature spectrum library, such as Euclidean distance and cosine similarity. The fault type corresponding to the fault feature spectrum with the highest similarity is selected as the simulation result. Finally, the simulated fault type is determined, which includes at least local overheating faults, developing arc faults, and mixed faults that have characteristics of both.

[0024] Preferably, the mechanical stress field of the fuel tank is the numerical distribution of tensile, compressive, and shear stresses borne by various spatial points on the fuel tank wall at different times in the forward simulation results. The simulated pressure field acts on the inner wall of the fuel tank and is calculated by stress-strain constitutive equations in combination with the fuel tank geometry and material elastic modulus. The fuel tank structural failure threshold refers to the tensile strength or yield strength of the fuel tank material, which is determined in advance through material tensile testing. When the mechanical stress at a certain point on the fuel tank wall exceeds this threshold, it is considered that the fuel tank has undergone plastic deformation or rupture. Then, starting from the current time, the maximum value of the fuel tank mechanical stress field is monitored hour by hour along the forward simulation time axis. The moment when the maximum mechanical stress first reaches or exceeds the fuel tank structural failure threshold is found, and the time difference between this moment and the current moment is calculated to determine the first remaining time parameter. If no intervention measures are taken, the fuel tank will experience structural failure after this time, such as rupture or explosion. The thermal aging rate field of the insulating material is the rate distribution of thermal degradation of the insulating material at each spatial point in the forward simulation result under the influence of temperature. The thermal aging rate is usually calculated using the Arrhenius model, and this rate value changes dynamically with time and temperature. Then, from the fault initiation time to each moment within the current forward simulation window, the thermal aging rate of each spatial point is integrated over time to obtain the cumulative damage. The upper limit of the integration is the current simulation time, reflecting the degree of irreversible chemical degradation that has occurred in the insulating material, which is directly related to the remaining mechanical strength and dielectric strength of the material. The cumulative thermal damage of the insulating material reaches the irreversible threshold. After the value is reached, the mechanical strength and dielectric strength of the material decrease to the point that they cannot meet the requirements for normal operation. This is determined in advance through thermal aging tests of the insulating material. Usually, the criterion is that the elongation at break decreases to 50% of the initial value or the breakdown voltage decreases to a certain percentage of the initial value. Then, starting from the current moment, the maximum value of the accumulated thermal damage of the insulating material is monitored hour by hour along the forward time axis. The moment when the accumulated damage value first reaches or exceeds the irreversible threshold is found. The time difference between this moment and the current moment is calculated to determine the second remaining time parameter. If no intervention measures are taken, the insulating material will undergo irreversible thermal damage failure after this time.

[0025] Step S300: Based on the first remaining time parameter, the second remaining time parameter, and the deduced fault type, dynamically match graded active intervention instructions from the pre-stored multi-level intervention strategy library, wherein the multi-level intervention strategy library includes at least energy cancellation strategy and arc surgery strategy.

[0026] Step S300 further includes constructing a dynamic decision plane with the first remaining time parameter and the second remaining time parameter as input coordinates; when the simulated fault type is a local overheating fault and the state point falls into the first region defined by the second remaining time parameter in the decision plane, matching the graded active intervention command corresponding to the energy cancellation strategy; when the simulated fault type is a developing arc fault and the state point falls into the second region defined by the first remaining time parameter in the decision plane, matching the graded active intervention command corresponding to the arc surgery strategy; wherein the first region and the second region partially overlap.

[0027] Preferably, a two-dimensional coordinate system is established, with the horizontal axis representing one remaining time parameter and the vertical axis representing another remaining time parameter. For each point in the coordinate system, the horizontal coordinate value represents the first remaining time parameter and the vertical coordinate value represents the second remaining time parameter. This determines the dynamic decision plane, which is the entire decision space spanned by the two-dimensional coordinate system. The boundary line position and region division threshold of the plane are adjusted in real time based on the current operating conditions of the transformer, the structural parameters of the transformer, the signal quality of the sensor network, and the remaining time parameters recalculated after the last state update. That is, after each forward deduction and calculation of new first and second remaining time parameters, the decision plane and its region division are updated accordingly. If the inferred fault type is a local overheating fault, the rate of thermal damage accumulation of the insulation material is much faster than the rate of pressure rise in the tank, meaning that thermal failure occurs before structural failure. In the dynamic decision plane, a spatial region is defined based on the magnitude of the second remaining time parameter. For example, the second remaining time parameter is less than the first threshold (e.g., 20 milliseconds) and the second remaining time parameter is dominant relative to the first remaining time parameter, while also satisfying that the second remaining time parameter ≤ the first remaining time parameter × the proportionality coefficient (e.g., 0.5 or 0.8). When the fault type is local overheating and the state point falls into the first region, it indicates that the risk of insulation material damage is dominant, meaning that the risk of thermal damage to the insulation material due to long-term high temperature accumulation is higher than the risk of the tank rupturing due to pressure impact. Therefore, an energy offsetting strategy aimed at heat management is given priority.

