Transformer health assessment method and system based on digital twinning and electrical characteristics

Abstract

The invention relates to the technical field of digital model transformer health assessment, and discloses a transformer health assessment method and system based on digital twinning and electrical characteristics, and the method comprises the steps: collecting the real-time electrical characteristic data and basic ledger information of a target transformer, and calling an initial digital twinning model; when the data meets updating conditions, generating model correction parameters or evaluation guide data by using multi-dimensional characteristics such as oil chromatography, partial discharge or winding deformation, and dynamically updating a digital twin model and a preset health evaluation process; and accurate health state assessment guidance is provided for operation and maintenance personnel through the assessment terminal. According to the method, personalized dynamic adaptation of the evaluation strategy is realized, and the intelligent level of transformer monitoring evaluation is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of digital model transformer health assessment technology, and in particular to a method and system for transformer health assessment based on digital twins and electrical characteristics. Background Technology

[0002] As a core piece of equipment in the power system, the operating status of transformers directly affects the safety and stability of the power grid. Currently, traditional health assessment methods for transformers mainly rely on periodic offline tests and alarm mechanisms with fixed thresholds. These methods struggle to identify potential transformer faults in their early stages, let alone dynamically track faulty parts of the transformer. Therefore, a more reasonable method is needed to assess transformer health, and digital twin technology is an excellent choice.

[0003] While some technologies currently employ digital twins for transformer health assessment, the integration between the transformer twin model and actual equipment data is superficial, hindering timely model updates. Furthermore, these technologies typically rely on pre-defined, standardized procedures for transformer health assessment, failing to adapt to different fault characteristics such as partial discharge, winding deformation, or insulation aging. For instance, in partial discharge monitoring, existing methods struggle to accurately pinpoint the source of the discharge and guide targeted inspections; similarly, in assessing winding mechanical condition, current methods cannot dynamically quantify the degree of deformation based on changes in frequency response characteristics to implement risk grading. Summary of the Invention

[0004] In view of the above-mentioned existing problems, the present invention proposes a transformer health assessment method and system based on digital twins and electrical characteristics, which can solve the problem that the depth of fusion between the digital twin model and the actual equipment data in the prior art is insufficient, resulting in the assessment conclusion deviating from the actual health status of the equipment.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0006] In a first aspect, the present invention provides a transformer health assessment method based on digital twins and electrical characteristics, comprising the following steps: The response data acquisition terminal collects the operating data of the target transformer, obtains the real-time electrical characteristic data and basic ledger information corresponding to the target transformer, and retrieves the initial digital twin model; When the real-time electrical feature data meets the model update conditions, the model state of the initial digital twin model is updated based on the model correction parameters generated from the corresponding real-time electrical feature data to obtain the current digital twin model. The assessment guidance data generated based on the corresponding basic ledger information updates the standard assessment steps included in the preset health assessment process, and controls the assessment terminal to guide the operation and maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps.

[0007] Preferably, updating the model state of the initial digital twin model based on the model correction parameters generated from the corresponding real-time electrical feature data to obtain the current digital twin model includes: The control data acquisition terminal performs multi-source data fusion on the target transformer to obtain a fused dataset; The fusion dataset is identified to contain characteristic gas components that indicate anomalies in dissolved gases in oil. Based on the gas concentration status of the corresponding characteristic gas components, the oil chromatography assessment level is determined, and the model correction coefficient and warning threshold corresponding to the oil chromatography assessment level are determined. The response determines that the target transformer does not meet the high-risk criteria based on the oil chromatography assessment level. Based on the model correction coefficient and the early warning threshold, the insulation aging module parameters of the initial digital twin model are updated. The assessment terminal is then controlled to guide the operation and maintenance personnel to perform a health status assessment of the target transformer based on the updated digital twin model. The response determines that the target transformer meets the high-risk criteria based on the oil chromatography assessment level. It then adds a special diagnostic step based on the model correction coefficient and the early warning threshold to the preset health assessment process. The assessment terminal guides the operation and maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment step and the special diagnostic step.

[0008] Preferably, determining the oil chromatography evaluation level based on the gas concentration state of the corresponding characteristic gas components includes: The concentration values ​​of each gas component constituting the characteristic gas are weighted and summed, and the current gas evaluation value is determined based on the obtained comprehensive gas index. Based on the preset gas ratio method, pattern recognition is performed on the characteristic gas components to obtain the fault feature vector located in the characteristic gas components; Obtain the vector dimension of the corresponding fault feature vector, and normalize the feature value of each dimension and calculate the entropy value of all the feature values ​​to obtain the feature normalization value and feature entropy value. The current mode evaluation value is determined based on the vector dimension, feature normalization value, and feature entropy value. The current rate evaluation value obtained based on the gas production rate of characteristic gas components is weighted and calculated together with the current model evaluation value and the current gas evaluation value, and the oil chromatography evaluation level is determined based on the obtained current comprehensive evaluation value.

[0009] Preferably, the method further includes: inputting the current comprehensive evaluation value into a pre-trained fault evolution prediction model to obtain the fault evolution duration of the corresponding characteristic gas component; If the fault evolution duration exceeds the preset safety margin duration, the current comprehensive evaluation value is updated based on the retrieved safety margin weights.

[0010] Preferably, the response determines that the target transformer does not meet the high-risk criteria based on the oil chromatography assessment level, updates the insulation aging module parameters of the initial digital twin model based on the model correction coefficient and the early warning threshold, and controls the assessment terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on the updated digital twin model, including: If the response oil chromatography assessment level is lower than the preset high-risk level, it is determined that the target transformer does not meet the high-risk determination criteria. Obtain the insulation aging module parameters of the initial digital twin model, and determine the standard aging rate and standard warning threshold of the insulation aging module parameters; The standard aging rate is corrected based on the model correction coefficient, and in response to the warning threshold being less than the standard warning threshold, the warning threshold is replaced with the standard warning threshold. The response is based on the assessment terminal determining that a three-dimensional model of the target transformer exists on the assessment display interface. The assessment terminal is then controlled to perform visualization rendering based on the updated digital twin model and generate a health status identification box to identify the target transformer.

[0011] Preferably, the response determines that the target transformer meets the high-risk criteria based on the oil chromatography assessment level, adds a specific diagnostic step based on the model correction coefficient and early warning threshold to the preset health assessment process, and controls the assessment terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps and the specific diagnostic steps, including: If the response oil chromatography assessment level is not lower than the preset high-risk level, it is determined that the target transformer meets the high-risk determination criteria. A diagnostic knowledge graph is generated, centered on the main characteristic gas of the corresponding characteristic gas component and associated with a preset fault type, and a preset diagnostic template is established. Fill the model correction coefficients and warning thresholds into the coefficient text slots and threshold text slots in the preset diagnostic template, and add the obtained special diagnostic steps to the preset health assessment process. The response is based on the assessment terminal determining that a three-dimensional model of the target transformer exists on the assessment display interface. The assessment terminal is then controlled to provide voice prompts based on the standard assessment steps and the special diagnostic steps, and to generate equipment identification boxes to identify the target transformer and risk area identification boxes to identify the fault feature vector.

[0012] Preferably, adding the obtained specialized diagnostic steps to the preset health assessment process includes: Obtain the standard assessment steps included in the preset health assessment process, and determine the standard diagnostic depth located in the standard assessment steps; The specific diagnostic step is added to the preset health assessment process. If the severity of the fault corresponding to the warning threshold is greater than the baseline severity corresponding to the standard diagnostic depth, the specific diagnostic step is adjusted to be placed before the standard assessment step. Otherwise, it is adjusted to be placed after the standard assessment step.

