A power equipment operation state monitoring method based on digital twinning

Abstract

This application relates to a method for monitoring the operating status of power equipment based on digital twins. The method includes: digitally coupling and modeling the operating status and deterioration status of the target power equipment based on its attribute data and operating data to obtain an initial digital twin; performing surface-enhanced Raman scattering analysis on furfural molecular information in insulating oil sample data to obtain measured concentration data and Raman spectral characteristic data; anchoring the oil-paper insulation aging state of the initial digital twin based on the measured concentration data and Raman spectral characteristic data to obtain a state-anchored digital twin; performing insulation aging trajectory offset analysis on the state-anchored digital twin based on historical operating data and real-time operating status data to obtain state drift characteristic quantities; and analyzing the operating status based on the state drift characteristic quantities to obtain equipment status characterization data. This method can improve the quality of power equipment condition monitoring.

Description

Technical Field

[0001] This application relates to the field of smart grid technology, and in particular to a method for monitoring the operating status of power equipment based on digital twins. Background Technology

[0002] The insulation system of a power transformer mainly consists of insulating oil and insulating paper. During the electrothermal aging process, the insulating paper produces aging characteristic substances such as furfural, acetone, and methanol, which subsequently dissolve in the insulating oil. Furfural, as one of the important indicators of oil-paper insulation aging, plays a crucial role in assessing the aging condition of transformers. According to DL / T596-1996 "Preventive Testing of Power Equipment," different levels of furfural content in mineral oil correspond to different aging states of insulating paper: 0.1 mg / L indicates mild aging, 0.5 mg / L indicates moderate aging, and 4 mg / L indicates severe aging. Currently, the main methods for detecting furfural content in oil both domestically and internationally include: high-performance liquid chromatography (HPLC), electrochemical methods, ultraviolet absorption spectroscopy, spectrophotometry, gas chromatography, and infrared spectroscopy. However, most of these methods require manual extraction and analysis in a laboratory environment, making them unsuitable for long-term monitoring, resulting in low quality of power equipment condition monitoring. Summary of the Invention

[0003] Therefore, it is necessary to provide a digital twin-based method for monitoring the operating status of power equipment that can improve the quality of power equipment condition monitoring, addressing the aforementioned technical problems.

[0004] This application provides a method for monitoring the operating status of power equipment based on digital twins, including: Based on the equipment attribute data and equipment operation data of the target power equipment, a digital coupling model is performed on the equipment operation status and equipment degradation status of the target power equipment to obtain an initial digital twin; Surface-enhanced Raman scattering analysis was performed on the furfural molecule information in the insulating oil sample data of the target power equipment to obtain the measured concentration data and Raman spectral characteristic data. Based on the measured concentration data and the Raman spectral characteristic data, the initial digital twin is anchored to the aging state of the paper insulation to obtain a state-anchored digital twin. Based on the historical operating data and real-time operating status data in the equipment operating data, insulation aging trajectory offset analysis is performed on the state anchoring digital twin to obtain state drift characteristic quantities. Based on the state drift characteristics, the operating state of the target power equipment is analyzed to obtain equipment state characterization data.

[0005] The aforementioned method for monitoring the operational status of power equipment based on digital twins constructs an initial digital twin using equipment attribute data and operational data. This allows the operational status, thermal aging process, and degradation evolution of the target power equipment to form a unified coupled model in virtual space, avoiding reliance on a single monitoring quantity for status assessment. Then, surface-enhanced Raman scattering analysis is performed on furfural molecules in insulating oil samples to obtain measured concentration data and Raman spectral characteristic data directly related to the aging of oil-paper insulation, improving the sensitivity and accuracy of detecting trace aging products. Next, the measured concentration data and Raman spectral characteristic data are used to anchor the oil-paper insulation aging status of the initial digital twin, ensuring that the digital twin is no longer confined to a theoretical modeling state but remains synchronized with the actual insulation aging status of the target power equipment. Further, insulation aging trajectory deviation analysis is conducted by combining historical operational data and real-time operational status data to obtain state drift characteristics, reflecting the degree, direction, and speed of deviation of the equipment degradation path relative to the normal aging trajectory. Finally, equipment status characterization data is generated based on the state drift characteristics, enabling a comprehensive assessment of the target power equipment's health status, insulation aging risk, remaining lifespan, and maintenance strategies. Overall, the technology upgrades from "single detection" to "simultaneous virtual and real monitoring", from "concentration judgment" to "deterioration trajectory analysis", and from "post-event diagnosis" to "predictive operation and maintenance" can effectively improve the real-time performance, accuracy, and proactive early warning capabilities of power equipment condition monitoring, thereby improving the quality of power equipment condition monitoring. Attached Figure Description

[0006] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0007] Figure 1 This is an application environment diagram of a power equipment operation status monitoring method based on digital twins in one embodiment; Figure 2 This is a flowchart illustrating a digital twin-based method for monitoring the operational status of power equipment in one embodiment. Figure 3 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0008] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0009] This application provides a method for monitoring the operating status of power equipment based on digital twins, which can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on other network servers. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0010] In one exemplary embodiment, such as Figure 2 As shown, a method for monitoring the operating status of power equipment based on digital twins is provided, which can be applied to... Figure 1 Taking the server in the example, the explanation includes the following steps 202 to 210. Wherein:

[0011] Step 202: Based on the equipment attribute data and equipment operation data of the target power equipment, perform digital coupling modeling on the equipment operation status and equipment degradation status of the target power equipment to obtain an initial digital twin; Step 204: Perform surface-enhanced Raman scattering analysis on the furfural molecule information in the insulating oil sample data of the target power equipment to obtain the measured concentration data and Raman spectral characteristic data. Step 206: Based on the historical operating data and real-time operating status data in the equipment operating data, perform insulation aging trajectory offset analysis on the state anchoring digital twin to obtain state drift characteristic quantities. Step 208: Based on historical operating data and real-time operating status data, perform insulation aging trajectory offset analysis on the state anchoring digital twin to obtain state drift characteristic quantities. Step 210: Analyze the operating status of the target power equipment based on the state drift characteristic quantities to obtain equipment state characterization data.

[0012] Among them, the target power equipment refers to the power equipment that needs to be monitored for operating status and assessed for insulation aging, preferably oil-immersed power transformers.

[0013] Among them, equipment attribute data refers to the inherent structure, materials, rated parameters and commissioning information of the target power equipment, including model, capacity, voltage level, winding structure, insulating oil type, insulating paper material and commissioning years.

[0014] Among them, equipment operation data refers to the operating condition data generated or collected by the target power equipment during operation, including load current, voltage, power, oil temperature, winding hot spot temperature, cooling status and ambient temperature.

[0015] Among them, the equipment operating status refers to the electrical, thermal and load operating conditions of the target power equipment under current or historical operating conditions.

[0016] Among them, the equipment deterioration state refers to the aging, performance degradation or risk accumulation of the insulating oil, insulating paper and related components of the target power equipment due to thermal, electrical, chemical or mechanical stress.

[0017] Digital coupling modeling refers to a processing method that correlates and models the structural parameters, operating conditions, thermal field changes, insulation aging, and risk evolution processes of target power equipment in digital space.

[0018] The initial digital twin refers to a virtual model of the target power equipment established based on equipment attribute data and equipment operation data, which has not yet been corrected for measured furfural concentration and Raman spectral characteristics.

[0019] Among them, insulating oil sample data refers to the oil sample information obtained from the insulating oil of the target power equipment and used for detection and analysis, including oil sample volume, sampling location, sampling time, oil temperature and oil sample spectral response.

[0020] Among them, furfural molecular information refers to information related to furfural molecule content, adsorption state, Raman response, and spectral peak changes in insulating oil. In one embodiment, 1 mL of insulating oil sample from the target power equipment is placed in a constant-temperature microreactor and brought into contact with a gold / silver composite nanostructure SERS substrate with an amino or 4-ATP trapping layer on its surface at 60°C for 3 min, allowing furfural molecules in the insulating oil to undergo selective adsorption or condensation reactions with the trapping layer; subsequently, a 532 nm Raman laser is used to scan multiple preset micro-regions on the substrate surface, with the laser power set to 30 mW and the integration time to 1... The cumulative integration was performed 10 times, and Raman spectra of blank insulating oil, standard furfural oil samples, and the insulating oil sample to be tested were collected. After dark current subtraction, fluorescence background stripping, baseline correction, hotspot drift compensation, and wavenumber alignment of the collected spectra, peak position, peak intensity, peak area, peak width, peak intensity ratio, and peak area ratio were calculated within a preset furfural candidate characteristic peak window. These values ​​were then matched with a preset concentration grid library to obtain specific information related to furfural molecule content, adsorption state, Raman response, and peak changes. For example, for a certain insulating oil sample, the furfural molecule content information could be a furfural concentration of 0.48 mg / L obtained through concentration grid quantization; the adsorption state information could be the characteristic peak areas of furfural molecules at 0 s, 60 s, and 180 s contact with the SERS substrate, corresponding to A0, A60, and A180 respectively, with A180 / A60 being 1.18, indicating that furfural molecules tend to be stably enriched at the substrate capture sites; and the Raman response information could be the furfural candidate characteristic peak at approximately 1502 cm⁻¹. -1 The peak intensity at this point is 12500 au, and the peak area is 2.8 × 10⁻⁶.5 With an au·cm⁻¹, a half-maximum width of 14 cm⁻¹, and a signal-to-noise ratio of 32, the spectral peak change information can be seen as the presence of an enhanced peak in the tested insulating oil sample relative to the blank insulating oil within the preset furfural peak window, a 2 cm⁻¹ shift in peak position from the corresponding position in the standard spectrum, and a 15% increase in peak area compared to the previous detection cycle. Thus, the furfural molecule information is obtained from the insulating oil sample after selective enrichment on a SERS substrate, Raman scanning, spectral correction, and concentration grid quantization.

[0021] Among them, surface-enhanced Raman scattering analysis refers to an analytical method that uses a metal nanostructure substrate with an enhancing effect to enhance the detection of the Raman signal of furfural molecules.

[0022] Among them, the measured concentration data refers to the simulated actual detection concentration results of furfural molecules in insulating oil obtained through surface-enhanced Raman scattering analysis.

[0023] Raman spectral characteristic data refers to spectral parameters such as peak position, peak intensity, peak area, peak width, peak intensity ratio, and peak topological relationship formed by furfural molecules during Raman detection.

[0024] Among them, the aging state anchoring of oil-paper insulation refers to the correction and positioning of the aging parameters of oil-paper insulation in the digital twin using measured concentration data and Raman spectral characteristic data.

[0025] Among them, the state-anchored digital twin refers to the digital twin that corresponds to the current aging state of the oil-paper insulation of the target power equipment after being corrected by measured furfural concentration data and Raman spectral characteristic data.

[0026] Historical operating data refers to the operating record data of the target power equipment before the current monitoring time, including historical load, voltage, current, oil temperature, winding hot spot temperature, cooling status, alarm records, maintenance records, and historical oil sample test results.

[0027] Among them, the real-time operating status data is the operating data collected by the target power equipment at the current monitoring time or within the current sampling period, including real-time load, voltage, current, oil temperature, winding hot spot temperature, ambient temperature, cooler status, partial discharge status, and current alarm information.

[0028] Among them, insulation aging trajectory offset analysis refers to the analysis process of comparing the difference between the actual aging evolution path of the target power equipment and the aging path predicted by the digital twin.

[0029] Among them, the state drift characteristic refers to the set of parameters used to describe the degree of deviation of the aging trajectory of oil paper insulation, including the offset direction, offset amplitude, offset rate, duration and source weight.

[0030] Among them, operational status analysis refers to making a comprehensive judgment on the current and future operational risks, insulation aging degree and operation and maintenance needs of the target power equipment based on the state drift characteristic quantities.

[0031] Among them, equipment status characterization data refers to the data used to output the status results of target power equipment, including health status level, insulation aging risk level, remaining lifespan, early warning level, and operation and maintenance strategy.

[0032] Specifically, the equipment attribute data of the target power equipment is structured, including equipment model, rated capacity, rated voltage, winding connection method, core structure parameters, cooling method, insulating oil type, insulating paper material parameters, oil-to-paper mass ratio, sensor deployment location, and years of operation. Then, the equipment operation data undergoes time synchronization and dimensional unification processing, including load current, voltage, active power, reactive power, top oil temperature, winding hot spot temperature, ambient temperature, cooler start / stop status, and historical alarm records. The structural parameters from the equipment attribute data are then written into the equipment geometric topology layer, and the electrical and thermal state quantities from the equipment operation data are written into the operation state layer. Multi-physics domain coupling relationships are established using a thermal field calculation model, a load loss model, an oil-to-paper insulation aging kinetic model, and a fault risk evolution model. Specifically, the thermal field calculation model is used to calculate the oil temperature and winding hot spot temperature distribution under different load conditions; the oil-to-paper insulation aging kinetic model is used to calculate the polymerization degree decay process of the insulating paper and the generation process of aging products; and the fault risk evolution model is used to calculate the coupling effects between thermal stress, electrical stress, and chemical degradation. By binding parameters in the aforementioned geometric topology layer, operating status layer, thermal aging layer, and risk evolution layer, an initial digital twin capable of changing with updated operating data is formed. Specific details regarding equipment attribute data include: structuring the equipment attribute data of the target power equipment, which includes equipment model, rated capacity, rated voltage, rated current, winding connection method, core structure parameters, cooling method, insulating oil type, insulating paper material parameters, oil-to-paper mass ratio, sensor placement location, service life, rated loss parameters, rated temperature rise parameters, and historical maintenance parameters; time synchronization and dimensional unification processing of equipment operating data, which includes load current, voltage, active power, reactive power, top oil temperature, winding hot spot temperature, ambient temperature, cooler start / stop status, and historical alarm records; and then writing the structural parameters, material parameters, and rated parameters from the equipment attribute data into the equipment geometric topology layer, and incorporating the load current, voltage, power, oil temperature, and winding parameters from the equipment operating data into the equipment geometric topology layer. Hot spot temperature and cooling status are written into the operating status layer, and a hierarchical coupling relationship of electrical domain, thermal domain, chemical aging domain, and risk domain is established between the operating status layer, thermal aging layer, and risk evolution layer. Specifically, the load loss model converts the current electrical load of the target power equipment into the heat intensity under the current operating conditions based on the load current, rated current, no-load loss, load loss, and cooler operating status, which is used to illustrate the impact of load changes on the heat source intensity inside the equipment. The thermal field calculation model calculates the top oil temperature change trend and winding hot spot temperature change trend of the target power equipment under the current load conditions based on the heat intensity, ambient temperature, cooling method, top oil temperature, winding hot spot temperature, and the thermal inertia relationship between oil and winding, which is used to illustrate how the electrical load further causes the oil temperature and winding hot spot temperature to rise through heat loss.The aging kinetic model for oil-paper insulation specifically calculates the process of decreasing the degree of polymerization of the insulation paper and the formation of aging products such as furfural based on winding hot spot temperature, top oil temperature, insulation paper material parameters, insulation oil type, water content, oxidation state, and operating duration. This model illustrates how thermal stress and chemical environment drive the deterioration of the oil-paper insulation system. The fault risk evolution model classifies the deterioration risk of the target power equipment based on changes in winding hot spot temperature, load impact intensity, degree of decrease in the degree of polymerization of the insulation paper, furfural formation level, and historical alarm information. This model illustrates the risk changes under the combined effects of thermal aging, electrical stress, and chemical deterioration. The multi-physics domain coupling relationship is not a series of separate, parallel models. Instead, it refers to the load current and power in the electrical domain being first converted into internal heat intensity through a load loss model, and then the thermal field calculation model in the thermal domain being used to calculate the heat intensity and cooling status. The system calculates oil temperature and winding hotspot temperature in the dynamic domain. In the chemical aging domain, the aging kinetics model of the oil-paper insulation calculates the polymerization degree decay and furfural formation of the insulation paper based on oil temperature and hotspot temperature. In the risk domain, the fault risk evolution model calculates equipment degradation risk based on temperature changes, polymerization degree decay, furfural formation, and load impact. This forms a continuous parameter transfer relationship of "operating load change—loss and heat generation change—oil temperature and hotspot temperature change—oil-paper insulation aging change—aging product formation change—equipment risk change." By associating and binding the equipment structure and material parameters in the geometric topology layer, the real-time electrical and thermal quantities in the operating state layer, the oil temperature, hotspot temperature, insulation paper polymerization degree, and furfural formation state in the thermal aging layer, and the risk level and degradation trend in the risk evolution layer at the same sampling time, an initial digital twin that can be synchronously updated with equipment operating data is obtained.