[0028] Preferably, if the deduced fault type is a developing arc fault, and the state point falls into the second region defined by the first remaining time parameter in the decision plane, the internal pressure rise rate of the tank is much faster than the cumulative rate of thermal damage to the insulation material, that is, structural failure occurs before thermal failure. In the dynamic decision plane, another spatial region is defined based on the magnitude of the first remaining time parameter. For example, the first remaining time parameter ≤ a preset second threshold (e.g., 10 milliseconds) and simultaneously satisfies the condition that the first remaining time parameter ≤ the second remaining time parameter × a proportional coefficient (e.g., 0.6 or 0.7). When the fault type is a developing arc fault and the state point falls into the second region, it indicates that the risk of mechanical failure of the tank is dominant, that is, the risk of the tank rupturing due to pressure wave impact is much higher than the risk of thermal damage to the insulation material. Therefore, an arc surgery strategy aimed at directly suppressing arc energy is given priority. The first and second regions partially overlap, meaning that both the rate of pressure rise in the tank and the rate of thermal damage accumulation in the insulation material are at dangerous levels. They are close in time, making it impossible to clearly distinguish which failure mode occurs first. For example, in a transformer internal fault, the arc energy simultaneously generates pressure waves and a high-temperature thermal field. When the fault energy level is in a certain intermediate range, the pressure rise time and the thermal damage accumulation time are similar. Factors such as sensor measurement noise, simulation model errors, and fault location uncertainties cause the estimated values ​​of both remaining time parameters to have error ranges.

[0029] Preferably, the graded active intervention commands are a set of control command sequences pre-stored in a multi-level intervention strategy library. Each sequence corresponds to a specific intervention intensity level, and is classified according to the invasiveness, energy level, scope of action, and execution time of the intervention. For example, Level 1 is low-energy intervention, Level 2 is medium-energy intervention, and Level 3 is high-energy intervention. Based on the position of the state point in the decision plane, the corresponding level of command sequence is selected. The graded active intervention commands corresponding to the energy cancellation strategy control the micro-pipeline release unit to release phase change cooling medium to the fault area, including at least the target release location coordinates, medium release flow rate, release duration, and selection of phase change medium type. The longer the remaining time, the lower the release flow rate and duration; the shorter the remaining time, the higher the release flow rate and duration. The graded active intervention commands corresponding to the arc surgery strategy sequentially control the electromagnetic intervention module to generate a pulsed strong magnetic field and control the medium injection device to inject composite functional medium, including at least the discharge voltage, discharge current, pulse width, and trigger time of the electromagnetic intervention module; and the injection sequence, injection pressure, and medium type of the medium injection device. The longer the remaining time, the lower the magnetic field strength and medium injection volume; the shorter the remaining time, the higher the magnetic field strength and medium injection volume are used.

[0030] Furthermore, step S300 also includes that when the state point falls into the overlapping part of the first region and the second region, and the deduced fault type has both overheating and arc characteristics, the graded active intervention instruction corresponding to the matched arc surgery strategy must include a preparatory cooling stage, which is executed before the generation of the pulsed strong magnetic field.

[0031] Preferably, the state point falling within the overlap of the first and second regions means that the time it takes for the tank pressure to rise to the structural failure threshold and the time it takes for the thermal damage to the insulation material to accumulate to the irreversible threshold are both within the danger range, and their values ​​are similar, making it impossible to clearly distinguish which failure mode will occur first. The coordinates of the state point in the dynamic decision plane simultaneously satisfy the boundary conditions of the first and second regions, i.e., the first remaining time parameter ≤ the upper limit value of the threshold condition in the first region, and the second remaining time parameter ≤ the upper limit value of the threshold condition in the second region. The deduced fault type has both overheating and arcing characteristics. For example, the overheating characteristic spectrum corresponds to the pyrolysis of the insulation material caused by local overheating, and the arcing characteristic spectrum corresponds to the cracking of oil molecules caused by a high-energy arc. When the gas production component ratio analysis results simultaneously satisfy the overheating characteristic condition and the arcing characteristic condition, the failure type is determined to be more likely to occur. When the arc characteristic condition is met, it is determined to be a mixed fault with both overheating and arc characteristics. For overlapping areas with both overheating and arc characteristics, the arc surgery strategy is selected as the basic strategy, which also includes a preparatory cooling stage. This stage is executed before the generation of the pulsed strong magnetic field, that is, the cooling stage is executed first. After the cooling stage is completed, the original pulsed strong magnetic field generation and medium injection in the arc surgery strategy are executed. In the preparatory cooling stage, the micro-pipeline release unit is controlled to release the phase change cooling medium to the estimated location of the fault point. After absorbing the heat in the fault area, the phase change cooling medium undergoes a phase change and carries away the heat from the fault area. The cooling parameters include at least the medium release flow rate, release duration and medium type selection, which are used to locally reduce the oil temperature, gas temperature and conductivity of the target fault area to suppress the development of overheating.