[0013] Preferably, the assessment guidance data generated based on the corresponding basic ledger information updates the standard assessment steps included in the preset health assessment process, and controls the assessment terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps, including: The sensor network located on the target transformer is controlled to collect the partial discharge pulse signal of the target transformer to obtain a partial discharge signal sequence; Time-frequency domain analysis was performed on the partial discharge signal sequence to obtain the phase distribution map of the corresponding discharge phase and the amplitude distribution map of the corresponding discharge amplitude. Discharge pulse points in adjacent phase intervals are clustered to obtain discharge clusters, and discharge clusters that are allowed to match the target transformer are determined as valid discharge clusters based on the basic ledger information. Obtain the standard assessment steps included in the preset health assessment process, and fill the text slots at the positions of the effective discharge clusters in the standard assessment steps with the winding positions corresponding to the winding positions of the effective discharge clusters. The response is based on the evaluation terminal determining that there is a three-dimensional model of the target transformer on the evaluation display interface. The evaluation terminal is then controlled to provide visual guidance based on standard evaluation steps and generate a partial discharge identification box to mark the winding area corresponding to the effective discharge cluster.

[0014] Preferably, the step of determining the discharge clusters that can match the target transformer as valid discharge clusters based on the basic ledger information includes: Obtain the equivalent frequency center for each discharge cluster, and determine all equivalent frequency centers that are allowed to match the target transformer as a primary screening group based on the rated voltage in the basic ledger information; Each frequency band range with a corresponding preset bandwidth is formed, which surrounds each equivalent frequency center located in the first screening group, and a preset discharge fingerprint database is retrieved based on the device type corresponding to the target transformer. Features of discharge pulse points within the frequency band are extracted to obtain each discharge feature subset; Each discharge feature subset is matched with a preset discharge fingerprint database for similarity, and based on the matching results, the discharge feature subsets with matching similarity greater than the preset fingerprint similarity are determined as typical discharge electron sets; Obtain the discharge energy of all typical discharge clusters located in the same frequency band, and determine the effective discharge cluster as the center of the equivalent frequency with the largest corresponding discharge energy.

[0015] Preferably, the assessment guidance data generated based on the corresponding basic ledger information updates the standard assessment steps included in the preset health assessment process, and controls the assessment terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps, further includes: The response data acquisition terminal applies a frequency sweep signal to the target transformer, obtains the frequency response characteristic curve corresponding to the target transformer, and retrieves historical frequency response characteristic curves. Correlation analysis was performed between the frequency response characteristic curve and the historical frequency response characteristic curve to obtain a correlation coefficient sequence; The correlation coefficient sequence is determined to have a low correlation coefficient interval that indicates winding deformation. The degree of winding deformation is determined based on the frequency band offset of the corresponding low correlation coefficient interval. The model deformation parameters and maintenance priority corresponding to the degree of winding deformation are also determined. The response determines that the target transformer does not meet the emergency maintenance conditions based on the degree of winding deformation. Based on the model deformation parameters and maintenance priority, the mechanical structure module of the current digital twin model is updated, and the evaluation terminal is controlled to guide the operation and maintenance personnel to perform a health status assessment of the target transformer based on the updated digital twin model. The response determines that the target transformer meets the emergency maintenance conditions based on the degree of winding deformation. The emergency handling steps established based on the model deformation parameters and maintenance priority are added to the preset health assessment process. The assessment terminal guides the operation and maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps and the emergency handling steps.

[0016] Preferably, determining the degree of winding deformation based on the frequency band offset of the corresponding low correlation coefficient interval includes: Gradient calculation is performed on the correlation coefficients of each frequency point that constitute the low correlation coefficient interval, and the current gradient evaluation value is determined based on the obtained frequency response gradient value. Energy spectrum analysis is performed on the low correlation coefficient interval based on wavelet packet decomposition to obtain the energy concentration frequency bands located in the low correlation coefficient interval; Obtain the number of frequency bands corresponding to each energy concentration frequency band, and calculate the variance of the center frequency of each energy concentration frequency band and sum the energy proportion of each energy concentration frequency band to obtain the frequency band variance and the total energy proportion. The current energy evaluation value is determined based on the number of frequency bands, frequency band variance, and total energy percentage. The current width evaluation value obtained based on the interval width of the low correlation coefficient interval is weighted and summed with the current energy evaluation value and the current gradient evaluation value, and the winding deformation degree is determined based on the obtained current comprehensive evaluation value.

[0017] Preferably, the response determines that the target transformer does not meet the emergency maintenance conditions based on the degree of winding deformation, updates the mechanical structure module of the current digital twin model based on the model deformation parameters and maintenance priority, and controls the evaluation terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on the updated digital twin model, including: If the degree of winding deformation is lower than a preset emergency threshold, it is determined that the target transformer does not meet the emergency maintenance conditions. Obtain the mechanical structure module of the current digital twin model, and determine the standard geometric dimensions and standard stress distribution of the mechanical structure module; The standard geometric dimensions are corrected based on the model deformation parameters, and the maintenance priority is updated to update the standard maintenance plan in response to the maintenance priority being higher than the preset normal priority. The response is based on the evaluation terminal determining that there is a three-dimensional model of the target transformer on the evaluation display interface. The evaluation terminal is then controlled to highlight the deformation based on the updated digital twin model, and a deformation area identification box is generated to identify the winding area corresponding to the low correlation coefficient interval.

[0018] Preferably, the response determines that the target transformer meets the emergency maintenance conditions based on the degree of winding deformation, adds the emergency handling steps established based on the model deformation parameters and maintenance priorities to the preset health assessment process, and controls the assessment terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps and the emergency handling steps, including: If the degree of winding deformation is not lower than a preset emergency threshold, the target transformer is determined to meet the emergency maintenance conditions. Based on the center frequency point of the corresponding low correlation coefficient interval, generate a treatment plan library that includes the low correlation coefficient interval and is associated with a preset deformation type, and establish a preset treatment template; Fill the model deformation parameters and maintenance priorities into the parameter text slots and priority text slots located in the preset treatment template, and add the obtained emergency treatment steps to the preset health assessment process; The response is based on the assessment terminal determining that a three-dimensional model of the target transformer exists on the assessment display interface. The assessment terminal is then controlled to issue alarm prompts based on the standard assessment steps and emergency handling steps, and to generate an equipment alarm box that identifies the target transformer and an emergency handling box that identifies the deformed area.

[0019] Preferably, the method further includes: responding to a load current exceeding a preset overload threshold in real-time electrical characteristic data, and generating a thermal accumulation evaluation factor based on the duration and peak value of the corresponding load current; The thermal path model of the current digital twin model is dynamically corrected based on the heat accumulation evaluation factor, and a temperature rise warning area corresponding to the heat accumulation evaluation factor is generated on the evaluation display interface of the evaluation terminal.

[0020] Preferably, the step of dynamically correcting the thermal circuit model of the current digital twin model based on the thermal accumulation evaluation factor includes: obtaining the top oil temperature calculation node and winding hot spot calculation node in the thermal circuit model of the current digital twin model; The heat dissipation coefficient of the top oil temperature calculation node is attenuated and corrected based on the heat accumulation evaluation factor, and the thermal resistance parameter of the winding hot spot calculation node is increased and corrected. Based on the corrected top oil temperature and winding hot spot temperature, the remaining life assessment value of the target transformer is recalculated and pushed to the assessment terminal.