[0033] Insulating oil samples from the target power equipment were introduced into a temperature-controlled microreactor, with the oil sample volume, reaction temperature, and contact time controlled to ensure full contact between the insulating oil sample and a functionalized surface-enhanced Raman scattering (SERS) substrate. The SERS substrate could be a noble metal nanostructure substrate with a molecular trapping layer on its surface capable of selectively interacting with furfural molecules. The microreactor was then sealed and temperature-controlled, allowing furfural molecules to be directionally enriched between the oil phase medium, the trapping layer, and the enhanced hotspots. A micro Raman spectroscopy module was used to perform multi-point, multi-temporal scanning of the substrate surface after the reaction. During the scanning process, laser wavelength, laser power, integration time, number of iterations, substrate temperature, and scanning position were recorded simultaneously. Subsequently, the obtained raw Raman spectra were processed by dark current subtraction, cosmic ray elimination, fluorescence background correction, oil matrix interference correction, and hot spot drift compensation to obtain furfural purified Raman spectrum data. Then, based on the preset furfural calibration spectrum, concentration grid library, and peak topological constraint relationship, the candidate peak regions in the furfural purified Raman spectrum data were subjected to peak topological locking and concentration grid quantization processing to obtain the measured concentration data of furfural molecules and Raman spectral characteristic data.

[0034] The measured concentration data of furfural molecules were used as the chemical anchoring quantification for the oil-paper insulation aging model, and Raman spectral characteristic data were used as the spectroscopic anchoring quantification for the microscopic aging information of the oil sample. Both were jointly incorporated into the insulation aging state layer of the initial digital twin. An anchoring objective function was established in the initial digital twin, which included at least a furfural concentration deviation term, a spectral topological similarity term, a temperature aging consistency term, and an operating load constraint term. The furfural concentration deviation term constrained the furfural level in the oil calculated by the digital twin; the spectral topological similarity term constrained the spectral changes of the oil sample aging products; the temperature aging consistency term constrained the relationship between hot spot temperature and aging rate; and the operating load constraint term constrained the influence of load impact on the insulation aging rate. Then, particle filtering, ensemble Kalman update, or multi-objective constraint optimization methods were used to update the equivalent degree of polymerization of the insulating paper, furfural formation coefficient, furfural-oil paper distribution coefficient, aging reaction rate constant, and hot spot aging acceleration factor in the initial digital twin, so that the oil-paper insulation aging state of the digital twin was consistent with the results of this SERS detection. After the update is completed, a state-anchored digital twin with the current aging state of the oil-paper insulation is obtained.

[0035] Historical operating data and real-time operating status data are processed to a unified time scale, mapping historical load, historical oil temperature, historical winding hot spot temperature, historical cooling status, and real-time load, real-time oil temperature, real-time winding hot spot temperature, and real-time ambient temperature onto the same operating time axis. Using the current oil-paper insulation aging state in the state-anchored digital twin as the starting boundary, historical stress accumulation paths and real-time stress driving paths are calculated within the digital twin. Furthermore, the historical stress accumulation path is used to calculate the insulation aging trajectory of the equipment under past operating conditions, and the real-time stress driving path is used to calculate the short-term aging trajectory under the continued action of current operating conditions. Then, time axis elastic registration and aging phase alignment are performed on the two types of trajectories, calculating the trajectory direction difference, trajectory amplitude difference, trajectory curvature difference, furfural growth rate difference, and hot spot aging acceleration difference under the same aging phase. These differences are then combined into a state drift vector, and state drift characteristic quantities are constructed using drift direction, drift amplitude, drift rate, and drift duration. This state drift characteristic includes not only the shift caused by changes in furfural concentration, but also the combined shift caused by load impact, thermal aging accumulation, and changes in the degradation path of oil-paper insulation.To address the differences in trajectory direction, amplitude, curvature, furfural growth rate, and hotspot aging acceleration under the same aging phase, both historical stress accumulation paths and real-time stress-driven paths are transformed into a unified aging phase coordinate system. The aging phase coordinate system is jointly determined by the degree of decrease in the equivalent degree of polymerization of the insulating paper, the cumulative change in furfural concentration, the cumulative effect of hotspot temperature, and the normalization progress of the running time. Specifically, for each sampling point in both types of trajectories, the running time, furfural concentration, furfural concentration change, hotspot temperature, hotspot temperature change, cumulative insulation aging, and corresponding aging phase value are recorded. Then, the aging phase values ​​are used instead of the original time sequence. The two types of trajectories are rearranged. When the sampling time intervals of the two types of trajectories are inconsistent, linear interpolation or sliding window interpolation between adjacent aging phase points is used to adjust the trajectory points on the historical stress accumulation path and the trajectory points on the real-time stress driving path to the same aging phase position, thereby completing the time axis elastic registration and aging phase alignment. At the same aging phase position, the trajectory direction difference is determined by comparing the change direction of adjacent points before and after that phase of the two types of trajectories. That is, the growth direction and growth rate of the historical stress accumulation path and the real-time stress driving path in the three dimensions of furfural concentration, hot spot temperature, and insulation aging accumulation are determined respectively, and the degree of inconsistency in direction or growth rate between the two is classified. The trajectory direction difference is defined as follows: the trajectory amplitude difference is determined by comparing the magnitudes of the state variables of the two types of trajectories at the same aging phase position. Specifically, the differences between the furfural concentration, hotspot temperature, and cumulative insulation aging amount corresponding to the real-time stress-driven path at that phase and the corresponding values ​​of the historical stress accumulation path are calculated, and the differences of each state variable are combined according to a preset weight to form the trajectory amplitude difference. The trajectory curvature difference is determined by comparing the curvature changes of the two types of trajectories at multiple consecutive aging phase positions. Specifically, the state changes between the previous phase, the current phase, and the next phase are observed to see whether they show a steady increase, an accelerated increase, or a decelerated increase, and the difference in the degree of curvature between the historical stress accumulation path and the real-time stress-driven path is used as the trajectory curvature difference. The difference in furfural growth rate is determined by comparing the increase in furfural concentration with phase advancement within the same aging phase window for the two types of trajectories. Specifically, the increase rate of furfural concentration in adjacent aging phase windows is calculated for the historical stress accumulation path and the real-time stress-driven path, and the difference in their increase rates is taken as the furfural growth rate difference. The difference in hot spot aging acceleration is determined by comparing the degree of acceleration of the hot spot temperature aging contribution of the two types of trajectories within a continuous aging phase window. Specifically, the change in the contribution of hot spot temperature to the cumulative amount of insulation aging is calculated first, and then the change in this contribution is compared between adjacent windows to see if it accelerates. The difference in the degree of acceleration between the historical stress accumulation path and the real-time stress-driven path is taken as the hot spot aging acceleration difference.

[0036] The state drift characteristics are input into the risk assessment layer of the state-anchored digital twin. Risk levels are calculated for the aging drift direction, amplitude, rate, and duration to form a drift risk level map. Then, based on the drift risk level map, the operating state of the target power equipment is divided into multi-state gating partitions, categorizing the current state into normal operation, mild aging, accelerated degradation, latent fault, or critical risk zones. Next, a twin shadow simulation model is run within each state partition to simulate the subsequent degradation paths of the target power equipment under different maintenance strategies, such as maintaining the current load, reducing the load, enhancing cooling, shortening the inspection cycle, and scheduling maintenance, resulting in candidate risk evolution path clusters. Finally, risk path resonance screening and state result folding encoding are performed on the candidate risk evolution path clusters, encoding the health status level, insulation aging risk level, remaining lifespan, warning level, recommended inspection cycle, and recommended maintenance strategy into unified equipment state characterization data.

[0037] In the aforementioned method for monitoring the operational status of power equipment based on digital twins, an initial digital twin is constructed using equipment attribute data and equipment operation data. This allows the operational status, thermal aging process, and degradation evolution process of the target power equipment to form a unified coupled model in virtual space, avoiding reliance on a single monitoring quantity for status judgment. Then, surface-enhanced Raman scattering analysis is performed on furfural molecular information in insulating oil samples to obtain measured concentration data and Raman spectral characteristic data directly related to the aging of oil-paper insulation, improving the sensitivity and accuracy of detecting trace aging products. Finally, the measured concentration data and Raman spectral characteristics are used... The initial digital twin is anchored to the aging state of its paper insulation using data, ensuring it no longer remains merely a theoretical model but synchronizes with the actual insulation aging state of the target power equipment. Further analysis of insulation aging trajectory deviation, combining historical and real-time operational data, yields state drift characteristics, reflecting the degree, direction, and speed of deviation of the equipment's degradation path relative to the normal aging trajectory. Finally, equipment condition characterization data is generated based on these characteristics, enabling a comprehensive assessment of the target power equipment's health status, insulation aging risk, remaining lifespan, and maintenance strategies. Overall, this represents a technological upgrade from "single detection" to "synchronous virtual-real monitoring," from "concentration judgment" to "degradation trajectory analysis," and from "post-event diagnosis" to "predictive maintenance." This effectively improves the real-time performance, accuracy, and proactive early warning capabilities of power equipment condition monitoring, thereby enhancing the overall quality of power equipment condition monitoring.

[0038] In an exemplary embodiment, based on historical operating data and real-time operating status data, insulation aging trajectory offset analysis is performed on the state-anchored digital twin to obtain state drift characteristic quantities, including steps 302 to 306. Wherein:

[0039] Step 302: Based on historical operating data and real-time operating status data, perform trend analysis on the aging evolution process of oil-paper insulation in the state-anchored digital twin to obtain insulation aging trend data.

[0040] Step 304: Based on the insulation aging trend data, predict and analyze the evolution process of furfural concentration in the state-anchored digital twin to obtain furfural predicted concentration data.

[0041] Step 306: Based on the predicted concentration data and the measured concentration data of furfural, perform prediction bias analysis on the state-anchored digital twin to obtain the state drift characteristic quantity.

[0042] Among them, the aging evolution process of oil-paper insulation refers to the gradual performance degradation, aging product generation, and decrease in insulation capacity of insulating paper and insulating oil in target power equipment under the action of thermal stress, electrical stress, oxidation reaction, and operating load.

[0043] Trend analysis refers to the process of calculating the direction, rate, and future range of aging changes in oil-paper insulation based on historical operating data, real-time operating status data, and the aging model in the state-anchored digital twin.

[0044] Among them, insulation aging trend data refers to data used to describe the future changes in the aging of oil paper insulation, including the direction of aging trend, the rate of aging growth, the aging acceleration, the duration of the trend, and the range of trend changes.

[0045] Among them, the furfural concentration evolution process refers to the dynamic process of furfural molecules undergoing generation, dissolution, migration, accumulation and concentration change during the aging process of oil paper insulation.

[0046] Predictive analysis refers to the process of calculating future furfural concentration changes based on insulation aging trend data and furfural generation, dissolution, and migration models in state-anchored digital twins.

[0047] Among them, furfural predicted concentration data refers to the results of furfural concentration changes in the future time period calculated by the state-anchored digital twin, including predicted concentration value, predicted concentration range, concentration growth rate and predicted peak time.

[0048] Among them, prediction deviation analysis refers to the process of comparing the predicted concentration data of furfural with the measured concentration data and calculating the differences between the two in terms of concentration value, rate of change, phase of change and duration.

[0049] Specifically, historical and real-time operating data are normalized along the time axis using a unified sampling period. Load current, top oil temperature, winding hotspot temperature, ambient temperature, cooler operating status, overload duration, and power outage / restoration conditions are normalized to ensure that operating quantities from different sources are included in the same calculation scale. The normalized historical operating data is written into the historical stress accumulation channel of the state-anchored digital twin, and the real-time operating data is written into the current stress driving channel of the state-anchored digital twin. The effects of thermal stress, electrical stress, and operating stress on the rate of decrease in the degree of polymerization of the insulating paper, the oxidation rate of the insulating oil, and the aging rate of the oil-paper interface are calculated in the oil-paper insulation aging model. The calculations include using the Arrhenius thermal aging equation to calculate the accelerating effect of temperature on the aging reaction rate, using the load loss model to calculate the hotspot temperature changes under different load conditions, and using the oil-paper insulation aging kinetic equation to calculate the degradation process of the insulating paper. The transient aging increments corresponding to the real-time operating status are then superimposed onto the historical aging accumulation results. Then, time window sliding calculations are performed on the aging evolution curve of the paper insulation in the state-anchored digital twin. The aging rate, aging acceleration, aging curvature and aging accumulation are calculated in the current window, adjacent historical window and long-term reference window respectively. Then, insulation aging trend data including aging trend direction, trend growth magnitude, trend duration and trend change intensity are generated.