[0032] Step S400: Based on the confidence parameters of the inferred fault type and the health status parameters of the physical intervention device, the graded active intervention instructions are adaptively fine-tuned to generate targeted intervention instructions.

[0033] Step S400 further includes: obtaining a confidence parameter characterizing the reliability of the current simulation results of the multiphysics coupling simulation model, and a health status parameter characterizing the current availability of the target physical intervention device; scaling the preset intervention intensity level in the graded active intervention instruction based on the confidence parameter; and extracting the control parameters related to the target physical intervention device from the scaled graded active intervention instruction for redundancy backup or downgraded execution adjustment to generate the targeted intervention instruction.

[0034] Preferably, the confidence parameter characterizes the reliability of the current inference results of the multiphysics coupling simulation model, and its value range is usually between 0 and 1, where 0 indicates completely unreliable and 1 indicates completely reliable. It is determined comprehensively based on the ratio of the number of effectively triggered sensors to the total number of sensors in the piezoelectric thin film sensor array, the signal-to-noise ratio of the fiber optic sensor network, the ratio of signal amplitude to background noise amplitude, the spatial distribution volume of the initial fault source set, the consistency of the intersection region obtained by the dual constraints, and the residual size of the numerical solution of the fluid-thermal-chemical coupling equation set. The health status parameter characterizes the current usability of the target physical intervention device, and its value range is usually between 0 and 1, where 0 indicates completely unusable and 1 indicates completely normal. It is determined comprehensively based on the ratio of the remaining storage of the phase change cooling medium to the rated storage, the health of the micro-pipe release unit, the health of the electromagnetic intervention module, and the health of the medium injection device. Based on the estimated location of the fault point, one or a group of devices with the closest spatial distance or the best intervention effect are selected as the target physical intervention device from multiple physical intervention devices installed at multiple different locations on the transformer.

[0035] Preferably, the preset intervention intensity level in the graded active intervention instruction is scaled proportionally according to the confidence level parameter. That is, the actual intensity parameter is obtained by multiplying the confidence level parameter value by the preset intensity level parameter. For example, the actual release flow rate = preset release flow rate × confidence level parameter, the actual pulse magnetic field strength = preset magnetic field strength × confidence level parameter, and the actual injection volume = preset injection volume × confidence level parameter. The lower limit of the scaled intensity parameter is set according to the preset minimum threshold (e.g., 0.3) if the confidence level parameter is lower than the preset minimum threshold, so as to avoid the intervention intensity approaching zero due to excessively low confidence. The upper limit of the scaled intensity parameter does not exceed the preset maximum safe intensity to prevent excessive execution due to errors in the calculation of the confidence level parameter. This achieves the matching of intervention intensity and decision information quality. From the scaled-down hierarchical active intervention commands, a subset of control parameters directly related to specific physical intervention devices is extracted. For micro-pipeline release units, the target release position coordinates, medium release flow rate, and release duration are extracted. For electromagnetic intervention modules, the discharge voltage, discharge current, pulse width, and trigger time are extracted. For medium injection devices, the injection sequence, such as the order and interval between the first and second injectors, injection pressure, and medium type selection, is extracted.

[0036] Preferably, when the health status parameter shows that the health of the target physical intervention device is between 0.4 and 0.7 and there are other available backup devices, a redundancy backup mechanism is activated, and the intervention command originally assigned to a single target device is split and assigned to multiple backup devices. When the health status parameter shows that the health of the target physical intervention device is below 0.4 and there are no available backup devices, a downgrade adjustment is performed, reducing the execution parameters of the intervention action to a level that the device can currently reliably execute. The downgraded intervention command still needs to meet the minimum intervention effectiveness threshold. If the threshold cannot be reached after downgrading, an alarm signal is output and the device enters a preset safety failure mode. When the health of all physical intervention devices corresponding to the originally matched graded active intervention command is below a preset unavailable threshold (e.g., 0.2), the next level of backup strategy is selected from the multi-level intervention strategy library. Finally, the control parameters after confidence scaling and health redundancy backup / downgrade adjustment are encapsulated into a targeted intervention command that can be directly executed by the physical intervention device, which includes at least the target device identifier, action type, execution parameters, and timestamp.

[0037] Step S500: According to the targeted intervention command, control the physical intervention device deployed at the estimated location of the fault point to perform targeted cancellation or blocking of the fault energy development process.