[0021] Preferably, the method further includes: periodically comparing the output evaluation results of the current digital twin model with the offline test data of the target transformer to obtain a comparison deviation value; When the comparison deviation value exceeds the preset allowable deviation, a global calibration process for the current digital twin model is triggered. The global calibration process includes inverting and optimizing the core physical parameters of the model using offline experimental data.

[0022] Secondly, the present invention provides a transformer health assessment system based on digital twins and electrical characteristics, comprising: The model retrieval module is configured to respond to the data acquisition terminal collecting the operating data of the target transformer, obtain the real-time electrical characteristic data and basic ledger information corresponding to the target transformer, and retrieve the initial digital twin model. The model update module is configured to update the model state of the initial digital twin model based on the model correction parameters generated from the corresponding real-time electrical feature data in response to the real-time electrical feature data meeting the model update conditions, so as to obtain the current digital twin model. The assessment guidance module is configured to update the standard assessment steps included in the preset health assessment process based on the assessment guidance data generated from the corresponding basic ledger information, and control the assessment terminal to guide the operation and maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps.

[0023] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention proposes a transformer health assessment method and system based on digital twins and electrical characteristics. This invention collects real-time electrical characteristic data and basic ledger information of the target transformer to retrieve an initial digital twin model. When the data meets the update conditions, it uses multi-dimensional features such as oil chromatography, partial discharge, or winding deformation to generate model correction parameters or assessment guidance data, dynamically updating the digital twin model and the preset health assessment process. Furthermore, it provides accurate health status assessment guidance to maintenance personnel through an assessment terminal. This invention achieves personalized dynamic adaptation of assessment strategies, significantly improving the intelligence level of transformer monitoring and assessment. Attached Figure Description

[0024] The accompanying drawings, which form part of this invention, are provided to further illustrate the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention, but do not constitute an undue limitation of the invention.

[0025] Figure 1 This is a flowchart illustrating a transformer health assessment method based on digital twins and electrical characteristics, provided as an embodiment of the present invention.

[0026] Figure 2 The timing flowchart is provided for a transformer health assessment method based on digital twins and electrical characteristics, according to an embodiment of the present invention.

[0027] Figure 3 This is a schematic diagram of an assessment terminal for a transformer health assessment method based on digital twins and electrical characteristics, provided in one embodiment of the present invention.

[0028] Figure 4 A flowchart illustrating the construction process of a gas evaluation value mapping table for a transformer health assessment method based on digital twins and electrical characteristics, provided in one embodiment of the present invention.

[0029] Figure 5 This is a schematic diagram illustrating the execution process of a transformer health assessment method based on digital twins and electrical characteristics, which performs pattern recognition of characteristic gas components based on a preset gas ratio method, according to an embodiment of the present invention. Detailed Implementation

[0030] The embodiments of the present invention will now be described in further detail with reference to the accompanying drawings. Through this description, the features and advantages of the present invention will become clearer and more apparent.

[0031] Reference Figure 1 , Figure 2 As shown, an embodiment of the invention illustrates a transformer health assessment method based on digital twin and electrical characteristics, including: S1, responding to the data acquisition terminal to collect the operating data of the target transformer, obtaining real-time electrical characteristic data and basic ledger information corresponding to the target transformer, and retrieving the initial digital twin model.

[0032] In some embodiments, the specific steps in step S1 of updating the model state of the initial digital twin model based on the model correction parameters generated from the corresponding real-time electrical feature data to obtain the current digital twin model may include:

[0033] Step S11: Control the data acquisition terminal to perform multi-source data fusion on the target transformer to obtain a fused dataset.

[0034] Specifically, the dissolved gas concentration data in the oil, the partial discharge pulse sequence, the frequency response characteristic curve, and the real-time load electrical parameters can be obtained by synchronously calling the oil chromatography online monitoring device, the partial discharge high-frequency current sensor, the winding deformation sweep frequency response unit, and the operating current and voltage transformer configured on the target transformer.

[0035] Furthermore, the multi-source data obtained above are aligned according to a unified timestamp and normalized to obtain a structured fused dataset.

[0036] It should be noted that in the above fusion process, different weights can be assigned to data from different sources based on the basic ledger information of the target transformer. The basic ledger information of the target transformer mentioned here includes at least the equipment model, rated capacity, and years of operation.

[0037] Step S12: Identify characteristic gas components in the fused dataset that indicate abnormal dissolved gas in the oil, determine the oil chromatography assessment level based on the gas concentration status of the corresponding characteristic gas components, and determine the model correction coefficient and warning threshold for the corresponding oil chromatography assessment level.

[0038] Step S13: In response to the determination that the target transformer does not meet the high-risk criteria based on the oil chromatography assessment level, the insulation aging module parameters of the initial digital twin model are updated based on the model correction coefficient and the early warning threshold. The assessment terminal then guides maintenance personnel to perform a health status assessment of the target transformer based on the updated digital twin model. A schematic diagram of the assessment terminal is shown below. Figure 3 As shown.

[0039] Step S14: In response to the determination that the target transformer meets the high-risk judgment conditions based on the oil chromatography evaluation level, add the special diagnostic steps established based on the model correction coefficient and the early warning threshold to the preset health assessment process, and control the assessment terminal to guide the operation and maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps and the special diagnostic steps.

[0040] In some embodiments, the specific steps of determining the oil chromatography evaluation level based on the gas concentration state of the corresponding characteristic gas components in step S12 may include: step S121, performing weighted summation on the gas concentration values ​​of each component of the characteristic gas components, and determining the current gas evaluation value based on the obtained comprehensive gas index.

[0041] Specifically, after analyzing the dissolved gases in transformer oil, seven key gases can be detected: hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, and carbon dioxide. Then, the corresponding weights are assigned based on the differences in the sensitivity of these gases under different fault types.

[0042] For example, in a health assessment scenario of a 220kV oil-immersed main transformer, acetylene is highly sensitive to arc discharge, while methane is sensitive to local overheating. For transformers, arc discharge is more harmful than local overheating. Therefore, the weighting coefficient of acetylene needs to be higher than that of methane. Relevant technicians can allocate different weights according to this judgment standard, and this invention does not limit it.

[0043] Furthermore, the actual concentration values ​​of each gas are multiplied by their corresponding weights and then summed to generate a comprehensive gas index. This comprehensive gas index reflects the overall level of gas production activity within the transformer. Based on a preset gas evaluation value mapping table, the comprehensive gas index can be converted into a current gas evaluation value. This current gas evaluation value is used to quantify the initial risk level of insulation system anomalies, and its value ranges from 0 to 1.

[0044] In this embodiment of the invention, the process of constructing the gas evaluation value mapping table is as follows: Figure 4 As shown, the specific steps may include: Step 1.1: Collect a sample set of historical oil chromatography data of the target transformer under different typical fault types. The sample set covers standard fault categories such as arc discharge, partial discharge, high temperature overheating, medium temperature overheating, and low temperature overheating, and ensures that each type of sample has been verified by offline testing and fault disassembly. Step 1.2: Normalize the concentration of each characteristic gas in each type of fault sample to eliminate dimensional differences, and calculate the frequency of occurrence and concentration contribution of each gas in the same type of fault. Step 1.3: Based on the association rules between characteristic gases and fault types in the power industry standards, assign an initial sensitivity level to each gas. Acetylene is given the highest sensitivity because it is only significantly generated under high-energy discharge, while methane has a lower sensitivity because it is widely present in a variety of thermal faults. Step 1.4 introduces an expert scoring mechanism, in which multiple senior condition-based maintenance engineers independently score the importance of various gases in different faults. The Delphi method is used for iterative convergence to form a consensus sequence of gas weights. The convergence condition can be that the standard deviation of each gas weight is less than a preset standard deviation threshold, or that more than a certain proportion of experts' scores for a certain gas weight fall in the same range, or the iteration rounds can be directly set. Relevant technical personnel can set it according to actual needs, and this invention does not limit it. Step 1.5: The gas weight consensus sequence is weighted and fused with the concentration contribution in Step 1.2 to generate a draft of the comprehensive evaluation value of each gas under each fault type. Step 1.6: Use the historical fault case library to retrospectively verify the draft. If the evaluation value of a certain gas causes the misjudgment rate to exceed the preset upper limit, return to step 1.4 to adjust the weight; otherwise, proceed to step 1.7. Here, the misjudgment rate is the ratio of the number of samples in the historical fault case library whose evaluation results are inconsistent with the actual fault type to the total number of verification samples. The preset upper limit can be set according to the actual needs of relevant technical personnel, combined with industry experience or risk tolerance. This invention does not limit it. Step 1.7: Organize the verified gas-fault combination evaluation values ​​into a structured mapping relationship according to gas type to form a gas evaluation value mapping table. This gas evaluation value mapping table is indexed by gas name, and the output is the quantitative evaluation value corresponding to the current gas combination.