[0050] Insulation aging trend data is input into the furfural generation-dissolution-migration coupled model in the state-anchored digital twin, with the insulation paper degradation rate, hot spot temperature change rate, oil-paper interface aging intensity, and aging accumulation as driving terms for the furfural generation process. However, the furfural generation process in the coupled model is established based on the pyrolysis and oxidative cracking reactions of the insulation paper, the furfural dissolution process is established based on the oil-paper two-phase distribution coefficient, oil temperature, and oil sample water content, and the furfural migration process is established based on the oil flow velocity, oil channel structure, convection diffusion coefficient, and local temperature gradient. Multiple future operating time steps are set in the state-anchored digital twin, and within each time step, the furfural generation, insulation paper retention, insulation oil dissolution, and oil channel migration are calculated sequentially according to their respective equations in the coupled model. For real-time operating disturbances such as load abrupt changes, cooler switching, and short-term overloads, pulse disturbance terms are superimposed onto the furfural generation rate equation, enabling the prediction process to simultaneously consider long-term thermal aging and short-term operating impacts. The furfural concentration calculation results obtained at each time step are subject to continuity constraints, physical boundary constraints, and oil-paper distribution balance constraints to avoid non-physical abrupt changes in the prediction results, and the predicted furfural concentration data corresponding to the future time series is output. The furfural generation-dissolution-migration coupled model is a discrete-time recursive model set in a state-anchored digital twin, which includes sequentially coupled furfural generation sub-models, furfural dissolution sub-models, and furfural migration sub-models. The remaining degradable amount of insulating paper, furfural retention in insulating paper, furfural dissolution in insulating oil, and furfural concentration distribution at various locations in the oil path at the previous future operating time step are used as the previous state, and the furfural generation amount, insulating paper retention amount, insulating oil dissolution amount, and oil path migration amount at the next future operating time step are used as the output state. The generation rate equation in the furfural generation sub-model is based on the calculation rule for the generation amount per unit time step established by the thermal and oxidative pyrolysis process of insulating paper cellulose. Specifically, within each future operating time step, the degradation rate of insulating paper in that time step is first determined based on the insulating aging trend data. The effective degradation amount of the insulating paper participating in the aging reaction is determined by the thermal decomposition enhancement degree based on the hot spot temperature change rate and the preset temperature acceleration coefficient. Then, the oxidative decomposition enhancement degree is determined based on the aging strength of the oil-paper interface, the cumulative aging amount, the moisture content correction coefficient, and the load disturbance correction coefficient. Finally, the effective degradation amount of the insulating paper, the thermal decomposition enhancement degree, and the oxidative decomposition enhancement degree are combined in a product or weighted manner to obtain the amount of newly generated furfural in the insulating paper phase within this time step. The amount of insulating paper retention refers to the portion of the newly generated furfural that has not yet entered the insulating oil phase but remains in the pores of the insulating paper or near the oil-paper interface. It is calculated by adding the amount of newly generated furfural in the current time step to the amount of furfural remaining in the insulating paper in the previous time step, and then subtracting the amount of furfural that can cross the oil-paper interface and enter the insulating oil in the current time step to obtain the updated amount of insulating paper retention.The distribution delay equation in the furfural dissolution sub-model is a time-step release rule established based on the oil-paper two-phase distribution and interfacial diffusion lag. Specifically, it determines the equilibrium distribution ratio of furfural between the insulating paper phase and the insulating oil phase based on the oil-paper two-phase distribution coefficient, corrects the furfural cross-interfacial diffusion rate based on oil temperature and oil sample water content, determines the maximum allowable release amount into the oil phase at this time step based on the oil-paper interface contact area and interfacial aging intensity, and takes the portion of the insulating paper retention that meets the release conditions as the insulating oil dissolution amount at the current time step. The migration diffusion equation in the furfural migration sub-model is a concentration update rule established based on oil convection, molecular diffusion, and temperature gradient migration. Specifically, it sets the insulating oil dissolution amount as... For new source terms at the oil inlet or oil-paper interface, the convective transport distance of the new source term in the main oil passage, winding gap oil passage, cooler circuit, and local stagnant zone is determined based on the oil flow velocity. The distribution ratio of different oil passage branches is determined based on the oil passage structure. The diffusion exchange between adjacent oil passage units is determined based on the convective diffusion coefficient. The additional correction amount when furfural migrates from the high-temperature region to the low-temperature region is determined based on the local temperature gradient. Then, the amount of furfural migrating from the current oil passage unit to the adjacent oil passage unit within this time step is calculated, and the furfural concentration distribution at each oil passage location is updated. When multiple consecutive future running time steps are set in the state-anchored digital twin, furfural generation is first performed in the first future running time step. The process begins with a furfural generation sub-model to obtain the newly generated furfural amount at that time step. Then, a furfural dissolution sub-model is executed, converting the portion satisfying the oil-paper distribution and interfacial diffusion conditions into insulating oil dissolution, while retaining the unreleased portion as insulating paper retention. Next, a furfural migration sub-model is executed, distributing the insulating oil dissolution to various oil path locations according to the oil flow path and diffusion rules, obtaining the oil path migration amount and oil path concentration distribution for that time step. This data is then used as the initial state for the next time step, and the generation, dissolution, and migration calculations are repeated until all future operating time steps are recursively calculated, thus obtaining a continuously changing data over time. The data includes predicted furfural concentrations. Real-time operational disturbances caused by load abrupt changes, cooler switching, and short-term overloads do not alter the model structure individually. Instead, they are added as corrections to the hotspot temperature change rate, temperature acceleration coefficient, oil flow velocity, and oil-paper interface release rate. This allows the same coupled model to simultaneously calculate stable furfural generation due to long-term thermal aging, transient furfural increase due to short-term thermal shock, insulation paper retention due to oil-paper distribution lag, and concentration delay at detection locations due to differences in oil transport. This clarifies the composition of the furfural generation-dissolution-migration coupled model, the establishment of the generation rate equation, and the calculation logic for furfural generation, insulation paper retention, insulating oil dissolution, and oil migration.

[0051] The predicted and measured furfural concentration data are mapped to the same aging phase coordinate system in the state-anchored digital twin, and time-aligned according to the detection time, prediction time step, and oil sample reaction time. Using the measured concentration data as the anchoring benchmark, the deviation of the predicted furfural concentration data is calculated to obtain the absolute concentration deviation, relative concentration deviation, concentration growth rate deviation, and concentration change curvature deviation. These deviations are then fed back into the oil-paper insulation aging model, furfural generation model, and oil-paper distribution model in the state-anchored digital twin to calculate the contribution weights of deviation sources in the aging reaction rate, furfural generation intensity, oil-paper distribution delay, and oil path migration process, respectively. For deviations that continuously increase over multiple detection cycles, a higher drift persistence weight is assigned; for local deviations caused by a single detection, suppression is achieved through detection confidence and spectral quality parameters. Finally, the concentration deviation amplitude, deviation direction, deviation change rate, deviation duration, and deviation source weights are vectorized and combined to obtain the state drift characteristic quantity.In calculating the contribution weights of deviation sources in the aging reaction rate, furfural generation intensity, oil-paper distribution delay, and oil migration process, the absolute concentration deviation, concentration growth rate deviation, concentration change curvature deviation, and detection time phase deviation formed between the predicted and measured furfural concentration data are first used as the total deviation benchmark. The oil-paper insulation aging model, furfural generation model, oil-paper distribution model, and oil migration model currently used by the state-anchored digital twin are maintained as the benchmark calculation state. Subsequently, correction channels for aging reaction rate, furfural generation intensity, oil-paper distribution delay, and oil migration are established respectively. Only one of these channels is allowed in each correction calculation. In one channel, parameters are adjusted within a preset physical range, while parameters in other channels remain unchanged. For example, in the aging reaction rate correction channel, only the rate of decrease in the equivalent degree of polymerization of the insulating paper, the hot spot temperature aging acceleration coefficient, and the aging reaction time constant are adjusted, and the remaining deviation between the predicted and measured furfural concentration curves is recalculated. In the furfural generation intensity correction channel, only the furfural generation source term intensity, generation duration, and generation growth slope are adjusted, and the remaining deviation is recalculated. In the oil-paper distribution delay correction channel, only the hysteresis time of furfural entering the insulating oil phase from the insulating paper phase, the oil-paper distribution coefficient, and the interface release retardation coefficient are adjusted, and the remaining deviation is recalculated. In the oil migration correction channel... In each channel, only the oil flow velocity, diffusion attenuation coefficient, oil path propagation time, and detection point arrival time are adjusted, and the remaining deviation is recalculated. The reduction in the remaining deviation before and after adjustment for each corrected channel is then used as the channel's explanatory power for the total deviation. This is further corrected by considering the consistency between the predicted and measured curves after correction in terms of direction of change, peak time, growth slope, and curvature changes. If a channel adjustment simultaneously reduces concentration deviation, growth rate deviation, and time phase deviation, the explanatory power of that channel is increased. If a channel only reduces a single deviation but increases other deviations, the explanatory power of that channel is decreased. For cases where multiple channels reduce deviation, the principle of "first one channel, then the others" applies. The process involves iterative comparisons of single-channel correction, pairwise combined correction, and finally multi-channel joint correction. The bias already explained by a single channel is deducted first, and then the remaining bias is allocated to other channels to avoid the same bias being counted repeatedly from multiple sources. Finally, the explained values ​​of each channel are normalized according to the total explained value, and the reliability is corrected by considering the confidence level of this SERS detection, the quality weight of the Raman spectrum, and the completeness of the equipment operation data. This yields the contribution weights corresponding to the aging reaction rate, furfural formation intensity, oil-paper partitioning delay, and oil migration process. A higher contribution weight indicates a greater impact of this process on the discrepancy between the predicted and measured concentration data.

[0052] In this embodiment, by first performing trend analysis on the aging evolution process of oil-paper insulation, then predicting the furfural concentration evolution process based on the insulation aging trend data, and finally performing prediction deviation analysis between the predicted furfural concentration data and the measured concentration data, the equipment status judgment can be transformed from static concentration comparison to dynamic trajectory verification, avoiding misjudgment caused by fluctuations in a single detection, local differences in oil samples, or short-term operational disturbances. At the same time, the insulation aging trend data provides a continuous aging evolution basis for furfural concentration prediction, and the predicted furfural concentration data provides a comparable twin benchmark for the measured concentration data, enabling the state drift characteristic quantity to reflect the degree of deviation between the actual aging path and the expected aging path of the equipment, thereby improving the ability to detect potential accelerated aging, abnormal furfural release, and inflection points of oil-paper insulation deterioration.

[0053] In an exemplary embodiment, based on historical operating data and real-time operating status data, a trend analysis is performed on the aging evolution process of the oil-paper insulation in the state-anchored digital twin to obtain insulation aging trend data, including steps 402 to 406. Wherein:

[0054] Step 402: Based on the real-time operating status data, perform transient aging disturbance identification on the current oil-paper insulation status in the state-anchored digital twin to obtain the current aging disturbance characteristic data.

[0055] Step 404: Based on the current aging disturbance characteristic data, perform reverse matching on the historical aging evolution segments in the historical operation data to obtain historical homogeneous aging trajectory data.

[0056] Step 406: Based on historical homogeneous aging trajectory data and current aging disturbance characteristic data, the forward trend reconstruction of the aging evolution process of oil-paper insulation is performed to obtain insulation aging trend data.

[0057] The current oil-paper insulation status refers to the aging degree, thermal stress level, furfural generation level, and insulation performance status of the insulating oil and insulating paper of the target power equipment under the current operating conditions.

[0058] Transient aging disturbance identification refers to the process of calculating and determining the short-term aging acceleration impact caused by changes in real-time operating status.

[0059] Among them, the current aging disturbance characteristic data refers to the data used to describe the transient impact of the current operating conditions on the aging of oil-paper insulation, including disturbance amplitude, disturbance duration, disturbance phase, disturbance cumulative intensity, and temperature rise hysteresis.

[0060] Among them, the historical aging evolution segment refers to the aging change process of oil paper insulation corresponding to a certain continuous operation stage in the historical operation data.

[0061] Among them, reverse matching refers to a processing method that uses the current aging perturbation feature data as the terminal condition and matches similar aging change processes in historical aging evolution segments in reverse chronological order.

[0062] Among them, historical homogeneous aging trajectory data refers to historical aging evolution trajectory data that has similar aging phase, operating stress changes and deterioration paths to the current aging disturbance characteristic data.

[0063] Forward trend reconstruction refers to the process of recalculating the direction, speed, and range of subsequent aging changes in oil-paper insulation based on historical homogeneous aging trajectory data and current aging disturbance characteristic data.

[0064] Specifically, real-time operating status data is written into the real-time stress calculation channel of the state-anchored digital twin according to a preset sampling period. The written real-time operating status data includes load current, winding hot spot temperature, top oil temperature, ambient temperature, cooler operating status, oil flow velocity, short-term overload duration, and voltage fluctuation amplitude. Then, the thermal stress field, electric stress field, and oil flow disturbance field at the current moment are established in the state-anchored digital twin, and each stress field is coupled with the thermal aging reaction rate, electric field acceleration factor, and oil-paper interface diffusion coefficient in the oil-paper insulation aging model for calculation. That is, for transient operating changes such as load surges, cooling state switching, rapid rise in hot spot temperature, and lag changes in oil temperature in the real-time operating status data, corresponding disturbance response functions are set respectively, and the increments of aging rate, aging acceleration, hot spot temperature rise lag, oil-paper interface release intensity, and furfural generation potential caused by them within the current time window are calculated. The above calculation results are subjected to time window compression and phase normalization to unify the transient effects caused by different operating quantities into the phase coordinates of oil-paper insulation aging. The current aging disturbance characteristic data are constructed by disturbance amplitude, disturbance duration, disturbance rise slope, disturbance decay slope and disturbance cumulative intensity.The aforementioned disturbance response functions are pre-established segmented response rules for different transient operational changes in real-time operating status data. Each disturbance response function includes a disturbance triggering condition, a disturbance reference value, a disturbance response amplitude, a response duration, a response rise segment, a peak holding segment, and a decay recovery segment. Specifically, for load surge disturbances, the average load, average winding hot spot temperature, and average top-layer oil temperature within a preset stable time window before the disturbance occur are first used as the reference state. When the current load exceeds the reference state and reaches the preset load increase threshold within a series of sampling points, the load surge disturbance is determined to be triggered. Then, the disturbance amplitude is determined according to the ratio of the load increase to the rated load, the disturbance duration is determined according to the duration of the load surge, and the hot spot temperature rise lag and thermal aging reaction rate increment caused by the load surge within the current time window are calculated according to the thermal inertia delay relationship between hot spot temperature and load change. For cooling state switching disturbances, the time from shutdown to startup or from startup of the cooler is first recorded. At the time of shutdown switching, the cooling response intensity is determined based on the rate of change of oil temperature, the rate of change of hot spot temperature, and the rated heat dissipation capacity of the cooler before and after switching. The cooling response intensity is then applied to the thermal stress field, so that the aging rate increment corresponding to enhanced cooling is reduced and the aging rate increment corresponding to weakened cooling is amplified. For the disturbance of rapid rise in hot spot temperature, the hot spot temperature rise rate within adjacent sampling points or sliding time windows is used as the judgment criterion. When the hot spot temperature rise rate exceeds the preset temperature rise threshold, the aging acceleration increment is calculated according to the temperature rise rate, peak temperature, and duration. This aging acceleration increment is then applied synchronously to the change in release intensity at the oil-paper interface and the change in furfural generation potential. For the disturbance of hysteretic change in oil temperature, the response time difference between the load change curve and the top oil temperature change curve is first compared. Then, the oil temperature hysteresis coefficient is determined by combining the oil flow velocity, cooling state, and oil channel thermal inertia. The time delay of furfural release and migration from the insulating paper phase to the insulating oil phase is corrected based on the oil temperature hysteresis coefficient.Furthermore, the specific calculation process of the disturbance response function is as follows: For each transient operational change, the deviation of the current sampled value from the corresponding reference state is first calculated. Then, the deviation is normalized according to the rated capacity, rated temperature rise, upper limit of hot spot temperature, or allowable oil temperature range to obtain the dimensionless disturbance intensity. Then, according to the disturbance type, the corresponding response rule is called to convert the dimensionless disturbance intensity into aging rate increment, aging acceleration increment, hot spot temperature rise lag, change in oil-paper interface release intensity, and change in furfural generation potential. Among them, the aging rate increment is determined according to the range of hot spot temperature and the preset thermal aging rate level table; the aging acceleration increment is determined according to the rate of change of aging rate increment within the continuous time window; the hot spot temperature rise lag is determined according to the time difference between the moment of load change and the moment when the hot spot temperature reaches the response peak; the change in oil-paper interface release intensity is determined jointly according to the oil-paper interface temperature gradient, oil flow disturbance intensity, and aging degree of insulating paper; and the change in furfural generation potential is determined according to the aging rate increment, aging acceleration increment, and aging acceleration increment. The duration and furfural generation coefficient in the state-anchored digital twin are jointly determined. Finally, the above-mentioned disturbance calculation results are compressed into a time window according to the same time window, that is, the disturbance results in multiple sampling points are compressed into the maximum disturbance value, average disturbance value, duration and cumulative disturbance amount in the time window, and phase normalization is performed. That is, the disturbance start time is taken as the phase start point, the time when the disturbance reaches the peak value is taken as the peak phase, and the time when the disturbance recovers to the reference state is taken as the decay phase. The transient effects caused by different operating quantities such as load, temperature, cooling and oil flow are uniformly mapped to the phase coordinate of oil paper insulation aging. On this basis, the disturbance amplitude, disturbance duration, disturbance rise slope, disturbance decay slope, disturbance cumulative intensity, hot spot temperature rise lag, interface release intensity change and furfural generation potential change corresponding to each disturbance type are combined in a preset order, and the combined results of multiple disturbance sources in the same time window are weighted and normalized and outlier limiting is performed to obtain the current aging disturbance characteristic data.

[0065] Historical operating data is divided into multiple continuous aging evolution segments according to the equipment's historical operating cycle. Each historical aging evolution segment corresponds to a complete operating condition change process, including load change process, hot spot temperature change process, oil temperature response process, cooler operation process, and aging change process corresponding to historical oil sample test results. A disturbance phase sequence is calculated for each historical aging evolution segment, which includes load disturbance phase, thermal response phase, oil-paper interface diffusion phase, and aging accumulation phase. Then, using the current aging disturbance characteristic data as the matching terminal condition, the historical aging evolution segments are reversed in time and their final phases are aligned, ensuring that the aging state at the end of the historical segment is at the same or similar aging phase position as the current aging disturbance characteristic data. Next, the disturbance amplitude distance, disturbance duration distance, thermal response hysteresis distance, oil-paper interface diffusion distance, and furfural formation potential distance between the historical aging evolution segments and the current aging disturbance characteristic data are calculated. Weighted dynamic time warping or twin phase distance measurement methods are used to screen historical segments that meet the matching requirements. Finally, the historical aging evolution segments that meet the matching conditions are spliced ​​together according to phase continuity, aging path similarity and operating condition compatibility to obtain historical homogeneous aging trajectory data.