[0038] Step S500 further includes, when the targeted intervention command originates from the energy cancellation strategy, controlling the microchannel release unit located at the presumed fault point to release phase change cooling medium into the target area; when the targeted intervention command originates from the arc surgery strategy, sequentially controlling the electromagnetic intervention module located at the presumed fault point to generate a pulsed strong magnetic field, and the integrated medium injection device to inject composite functional medium into the constrained arc channel.

[0039] Preferably, when the targeted intervention command originates from an energy cancellation strategy, i.e., the deduced fault type is a local overheating fault, and the state point falls within the first region defined by the second remaining time parameter, the micro-pipe release unit located at the estimated fault point position is controlled to release the phase change cooling medium into the target region. Specifically, the estimated fault point position is determined by weighted averaging of multiple candidate positions in the initial fault source set or by selecting the candidate position with the highest confidence. The micro-pipe release unit includes one or more capillary tubes extending from the outside of the transformer through the tank wall to a predetermined internal position, one or more solenoid valves or piezoelectric valves for controlling the on / off state and flow rate of the medium, and a medium storage tank for storing the pressurized phase change cooling medium. The micro-pipe release units are distributed in multiple numbers at different spatial locations inside the transformer, each unit corresponding to a specific protection area. The valve drive circuit of the micro-pipe release unit sends... The power supply signal, such as a voltage pulse or PWM signal, includes parameters including at least the valve opening duration, valve opening degree, and trigger time. The target area is a spherical or ellipsoidal region with a preset radius (e.g., 20 mm to 100 mm) centered on the estimated fault location. A phase change cooling medium, which is a gaseous substance at normal temperature and pressure and a liquid or supercritical substance under pressurized storage conditions, such as liquid nitrogen, liquid carbon dioxide, or fluorocarbons, is released into the target area. After the valve opens, the pressurized liquid phase change medium is transported to the injection port via a capillary tube. The medium experiences a sudden pressure drop after being ejected from the injection port, resulting in vaporization and phase change. During vaporization, it absorbs a large amount of heat from the surrounding environment, lowering the temperature of the target area. The vaporized gaseous medium diffuses naturally inside the transformer or is discharged through a pressure difference. Control parameters include the release flow rate, release duration, total release volume, and release mode.

[0040] Preferably, when the targeted intervention command originates from the arc surgery strategy, the deduced fault type is a progressive arc fault, and the state point falls into the second region defined by the first remaining time parameter, the electromagnetic intervention module located at the estimated fault point position is sequentially controlled to generate a pulsed strong magnetic field. The electromagnetic intervention module is a pulsed magnetic field generating device, including an energy storage capacitor bank, a discharge switch, and a pulsed magnetic field coil. Multiple electromagnetic intervention modules are distributed in different spatial positions inside the transformer, and each module corresponds to a specific protection area. Specifically, the discharge switch is controlled to be turned on, the energy storage capacitor bank discharges to the pulsed magnetic field coil, and the discharge current generates a transient magnetic field in the space around the coil. The magnetic field strength is proportional to the discharge current amplitude. The parameters of the pulsed magnetic field include at least the peak magnetic field strength, the pulse rise time, and the pulse duration.

[0041] Preferably, the medium injection device and the electromagnetic intervention module are mechanically fixed to the same mounting base, or their injection ports are aligned with the central axis of the magnetic field coil. The injection direction of the medium injection device has a preset spatial geometric relationship with the constraint direction of the pulsed magnetic field. The medium injection device includes a first medium storage tank and a first injector, a second medium storage tank and a second injector, and an injection drive mechanism. The constrained arc channel refers to the fault arc plasma channel after radial compression by the pulsed strong magnetic field. The diameter of the compressed arc channel is smaller than that of the original arc channel, and the plasma density and temperature within the channel increase, making it easier to accept the injection and interaction of external media. The composite functional medium includes two types of media with different functions, injected sequentially in chronological order. The first type of electron affinity medium may be sulfur hexafluoride, perfluoroketone, perfluoroisobutyronitrile, etc., used to reduce the conductivity of the arc plasma. The second type of rapid curing medium may be room temperature curing silicone rubber, fast-drying epoxy resin, cyanoacrylate, etc., used to form a physical isolation layer within the arc channel. The control parameters include at least the injection start time, injection pressure, and injection duration of the first and second injectors.

[0042] Furthermore, step S500 also includes controlling the target electromagnetic intervention module to generate a pulsed strong magnetic field that meets the preset field strength and pulse width requirements in the spatial normal plane at the estimated location of the fault point, so as to produce a radial compression effect on the fault arc channel; during the peak effect stage of the pulsed strong magnetic field, the control medium injection device is activated to sequentially inject a first type of electron affinity medium and a second type of rapid curing medium into the compressed fault arc channel.