[0045] The historical oil chromatography data sample set mentioned above refers to a transformer oil sample analysis database labeled with fault types and containing complete records of the concentrations of several component gases. Sensitivity level refers to the strength of a gas's specific indicative power for a particular fault type. The gas weight consensus sequence refers to the gas importance ranking result reached through multiple rounds of anonymous expert review.

[0046] Step S122: Based on the preset gas ratio method, perform pattern recognition on the characteristic gas components to obtain the fault feature vector located in the characteristic gas components.

[0047] It should be noted that the proportion of dissolved gases generated in the oil varies significantly when different types of faults occur inside the transformer. Therefore, it is necessary to classify and identify the characteristic gas components using standardized gas ratio rules.

[0048] In this embodiment of the invention, pattern recognition of characteristic gas components is performed based on a preset gas ratio method to obtain fault feature vectors located in the characteristic gas components. The execution process is as follows: Figure 5 As shown, the specific steps may include: Step 2.1: Extract the concentration values ​​of seven characteristic gases (hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, and carbon dioxide) from the real-time oil chromatography analysis results, and verify whether the concentration of each gas is higher than the detection limit. If all are higher than the detection limit, proceed to step 2.2; otherwise, proceed to step 2.6. The detection limit can be obtained from the equipment manual or industry standards. Step 2.2: Calculate three sets of core ratios according to the preset gas ratio method, including the ratio of acetylene to ethylene, the ratio of ethylene to ethane, and the ratio of methane to hydrogen. Determine whether each ratio is within the valid calculation range. If all are within the valid range, proceed to step 2.3; otherwise, proceed to step 2.6. The valid calculation range here refers to the range of gas ratios within which the physical meaning is clear and the fault indication is valid. Ratios outside this range may fail due to background interference, measurement errors, or non-fault gas generation. Relevant technical personnel can set these ratios based on actual needs and power industry experience and standards; this invention does not impose any limitations. Step 2.3: Match the three sets of ratios obtained by calculation with the preset fault coding rule table item by item to determine the corresponding fault type code. If there is a unique match, proceed to step 2.4. If there are multiple matches or no match, proceed to step 2.5. The preset fault coding rule table here is a structured rule base that maps the combination of three ratios to fault type codes. It is derived from the classic three-ratio coding method and may be optimized by combining local experience. Step 2.4: Based on the successfully matched fault type code, retrieve the corresponding standardized fault feature vector from the fault feature template library. This vector contains the typical relative weight distribution of each feature gas under the fault, complete the pattern recognition, and proceed to step 2.7. Step 2.5: Activate the auxiliary criterion fusion mechanism, introduce the carbon monoxide to carbon dioxide ratio and total hydrocarbon growth rate as supplementary features, reconstruct the extended ratio combination, and match the fault coding rule table again. If the match is successful, proceed to step 2.4; otherwise, proceed to step 2.6. Step 2.6: Mark the current gas component as "mode not recognized", generate a default fault feature vector, where the weight of each gas is set according to the background distribution under the fault-free state of similar equipment in history, and proceed to step 2.7; Step 2.7: Output the final fault feature vector. The fault feature vector obtained here will be used in subsequent fusion evaluation with multi-source features such as discharge fingerprint and thermal accumulation factor.

[0049] The fault feature template library mentioned above refers to a vector set that stores the characteristic gas weight distribution under various standard faults.

[0050] Step S123: Obtain the vector dimension of the corresponding fault feature vector, and perform normalization calculation on the feature value of each dimension and entropy calculation on the feature value of all dimensions to obtain the feature normalization value and feature entropy value.

[0051] Specifically, the fault feature vector generated in step S122 can be read first. The fault feature vector designed in this invention is usually a three-dimensional or five-dimensional structure. Each dimension of the fault feature vector corresponds to the confidence value of a typical fault mode. Then, the original feature value of each dimension is normalized by the min-max normalization method to map the original feature value to the interval between 0 and 1 to obtain the feature normalization value. At the same time, the normalized feature values ​​of all dimensions are used as the probability distribution input into the Shannon entropy formula to calculate the corresponding information entropy. The information entropy obtained in this invention is used to measure the degree of uncertainty in the current fault mode judgment. The lower the entropy value, the clearer the fault type. The higher the entropy value, the more ambiguous the gas characteristics and the possibility of multiple fault superposition. This information entropy is the feature entropy value mentioned above.

[0052] Step S124: Determine the current mode evaluation value based on the vector dimension, feature normalization value, and feature entropy value.

[0053] Specifically, you can first obtain the number of dimensions of the fault feature vector, for example, a three-dimensional vector corresponds to three preset fault types.

[0054] Furthermore, a weighted concentration index is calculated by combining the normalized values ​​of features from each dimension; for example, the maximum normalized value is taken as the dominant fault intensity. Simultaneously, feature entropy is introduced as a dispersion penalty factor, where a higher entropy value results in a greater reduction in concentration.

[0055] Furthermore, through a pre-defined nonlinear mapping function, the vector dimension, dominance strength, and entropy value are jointly mapped to a current mode evaluation value between 0 and 1. This pre-defined nonlinear mapping function can be any function that maps the vector dimension, dominance strength, and entropy value to a current mode evaluation value between 0 and 1. For example, based on an exponentially decaying concentration-entropy fusion model, the expression is: , This is the Shannon entropy; the Shannon entropy varies depending on the vector dimension. This is the position... Represents the entropy sensitivity coefficient. , The normalized fault feature vector can be used as a reference. Relevant technicians can preset a nonlinear mapping function according to actual needs. This invention does not limit this function.

[0056] Step S125: The current rate evaluation value obtained based on the gas production rate of characteristic gas components is weighted and calculated with the current model evaluation value and the current gas evaluation value, and the oil chromatography evaluation level is determined based on the obtained current comprehensive evaluation value.

[0057] It should be noted that in the health status assessment of transformers, relying solely on a single type of static gas index is unlikely to fully reflect the true development trend of the fault. For example, high concentrations of gas may originate from historical accumulation rather than being caused by current active faults. Ignoring the dynamic dimension of gas production rate will lead to serious misjudgments, which will affect the operation of the digital twin model.

[0058] Specifically, the current gas evaluation value, the current model evaluation value, and the current rate evaluation value are obtained separately. The current rate evaluation value is obtained by calculating the rate of change of the concentration of each characteristic gas in the two most recent sampling periods, and then normalizing it after weighting it with gas type sensitivity.