[0066] Historical homogeneous aging trajectory data is written into the historical trajectory constraint channel of the state-anchored digital twin, and current aging disturbance characteristic data is written into the real-time disturbance driving channel, with the current oil-paper insulation aging state as the initial boundary for forward reconstruction. The aging rate change, hotspot temperature response, oil-paper interface diffusion delay, furfural formation lag, and aging accumulation slope in the historical homogeneous aging trajectory data are parameterized to form historical trajectory constraint terms that can participate in forward calculation. Then, the disturbance amplitude, disturbance duration, disturbance phase, and disturbance accumulation intensity in the current aging disturbance characteristic data are superimposed onto the oil-paper insulation aging model to progressively calculate the changes in insulation paper polymerization degree, oil-paper interface aging rate, furfural formation potential, and thermal aging acceleration coefficient over several future time steps. During the recursive calculation, the historical trajectory constraint terms are used to limit non-physical abrupt changes in the future aging path, while the real-time disturbance driving terms are used to correct the lag error caused by relying solely on the historical path. Finally, the calculation results of multiple future time steps are processed by trajectory smoothing, trend boundary constraints and confidence interval convergence to form insulation aging trend data that includes aging trend direction, aging growth rate, aging acceleration, trend duration and trend change range.

[0067] In this embodiment, by first identifying transient aging disturbances in the current state of the oil-paper insulation, then using the current aging disturbance characteristic data as the terminal condition to perform reverse matching on historical aging evolution segments, and combining historical homogeneous aging trajectory data with the current aging disturbance characteristic data to perform forward trend reconstruction, the traditional processing method of extrapolating aging trends according to historical time sequence can be changed. This transforms trend calculation from "past to future" to "current disturbance backtracking to homogeneous history, and then constraining future evolution by homogeneous trajectory". This can reduce the interference of different operating conditions of different equipment on trend prediction, and enable transient factors such as short-term overload, cooling switching, and hot spot temperature rise lag to participate in aging trend calculation, and improve the adaptability to discontinuous aging, sudden accelerated aging, and phased deterioration transitions.

[0068] In an exemplary embodiment, based on historical homogeneous aging trajectory data and current aging disturbance characteristic data, a forward trend reconstruction of the aging evolution process of oil-paper insulation is performed to obtain insulation aging trend data, including steps 502 to 508. Wherein:

[0069] Step 502: Based on the current aging disturbance characteristic data, perform a virtual annealing state search on the aging evolution process of the oil-paper insulation to obtain the aging evolution boundary set.

[0070] Step 504: Based on the aging evolution boundary set, perform reverse time-series trajectory stripping on the historical homogeneous aging trajectory data to obtain the key aging driving factor sequence.

[0071] Step 506: Based on the key aging driving factor sequence and the current aging disturbance characteristic data, perform cross-timescale disturbance propagation simulation of the aging evolution process of oil-paper insulation to obtain the aging trajectory reconstruction sequence.

[0072] Step 508: Based on the aging trajectory reconstruction sequence, twin bifurcation path screening is performed on the aging evolution process of oil paper insulation to obtain insulation aging trend data.

[0073] Among them, virtual annealing state search refers to simulating the annealing cooling parameter optimization process within a state-anchored digital twin, and calculating the stable evolution boundary that the oil-paper insulation aging may reach with the current aging disturbance characteristic data as constraints.

[0074] Among them, the aging evolution boundary set refers to multiple candidate aging boundary states obtained by virtual annealing state search that satisfy the constraints of insulating paper polymerization degree, furfural generation rate, oil-paper distribution relationship and hot spot temperature.

[0075] Among them, reverse time-series trajectory stripping refers to using the aging evolution boundary set as the terminal constraint, and reducing the operational stress contribution in the historical homogeneous aging trajectory in reverse time order to determine the driving factors that play a major role in the formation of the current aging state.

[0076] Among them, the key aging driving factor sequence refers to the combination of factors such as load stress, thermal stress, diffusion effect at the oil-paper interface, and furfural formation effect, which are arranged in chronological order and have the main driving effect on the aging evolution of oil-paper insulation.

[0077] Among them, cross-timescale disturbance propagation simulation refers to the simulation process of jointly calculating the transmission relationship between short-term operational disturbances and long-term aging accumulation at different time scales such as minute, hour, day or month.

[0078] Among them, the aging trajectory reconstruction sequence refers to the sequence of changes in the aging state of oil-paper insulation, which is obtained by simulation of perturbation propagation across time scales and arranged in future time sequence.

[0079] Among them, twin bifurcation path screening refers to the process of generating multiple aging evolution paths corresponding to different future operating conditions in the state-anchored digital twin, and selecting a reliable path based on physical constraints, historical homology and current disturbance consistency.

[0080] Specifically, a virtual annealing search space for the aging evolution process of oil-paper insulation is set in the state-anchored digital twin. This search space includes the equivalent polymerization degree range of the insulation paper, the furfural generation rate range, the oil-paper distribution coefficient range, the hot spot temperature acceleration factor range, the oil diffusion coefficient range, and the load stress amplification factor range. Then, the disturbance amplitude, disturbance duration, disturbance phase, disturbance cumulative intensity, and temperature rise hysteresis in the current aging disturbance characteristic data are written into the annealing energy function for calculation. During the calculation, the annealing energy function takes the thermal stress deviation, electrical stress deviation, furfural generation potential deviation, and oil-paper interface diffusion deviation between the candidate aging state and the current disturbance state as components of the energy function, and sets the initial annealing temperature, cooling coefficient, and state disturbance step size. In each round of annealing iteration, the model parameters of the candidate aging state are randomly perturbed, the virtual annealing energy value after perturbation is calculated, and candidate states that meet the energy reduction requirement or the probability acceptance condition are retained according to the annealing acceptance criterion. After multiple rounds of cooling iterations, the candidate states that converge to the low-energy region and satisfy physical constraints (including the monotonically decreasing degree of polymerization of insulating paper, the non-negative concentration of furfural, the balanced distribution of oil and paper, and the upper limit of hot spot temperature) are combined into the aging evolution boundary set.

[0081] The aging evolution boundary set is set as the terminal constraint of the historical homogeneous aging trajectory data. This historical homogeneous aging trajectory data is written into the reverse time-series calculation channel in reverse chronological order. Within each reverse time step, driving contribution terms are established for load stress, hotspot temperature stress, oil temperature hysteresis stress, cooling switching effect, oil-paper interface diffusion effect, and furfural generation effect in the historical homogeneous aging trajectory data. Then, keeping the terminal aging boundary unchanged, individual driving contribution terms are progressively reduced, and the residual changes between the aging trajectory and the aging evolution boundary set before and after reduction are compared. Driving contribution terms with larger residual changes are assigned higher weights, while those with smaller residual changes are assigned lower weights. Next, phase continuity correction and time compression are performed on the driving contribution terms in consecutive reverse time steps to ensure that the driving intensity changes between adjacent time steps satisfy the continuity constraint of the oil-paper insulation aging process. Finally, driving contribution terms whose weights meet the threshold requirements, whose time sequence meets the phase continuity requirements, and which can maintain the stability of the terminal aging boundary are rearranged according to the reverse calculation results to obtain the key aging driving factor sequence.

[0082] The key aging driving factor sequence is divided into a slow-change driving channel and a fast-change disturbance channel. The slow-change driving channel is used to calculate the aging evolution caused by long-term load accumulation, long-term temperature accumulation, and continuous degradation of the insulating paper. The fast-change disturbance channel is used to calculate the local aging changes caused by short-term overload, cooling switching, sudden temperature rise of hot spots, and oil flow disturbance. Then, the current aging disturbance characteristic data is used as the initial input of the fast-change disturbance channel, and the key aging driving factor sequence is used as the constraint input of the slow-change driving channel. The joint calculation is performed in a state-anchored digital twin using a multi-time-step integration method. During the calculation, for short-term disturbances at the minute or hour level, a smaller integration step size is used to calculate the hot spot temperature response, oil-paper interface diffusion changes, and furfural release intensity changes; for long-term aging processes at the daily or monthly level, a larger integration step size is used to calculate the decrease in the degree of polymerization of the insulating paper, the cumulative formation of furfural, and the changes in the oil-paper two-phase distribution. Furthermore, the calculation results of the fast-change disturbance channel are periodically written back to the slow-change drive channel, so that the local aging increment caused by short-term disturbances can participate in the long-term aging accumulation; at the same time, the aging benchmark of the slow-change drive channel is fed back to the fast-change disturbance channel, so that the short-term disturbance calculation does not deviate from the existing aging level of the equipment. After cyclic propagation calculations at multiple time scales, an aging trajectory reconstruction sequence arranged in future time order is obtained.

[0083] Using each future time node in the aging trajectory reconstruction sequence as a bifurcation node, multiple candidate twin bifurcation paths are generated in the state-anchored digital twin. Each candidate twin bifurcation path corresponds to different combinations of operating conditions (including maintaining the current load, short-term overload, load reduction operation, enhanced cooling, increased ambient temperature, and changes in detection cycle). Then, for each candidate twin bifurcation path, the rate of change of the degree of polymerization of insulating paper, furfural formation rate, hot spot temperature duration, oil-paper interface diffusion delay, and risk energy level growth are calculated. The calculation results are then checked for boundary consistency with the aging evolution boundary set. Candidate paths that do not meet the monotonic aging constraint, thermal field boundary constraint, or furfural formation physical constraint are eliminated. For the remaining candidate paths, a path score is then calculated based on the weight distribution of the key aging driving factor sequence and the perturbation phase of the current aging perturbation characteristic data. Candidate paths that are closer to the current equipment operating state, more continuous with historical aging trajectories, and have more stable physical constraints receive higher scores. Finally, the candidate paths that meet the scoring requirements are merged to form insulation aging trend data, which includes aging trend direction, aging growth rate, aging acceleration, trend duration, and trend confidence interval.

[0084] In this embodiment, the reachable boundary of the aging evolution process of oil-paper insulation is first defined by virtual annealing state search. Then, the reverse time-series trajectory stripping of historical homogeneous aging trajectory data is constrained by this boundary. This reduces the interference of irrelevant historical operating conditions on trend calculation, making the key aging driving factor sequence more focused on the core operating factors that drive the formation of the current deterioration state. On this basis, through cross-timescale disturbance propagation simulation, the local aging increment caused by short-term disturbance is linked with the cumulative aging process caused by long-term operation for calculation. Then, through twin bifurcation path screening, the aging path under different future operating conditions is constrained and selected. This avoids the trend rigidity caused by single path extrapolation, and makes the insulation aging trend data take into account the current disturbance, historical causes and future operating condition changes, thereby improving the stability of the judgment on the aging evolution direction and deterioration acceleration range under complex operating conditions.

[0085] In an exemplary embodiment, based on insulation aging trend data, the evolution process of furfural concentration in the state-anchored digital twin is predicted and analyzed to obtain furfural predicted concentration data, including steps 602 to 606. Wherein:

[0086] Step 602: Based on the insulation aging trend data, the furfural release potential is mapped to the aging energy level of the oil paper insulation in the state-anchored digital twin to obtain the furfural release potential sequence.

[0087] Step 604: Based on the furfural release potential sequence, perform virtual pulse reinjection simulation on the furfural generation process, furfural dissolution process, and furfural migration process in the state-anchored digital twin to obtain a multi-branch furfural concentration evolution cluster.

[0088] Step 606: Based on the multi-branch furfural concentration evolution cluster, perform concentration manifold folding prediction on the furfural concentration evolution process to obtain furfural predicted concentration data.

[0089] Among them, the aging energy level of oil-paper insulation refers to the aging strength level calculated in the state-anchored digital twin based on the degree of degradation of the insulation paper, hot spot temperature, load stress, degree of oxidation and degree of release hindrance at the oil-paper interface.

[0090] Among them, furfural release potential mapping refers to the calculation process of converting the aging energy level of oil paper insulation into the potential release driving force of furfural molecules from the insulating paper phase to the insulating oil phase.

[0091] Among them, the furfural release potential sequence refers to the data on the potential driving forces of furfural release arranged in chronological order, which is used to describe the intensity and rhythm of furfural release at different operating stages.

[0092] The furfural generation process refers to the process by which cellulose in insulating paper degrades under the influence of heat, electricity, oxidation, and moisture, producing furfural molecules.

[0093] The furfural dissolution process refers to the process by which furfural molecules generated in the insulating paper enter the insulating oil through the oil-paper interface and form a detectable concentration in the oil phase.

[0094] Among them, furfural migration process refers to the process by which furfural molecules dissolved in insulating oil move and redistribute in the equipment oil circuit with the oil flow circulation, convection diffusion and temperature gradient.

[0095] Virtual pulse reinjection simulation refers to the process of converting furfural release potential into multiple virtual release pulses in a state-anchored digital twin and loading them into a furfural generation, dissolution and migration model for parallel simulation.

[0096] Among them, the multi-branch furfural concentration evolution cluster refers to the set of multiple candidate furfural concentration change trajectories formed by various furfural generation intensities, dissolution hysteresis and oil migration paths.

[0097] Among them, concentration manifold folding prediction refers to the process of mapping multiple furfural concentration evolution branches to a unified concentration manifold space and obtaining a continuous predicted concentration curve through folding and fusion.

[0098] Specifically, insulation aging trend data is written into the aging energy level calculation layer of the state-anchored digital twin, and the decrease in the degree of polymerization of insulating paper, the duration of hot spot temperature, the aging rate of the oil-paper interface, the aging acceleration, and the cumulative aging intensity are discretized according to future time steps. Then, an oil-paper insulation aging energy level function is established in the state-anchored digital twin, which includes thermal decomposition energy level terms, oxidative degradation energy level terms, oil-paper interface release energy level terms, and load impact amplification terms. The aging trend changes under different future time steps are written into the corresponding energy level terms of the oil-paper insulation aging energy level function, and the oil-paper insulation aging energy level is corrected using temperature acceleration coefficient, insulating paper moisture content correction coefficient, oil-paper distribution correction coefficient, and historical aging accumulation correction coefficient. On this basis, a furfural release potential mapping function is established to convert the oil-paper insulation aging energy level into the potential driving force for furfural molecules to be released from the insulating paper to the insulating oil. For low aging energy levels, a slow release region is corresponding to the aging energy level; for high aging energy levels, a rapid release region is corresponding to the aging energy level; and for energy level abrupt change regions, a potential abnormal release region is corresponding to the aging energy level. Finally, the release potential values ​​corresponding to each future time step are arranged in chronological order, and a furfural release potential sequence is generated by combining the release duration, release growth rate, and release transition amplitude.

[0099] The furfural release potential sequence was converted into multiple virtual release pulses with time position, release intensity, and duration. These virtual release pulses were then applied to the aging region of the insulating paper, the oil-paper contact interface region, and the oil flow migration region in the state-anchored digital twin. For the furfural generation process, the furfural generation source term corresponding to each virtual release pulse was calculated based on the insulating paper degradation reaction rate, hot spot temperature, oxidation degree, and decrease in polymerization degree. For the furfural dissolution process, the response process of furfural entering the oil phase from the insulating paper phase was calculated based on the oil-paper two-phase distribution coefficient, oil temperature, water content, dissolution lag time, and oil-paper interface contact area. For the furfural migration process, the diffusion and transport process of furfural in the insulating oil was calculated based on oil flow velocity, oil channel structure, convective diffusion coefficient, local temperature gradient, and oil circulation path. Next, multiple parallel simulations were performed in the digital twin for different combinations of virtual release pulses. Each simulation corresponded to a combination of furfural generation intensity, dissolution lag, and migration / diffusion path. For sudden increases in release potential caused by load abrupt changes, cooler switching, or rapid changes in hotspot temperature, additional pulse perturbations were applied at the corresponding time steps to ensure the simulation results covered both conventional aging release and anomalously accelerated release scenarios. Finally, the furfural concentration time series obtained from each simulation were branched and merged according to the generation source term, dissolution response, and migration path to obtain multi-branch furfural concentration evolution clusters.