[0043] Preferably, from multiple electromagnetic intervention modules installed at different spatial locations inside the transformer, the module closest to the estimated fault location or with the optimal spatial geometric relationship between the magnetic field direction and the estimated fault location is selected as the target module. A trigger pulse signal is sent to its discharge switch. A two-dimensional plane perpendicular to the principal axis of the fault arc channel is constructed, with the estimated fault location as the origin and the principal axis direction as the normal direction, thus determining the spatial normal plane. The energy storage capacitor bank discharges into the pulsed magnetic field coil, and a transient large current flows through the coil. According to the Biot-Savart law, the current-carrying coil generates a time-varying magnetic field in the space surrounding it. At the estimated fault location, the magnetic field strength vector lies within the pre-designed spatial normal plane. The peak magnetic induction intensity of the magnetic field at the estimated location of the fault point must reach a preset threshold, such as 0.5 Tesla to 5 Tesla. The duration of the pulsed magnetic field must reach a preset width, typically 200 microseconds to 2 milliseconds. The magnetic field lines of the pulsed strong magnetic field are perpendicular to the arc axis. When the direction of motion of the charged particles is perpendicular to the direction of the magnetic field, the direction of the Lorentz force is perpendicular to both the velocity direction and the magnetic field direction, that is, it points to the radial inward side of the arc channel. The direction of this radial force is centripetal, pressing all charged particles toward the central axis of the arc channel. Under the action of the radial centripetal force, the charged particles gather toward the central axis, resulting in a reduction in the diameter of the arc channel. Furthermore, the resistance of the arc channel is inversely proportional to the cross-sectional area of ​​the channel, and compression leads to an increase in resistance.

[0044] Furthermore, step S500 also includes the following: the medium injection device includes an independent first medium storage tank and a first injector, and a second medium storage tank and a second injector; firstly, the first injector is controlled to spray the first type of electron affinity medium into the fault arc channel in an atomized form; after a preset delay time, the second injector is controlled to inject the second type of rapid curing medium into the fault arc channel.

[0045] Preferably, the peak effect phase is the time interval during which the magnetic field strength is higher than the preset peak threshold during the entire process from the start of the pulsed strong magnetic field to the end of its descent. During this time, the Lorentz force is at its maximum, and the radial compression effect is strongest. At the preset trigger moment within the peak effect phase, a start signal is sent to the drive circuit of the medium injection device, and a first type of electron affinity medium and a second type of rapid solidification medium are sequentially injected into the compressed fault arc channel. When the liquid or supercritical first type of medium passes through the nozzle, it is broken into a large number of tiny droplets due to the pressure drop and the special geometry of the nozzle, increasing the contact surface area between the medium and the arc plasma. The atomized medium droplets are ejected from the nozzle at high speed. The droplet jet passes through the thermal boundary layer outside the compressed arc channel and enters the core region of the arc plasma. The droplets are rapidly heated in the high-temperature plasma, and the electron affinity molecules capture free electrons in the arc channel to form negative ions, thereby reducing the conductivity. The preset delay time is a fixed time interval between the opening of the first injector and the opening of the second injector. It is calibrated in advance through experiments or simulations, with a typical value of 0.1 milliseconds to 5 milliseconds. When the preset delay time ends, an opening signal is sent to the valve drive circuit of the second injector. The second type of fast-curing medium is ejected from the nozzle of the second injector in the form of a liquid jet (not necessarily atomized). The jetting direction is also aligned with the central axis of the compressed arc channel. The fast-curing medium enters the arc channel, which has been weakened by the first type of medium, and undergoes a phase change or polymerization reaction at high temperature. The solid products fill the space of the arc channel, forming a physical isolation layer, sealing the fault point, and preventing the insulating oil from flowing back into the fault area.

[0046] In the above text, refer to Figure 1 An emergency protection method for an oil-immersed transformer according to an embodiment of the present invention is described in detail. Next, reference will be made to... Figure 2 An emergency protection device for an oil-immersed transformer according to an embodiment of the present invention is described.

[0047] The abnormal emergency protection device for oil-immersed transformers according to embodiments of the present invention addresses the technical problems in the prior art, such as delayed protection action response, inability to accurately locate and actively intervene in the nascent stage of fault energy, leading to structural damage to the transformer during internal arc faults. It achieves the technical effect of improving the accuracy and timeliness of graded active intervention and the transformer's ability to autonomously recover to a healthy state after fault disturbances. Figure 2 As shown, the emergency protection device for oil-immersed transformers includes: a signal acquisition module 10, a signal analysis module 20, a command matching module 30, a command fine-tuning module 40, and a fault handling module 50.