[0059] Furthermore, the three evaluation values ​​are weighted and summed according to the preset weight coefficients. The weight allocation mentioned here follows three allocation principles: dynamic priority, mode-driven, and concentration-based. For example, the rate weight is set to 0.4, the mode weight to 0.35, and the concentration weight to 0.25.

[0060] Furthermore, the weighted results are mapped to a predefined four-level oil chromatography assessment grade range, which can be set as follows: 0 to 0.3 is the normal grade, 0.3 to 0.6 is the attention grade, 0.6 to 0.85 is the abnormal grade, and above 0.85 is the high-risk grade.

[0061] In some embodiments, the concentration changes of key gases such as acetylene and ethylene are first extracted from historical and current oil chromatography data. Then, the average gas production rate per unit time is calculated and weighted to synthesize the current rate evaluation value. For example, in the aforementioned 220kV main transformer scenario, if the current gas evaluation value is 0.72, the current mode evaluation value is 0.61, and the current rate evaluation value is 0.78, the weighted calculation yields 0.72×0.25+0.61×0.35+0.78×0.4=0.7055. Based on threshold division, this value is determined to fall within the 0.6 to 0.85 range, and the oil chromatography assessment level can then be determined as "abnormal." Finally, this "abnormal" level directly triggers parameter correction or the insertion of a special diagnostic process into the digital twin model's insulation aging module. The oil chromatography assessment level refers to one of four health status levels based on the current comprehensive evaluation value.

[0062] In some embodiments, the method further includes: step S1251, inputting the current comprehensive evaluation value into a pre-trained fault evolution prediction model to obtain the fault evolution duration of the corresponding characteristic gas component.

[0063] Specifically, a time-series regression model trained on a historical transformer fault case library can be invoked. This time-series regression model uses the current comprehensive evaluation value as the core input, while integrating the basic ledger information of the target transformer as auxiliary features. The auxiliary features mentioned here include the integration of the commissioning year, rated capacity, cooling method, and recent load fluctuation characteristics.

[0064] Furthermore, by using deep neural networks or gradient boosting tree structures to learn the nonlinear mapping relationship between the degree of gas anomaly and the time required for the fault to develop from its current state to critical failure, the fault evolution time is output in hours or days. This fault evolution time indicates how long it is expected that the internal defects indicated by the characteristic gas components will evolve into a serious fault requiring immediate shutdown, assuming the current operating conditions remain unchanged.

[0065] Step S1252: If the response fault evolution duration is greater than the preset safety margin duration, update the current comprehensive evaluation value based on the retrieved safety margin weight.

[0066] It should be noted that the safety margin duration is the minimum acceptable early warning window set based on the power grid dispatch cycle, maintenance resource preparation cycle, and equipment redundancy configuration. In this invention, it is set to 7 days or 14 days. Relevant technicians can make adaptive designs according to actual needs, and this invention does not impose any limitations. When the fault evolution duration exceeds the preset safety margin duration, it indicates that the current fault development speed is relatively slow, and there is still sufficient time to arrange planned maintenance. At this time, the corresponding safety margin weight is automatically retrieved from the strategy configuration library. This safety margin weight is a decay coefficient less than 1, used to moderately reduce the current comprehensive evaluation value. Setting this safety margin weight is to avoid excessive triggering of high-risk responses due to high static gas indicators.

[0067] Furthermore, the current comprehensive evaluation value is multiplied by the corresponding safety margin weight to obtain the updated comprehensive evaluation value. This updated comprehensive evaluation value can serve as the final basis for subsequent oil chromatography assessment grade determination.

[0068] In some embodiments, the specific steps in step S13, which involve responding to the determination that the target transformer does not meet the high-risk criteria based on the oil chromatography assessment level, updating the insulation aging module parameters of the initial digital twin model based on the model correction coefficient and the early warning threshold, and controlling the assessment terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on the updated digital twin model, may include:

[0069] S131, the response oil chromatography assessment level is lower than the preset high-risk level, indicating that the target transformer does not meet the high-risk determination criteria.

[0070] Specifically, first read the oil chromatography evaluation level output by step S125. The output oil chromatography evaluation level mentioned above is a certain level in the four-level system.

[0071] Furthermore, it is compared with the system's preset high-risk level threshold, which usually corresponds to the "high-risk" level (i.e., the fourth level), with a value range of 0.85 to 1.0.

[0072] If the current oil chromatography assessment level is "normal", "caution" or "abnormal" (i.e., the comprehensive evaluation value is less than 0.85), the target transformer is determined not to meet the high-risk judgment conditions. There is no need to insert a special diagnostic step. Instead, it enters the parameter correction process of the insulation aging module.

[0073] Step S132: Obtain the insulation aging module parameters of the initial digital twin model, and determine the standard aging rate and standard early warning threshold of the insulation aging module parameters.

[0074] Specifically, the complete parameter set of the insulation aging module is first retrieved from the initial digital twin model corresponding to the target transformer. The aforementioned insulation aging module is usually constructed based on the Arrhenius thermal aging equation or the Monsinger model.

[0075] Furthermore, two key benchmark parameters are extracted from the complete parameter set obtained above: the standard aging rate and the standard warning threshold. The standard aging rate is the percentage decrease in the degree of polymerization of the insulating paper per year under standard operating conditions such as rated load, top oil temperature of 98℃, and hot spot temperature rise of 25K. The standard warning threshold is the set point at which the system should issue a warning when the degree of polymerization drops to a certain critical value (e.g., 250 here) or the furfural concentration exceeds a certain limit (e.g., 2 mg / L here).

[0076] Step S133: Correct the standard aging rate based on the model correction coefficient, and replace the standard warning threshold with the warning threshold if the warning threshold is less than the standard warning threshold.

[0077] Specifically, the model correction coefficient corresponding to the current oil chromatography evaluation level is first obtained from step S12. The correction coefficient mentioned here is derived from the degree of gas anomaly and the value range of the correction coefficient is usually 0.8 to 1.5.

[0078] Furthermore, by multiplying the standard aging rate in the initial digital twin model by the corresponding correction coefficient, the corrected actual aging rate can be obtained. Simultaneously, it is determined whether the warning threshold determined by oil chromatography is less than the standard warning threshold. If the condition is met, the more stringent warning threshold is used to override the original standard warning threshold. The standard warning threshold refers to the preset critical value of the insulation aging index in the digital twin model that triggers a health alarm.

[0079] In step S134, the response is based on the assessment terminal determining that there is a three-dimensional model of the target transformer on the assessment display interface, and controls the assessment terminal to perform visualization rendering based on the updated digital twin model, and generate a health status identification box to identify the target transformer.

[0080] Specifically, the first step is to check whether the current user interface of the evaluation terminal has loaded the three-dimensional digital model of the target transformer. In this invention, the three-dimensional digital model is embedded in the operation and maintenance platform in a lightweight BIM or CAD format, and the evaluation terminal refers to an interactive human-machine interface deployed in the substation monitoring center or mobile operation and maintenance equipment.

[0081] Furthermore, once the existence of the 3D model is confirmed, the insulation aging status data in the updated digital twin model can be called, and the graphics engine can be driven to perform dynamic color rendering operations on the transformer body. For example, the winding area can be displayed in a gradient from green to red according to the degree of aging. The insulation aging status data mentioned here includes the corrected aggregation degree distribution, the temperature field of hot spot areas, and the predicted value of remaining life.

[0082] Furthermore, a semi-transparent rectangular frame can be generated around the 3D model as a health status indicator frame, with text information displayed in real time inside the frame. The displayed text information can be key indicators such as "Insulation aging level: Caution", "Remaining life: 8.2 years", and "Current comprehensive evaluation value: 0.60".