[0100] Each concentration evolution branch in the multi-branch furfural concentration evolution cluster is mapped to a unified concentration manifold space, which uses prediction time, furfural concentration, concentration growth rate, concentration change curvature, release potential intensity, and oil-paper distribution delay as coordinate dimensions. Within this concentration manifold space, the neighborhood distance, directional consistency, curvature continuity, and physical constraint satisfaction among each concentration evolution branch are calculated. Branches that do not satisfy the furfural concentration non-negativity constraint, oil-paper distribution balance constraint, generation rate upper limit constraint, and migration continuity constraint are removed. Then, the remaining concentration evolution branches are compressed into several continuous prediction bands according to their change direction between adjacent time steps. Within each prediction band, the central concentration trajectory, upper boundary concentration trajectory, lower boundary concentration trajectory, and abrupt change inflection point location are calculated. For regions where multiple prediction bands overlap, folding and fusion are performed based on the intensity ranking of the furfural release potential sequence, the correction weight of historical detection results, and the physical boundary constraints of the current state-anchored digital twin, forming a continuous, stable furfural concentration prediction curve that satisfies the oil-paper insulation aging mechanism. Finally, the predicted concentration value, predicted concentration range, concentration growth rate, predicted peak time, and abnormal release probability are combined to form the furfural predicted concentration data.

[0101] In this embodiment, by converting insulation aging trend data into a furfural release potential sequence, furfural concentration prediction no longer relies solely on the extension of historical concentration curves, but establishes a correspondence with the changes in the aging energy level of the oil-paper insulation. Furthermore, through virtual pulse reinjection simulation, the generation, dissolution, and migration processes of furfural are incorporated into the same prediction chain, allowing for the synchronous calculation of concentration changes caused by lag in release within the insulation paper, delays in oil-paper interface distribution, and oil diffusion transport. Finally, through concentration manifold folding prediction, multiple possible furfural concentration evolution paths are continuously compressed and converged at their boundaries. This reduces the inadequacy of a single prediction model to adapt to abnormal releases, short-term thermal shocks, or oil flow disturbances, making the obtained furfural predicted concentration data more suitable for subsequent prediction bias analysis and state drift judgment.

[0102] In an exemplary embodiment, based on the furfural release potential sequence, a virtual pulse reinjection simulation is performed on the furfural generation process, furfural dissolution process, and furfural migration process in the state-anchored digital twin to obtain a multi-branch furfural concentration evolution cluster, including steps 702 to 712. Wherein:

[0103] Step 702: Based on the furfural release potential sequence, perform energy state discrete reconstruction on the furfural potential release field in the state-anchored digital twin to obtain the furfural release energy cell set.

[0104] Step 704: Based on the furfural release energy cell set, perform twin crack source domain mapping on the oil-paper insulation interface in the state-anchored digital twin to obtain furfural virtual reinjection source domain data.

[0105] Step 706: Based on the furfural virtual reinjection source domain data and the furfural release energy cell set, perform source domain phase perturbation arrangement on the furfural generation process to obtain the furfural generation phase source term sequence.

[0106] Step 708: Based on the furfural generation phase source term sequence, the furfural dissolution process is remodeled by dissolution hysteresis wavefront to obtain the furfural dissolution response sequence.

[0107] Step 710: Based on the furfural dissolution response sequence, perform multi-scale convection-diffusion-bifaction propagation simulation of the furfural migration process to obtain candidate furfural concentration evolution branches.

[0108] Step 712: Based on the candidate furfural concentration evolution branches, perform branch clustering and recombination on the furfural concentration evolution paths in the state-anchored digital twin to obtain multi-branch furfural concentration evolution clusters.

[0109] Among them, the furfural potential release field refers to the spatialized computational field constructed in the state-anchored digital twin for calculating the possible release of furfural molecules in different regions of insulating paper at different aging energy levels.

[0110] Among them, discrete energy state reconstruction refers to the process of dividing the continuous furfural potential release field into multiple discrete release energy units according to furfural release intensity, release time, hot spot temperature and oil-paper interface hindrance conditions.

[0111] Among them, the furfural release energy cell set refers to a data set composed of multiple furfural release energy cells, each cell corresponding to a furfural release unit with spatial location, release intensity, duration width and energy level.

[0112] Among them, the paper insulation interface refers to the interface region in the target power equipment where the insulating oil and the insulating paper come into contact and furfural is released, dissolved, and migrated and exchanged.

[0113] Among them, twin crack source domain mapping refers to mapping furfural release energy cells in a digital twin to a virtual release source region at the oil-paper insulation interface to simulate the path of furfural from the insulating paper into the insulating oil.

[0114] Among them, furfural virtual reinjection source domain data refers to data used to describe the location, area, connectivity, upper limit of release flux, and interphase resistance coefficient of each virtual release source region.

[0115] Among them, source domain phase perturbation orchestration refers to the process of organizing furfural generation source terms in a time sequence according to the start time, peak time, decay time and operational perturbation effects of different virtual release source regions.

[0116] Among them, the furfural generation phase source term sequence refers to the furfural generation source term data arranged in chronological order, which is used to describe the furfural generation intensity and duration in different virtual release source regions at different phases.

[0117] Among them, dissolution hysteresis wavefront reshaping refers to the process of recalculating the hysteresis release boundary of furfural during the process of entering insulating oil from insulating paper based on the oil-paper distribution coefficient, interfacial hindrance, and source term phase difference.

[0118] Among them, the furfural dissolution response sequence refers to the response data of furfural entering the insulating oil phase arranged in chronological order, including the amount of oil phase entering, the amount of delayed release, the amount of interfacial residue, and the dissolution increment.

[0119] Among them, the multi-scale convection diffusion bifurcation propagation simulation refers to simulating the process of furfural convection, molecular diffusion and path bifurcation propagation with oil flow at different scales such as oil-paper interface, winding oil passage and whole machine oil circulation.

[0120] Among them, the candidate furfural concentration evolution branch refers to the candidate furfural concentration change trajectory formed under the influence of different release source domains, dissolution hysteresis and oil propagation paths.

[0121] Among them, the furfural concentration evolution path refers to the continuous calculation path in which the furfural concentration changes with time, release location, dissolution response and oil migration process.

[0122] Among them, branch clustering and recombination refers to the process of classifying, merging and rearranging multiple candidate furfural concentration evolution branches according to the similarity of concentration changes, the consistency of propagation paths and the phase correlation of source domains.

[0123] Specifically, a furfural potential release field covering the aging region of the insulating paper, the oil-paper contact interface region, and the hot spot temperature region is established in the state-anchored digital twin, and divided into spatial grid cells according to the oil-paper insulation structure of the equipment. Then, the furfural release potential sequence is written into the corresponding grid cells according to time steps, establishing a correspondence between the release potential at each time step and the decrease in the degree of polymerization of the insulating paper, the duration of the hot spot temperature, the energy level transition, and the release inhibition at the oil-paper interface. Next, an energy state calculation function is set for each spatial grid cell. This energy state calculation function includes thermal decomposition contribution terms, oxidative degradation contribution terms, oil-paper interface release contribution terms, and load disturbance amplification terms, and the energy state calculation results are calibrated using temperature correction coefficients, oil-paper distribution correction coefficients, and historical aging cumulative correction coefficients. Based on this, the furfural potential release field is discretized into multiple release energy cells according to the energy state level. Each release energy cell contains spatial location, temporal location, release intensity, duration width, energy transition level, and corresponding oil-paper interface constraint parameters, thus obtaining the furfural release energy cell set.

[0124] A virtual interface layer for the oil-paper insulation interface is established in the state-anchored digital twin. This virtual interface layer includes the fiber pore region of the insulating paper, the oil-wetting region, the oil-paper contact boundary region, and the local thermal stress concentration region. Then, the furfural release energy cell set is mapped to the virtual interface layer according to its spatial location and energy transition level. The interface crack opening degree is calculated based on the porosity of the insulating paper, the oil immersion depth, the local temperature gradient, the oil flow shearing effect, and the degree of aging accumulation. Next, a virtual release source region simulating furfural entering the insulating oil phase from the insulating paper phase is constructed in the virtual interface layer as a twin crack source domain. Each twin crack source domain corresponds to one or more release energy cells and has source domain area, source domain depth, source domain connectivity, upper limit of release flux, and oil-paper phase hindrance coefficient. Finally, connectivity correction and boundary overlap elimination processing are performed on adjacent twin crack source domains to ensure that the virtual reinjection region is spatially continuous and meets the release intensity constraint in terms of energy, thus obtaining the furfural virtual reinjection source domain data.

[0125] For each furfural virtual reinjection source domain, a starting phase, peak phase, decay phase, and intermittent phase are set. The generation phase parameters, including the time length and intensity weight of each phase, are determined based on the release intensity, duration, and energy transition level of the furfural release energy cell set. The generation phases corresponding to different virtual reinjection source domains are then sorted according to the oil-paper interface diffusion path, hotspot conduction path, and oil flow migration path, ensuring that the furfural generation process between adjacent source domains satisfies the thermal response sequence and the oil-paper release lag relationship. Next, a phase perturbation function is introduced into the furfural generation rate equation to perturb the arrangement of source domain phases, allowing load abrupt changes, hotspot temperature rise, cooling state switching, and local oil flow changes to participate in the furfural generation calculation in the form of phase advance, phase delay, phase superposition, or phase weakening. Finally, the furfural generation intensity, generation duration, generation phase offset, and source domain flux constraint corresponding to each virtual reinjection source domain at different times are arranged in chronological order to form a furfural generation phase source term sequence.

[0126] The furfural generation phase source term sequence is input into the oil-paper two-phase distribution calculation layer, and a dissolution hysteresis model from the insulating paper phase to the insulating oil phase is established. This model includes the oil-paper distribution coefficient, insulating oil temperature, insulating paper moisture content, interfacial contact area, molecular diffusion coefficient, and oil phase dissolution capacity. For each generation phase source term, the required hysteresis time and dissolution rate for entering the insulating oil phase are calculated, and corresponding dissolution wavefronts are generated according to the location of the virtual reinjection source domain. Then, for multiple dissolution wavefronts, based on the release intensity, phase difference, and oil-paper interface hindrance degree of different source domains, the wavefront propagation velocity, wavefront thickness, wavefront superposition region, and wavefront attenuation region are recalculated to ensure that the process of furfural entering the insulating oil from the insulating paper is not simplified to instantaneous dissolution. Finally, the reconstructed dissolution wavefronts are accumulated in chronological order to obtain the amount of furfural entering the oil phase, the hysteresis release, the interfacial residue, and the dissolution increment at different time steps, thus forming a furfural dissolution response sequence.

[0127] The furfural dissolution response sequence was incorporated into the oil path migration calculation layer of the state-anchored digital twin, and an insulating oil circulation path model was established. This model includes the main oil path, winding gap oil path, cooler circuit, local stagnation zone, and sampling detection location. Migration calculation rules were set for different spatial scales: molecular diffusion near the oil-paper interface was calculated at the microscale, convective diffusion within the winding oil path was calculated at the mesoscale, and concentration transport in the entire equipment's oil circulation circuit was calculated at the macroscale. The furfural dissolution response sequence was used as the dynamic boundary condition for the release of furfural into the insulating oil from each virtual reinjection source domain. Diffusion coefficient, convection velocity, oil temperature gradient, oil path resistance, local mixing coefficient, and cooler heat transfer coefficient were set at the micro, meso, and macroscales, respectively, enabling simultaneous calculation of furfural molecular diffusion near the oil-paper interface, convective diffusion within the winding oil path, and cyclic transport in the entire equipment's oil circulation circuit. When changes occur in oil flow direction, cooler start / stop status, hotspot temperature distribution, or local oil passage resistance, corresponding bifurcation propagation nodes are generated in the multi-scale oil flow migration network. Based on the oil flow distribution ratio, propagation delay time, and diffusion attenuation coefficient of different bifurcation propagation nodes, multi-path propagation calculations are performed on the same furfural dissolution response. Finally, the furfural arrival time, concentration accumulation, diffusion attenuation, local retention, and detection node response of each propagation path are continuously iterated over in future time steps to obtain candidate furfural concentration evolution branches corresponding to different oil flow propagation paths.

[0128] Each candidate furfural concentration evolution branch is projected onto a unified prediction time axis, and a multidimensional similarity relationship between candidate branches is constructed using concentration amplitude, concentration growth rate, concentration change curvature, propagation arrival time, oil path propagation delay, and path physical constraint satisfaction as branch recombination parameters. Then, based on the continuity of concentration change, consistency of propagation path, and consistency of generation source domain phase, candidate branches are adaptively clustered, grouping those formed by the same type of furfural generation source term, dissolution response process, or oil path migration path into the same evolution group. Next, trajectory fusion and boundary convergence processing are performed on candidate branches within each evolution group, calculating the corresponding central concentration trajectory, upper envelope concentration trajectory, lower envelope concentration trajectory, branch duration, and branch confidence weight. For cases where different evolution groups intersect, overlap, or undergo short-term merging, the evolution groups are split, merged, or reordered based on the furfural generation phase source term sequence, furfural dissolution response sequence, and oil path convection-diffusion constraints to avoid erroneous merging of concentration evolution results corresponding to different generation source domains or different migration paths. Finally, the processed evolution groups were organized according to the concentration evolution types of conventional release, accelerated release, delayed release, and local accumulation, resulting in a multi-branched furfural concentration evolution cluster with multiple pathways, multiple time scales, and multiple release modes.

[0129] In this embodiment, by discretizing the potential furfural release field into release energy cells and further mapping it to twin crack source domains on the oil-paper insulation interface, the originally continuous and difficult-to-locate furfural release process can be transformed into a calculable source domain with spatial location, release intensity, and temporal phase. Based on this, through source domain phase perturbation arrangement and dissolution hysteresis wavefront reshaping, the time difference between furfural generation and entry into the insulating oil, interface hindrance, and hysteresis release process are incorporated into the concentration evolution calculation. Combined with multi-scale convection diffusion bifurcation propagation simulation and branch clustering recombination, different oil propagation paths, different release rhythms, and different local accumulation scenarios can be covered simultaneously, reducing the bias caused by simplifying furfural concentration changes to a single generation curve. This makes the multi-branch furfural concentration evolution cluster more suitable for concentration prediction under complex oil-paper structures and variable operating conditions.

[0130] In an exemplary embodiment, based on the predicted and measured concentration data of furfural, a prediction bias analysis is performed on the state-anchored digital twin to obtain state drift characteristics, including steps 802 to 806. Wherein:

[0131] Step 802: Based on the measured concentration data, the measured zero-drift datum is reconstructed for the furfural concentration observation state in the state-anchored digital twin to obtain the measured furfural anchoring datum.

[0132] Step 804: Based on the predicted furfural concentration data, perform concentration phase trajectory penetration mapping on the measured furfural anchoring datum to obtain the predicted measured misalignment tensor.

[0133] Step 806: Based on the predicted measured misalignment tensor, perform drift source type attribution analysis on the aging trajectory of the oil-paper insulation in the state-anchored digital twin to obtain the state drift characteristic quantity.

[0134] Among them, the furfural concentration observation state refers to the furfural detection benchmark state formed by measured concentration data, detection time, oil sample temperature, detection confidence weight, and Raman spectrum quality in the state-anchored digital twin.

[0135] Among them, the measured zero-drift datum reconstruction refers to the process of correcting the batch shift, SERS substrate enhancement fluctuation and Raman spectrum baseline drift in the measured concentration data and forming a stable measured comparison datum.

[0136] Among them, the furfural measured anchoring datum refers to the measured concentration reference surface formed after zero drift correction, which is used to compare the deviation with the predicted furfural concentration data.

[0137] Among them, concentration phase trajectory penetration mapping refers to the process of mapping the concentration evolution trajectory formed by the predicted concentration data of furfural to the measured anchoring datum of furfural, and calculating the misalignment relationship between the two in terms of concentration, time, phase and curvature.