[0048] The signal acquisition module 10 is used to simultaneously acquire a first type of high-frequency transient signal and a second type of multi-physics field evolution signal characterizing the energy release process of insulation faults through a sensor network embedded in the transformer body; the signal analysis module 20 is used to simultaneously input the first type of high-frequency transient signal and the second type of multi-physics field evolution signal into a built-in multi-physics field coupling simulation model, and dynamically analyze and output the deduced fault type, the first remaining time parameter and the second remaining time parameter characterizing the current fault state and future evolution risk; the instruction matching module 30 is used to dynamically match graded active intervention instructions from a pre-stored multi-level intervention strategy library based on the first remaining time parameter, the second remaining time parameter and the deduced fault type, wherein the multi-level intervention strategy library includes at least energy cancellation strategy and arc surgery strategy; the instruction fine-tuning module 40 is used to adaptively fine-tune the graded active intervention instructions according to the confidence parameter of the deduced fault type and the health status parameter of the physical intervention device, and generate targeted intervention instructions; the fault processing module 50 is used to control the physical intervention device deployed at the estimated location of the fault point according to the targeted intervention instructions, and to perform targeted cancellation or blocking of the fault energy development process.

[0049] The signal acquisition module 10 further includes: a piezoelectric thin film sensor array embedded in the winding insulation with a non-uniform topology to acquire first-type high-frequency transient signals reflecting the stress wave characteristics of partial discharge and arc initiation stage; and second-type multi-physics field evolution signals reflecting temperature field gradient changes, bubble dynamics behavior and local pressure disturbances through an optical fiber sensor network and an immersion micro differential pressure sensor embedded in key locations of the transformer body.

[0050] The signal analysis module 20 further includes: utilizing the spatially directional stress wavefront information in the first type of high-frequency transient signal, performing dual-constraint localization of potential fault sources based on time-of-arrival difference and waveform attenuation characteristics in the multiphysics coupling simulation model to obtain an initial set of fault sources; using the spatiotemporal sequence in the second type of multiphysics evolution signal as a dynamic observation vector to perform recursive state updates on the preset fluid-thermal-chemical coupling equations in the multiphysics coupling simulation model; after each state update, using the simulation state of the previous moment as the initial value, running the updated fluid-thermal-chemical coupling equations in the finite-time domain. The rapid forward simulation simulates the energy release and propagation process of potential faults in the initial fault source set, generating forward simulation results. It analyzes the proportional relationship of energy conversion to preset gas-generating components during the rapid forward simulation, compares it with preset fault characteristic spectra, and determines the type of fault being simulated. Based on the spatiotemporal extrema of the mechanical stress field of the tank and the thermal aging rate field of the insulating material in the forward simulation results, it calculates a first remaining time parameter and a second remaining time parameter. The first remaining time parameter is the time when the simulated pressure reaches the tank structure failure threshold, and the second remaining time parameter is the time when the simulated thermal damage to the insulating material accumulates to the irreversible threshold.

[0051] The instruction matching module 30 further includes: constructing a dynamic decision plane with the first remaining time parameter and the second remaining time parameter as input coordinates; when the simulated fault type is a local overheating fault and the state point falls into the first region defined by the second remaining time parameter in the decision plane, matching the graded active intervention instruction corresponding to the energy cancellation strategy; when the simulated fault type is a developing arc fault and the state point falls into the second region defined by the first remaining time parameter in the decision plane, matching the graded active intervention instruction corresponding to the arc surgery strategy; wherein the first region and the second region partially overlap.

[0052] Below, the instruction matching module 30 further includes: when the state point falls into the overlapping part of the first region and the second region, and the inferred fault type has both overheating and arc characteristics, the graded active intervention instruction corresponding to the matched arc surgery strategy must include a preparatory cooling stage, which is executed before the generation of the pulsed strong magnetic field.

[0053] The specific configuration of the instruction fine-tuning module 40 will be described in detail below. The instruction fine-tuning module 40 further includes: acquiring a confidence parameter characterizing the reliability of the current simulation results of the multiphysics coupling simulation model, and a health status parameter characterizing the current usability of the target physical intervention device; scaling the preset intervention intensity level in the graded active intervention instruction based on the confidence parameter; and extracting the control parameters related to the target physical intervention device from the scaled graded active intervention instruction based on the health status parameter, performing redundant backups or downgraded execution adjustments, and generating the targeted intervention instruction.

[0054] The fault handling module 50 further includes: when the targeted intervention command originates from the energy cancellation strategy, controlling the microchannel release unit located at the presumed fault point to release the phase change cooling medium into the target area; when the targeted intervention command originates from the electric arc surgery strategy, sequentially controlling the electromagnetic intervention module located at the presumed fault point to generate a pulsed strong magnetic field, and the integrated medium injection device to inject the composite functional medium into the constrained electric arc channel.