[0083] In some embodiments, step S14, in response to determining that the target transformer meets the high-risk criteria based on the oil chromatography assessment level, adds a specific diagnostic step based on the model correction coefficient and the early warning threshold to the preset health assessment process, and controls the assessment terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps and the specific diagnostic steps. The specific steps may include:

[0084] Step S141: If the response oil chromatography assessment level is not lower than the preset high-risk level, it is determined that the target transformer meets the high-risk judgment conditions.

[0085] Specifically, firstly, the current oil chromatography evaluation level output in step S125 is obtained, and then the obtained current oil chromatography evaluation level is compared with the preset high-risk level threshold. Here, the preset high-risk level threshold refers to the lower limit of the comprehensive evaluation value used to determine whether the current transformer has entered the "high-risk" state, which is 0.85 corresponding to the "high-risk" level in the above oil chromatography evaluation level.

[0086] Furthermore, when the oil chromatography assessment level is "high risk", it indicates that the concentration of characteristic gas components, gas production rate and typicality of fault modes have reached the critical state that may cause insulation breakdown or winding burnout. At this time, the target transformer is determined to meet the high risk assessment conditions, which can trigger subsequent special diagnosis and emergency response procedures.

[0087] Step S142: Using the main characteristic gas of the corresponding characteristic gas component as the center, generate a diagnostic knowledge graph that includes the main characteristic gas and is associated with preset fault types, and establish a preset diagnostic template.

[0088] Specifically, the gas with the highest concentration or the most discriminative gas for fault type is first identified from the characteristic gas components as the main characteristic gas. For example, acetylene dominates in arc discharge, and ethylene increases significantly in high-temperature overheating.

[0089] Furthermore, using the obtained main characteristic gas as the root node, a structured diagnostic knowledge graph is constructed by retrieving strongly associated fault types, typical gas production combinations, common occurrence locations, and historical cases from a pre-set power equipment fault knowledge base. This knowledge graph contains multiple entities and their semantic relationships, including main characteristic gas, auxiliary gas, fault mechanism, typical equipment structural location, and recommended detection methods.

[0090] Furthermore, a preset diagnostic template matching the corresponding fault type is invoked. This preset diagnostic template is a standardized text frame containing multiple fillable text slots.

[0091] For example, in the aforementioned 220kV main transformer scenario, when the acetylene concentration reaches 15 microliters per liter and the C2H2 / C2H4 ratio exceeds 0.5, acetylene can be identified as the main characteristic gas. A knowledge graph can be generated with "acetylene" as the center. The knowledge graph is associated with fault types such as "high-energy discharge", "inter-turn short circuit in windings", and "poor core grounding". At the same time, the knowledge graph can also link suggested measures such as "partial discharge location" and "winding deformation detection".

[0092] Continuing with the scenario from the above example, the system automatically loads a preset diagnostic template named "High Energy Discharge Special Diagnosis". This preset diagnostic template for high energy discharge special diagnosis will include text slots such as "Main Characteristic Gas: {Main Gas}", "Model Correction Coefficient: {Coefficient}", and "Suggested Priority Detection Area: {Area}".

[0093] Step S143: Fill the model correction coefficient and warning threshold into the coefficient text slot and threshold text slot located in the preset diagnostic template, and add the obtained special diagnostic steps to the preset health assessment process.

[0094] Step S144: Based on the determination of the three-dimensional model of the target transformer on the evaluation display interface by the evaluation terminal, the control evaluation terminal provides voice prompts based on the standard evaluation steps and the special diagnosis steps, and generates equipment identification boxes to identify the target transformer and risk area identification boxes to identify the fault feature vectors.

[0095] In some embodiments, the specific steps of adding the obtained specialized diagnostic steps to the preset health assessment process in step S143 may include: step S1431, obtaining the standard assessment steps included in the preset health assessment process, and determining the standard diagnostic depth located in the standard assessment steps.

[0096] Specifically, the system first retrieves a preset health assessment process applicable to the target transformer type from the system configuration library. This preset health assessment process is stored in the system in the form of an ordered task list. The preset health assessment process mentioned above includes standard assessment steps such as "infrared thermometry", "oil chromatography retest", "partial discharge location", and "winding deformation detection".

[0097] Furthermore, for each standard evaluation step, the metadata fields associated with these steps are read. The “standard diagnostic depth” indicates the level of fault identification precision that the corresponding step can cover under normal conditions. For example, the standard diagnostic depth for “partial discharge location” is “winding level”, which means that it can only locate the high-voltage or low-voltage winding, but cannot accurately locate the specific coil.

[0098] Step S1432: Add the special diagnostic step to the preset health assessment process, and if the severity of the fault corresponding to the warning threshold is greater than the baseline severity corresponding to the standard diagnostic depth, adjust the special diagnostic step to be placed before the standard assessment step, and vice versa.

[0099] Specifically, the specific diagnostic steps generated by step S143 are first inserted into the end of the preset health assessment process as the initial position.

[0100] Furthermore, the warning thresholds associated with the corresponding specific diagnostic steps are extracted, and these warning thresholds are mapped to quantified fault severity values. For example, the severity of a warning threshold corresponding to an aggregation degree of 280 is 0.88. Simultaneously, the standard diagnostic depth of core items such as "partial discharge detection" in the standard evaluation steps is obtained and converted into comparable baseline severity. For example, the baseline severity corresponding to the "winding level" positioning capability is 0.75.

[0101] Furthermore, if the severity of the specific fault diagnosis is greater than the baseline severity, the specific diagnostic step is moved before all standard evaluation steps; otherwise, the specific diagnostic step is retained after all standard evaluation steps. The aforementioned specific diagnostic step refers to a task unit generated for a specific high-risk fault type, and this task unit also includes the functions of diagnostic objectives and operational guidelines.

[0102] Step S2: In response to the real-time electrical feature data meeting the model update conditions, the model state of the initial digital twin model is updated based on the model correction parameters generated from the corresponding real-time electrical feature data to obtain the current digital twin model.

[0103] It should be noted that most existing transformer digital twin systems employ static or periodic offline update mechanisms. The parameters of the twin models established by existing methods are often set at the initial stage of equipment commissioning and then remain unchanged, or some may only undergo coarse-grained corrections based on annual preventative test data. Twin models established by existing methods cannot respond promptly to dynamic anomalies such as insulation degradation, winding deformation, or partial discharge that occur during operation.

[0104] Understandably, this solution automatically initiates the model correction process by setting explicit model update trigger conditions, when any of the real-time electrical characteristic data meets any of these conditions. For example, update trigger conditions can be set as an oil chromatography comprehensive evaluation value exceeding 0.6, a sudden increase of 50% in partial discharge, or a continuous over-limit load current. When any of these conditions is met, the model enters the correction process.

[0105] Furthermore, the specific operations of the aforementioned correction process may include the following steps: performing fusion analysis on multi-dimensional data such as oil chromatography, partial discharge, frequency response, and thermal load; and generating correction parameters for the insulation aging module, mechanical structure module, or thermal circuit model, respectively. These correction parameters may include aging rate correction coefficients, winding deformation displacement, and heat dissipation coefficient attenuation factors.

[0106] Furthermore, these parameters are injected into the corresponding physical sub-modules of the initial digital twin model to dynamically refresh the model's state, forming a current digital twin model that is highly consistent with the actual health status of the current device.

[0107] Step S3: Based on the assessment guidance data generated from the corresponding basic ledger information, update the standard assessment steps included in the preset health assessment process, and control the assessment terminal to guide the operation and maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps.