[0138] Among them, the predicted-measured misalignment tensor refers to the data structure used to describe the multidimensional deviation between the predicted and measured concentration data of furfural, including concentration misalignment, time misalignment, phase misalignment, and curvature misalignment.

[0139] Among them, the aging trajectory of oil-paper insulation refers to the evolution path of the target power equipment's oil-paper insulation system under operating stress over time, resulting in aging accumulation, furfural generation, and a decline in insulation performance.

[0140] Among them, drift source type attribution analysis refers to the process of calculating and attributing the source type of the aging trajectory deviation of oil-paper insulation based on the predicted measured misalignment tensor. The source types include accelerated thermal aging, abnormal furfural release, delayed oil-paper distribution, delayed oil path migration, and detection disturbance.

[0141] Specifically, the measured concentration data is written into the furfural observation layer of the state-anchored digital twin according to the detection time, oil sample temperature, sampling location, Raman spectroscopy acquisition batch, and detection confidence weight, and a concentration observation coordinate system is established in the furfural observation layer. This concentration observation coordinate system includes at least a detection time axis, a furfural concentration axis, a spectral quality weight axis, an oil sample temperature correction axis, and a sampling location correction axis. Based on the blank sample response, standard oil sample response, the fluctuation range of the substrate enhancement coefficient, and the Raman spectrum baseline drift, the zero-drift correction amount corresponding to the measured concentration data is calculated, and the zero-drift correction amount is used to standardize the concentration observation results under different detection batches. Then, in the state-anchored digital twin, a local datum surface of the furfural concentration observation state is constructed with the corrected measured concentration data as the reference point and the detection confidence weight and spectral quality weight as constraints. For the measured concentration changes within adjacent detection cycles, a smooth connection is established based on the constraints of concentration continuity, monotonicity of oil-paper aging, and detection error boundary constraints. This ensures that the local base surface remains continuous in the time dimension and satisfies the aging mechanism of oil-paper insulation in the concentration dimension. Finally, the local base surfaces corresponding to each detection cycle are spliced ​​and the boundary convergence is processed to obtain the furfural measured anchoring base surface that can serve as the benchmark for subsequent prediction deviation analysis.

[0142] Furfural predicted concentration data were converted into a concentration phase track sequence according to prediction time, predicted concentration value, predicted growth rate, predicted curvature, and prediction confidence interval. This concentration phase track sequence was then loaded into the same concentration observation coordinate system as the measured furfural anchoring datum. Penetration mapping coordinates were established using the normal direction, tangential direction, and time progression direction of the measured furfural anchoring datum. This ensured that each concentration phase track point at each predicted time step could form a spatial mapping relationship with the anchoring datum. The normal penetration, tangential slip, time lag, and curvature misalignment of the concentration phase track sequence relative to the measured furfural anchoring datum were then calculated. Based on this, the normal penetration was used to calculate the longitudinal deviation of the predicted concentration relative to the measured datum; the tangential slip was used to calculate the lateral deviation of the predicted trajectory during the aging stage; the time lag was used to calculate the phase difference between the predicted and measured changes; and the curvature misalignment was used to calculate the difference between the predicted concentration change pattern and the curvature of the measured datum. The aforementioned multidimensional deviations are then tensor-arranged according to the prediction time step, concentration change range, and detection confidence weight. Neighborhood constraints and physical boundary corrections are applied to abrupt deviations in continuous time steps to obtain the predicted and measured misalignment tensors that include concentration misalignment, time misalignment, phase misalignment, and curvature misalignment.

[0143] In a specific embodiment, after writing the predicted-measured misalignment tensor into the drift calculation layer of the state-anchored digital twin, the predicted-measured misalignment tensor is first decomposed into concentration misalignment components, time misalignment components, phase misalignment components, and curvature misalignment components. The concentration misalignment component describes the numerical difference between the predicted and measured furfural concentrations at the same detection time or aging phase, and is preferentially associated with the furfural generation model. The furfural generation model refers to a model in the state-anchored digital twin that calculates the furfural source term intensity generated after the degradation of cellulose in the insulating paper based on the decrease in the equivalent degree of polymerization of the insulating paper, the winding hot spot temperature, the aging energy level of the oil-paper insulation, the degree of oxidative aging, and the historical aging accumulation. Its model parameters include the furfural generation coefficient, temperature acceleration coefficient, etc. The model uses the following parameters: concentration misalignment component, insulation paper aging correction coefficient, and generation source term strength coefficient. Therefore, when the concentration misalignment component is consistently positive and the measured concentration is higher than the predicted concentration, the drift calculation layer increases the generation source term strength or aging correction coefficient in the furfural generation model to determine whether there is abnormal furfural release or localized accelerated degradation of the insulation paper. The time misalignment component is used to describe the advance or lag relationship between the arrival time of the predicted furfural concentration change and the occurrence time of the measured concentration change, and is preferentially associated with the oil-paper distribution model. The oil-paper distribution model refers to calculating the release ratio and retention of generated furfural from the insulation paper phase to the insulation oil phase in the state-anchored digital twin based on the oil-paper two-phase distribution coefficient, insulation oil temperature, insulation paper moisture content, oil-paper contact area, interface resistance degree, and dissolution lag time. The model for the proportional and delayed release process includes parameters such as the oil-paper distribution coefficient, the interface release retardation coefficient, the dissolution retardation coefficient, and the oil phase entry ratio. Therefore, when the time misalignment component shows that the measured concentration change lags behind the predicted concentration change, the drift calculation layer adjusts the dissolution retardation coefficient or the interface retardation coefficient in the oil-paper distribution model to determine whether furfural is delayed in entering the insulating oil due to slow release at the oil-paper interface. The phase misalignment component is used to describe the difference in response rhythm between furfural concentration changes and load disturbances, hot spot temperature changes, and oil flow circulation changes, and is preferentially associated with the oil path migration model. The oil path migration model refers to the oil path topology formed in the state-anchored digital twin based on the main oil path, winding gap oil path, cooler circuit, local stagnant zone, and sampling detection location. Combining oil flow velocity, convection diffusion coefficient, local temperature gradient, oil channel resistance, and cooler start-up and shutdown status, a model is calculated to depict the propagation, mixing, attenuation, and retention process of furfural in insulating oil from the release area to the detection area. The model parameters include the oil flow propagation delay coefficient, convection diffusion coefficient, local retention coefficient, and detection position response coefficient. Therefore, when the phase misalignment component shows that the predicted concentration change and the measured concentration change are inconsistent in the rhythm of the operational disturbance response, the drift calculation layer adjusts the propagation delay coefficient or local retention coefficient in the oil channel migration model to determine whether the misalignment originates from a change in the oil flow migration path or a lag in the sampling position response. The curvature misalignment component is used to describe the acceleration, deceleration, or inflection point morphology differences of the furfural concentration change curve and is preferentially associated with the thermal aging model.The aforementioned thermal aging model refers to a model in a state-anchored digital twin that calculates the aging rate, aging acceleration, and accumulated thermal stress of oil-paper insulation based on winding hot spot temperature, top oil temperature, ambient temperature, load loss, cooling status, and temperature duration. Its model parameters include a hot spot temperature acceleration factor, a thermal aging reaction rate coefficient, a temperature duration weight, and a cooling correction coefficient. Therefore, when the curvature misalignment component shows a change in measured concentration from gradual to rapid or an abnormal inflection point, the drift calculation layer adjusts the hot spot temperature acceleration factor or thermal aging reaction rate coefficient in the thermal aging model to determine whether there is accelerated aging caused by local overheating or sustained high temperatures. After completing the above-mentioned correlation, the drift calculation layer sets furfural abnormal release channels, oil-paper distribution lag channels, oil path migration delay channels, and thermal aging acceleration channels in the furfural generation model, oil-paper distribution model, oil path migration delay channel, and thermal aging model, respectively. It can also set up a detection disturbance suppression channel to mitigate non-equipment degradation deviations caused by Raman detection batch errors, SERS substrate enhancement fluctuations, or oil sample sampling fluctuations. Subsequently, the drift calculation layer changes the corresponding model parameters within each channel according to a preset step size and recalculates the remaining misalignment degree of the predicted measured misalignment tensor. If adjusting the parameters of a certain channel can simultaneously reduce the corresponding misalignment component and the overall misalignment tensor, then that channel is identified as the main drift source and its drift contribution weight is increased. If adjusting the parameters only reduces local misalignment without converging the overall misalignment, then the drift contribution weight of that channel is decreased. Finally, the drift contribution weights of each channel are combined with the deviation amplitude, deviation direction, duration, and rate of change in the predicted measured misalignment tensor to form drift direction, drift intensity, drift duration, and drift source weights, which are then normalized and encoded according to a unified vector format to obtain state drift characteristic quantities.

[0144] In this embodiment, by reconstructing the measured zero-drift datum, the influence of batch differences, SERS substrate enhancement fluctuations, and Raman spectrum baseline drift on the measured concentration benchmark can be eliminated first, making the measured concentration data form a stable comparison datum. Then, through concentration phase-track penetration mapping, the difference between the furfural predicted concentration data and the furfural measured anchoring datum is converted into a predicted-measured misalignment tensor that includes concentration misalignment, time misalignment, phase misalignment, and curvature misalignment, so that the deviation analysis is no longer limited to a single concentration difference. Finally, through drift source type attribution analysis, the misalignment tensor is correlated with degradation sources such as accelerated thermal aging, abnormal furfural release, delayed oil-paper distribution, and delayed oil path migration. This can distinguish whether the prediction error comes from detection disturbances or actual equipment degradation offset, thereby reducing the probability of false alarms and improving the ability to locate abnormal changes in the aging path of oil-paper insulation.

[0145] In an exemplary embodiment, based on the predicted furfural concentration data, a concentration phase trajectory penetration mapping is performed on the measured furfural anchoring datum to obtain the predicted measured misalignment tensor, including steps 902 to 908. Wherein:

[0146] Step 902: Based on the furfural predicted concentration data, perform phase orbit convolution encoding on the furfural predicted concentration evolution sequence in the state-anchored digital twin to obtain the predicted concentration phase orbit chain.

[0147] Step 904: Based on the measured anchoring datum of furfural, perform a datum normal penetration scan on the predicted concentration phase track chain to obtain the phase track penetration intersection set.

[0148] Step 906: Based on the phase orbit penetration intersection set, tensor-quantized rearrangement is performed on the concentration misalignment relationship and time misalignment relationship between the predicted concentration phase orbit chain and the measured furfural anchoring datum to obtain the concentration misalignment sub-tensor.

[0149] Step 908: Based on the concentration misalignment tensor, perform vortex nucleation stitching encoding on the misalignment propagation topology between the predicted concentration phase orbital chain and the measured furfural anchoring datum to obtain the predicted measured misalignment tensor.

[0150] Among them, the furfural predicted concentration evolution sequence refers to the furfural concentration prediction results arranged in the order of future prediction time, which is used to describe the continuous process of furfural concentration changing over time.

[0151] Among them, phase trajectory curvature coding refers to the process of converting the concentration value, growth rate and change curvature in the furfural predicted concentration evolution sequence into a phase space trajectory with direction, curvature and evolution speed.

[0152] Among them, the predicted concentration phase track chain refers to the predicted concentration change trajectory of furfural formed by multiple continuous phase track segments connected in series, which is used to describe the change path, direction and rhythm of the predicted concentration.

[0153] Among them, the basal plane normal penetration scan refers to the process of stepwise calculation of the intersection relationship between the predicted concentration phase orbital chain and the furfural measured anchoring basal plane along the normal direction of the measured anchoring basal plane.

[0154] Among them, the phase track penetration intersection set refers to the set of intersection data formed when the predicted concentration phase track chain intersects, projects, or effectively penetrates the measured anchoring datum of furfural.

[0155] Among them, the concentration misalignment relationship refers to the hierarchical deviation relationship between the predicted concentration phase track and the measured anchoring datum of furfural in the direction of concentration values.

[0156] Among them, the time misalignment relationship refers to the sequential deviation between the predicted concentration phase track and the measured anchoring datum of furfural at the predicted time and the measured time.

[0157] Among them, Tensor quantization rearrangement refers to the process of reorganizing the concentration misalignment relationship, time misalignment relationship, phase misalignment relationship and curvature misalignment relationship into a multi-dimensional data structure according to the prediction time step and the number of penetration intersection point.

[0158] Among them, the concentration misalignment subtensor is an intermediate tensor composed of multidimensional deviations such as concentration misalignment relation and time misalignment relation, which is used to describe the local misalignment state of the predicted concentration phase track relative to the measured anchoring datum of furfural.

[0159] Among them, the misaligned propagation topology refers to the propagation direction, aggregation region, diffusion path and continuity relationship formed by the amount of misalignment between the predicted concentration phase track and the measured anchoring datum of furfural over time.

[0160] Among them, vortex nucleation stitching coding refers to the process of encoding the misalignment aggregation center, swirling region, propagation break location and reconnection relationship in the misalignment propagation topology to form a continuous prediction of the measured misalignment structure.

[0161] Specifically, furfural predicted concentration data is written into the concentration evolution calculation layer of the state-anchored digital twin according to the predicted time step, and the predicted concentration value, concentration growth rate, concentration change curvature, prediction confidence interval, release potential intensity, and oil-paper distribution delay are configured for each predicted time step. Using prediction time as the main axis and the predicted concentration value, concentration growth rate, and concentration change curvature as phase space coordinates, a concentration phase space for the furfural predicted concentration evolution sequence is constructed. Then, the direction of concentration change, curvature change, and growth rate change between adjacent predicted time steps are continuously encoded in the concentration phase space, transforming the originally time-arranged predicted concentration points into phase track segments with direction, curvature, and evolution speed. Then, short-term concentration spikes, long-term slow increases, delayed release declines, and local accumulation plateaus in each phase track segment are converted into different curling radii, curling angles, and phase track pitch parameters, and the phase track segments are concatenated in chronological order to obtain the predicted concentration phase track chain.

[0162] A local geometric coordinate system is established on the furfural measured anchoring datum, including the datum tangential direction, datum normal direction, and detection time progression direction. The normal vector of each datum grid point is calculated based on the measured concentration change between adjacent detection cycles, zero drift correction, and detection confidence weight. The predicted concentration phase track chain is then projected into the local geometric coordinate system, and the phase track chain is sequentially pushed along the datum normal direction for penetration scanning according to the predicted time step. In each scan, the spatial intersection position, intersection time position, and normal penetration distance between the phase track point or phase track segment on the predicted concentration phase track chain and the furfural measured anchoring datum are calculated. For cases where the phase track chain intersects the datum multiple times, the primary intersection point is determined according to the time sequence and the principle of minimum normal distance, while secondary intersection points are retained for subsequent staggered layer relationship calculations. For cases where the phase track chain does not directly pass through the datum, an equivalent penetration point is generated according to the minimum normal projection distance. Finally, all primary intersection points, secondary intersection points, and equivalent penetration points are organized according to the predicted time sequence to obtain the phase track penetration intersection point set.

[0163] Using each penetration intersection point in the phase track penetration intersection set as a calculation node, the predicted concentration coordinates, measured datum concentration coordinates, predicted time coordinates, measured datum time coordinates, phase track tangential direction, and datum normal direction corresponding to that intersection point are calculated respectively. Simultaneously, the concentration layer difference is determined based on the normal distance between the predicted concentration coordinates and the measured datum concentration coordinates of the phase track chain and the measured datum anchoring surface of furfural; the time layer difference is determined based on the projected distance between the predicted time coordinates and the measured datum time coordinates of the phase track chain and the measured datum anchoring surface of furfural; the phase layer difference is determined based on the angle between the phase track tangential direction and the datum tangential direction of the phase track chain and the measured datum anchoring surface of furfural; and the local curvature layer difference is determined based on the difference between the phase track segment curvature and the local curvature of the datum surface between the phase track chain and the measured datum anchoring surface of furfural. Then, the above calculation results are tensor-quantized and arranged according to the prediction time step, penetration intersection number, concentration layer difference dimension, time layer difference dimension, phase layer difference dimension, and curvature layer difference dimension. Neighborhood smoothing and physical constraint correction are applied to abrupt layer differences in continuous time steps to ensure that the concentration misalignment relationship and the time misalignment relationship remain continuous between adjacent prediction time steps. Finally, the multidimensional misalignment calculation results after tensor-quantization rearrangement are combined into a concentration misalignment subtensor.