[0055] Below, the fault handling module 50 further includes: a control target electromagnetic intervention module, which generates a pulsed strong magnetic field that meets the preset field strength and pulse width requirements in the spatial normal plane at the estimated location of the fault point, so as to produce a radial compression effect on the fault arc channel; during the peak effect stage of the pulsed strong magnetic field, the control medium injection device is activated to sequentially inject a first type of electron affinity medium and a second type of rapid curing medium into the compressed fault arc channel.

[0056] The fault handling module 50 further includes: the medium injection device includes an independent first medium storage tank and a first injector, and a second medium storage tank and a second injector; firstly, the first injector is controlled to spray the first type of electron affinity medium into the fault arc channel in an atomized form; after a preset delay time, the second injector is controlled to inject the second type of rapid curing medium into the fault arc channel.

[0057] The abnormal emergency protection device for oil-immersed transformers provided in the embodiments of the present invention can execute the abnormal emergency protection method for oil-immersed transformers provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0058] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An emergency protection method for abnormal conditions in oil-immersed transformers, characterized in that, The method includes: By using a sensor network embedded in the transformer body, the first type of high-frequency transient signal and the second type of multi-physics field evolution signal characterizing the energy release process of insulation faults are collected simultaneously. The first type of high-frequency transient signal and the second type of multiphysics field evolution signal are synchronously input into a built-in multiphysics field coupling simulation model, and the model is dynamically analyzed and output to represent the current fault state and future evolution risk, the inferred fault type, the first remaining time parameter and the second remaining time parameter. Based on the first remaining time parameter, the second remaining time parameter, and the deduced fault type, hierarchical active intervention instructions are dynamically matched from the pre-stored multi-level intervention strategy library, wherein the multi-level intervention strategy library includes at least energy cancellation strategy and arc surgery strategy. Based on the confidence parameters of the inferred fault type and the health status parameters of the physical intervention device, the graded active intervention instructions are adaptively fine-tuned to generate targeted intervention instructions. According to the targeted intervention command, the physical intervention device deployed at the presumed location of the fault point is controlled to directionally cancel or block the development process of the fault energy.

2. The emergency protection method for an oil-immersed transformer as described in claim 1, characterized in that, Through a sensor network embedded in the transformer body, the system simultaneously acquires first-type high-frequency transient signals and second-type multi-physics evolution signals characterizing the energy release process of insulation faults, including: A piezoelectric thin-film sensor array with a non-uniform topology embedded inside the winding insulation is used to collect first-class high-frequency transient signals that reflect the stress wave characteristics of partial discharge and arc initiation stages. By using fiber optic sensor networks and immersion micro differential pressure sensors embedded in key locations of the transformer body, second-type multiphysics field evolution signals reflecting temperature field gradient changes, bubble dynamics, and local pressure disturbances are collected.

3. The emergency protection method for an oil-immersed transformer as described in claim 1, characterized in that, The first type of high-frequency transient signal and the second type of multiphysics field evolution signal are synchronously input into a built-in multiphysics field coupled simulation model. The model dynamically analyzes and outputs the deduced fault type, the first remaining time parameter, and the second remaining time parameter, which characterize the current fault state and future evolution risk. Using the spatially directional stress wavefront information in the first type of high-frequency transient signal, the potential fault source is located in a multi-physics coupled simulation model based on dual constraints of time of arrival difference and waveform attenuation characteristics, thus obtaining an initial set of fault sources. The spatiotemporal sequence in the second type of multiphysics evolution signal is used as a dynamic observation vector to perform recursive state updates on the preset fluid-thermal-chemical coupling equations in the multiphysics coupling simulation model. After each state update, using the simulation state of the previous moment as the initial value, the updated fluid-thermal-chemical coupling equations are run to perform a fast forward deduction in the finite time domain, simulating the energy release and propagation process of potential faults in the initial fault source set, and generating forward deduction results. Analyze the proportion of energy converted to the preset gas-producing components during the rapid forward simulation process, compare it with the preset fault characteristic spectrum, and determine the type of fault being simulated; Based on the spatiotemporal extrema of the mechanical stress field of the oil tank and the thermal aging rate field of the insulating material in the forward simulation results, the first remaining time parameter and the second remaining time parameter are calculated. The first remaining time parameter is the time when the simulated pressure reaches the failure threshold of the oil tank structure, and the second remaining time parameter is the time when the simulated thermal damage of the insulating material accumulates to the irreversible threshold.