[0108] In some embodiments, step S3, which updates the standard assessment steps included in the preset health assessment process based on the assessment guidance data generated according to the corresponding basic ledger information, and controls the assessment terminal to guide maintenance personnel to perform specific steps for assessing the health status of the target transformer based on the standard assessment steps, may include:

[0109] Step S31: Control the sensor network located on the target transformer to collect the partial discharge pulse signal of the target transformer and obtain the partial discharge signal sequence.

[0110] Specifically, the high-frequency current sensor array installed on the end screen of the bushing of the target transformer, the grounding wire of the iron core, and the neutral point lead-out terminal can be activated first. These high-frequency current sensor arrays have a bandwidth of more than 20MHz and a time resolution of nanosecond level.

[0111] Furthermore, a multi-channel high-speed acquisition card is synchronously triggered to perform parallel recording operations on the output signals of each high-frequency current sensor array at a sampling rate of no less than 100MSa / s, and to perform timestamp alignment operations based on the power frequency voltage phase, ultimately forming a structured partial discharge signal sequence containing discharge amplitude, arrival time, phase angle, and channel identifier.

[0112] Step S32: Perform time-frequency domain analysis on the partial discharge signal sequence to obtain the phase distribution map of the corresponding discharge phase and the amplitude distribution map of the corresponding discharge amplitude.

[0113] Specifically, the obtained partial discharge signal sequence is first segmented and aligned according to the power frequency cycle, and the angle information of each discharge pulse relative to the voltage phase is extracted. Simultaneously, the number of discharges within each phase interval (taking 10° intervals as an example) is counted, and a phase distribution diagram is plotted with phase as the horizontal axis and discharge frequency as the vertical axis. At the same time, histogram statistics are performed on the amplitude of all discharge pulses, dividing the data into several amplitude intervals, and an amplitude distribution diagram is generated with amplitude as the horizontal axis and occurrence frequency as the vertical axis.

[0114] Furthermore, by jointly analyzing the two types of distribution maps mentioned above, the discharge type can be preliminarily determined. For example, the clustered distribution with symmetry in the positive and negative half-cycles mostly corresponds to internal air gap discharge, while the single-sided discharge that only appears near the peak may originate from surface flashover or floating potential.

[0115] Step S33: Cluster the discharge pulse points in adjacent phase intervals to obtain each discharge cluster, and determine the discharge clusters that are allowed to match the target transformer as valid discharge clusters based on the basic ledger information.

[0116] Step S34: Obtain the standard evaluation steps included in the preset health assessment process, and fill the winding position of the corresponding effective discharge cluster into the position text slot located in the standard evaluation step.

[0117] Specifically, the process first analyzes the various standard assessment steps included in the preset health assessment process and identifies the text slots used to record partial discharge location information.

[0118] Furthermore, based on the winding position information corresponding to the previously determined effective discharge clusters, the specific location description of the winding position information is filled into the corresponding text slot in the above steps.

[0119] For example, if the effective discharge cluster found during the partial discharge detection of a 220kV main transformer is located on the side of the B-phase low-voltage winding near the tank wall, then when performing the preset health assessment procedure, the description of the side of the B-phase low-voltage winding near the tank wall needs to be accurately filled into the location text slot in the standard assessment step.

[0120] Step S35: Based on the determination of the evaluation terminal that there is a three-dimensional model of the target transformer on the evaluation display interface, the evaluation terminal is controlled to provide visual guidance based on the standard evaluation steps, and a partial discharge identification box is generated to mark the winding area corresponding to the effective discharge cluster.

[0121] Specifically, the first step is to check whether the target transformer's three-dimensional digital model has been loaded into the current display interface of the evaluation terminal. The three-dimensional digital model mentioned here is a three-dimensional display diagram that includes the geometric structure and hierarchical relationship of key components such as high-voltage winding, low-voltage winding, iron core, and oil tank.

[0122] Furthermore, once the existence of the 3D model is confirmed, it is necessary to read the winding position information that has been filled in the standard evaluation steps and map it to the corresponding physical region in the 3D model.

[0123] Furthermore, a high-brightness coloring or semi-transparent overlay layer is superimposed on the surface of the corresponding area to achieve visual guidance. At the same time, a partial discharge identification box with an arrow pointing to the outside of the model is generated. The text displayed in the box can include key diagnostic information such as "effective discharge cluster: B phase low voltage winding near the oil tank side", "discharge frequency: 89 times / cycle", and "main frequency band: 100–500kHz".

[0124] It should be noted that the linkage between the above-mentioned identification frame design and the three-dimensional highlight area design allows maintenance personnel to intuitively locate the defect.

[0125] In some embodiments, the specific steps in step S33 of determining the discharge clusters that allow matching the target transformer as valid discharge clusters based on the basic ledger information may include:

[0126] Step S331: Obtain the equivalent frequency center for each discharge cluster, and determine all equivalent frequency centers that are allowed to match the target transformer as a primary screening group based on the rated voltage in the basic ledger information.

[0127] Step S332: Form a frequency band range corresponding to the preset frequency band width that surrounds each equivalent frequency center located in the primary screening group, and retrieve the preset discharge fingerprint database based on the equipment type of the corresponding target transformer. The preset frequency band width can be designed according to the actual needs of relevant technical personnel, combined with industry experience or on-site testing. It is generally 400–600kHz, but this invention does not limit it.

[0128] Step S333: Extract features from the discharge pulse points within the frequency band to obtain each discharge feature subset;

[0129] Step S334: Perform similarity matching between each discharge feature subset and the preset discharge fingerprint database, and determine the discharge feature subset with a matching similarity greater than the preset fingerprint similarity as the typical discharge electron set based on the matching results;

[0130] Step S335: Obtain the discharge energy of all typical discharge clusters located in the same frequency band, and determine the equivalent frequency center with the largest corresponding discharge energy as the effective discharge cluster.

[0131] For example, in the aforementioned 220kV main transformer scenario, the system identifies three discharge clusters from the partial discharge signal sequence. The equivalent frequency centers of the three discharge clusters are 85kHz, 420kHz, and 1.8MHz, respectively. Based on the typical frequency band upper limit of 1.5MHz for 220kV equipment, the 1.8MHz center is eliminated, and 85kHz and 420kHz are retained to form a primary screening group. A 65–105kHz frequency band is constructed with 85kHz as the center, and a 220–620kHz frequency band is constructed with 420kHz as the center. The system then retrieves the "220kV" signal. After establishing a dedicated discharge fingerprint database for "oil-immersed transformers," it was found that the discharge feature subset within the 420kHz frequency band had a similarity of 0.89 with the "internal air gap discharge" template, while the 85kHz subset only matched the "external interference" template with a similarity of 0.76. The matching similarity did not reach the "preset fingerprint similarity threshold." Therefore, only the 420kHz frequency band generated a typical discharge cluster, with a total discharge energy of 12.4mJ, far exceeding other frequency bands. The system ultimately identified the discharge cluster corresponding to 420kHz as a valid discharge cluster and associated it with the central region of the A-phase high-voltage winding. The preset fingerprint similarity threshold is set based on the equipment type and historical verification data. For example, for a 220kV oil-immersed transformer, a typical value is 0.85. Relevant personnel can set the fingerprint similarity threshold according to actual needs; this invention does not limit this setting. Only when the matching similarity is higher than this fingerprint similarity threshold is the corresponding discharge feature subset considered a typical discharge cluster with diagnostic value. The equivalent frequency center mentioned above refers to the weighted average frequency point of the energy distribution of a discharge cluster in the frequency domain. The equivalent frequency center is used to characterize the dominant frequency band of the cluster. The preset discharge fingerprint database refers to a template database that is pre-established for a specific equipment type and contains multi-dimensional features of various typical partial discharge modes. The typical discharge cluster refers to a subset of discharge features whose similarity to a certain type of template in the preset discharge fingerprint database exceeds a threshold. The effective discharge cluster refers to the discharge cluster with the largest discharge energy among all typical discharge clusters and is most likely to represent the current major insulation defect.