[0164] The concentration layer difference, time layer difference, phase layer difference, and curvature layer difference in the concentration misalignment subtensor are mapped to the misalignment propagation topology space. A misalignment propagation network is constructed relative to the measured furfural anchoring datum, with the predicted time step as the propagation direction and the concentration misalignment change rate and time misalignment change rate as the topological flow direction. The aggregation center, diffusion direction, and closed loop region of the misalignment are calculated in the misalignment propagation network. Regions where the misalignment continuously increases and the propagation direction is stable are designated as vortex core regions, while regions where the misalignment changes periodically or loops locally are designated as stitching candidate regions. Vortex core regions are vortex-nucleated, with the vortex core center position, vortex core radius, misalignment rotation intensity, misalignment propagation direction, and duration written into the topology coding unit. Simultaneously, stitching candidate regions are stitched, with the break position, reconnection direction, phase compensation, and concentration compensation between adjacent misalignment propagation segments written into the stitching coding unit. Finally, the topological coding unit and the stitching coding unit are combined according to the prediction time order, concentration misalignment intensity, and phase misalignment intensity to form a predicted measured misalignment tensor that can simultaneously describe the misalignment amplitude, misalignment direction, misalignment propagation path, misalignment clustering region, and misalignment continuity.

[0165] In this embodiment, the furfural predicted concentration evolution sequence is converted into a predicted concentration phase track chain with direction, curvature, and evolution speed through phase track curling encoding, which can preserve the growth rhythm, delayed release, and local accumulation characteristics during the concentration change process. Then, through basal plane normal penetration scanning, the correspondence between the predicted concentration phase track chain and the measured furfural anchoring basal plane is converted into a phase track penetration intersection set, so that the prediction result and the measured benchmark form a continuous spatial comparison relationship. Subsequently, through tensor rearrangement, the concentration misalignment relationship and the time misalignment relationship are organized into a concentration misalignment subtensor, avoiding the evaluation of prediction deviation based solely on single-point concentration difference. Finally, through vortex core stitching encoding, the misalignment aggregation, misalignment diffusion, and misalignment breakage and reconnection processes in the misalignment propagation topology are uniformly encoded, so that the predicted measured misalignment tensor can simultaneously reflect the deviation intensity, deviation direction, deviation persistence, and deviation propagation path, thereby improving the ability to distinguish between furfural prediction anomalies, detection lag, and aging trajectory deviations.

[0166] In an exemplary embodiment, surface-enhanced Raman scattering analysis is performed on furfural molecular information in insulating oil sample data of the target power equipment to obtain measured concentration data and Raman spectral characteristic data, including steps 1002 to 1010. Wherein:

[0167] Step 1002: Construct an interference spectral field for the fluorescence background, oil viscosity disturbance, and interference from non-target aging products in the insulating oil sample data to obtain the oil sample interference spectral field data.

[0168] Step 1004: Based on the interference spectral field data of the oil sample, reverse selective activation is performed on the furfural trapping sites in the surface-enhanced Raman scattering substrate to obtain the furfural directional enrichment interface.

[0169] Step 1006: Perform micro-region step-Raman excitation scanning on the furfural molecule information in the furfural directional enrichment interface to obtain multi-temporal enhanced Raman spectrum data.

[0170] Step 1008: Based on the multi-temporal enhanced Raman spectral data and the oil sample interference spectral field data, background stripping and hotspot drift compensation are performed on the furfural molecular information to obtain furfural purification Raman spectral data.

[0171] Step 1010: Based on the furfural purification Raman spectrum data, perform peak topology locking and concentration grid quantization on the furfural candidate characteristic peak data in the furfural purification Raman spectrum data to obtain the measured concentration data and Raman spectral characteristic data.

[0172] Among them, the fluorescent background refers to the broadband fluorescence signal generated by insulating oil and its impurities under Raman excitation light, which will raise the spectrum baseline and weaken the discernibility of furfural Raman peak.

[0173] Among them, oil viscosity perturbation refers to the effect of changes in insulating oil viscosity on molecular diffusion, substrate contact efficiency, and Raman scattering peak shape.

[0174] Interference from non-target aging products refers to the overlapping peaks, background peaks, or competitive adsorption effects of alcohols, ketones, acids, aldehydes, or oxidation byproducts other than furfural in insulating oil during Raman detection.

[0175] Among them, interference spectral field construction refers to organizing the fluorescence background, oil viscosity disturbance and non-target aging product interference into a spectroscopic interference model that can be used for correction according to wavenumber, intensity, time and detection conditions.

[0176] Among them, the oil sample interference spectral field data refers to the data constructed from the interference spectral field, which is used to describe the background distribution, peak position interference, peak shape disturbance and baseline drift of the insulating oil sample in Raman detection.

[0177] The substrate refers to the simulated detection carrier used to generate the surface-enhanced Raman scattering effect, which typically includes simulated gold, silver, or gold-silver composite nanostructures and their surface modification layers.

[0178] Among them, furfural capture sites refer to functional regions on the substrate surface that can selectively adsorb, condense, coordinate, or hydrogen bond with furfural molecules.

[0179] Among them, reverse selective activation refers to a treatment method that first suppresses the base region that is prone to adsorbing interfering substances based on the interference spectral field data of the oil sample, and then activates the capture sites suitable for furfural molecule enrichment.

[0180] Among them, the furfural directional enrichment interface refers to the interface formed after reverse selective activation, which can promote furfural molecules to preferentially enter the substrate surface functional interface that effectively enhances the hot spot region.

[0181] Among them, furfural molecular information refers to information related to the content, adsorption state, Raman response, peak position and peak intensity of furfural molecules.

[0182] Among them, micro-area step-type Raman excitation scanning refers to a detection method that uses Raman excitation to collect furfural molecules in different tiny regions of the substrate according to progressively varying laser power, integration time, or scanning sequence.

[0183] Among them, multi-temporal enhanced Raman spectral data refers to the collection of enhanced Raman spectra obtained at different scanning times, different excitation stages, and different substrate microregions.

[0184] Background stripping refers to the process of subtracting or reducing the fluorescence background, baseline rise, and non-target interference peaks in the Raman spectrum based on the interference spectral field data of the oil sample.

[0185] Hotspot drift compensation refers to the process of correcting signal intensity fluctuations caused by changes in scanning position, temperature, time, or adsorption state of enhanced hotspots in the SERS substrate.

[0186] Among them, furfural purified Raman spectrum data refers to the spectrum data obtained after background stripping and hotspot drift compensation, which has a more stable furfural Raman response and reduced interference.

[0187] Among them, furfural candidate characteristic peak data refers to the peak position, peak intensity, peak area, peak width, and peak shape data that may correspond to the vibrational response of furfural molecules in the furfural purification Raman spectrum data.

[0188] Peak topology locking refers to the process of determining a stable furfural peak combination based on the peak position order, peak spacing, peak intensity ratio, peak shape continuity, and vibration response law among the candidate characteristic peaks of furfural.

[0189] Concentration grid quantization refers to the process of matching the locked furfural peak combination with a preset concentration grid library and calculating the measured concentration of furfural accordingly.

[0190] Specifically, the insulating oil sample data was simulated within a constant-temperature detection chamber, and the oil sample temperature, thickness, optical path distance, and Raman excitation power were controlled within preset ranges to reduce the impact of fluctuations in acquisition conditions on subsequent spectral field calculations. Without contact with the surface-enhanced Raman scattering substrate, the insulating oil sample data underwent multi-wavenumber interval pre-scanning, recording the fluorescence background intensity, baseline rise, scattering broadening caused by oil viscosity, and the positions of interference peaks corresponding to non-target aging products. An interference correction matrix was established using blank insulating oil, standard furfural oil samples, and control oil samples containing non-target aging products. The fluorescence background term, viscosity broadening term, non-target peak overlap term, and temperature drift term were uniformly written into the same wavenumber-intensity-time three-dimensional coordinate system. For baseline differences between different scanning batches, dark current subtraction, laser power normalization, and local baseline smoothing were used for correction. The corrected fluorescence background distribution, viscosity perturbation distribution, and non-target aging product interference distribution were combined to form the oil sample interference spectral field data.

[0191] Surface-enhanced Raman scattering (SERS) substrate data (obtained from actual substrates using gold, silver, or gold-silver composite nanostructures through modeling) were simulated and placed on a substrate platform in a temperature-controlled microreactor. A furfural trapping molecular layer containing amino, thiol, or aptamer structures was then modified onto its surface. Based on the interference peak positions and fluorescence background intensity of non-target aging products in the oil sample interference spectral field data, the non-specific adsorption regions on the substrate surface were simulated for shielding. For example, short-chain thiols, inert polymer brushes, or oleophobic barrier molecules were used to passivate hotspot regions prone to background adsorption. Temperature, potential, or microenvironment polarity adjustments were simulated to the furfural trapping sites that still retained reactivity, preferentially maintaining condensation, hydrogen bonding, or coordination with furfural molecules, while reducing their adsorption capacity for ketones, alcohols, acids, and macromolecular matrices in the oil. However, this process did not directly enhance all SERS hotspots; instead, it first suppressed hotspot regions easily occupied by interfering substances and then activated trapping sites that reacted with furfural. Finally, a furfural-oriented enrichment interface consisting of an activation capture region, a passivation isolation region, and an enhanced detection region is formed on the simulated substrate surface.

[0192] Insulating oil sample data was injected into a simulated microreactor containing a furfural-oriented enrichment interface. The reaction temperature, reaction time, and oil flow rate were simulated and controlled to ensure stable adsorption of furfural molecules at the interface. A Raman laser was then focused onto multiple pre-defined microregions on the substrate surface. Each microregion corresponded to a different enhanced hotspot density and capture site distribution. A step-excitation method was simulated, involving low-power pre-irradiation, medium-power stable scanning, and target-power enhanced acquisition. In practice, a lower laser power was used in the first step-excitation stage to record the substrate background and initial adsorption state. In the second step-excitation stage, the laser power was increased and the integration time shortened to record the dynamic enhancement signal during furfural adsorption. In the third step-excitation stage, the target excitation power and multiple cumulative integration methods were used to record the stable enhancement signal. Simultaneously, multiple scans were performed sequentially within each microregion to record Raman signals at different adsorption times, hotspot locations, and excitation powers. Finally, the enhanced Raman spectra corresponding to each microregion, each step power, and each acquisition time were registered according to the acquisition sequence to obtain multi-temporal enhanced Raman spectral data.

[0193] Multi-temporal enhanced Raman spectral data and oil sample interference spectral field data were placed under the same wavenumber coordinates and normalized according to laser power, integration time, substrate temperature, and scanning micro-region location. The slowly changing baseline in the multi-temporal enhanced Raman spectral data was subtracted using the fluorescence background distribution in the oil sample interference spectral field data. Peak broadening and slight peak position drift were corrected using the viscosity perturbation distribution. Interference peaks in overlapping wavenumber intervals were weighted and reduced using the non-target aging product interference distribution. Then, signal differences caused by inconsistent SERS hotspot intensities between different micro-regions were compensated. Compensation was calculated based on the substrate-embedded reference peak, the ratio of enhancement intensities of adjacent scanning points, and the hotspot response stability coefficient at the same temporal phase. This compensation factor was applied to the Raman spectrum of the corresponding micro-region. After background stripping and hotspot compensation, the spectra were subjected to wavenumber alignment, peak smoothing, and abnormal peak suppression to maintain comparability of furfural-related peaks across different time phases and scanning micro-regions, ultimately yielding furfural purification Raman spectral data.

[0194] A candidate peak search window was established within a preset wavenumber range of the furfural-purified Raman spectrum data. Combined with standard furfural-enhanced Raman spectra, substrate-modified molecular response spectra, and insulating oil blank spectra, the allowable peak position offset range, allowable peak tolerance range, peak intensity ratio constraint, and peak spacing constraint of the furfural candidate characteristic peak data in the furfural-purified Raman spectrum data were determined. Then, topological correlation calculations were performed on the peak position order, peak spacing, peak intensity ratio, peak area ratio, and peak shape continuity among the furfural candidate characteristic peak data. This ensured that multiple candidate peaks conforming to the vibrational response law of furfural molecules formed spectral peak topological units, and residual fluorescence peaks, oil matrix interference peaks, and random noise peaks that did not meet the topological correlation constraints were excluded from the quantization calculations. The spectral peak topological units were input into a preset concentration grid library, which consisted of peak intensity combinations, peak area combinations, peak intensity ratio combinations, peak width combinations, and topological stability parameters under different known furfural concentrations. The target concentration grid and the weights of adjacent concentration grids were determined according to the multi-parameter distances between the spectral peak topological units and each concentration grid. Finally, furfural concentration values ​​are calculated based on the target concentration grid points and the weights of adjacent concentration grid points to obtain measured concentration data. At the same time, the peak position, peak intensity, peak area, peak width, peak intensity ratio, topological connectivity, grid point weight, and spectral confidence weight corresponding to the spectral peak topological unit are combined into Raman spectral characteristic data.

[0195] In this embodiment, by first constructing oil sample interference spectral field data, the interference from fluorescence background, oil viscosity disturbance, and non-target aging products is incorporated into the subsequent analysis constraints, which can reduce the masking and peak position interference of the complex insulating oil matrix on the furfural Raman signal. Then, based on the oil sample interference spectral field data, the furfural capture sites in the SERS substrate are selectively activated in reverse, so that the substrate preferentially forms a directional enrichment interface adapted to furfural molecules, reducing the competition for enhancement hotspots by ketones, alcohols, acids, and macromolecular matrices in the oil. Subsequently, multi-temporal enhanced Raman spectrum data are obtained by micro-area step Raman excitation scanning, so that the furfural response under different adsorption stages, different enhancement hotspots, and different excitation powers can be continuously recorded. Further, background stripping and hotspot drift compensation are performed by combining the oil sample interference spectral field data, which can weaken the measurement deviation caused by baseline drift, fluorescence background, and SERS hotspot inhomogeneity. Finally, by peak topology locking and concentration grid quantization, the furfural candidate characteristic peak data are transformed into stable concentration results and spectral characteristic results, thereby improving the anti-interference ability, repeatability, and quantitative stability of furfural in-situ detection under complex insulating oil environment.

[0196] In an exemplary embodiment, the operating state of the target power equipment is analyzed based on state drift characteristics to obtain equipment state characterization data, including steps 1102 to 1108. Wherein:

[0197] Step 1102: Based on the state drift characteristics, calculate the risk level of the aging drift direction, aging drift amplitude, and aging drift rate in the state-anchored digital twin to obtain the drift risk level map.

[0198] Step 1104: Based on the drift risk energy level map, perform multi-state gating partitioning on the operating status of the target power equipment to obtain the operating status gating label.

[0199] Step 1106: Based on the operating state gating label and state drift characteristics, perform twin shadow simulation analysis on the deterioration path of the target power equipment under future operating conditions to obtain a cluster of candidate risk evolution paths.

[0200] Step 1108: Based on the candidate risk evolution path cluster, perform risk path resonance screening and state characterization folding coding on the operating state of the target power equipment to obtain equipment state characterization data.

[0201] Among them, the aging drift direction refers to the direction of change when the aging trajectory of the oil-paper insulation of the target power equipment deviates from the expected aging trajectory of the state-anchored digital twin.

[0202] Among them, the aging drift amplitude refers to the deviation between the actual aging trajectory and the expected aging trajectory of the target power equipment.

[0203] Among them, the aging drift rate refers to the rate at which the aging drift amplitude increases or decreases over time.

[0204] Among them, risk level conversion refers to the process of converting the aging drift direction, aging drift amplitude, aging drift rate and duration into a gradeable risk value.