4. The emergency protection method for an oil-immersed transformer as described in claim 1, characterized in that, Based on the first remaining time parameter, the second remaining time parameter, and the simulated fault type, hierarchical active intervention instructions are dynamically matched from a pre-stored multi-level intervention strategy library. The multi-level intervention strategy library includes at least energy cancellation strategies and arc surgery strategies, including: Construct a dynamic decision plane with the first remaining time parameter and the second remaining time parameter as input coordinates; When the simulated fault type is a local overheating fault, and the state point falls into the first region defined by the second remaining time parameter in the decision plane, the graded active intervention command corresponding to the energy offsetting strategy is matched. When the inferred fault type is a progressive arc fault, and the state point falls into the second region defined by the first remaining time parameter in the decision plane, the graded active intervention instruction corresponding to the arc surgery strategy is matched. The first region and the second region partially overlap.

5. The emergency protection method for an oil-immersed transformer as described in claim 4, characterized in that, When the state point falls into the overlapping part of the first region and the second region, and the inferred fault type has both overheating and arc characteristics, the graded active intervention instruction corresponding to the matched arc surgery strategy must include a preparatory cooling stage, which is executed before the generation of the pulsed strong magnetic field.

6. The emergency protection method for an oil-immersed transformer as described in claim 1, characterized in that, Based on the confidence parameters of the inferred fault type and the health status parameters of the physical intervention device, the graded active intervention instructions are adaptively fine-tuned to generate targeted intervention instructions, including: Obtain the confidence level parameter characterizing the reliability of the current inference results of the multiphysics coupling simulation model, and the health status parameter characterizing the current usability of the target physical intervention device; Based on the confidence level parameter, the preset intervention intensity level in the graded active intervention instruction is scaled proportionally. Based on the health status parameters, the control parameters related to the target physical intervention device are extracted from the scaled-down hierarchical active intervention instructions, and redundant backups or downgraded execution adjustments are performed to generate the targeted intervention instructions.

7. The emergency protection method for an oil-immersed transformer as described in claim 1, characterized in that, According to the targeted intervention command, control the physical intervention device deployed at the presumed location of the fault point to directionally cancel or block the development process of fault energy, including: When the targeted intervention command originates from the energy cancellation strategy, the micro-pipe release unit located at the estimated fault point is controlled to release the phase change cooling medium into the target area. When the targeted intervention command originates from the electric arc surgery strategy, the electromagnetic intervention module located at the presumed fault point is sequentially controlled to generate a pulsed strong magnetic field, and the integrated media injection device injects a composite functional medium into the constrained electric arc channel.

8. The emergency protection method for an oil-immersed transformer as described in claim 7, characterized in that, The electromagnetic intervention module located at the presumed fault point is sequentially controlled to generate a pulsed strong magnetic field, and the integrated media injection device is controlled to inject a composite functional medium into the constrained arc channel, including: The electromagnetic intervention module of the control target is made to generate a pulsed strong magnetic field that meets the preset field strength and pulse width requirements in the spatial normal plane at the estimated location of the fault point, so as to produce a radial compression effect on the fault arc channel. During the peak phase of the pulsed strong magnetic field, the control medium injection device is activated, sequentially injecting a first type of electron affinity medium and a second type of rapid curing medium into the compressed fault arc channel.

9. The emergency protection method for an oil-immersed transformer as described in claim 8, characterized in that, A first-type electron affinity medium and a second-type rapid curing medium are sequentially injected into the compressed fault arc channel, including: The medium injection device includes a separate first medium storage tank and a first injector, and a second medium storage tank and a second injector; First, control the first injector to spray the first type of electron affinity medium into the fault arc channel in an atomized form; After a preset delay time, the second injector is controlled to inject the second type of rapid curing medium into the fault arc channel.

10. An emergency protection device for abnormal conditions in an oil-immersed transformer, characterized in that, The device is used to implement the abnormal emergency protection method for an oil-immersed transformer according to any one of claims 1 to 9, the device comprising: The signal acquisition module is used to simultaneously acquire the first type of high-frequency transient signal and the second type of multi-physics field evolution signal, which characterize the energy release process of insulation faults, through the sensing network embedded in the transformer body. The signal analysis module is used to synchronously input the first type of high-frequency transient signal and the second type of multiphysics field evolution signal into a built-in multiphysics field coupling simulation model, and dynamically analyze and output the deduced fault type, the first remaining time parameter and the second remaining time parameter that characterize the current fault state and the future evolution risk. The instruction matching module is used to dynamically match graded active intervention instructions from a pre-stored multi-level intervention strategy library based on the first remaining time parameter, the second remaining time parameter and the simulated fault type. The multi-level intervention strategy library includes at least energy cancellation strategy and arc surgery strategy. The instruction fine-tuning module is used to adaptively fine-tune the graded active intervention instruction based on the confidence parameter of the inferred fault type and the health status parameter of the physical intervention device, and generate a targeted intervention instruction. The fault handling module is used to control a physical intervention device deployed at the presumed location of the fault point according to the targeted intervention command, so as to directionally cancel or block the development process of the fault energy.

Images