[0205] Among them, the drift risk level map refers to the distribution result of aging drift risk organized according to time, risk source, risk level and operating conditions.

[0206] Among them, the operating status refers to the electrical load, thermal condition, insulation aging degree and risk level of the target power equipment under the current or future operating conditions.

[0207] Among them, multi-state gating zoning refers to the process of classifying target power equipment into different state ranges such as normal operation, mild aging, accelerated deterioration, latent fault or critical risk according to the drift risk energy level map.

[0208] Among them, the operating status gating label refers to the status mark obtained by multi-state gating partitioning, which is used to record the current state interval, risk source type, duration and transition direction of the target power equipment.

[0209] The future operating conditions refer to the load level, cooling method, ambient temperature, testing cycle, or maintenance intervention conditions that the target power equipment may encounter during subsequent operation.

[0210] Among them, the degradation path refers to the process by which the insulation aging, furfural concentration changes, risk level rises and falls, and remaining life changes of the target power equipment evolve over time under specific operating conditions.

[0211] Among them, twin shadow simulation analysis refers to the process of replicating multiple shadow twins in a state-anchored digital twin and simulating the equipment degradation path under different future operating conditions.

[0212] Among them, the candidate risk evolution path cluster refers to the set of risk change paths formed by multiple shadow twins under different future operating conditions.

[0213] Among them, risk path resonance screening refers to the process of screening out risk paths that repeatedly appear, change in the same direction, and persist in multiple future operating conditions from the candidate risk evolution path cluster.

[0214] Among them, the state characterization folding coding refers to the process of compressing and coding the health level, risk level, remaining life, warning level and operation and maintenance recommendations in multiple risk evolution paths into unified equipment state characterization data.

[0215] Specifically, state drift characteristics are written into the risk calculation layer of the state-anchored digital twin, and an aging drift coordinate system is established in this risk calculation layer. This aging drift coordinate system includes an aging drift direction axis, an aging drift amplitude axis, an aging drift rate axis, a drift duration axis, and a drift source weight axis. Then, risk energy level conversion functions are set for thermal aging acceleration, abnormal furfural release, oil-paper distribution lag, oil path migration delay, and detection disturbance suppression, respectively, converting the drift direction, drift amplitude, and drift rate corresponding to different drift sources into corresponding risk energy level values. During the conversion process, a higher directional weight is assigned to the drift direction that continuously moves towards the severely aging area, a higher intensity weight is assigned to the drift amplitude that increases rapidly in a short period of time, and a higher time weight is assigned to the drift rate that continues to increase for multiple consecutive detection cycles. At the same time, risk energy level boundaries are set using the equipment rated load, hot spot temperature upper limit, insulation paper aging threshold, and furfural concentration safety limit to avoid excessive amplification of risk levels due to a single abnormal quantity. Finally, the risk energy level values ​​under different drift sources, different time windows, and different operating conditions are arranged in a grid, and a drift risk energy level map is formed according to the energy level.

[0216] A multi-state gating system is established within the state-anchored digital twin, comprising gating zones for normal operation, mild aging, accelerated degradation, latent faults, and critical risks. The risk level values, growth rates, durations, and distribution ranges from the drift risk level map are then input into the gating system calculation unit and compared with the entry, exit, and hysteresis conditions for each gating zone. For cases where the risk level rises briefly but for insufficient duration, the gating system calculation unit restricts direct entry into the high-risk gating zone using hysteresis conditions. For cases where the risk level amplitude is low but the growth rate continues to accelerate, the gating system calculation unit includes it in the accelerated degradation gating zone using trend thresholds. For cases where the risk level is concentrated at sources of accelerated thermal aging or abnormal furfural release, the gating system calculation unit further records the corresponding source type. After the above gating calculations, the gating zone number, source type, duration, and transition direction corresponding to the current operating state of the target power equipment are combined into an operating state gating label.

[0217] Multiple shadow twins consistent with the current equipment state are replicated in the state-anchored digital twin. Each shadow twin corresponds to a future operating condition combination, which includes maintaining the current load, short-term overload, load reduction operation, enhanced cooling, increased ambient temperature, extended detection cycle, and early maintenance intervention. Then, the operating state gating label is used as the initial risk boundary for each shadow twin, and state drift characteristics are used as degradation driving inputs. An oil-paper insulation aging model, a furfural generation-dissolution-migration model, a thermal field model, and a risk evolution model are run in each shadow twin. After running the models in each shadow twin, each shadow twin advances the time step according to the corresponding future operating condition, calculating the oil-paper insulation aging rate, furfural concentration change, hotspot temperature duration, drift risk level change, and gating zone transition. For shadow twins entering the high-risk gating zone, their risk expansion rate and reversible operating window are calculated; for shadow twins whose risk level decreases after load reduction or enhanced cooling, their risk fallback time and operating constraints are recorded. Finally, the risk evolution trajectories obtained from multiple shadow twins are organized according to future operating conditions, risk source types, and gating transition results to obtain a cluster of candidate risk evolution paths.

[0218] Each risk evolution path in the candidate risk evolution path cluster is projected onto a unified risk phase space that includes risk level, risk growth rate, number of gate transitions, furfural concentration change rate, hotspot temperature duration, remaining lifetime change, and maintenance intervention response. The synchronization degree of risk level, gate transition, furfural change, and thermal aging change among each risk evolution path is calculated. Paths that repeatedly appear under different future operating conditions and have consistent risk change directions are identified as resonant risk paths, while paths that only appear briefly under a single operating condition and lack continuity are downweighted. Then, the health level, aging risk level, remaining lifetime range, warning level, recommended detection cycle, recommended load adjustment method, and recommended maintenance window of multiple resonant risk paths at different time steps are compressed into the same state coding framework. For cases where coding results conflict, decision ranking is performed according to the rules of risk level priority, duration priority, and reversibility priority. Finally, the folded coded health status level, insulation aging risk level, remaining lifetime range, risk source type, warning level, recommended detection cycle, recommended operating constraints, and recommended maintenance strategy are combined into equipment state characterization data.

[0219] In this embodiment, by performing risk level conversion on the aging drift direction, aging drift amplitude, and aging drift rate in the state drift characteristic quantities, the abstract trajectory offset can be transformed into a graded and comparable drift risk level map. Then, by using the drift risk level map for multi-state gating partitioning, frequent jumps in equipment operating states between adjacent risk levels can be avoided, clearly distinguishing between normal fluctuations, continuous degradation, and critical risks. Subsequently, through twin shadow extrapolation analysis, multiple possible degradation paths are simultaneously calculated under different future operating conditions, allowing for early comparison of risk changes caused by different strategies such as load reduction, enhanced cooling, continued operation, or maintenance intervention. Finally, through risk path resonance screening and state representation folding encoding, repeatedly occurring, consistent, and persistent risk results in multiple risk evolution paths are compressed into unified equipment state representation data, thereby reducing the impact of occasional operating condition disturbances on state conclusions and improving the stability and executability of operation and maintenance decisions.

[0220] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise expressly stated herein, there is no strict order restriction on the execution of these steps, and they may be executed in other orders.

[0221] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 3As shown. This computer device includes a processor, memory, input / output interfaces (I / O), and communication interfaces.

[0222] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0223] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0224] In one embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0225] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and executes the computer instructions, causing the computer device to perform the steps in the above method embodiments.

[0226] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0227] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods.

[0228] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0229] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for monitoring the operating status of power equipment based on digital twins, characterized in that, The method includes: Based on the equipment attribute data and equipment operation data of the target power equipment, a digital coupling model is performed on the equipment operation status and equipment degradation status of the target power equipment to obtain an initial digital twin; Surface-enhanced Raman scattering analysis was performed on the furfural molecule information in the insulating oil sample data of the target power equipment to obtain the measured concentration data and Raman spectral characteristic data. Based on the measured concentration data and the Raman spectral characteristic data, the initial digital twin is anchored to the aging state of the paper insulation to obtain a state-anchored digital twin. Based on the historical operating data and real-time operating status data in the equipment operating data, insulation aging trajectory offset analysis is performed on the state anchoring digital twin to obtain state drift characteristic quantities. Based on the state drift characteristics, the operating state of the target power equipment is analyzed to obtain equipment state characterization data.

2. The method according to claim 1, characterized in that, The process involves performing insulation aging trajectory offset analysis on the state-anchored digital twin based on historical operating data and real-time operating status data from the equipment operating data to obtain state drift characteristic quantities, including: Based on the historical operating data and the real-time operating status data, a trend analysis is performed on the aging evolution process of the oil-paper insulation in the state-anchored digital twin to obtain insulation aging trend data. Based on the insulation aging trend data, the evolution process of furfural concentration in the state-anchored digital twin is predicted and analyzed to obtain furfural predicted concentration data. Based on the predicted furfural concentration data and the measured concentration data, a prediction bias analysis is performed on the state-anchored digital twin to obtain the state drift characteristic quantity.

3. The method according to claim 2, characterized in that, The step involves performing trend analysis on the aging evolution process of the oil-paper insulation in the state-anchored digital twin based on the historical operating data and the real-time operating status data, to obtain insulation aging trend data, including: Based on the real-time operating status data, transient aging disturbance identification is performed on the current oil-paper insulation state in the state-anchored digital twin to obtain the current aging disturbance characteristic data; Based on the current aging disturbance characteristic data, the historical aging evolution segments in the historical operation data are reverse matched to obtain historical homogeneous aging trajectory data. Based on the historical homogeneous aging trajectory data and the current aging disturbance characteristic data, the forward trend reconstruction of the aging evolution process of the oil-paper insulation is performed to obtain the insulation aging trend data.

4. The method according to claim 3, characterized in that, The step of reconstructing the forward trend of the aging evolution process of the oil-paper insulation based on the historical homogeneous aging trajectory data and the current aging disturbance characteristic data to obtain the insulation aging trend data includes: Based on the current aging disturbance characteristic data, a virtual annealing state search is performed on the aging evolution process of the oil-paper insulation to obtain the aging evolution boundary set; Based on the aging evolution boundary set, the historical homogeneous aging trajectory data is stripped in reverse time sequence to obtain the key aging driving factor sequence. Based on the key aging driving factor sequence and the current aging disturbance characteristic data, the aging evolution process of the oil-paper insulation is simulated across time scales to obtain the aging trajectory reconstruction sequence. Based on the aging trajectory reconstruction sequence, twin bifurcation path screening is performed on the aging evolution process of the oil-paper insulation to obtain the insulation aging trend data.

5. The method according to claim 2, characterized in that, The step involves predicting and analyzing the furfural concentration evolution process in the state-anchored digital twin based on the insulation aging trend data to obtain furfural predicted concentration data, including: Based on the insulation aging trend data, furfural release potential mapping is performed on the oil paper insulation aging energy level in the state-anchored digital twin to obtain the furfural release potential sequence. Based on the furfural release potential sequence, a virtual pulse reinjection simulation was performed on the furfural generation process, furfural dissolution process, and furfural migration process in the state-anchored digital twin to obtain a multi-branch furfural concentration evolution cluster. Based on the multi-branch furfural concentration evolution cluster, the furfural concentration evolution process is predicted by concentration manifold folding to obtain the predicted furfural concentration data.

6. The method according to claim 5, characterized in that, The process involves virtual pulse reinjection simulation of the furfural generation, dissolution, and migration processes in the state-anchored digital twin based on the furfural release potential sequence, resulting in a multi-branch furfural concentration evolution cluster, including: Based on the furfural release potential sequence, the energy state discrete reconstruction of the furfural potential release field in the state-anchored digital twin is performed to obtain the furfural release energy cell set. Based on the furfural release energy cell set, twin crack source domain mapping is performed on the oil-paper insulation interface in the state-anchored digital twin to obtain furfural virtual reinjection source domain data. Based on the furfural virtual reinjection source domain data and the furfural release energy cell set, the source domain phase perturbation arrangement of the furfural generation process is performed to obtain the furfural generation phase source term sequence. Based on the furfural generation phase source term sequence, the furfural dissolution process is remodeled by dissolution hysteresis wavefront reconstruction to obtain the furfural dissolution response sequence. Based on the furfural dissolution response sequence, a multi-scale convection-diffusion-bifaction propagation simulation of the furfural migration process was performed to obtain candidate furfural concentration evolution branches. Based on the candidate furfural concentration evolution branches, the furfural concentration evolution paths in the state-anchored digital twin are subjected to branch clustering and recombination to obtain the multi-branch furfural concentration evolution cluster.

7. The method according to claim 2, characterized in that, The step of performing prediction bias analysis on the state-anchored digital twin based on the predicted furfural concentration data and the measured concentration data to obtain state drift characteristic quantities includes: Based on the measured concentration data, the measured zero-drift datum plane is reconstructed for the furfural concentration observation state in the state-anchored digital twin to obtain the measured furfural anchoring datum plane. Based on the furfural predicted concentration data, concentration phase trajectory penetration mapping is performed on the furfural measured anchoring base to obtain the predicted measured misalignment tensor. Based on the predicted measured misalignment tensor, a drift source type attribution analysis is performed on the aging trajectory of the oil-paper insulation in the state-anchored digital twin to obtain the state drift characteristic quantity.

8. The method according to claim 7, characterized in that, The step of performing concentration phase-track penetration mapping on the measured anchoring datum of furfural based on the predicted furfural concentration data to obtain the predicted measured misalignment tensor includes: Based on the furfural predicted concentration data, the furfural predicted concentration evolution sequence in the state-anchored digital twin is subjected to phase orbit convolution encoding to obtain the predicted concentration phase orbit chain. Based on the measured anchoring datum plane of furfural, the predicted concentration phase track chain is subjected to a datum plane normal penetration scan to obtain the phase track penetration intersection set. Based on the phase track penetration intersection set, tensor rearrangement is performed on the concentration misalignment relationship and time misalignment relationship between the predicted concentration phase track chain and the measured furfural anchoring datum to obtain the concentration misalignment subtensor. Based on the concentration misalignment subtensor, the misalignment propagation topology between the predicted concentration phase orbital chain and the measured furfural anchoring datum is vortex-nucleated and stitched together to obtain the predicted measured misalignment tensor.

9. The method according to claim 1, characterized in that, The surface-enhanced Raman scattering analysis of furfural molecular information in the insulating oil sample data of the target power equipment yields measured concentration data and Raman spectral characteristic data, including: Interference spectral field data of oil sample is obtained by constructing interference spectral field data for fluorescence background, oil viscosity disturbance and non-target aging product interference in the insulating oil sample data. Based on the interference spectral field data of the oil sample, the furfural trapping sites in the surface-enhanced Raman scattering substrate are selectively activated in reverse to obtain a furfural-oriented enrichment interface. Micro-region step-Raman excitation scanning was performed on the furfural molecule information in the furfural directional enrichment interface to obtain multi-temporal enhanced Raman spectrum data. Based on the multi-temporal enhanced Raman spectral data and the oil sample interference spectral field data, background stripping and hotspot drift compensation were performed on the furfural molecular information to obtain furfural purified Raman spectral data. Based on the furfural purification Raman spectrum data, peak topology locking and concentration grid quantization are performed on the furfural candidate characteristic peak data in the furfural purification Raman spectrum data to obtain the measured concentration data and the Raman spectral characteristic data.

10. The method according to claim 1, characterized in that, The step of analyzing the operating state of the target power equipment based on the state drift characteristic quantity to obtain equipment state characterization data includes: Based on the state drift characteristics, the risk level of the aging drift direction, aging drift amplitude and aging drift rate in the state anchored digital twin is calculated to obtain a drift risk level map. Based on the drift risk energy level map, the operating status of the target power equipment is divided into multi-state gating partitions to obtain operating status gating labels. Based on the operating state gating label and the state drift feature, twin shadow simulation analysis is performed on the deterioration path of the target power equipment under future operating conditions to obtain a cluster of candidate risk evolution paths. Based on the candidate risk evolution path cluster, the operating state of the target power equipment is subjected to risk path resonance screening and state characterization folding encoding to obtain the equipment state characterization data.

Images