[0132] In some embodiments, step S3, which updates the standard assessment steps included in the preset health assessment process based on the assessment guidance data generated according to the corresponding basic ledger information, and controls the assessment terminal to guide maintenance personnel to perform specific steps for assessing the health status of the target transformer based on the standard assessment steps, may also include:

[0133] Step S36: The response data acquisition terminal applies a frequency sweep signal to the target transformer, obtains the frequency response characteristic curve corresponding to the target transformer, and retrieves the historical frequency response characteristic curve.

[0134] Specifically, the data acquisition end first injects a sinusoidal sweep frequency signal with logarithmic intervals in the range of 1kHz to 2MHz into the target transformer winding through a dedicated sweep frequency impedance analyzer. At the same time, the voltage and current responses at the input port are measured, the amplitude and phase of the transfer function at each frequency point are calculated, and finally the current frequency response characteristic curve is formed.

[0135] Furthermore, retrieve the historical frequency response characteristic curve of the corresponding transformer under the most recent abnormal condition from the equipment health record database. The test conditions for obtaining these two curves need to be consistent (the test conditions mentioned here can be the same winding pair, the same wiring method, and similar oil temperature).

[0136] Step S37: Perform correlation analysis based on the frequency response characteristic curve and the historical frequency response characteristic curve to obtain the correlation coefficient sequence.

[0137] Specifically, the frequency response characteristic curve and the historical frequency response characteristic curve can be further subdivided into three typical frequency bands according to international standards, each sub-band having a width of 50kHz.

[0138] Furthermore, within each sub-band, the Pearson correlation coefficient is calculated for the amplitude sequences of the two curves, generating a correlation coefficient sequence aligned with the frequency axis. The closer the value is to 1, the more stable the winding state. The correlation coefficient sequence refers to the numerical sequence formed by calculating the amplitude similarity between the current curve and the historical curve segment by segment within the preset sub-band, used to quantify the frequency domain distribution characteristics of winding structure changes.

[0139] Step S38: Determine that there is a low correlation coefficient interval in the correlation coefficient sequence that indicates winding deformation, determine the degree of winding deformation based on the frequency band offset of the corresponding low correlation coefficient interval, and determine the model deformation parameters and maintenance priority corresponding to the degree of winding deformation.

[0140] Step S39: In response to the determination that the target transformer does not meet the emergency maintenance conditions based on the degree of winding deformation, the mechanical structure module of the current digital twin model is updated based on the model deformation parameters and maintenance priority, and the evaluation terminal is controlled to guide the operation and maintenance personnel to perform a health status assessment of the target transformer based on the updated digital twin model.

[0141] Step S310: In response to determining that the target transformer meets the emergency maintenance conditions based on the degree of winding deformation, the emergency handling steps established based on model deformation parameters and maintenance priorities are added to the preset health assessment process. The assessment terminal is then controlled to guide maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps and the emergency handling steps.

[0142] In some embodiments, the specific steps in step S38 of determining the degree of winding deformation based on the frequency band offset of the corresponding low correlation coefficient interval may include:

[0143] Step S381: Calculate the gradient of the correlation coefficients of each frequency point that makes up the low correlation coefficient interval, and determine the current gradient evaluation value based on the obtained frequency response gradient value.

[0144] Step S382: Based on wavelet packet decomposition, perform energy spectrum analysis on the low correlation coefficient interval to obtain the energy concentration frequency bands located in the low correlation coefficient interval;

[0145] Step S383: Obtain the number of frequency bands corresponding to each energy concentration frequency band, and calculate the variance of the center frequency of each energy concentration frequency band and sum the energy proportion of each energy concentration frequency band to obtain the frequency band variance and the total energy proportion.

[0146] Step S384: Determine the current energy evaluation value based on the number of frequency bands, frequency band variance, and total energy percentage;

[0147] Step S385: The current width evaluation value obtained based on the interval width of the low correlation coefficient interval is weighted and summed with the current energy evaluation value and the current gradient evaluation value, and the winding deformation degree is determined based on the obtained current comprehensive evaluation value.

[0148] Specifically, frequency bands that are continuously lower than a preset correlation threshold can be identified from the correlation coefficient sequence output in step S37 to form a low correlation coefficient interval. The correlation threshold here is the lower limit of the Pearson correlation coefficient, which is used to identify abnormal frequency bands in the winding structure. The specific value can be set according to the actual needs of relevant technical personnel, and this invention does not limit it.

[0149] Furthermore, a first-order difference operation is performed on the correlation coefficient of each frequency point within the interval to obtain a frequency response gradient value sequence, and the mean of its absolute values ​​is calculated as the current gradient evaluation value. The larger the gradient, the more severe the curve distortion, which may correspond to a local abrupt change in the winding.

[0150] Furthermore, the original frequency response amplitude signal in the low correlation coefficient range is decomposed into three-level wavelet packet decomposition to extract the energy distribution of each sub-band, and the sub-band with an energy ratio of more than 5% is selected as the energy concentration band.

[0151] Furthermore, the total number of energy-concentrated frequency bands is counted, and the standard deviation of all center frequencies is calculated as the frequency band variance. At the same time, the energy proportion of each energy-concentrated frequency band is accumulated to obtain the total energy proportion. The frequency band variance mentioned above can reflect the dispersion of abnormal frequency bands.

[0152] Furthermore, based on these three indicators, the current energy evaluation value is generated through a preset mapping function, which can use linear weighting or a lookup table method.

[0153] Furthermore, the current width evaluation value, obtained by normalizing the physical width (in kHz in this invention) of the low correlation coefficient interval, is weighted and summed with the current gradient evaluation value and the current energy evaluation value with weights of 0.3, 0.4, and 0.3 to form the current comprehensive evaluation value. The degree of winding deformation is determined based on the interval in which this value falls (here, 0–0.4 is set as no deformation, 0.4–0.7 as slight deformation, and 0.7–1.0 as severe deformation).

[0154] The frequency response gradient value mentioned above refers to the rate of change of the correlation coefficient between adjacent frequency points within the low correlation coefficient interval, reflecting the local steepness of the curve; the current gradient evaluation value characterizes the severity of distortion; the energy-concentrated frequency band refers to the sub-band whose energy proportion is significantly higher than the background noise after wavelet packet decomposition; the total energy proportion refers to the proportion of the sum of the energy of all energy-concentrated frequency bands to the total energy of the low correlation coefficient interval; the current energy evaluation value refers to the structural anomaly energy dimension score generated by fusing the number of frequency bands, frequency band variance, and total energy proportion; the current width evaluation value refers to the scale score corresponding to the normalized physical width of the low correlation coefficient interval; the winding deformation degree refers to the mechanical structure deterioration level classified according to the current comprehensive evaluation value.

[0155] In some embodiments, the specific steps in step S39, which involve responding to the determination that the target transformer does not meet the emergency maintenance conditions based on the degree of winding deformation, updating the mechanical structure module of the current digital twin model based on the model deformation parameters and maintenance priority, and controlling the evaluation terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on the updated digital twin model, may include:

[0156] Step S391: If the degree of winding deformation is lower than the preset emergency threshold, it is determined that the target transformer does not meet the emergency maintenance conditions.

[0157] Step S392: Obtain the mechanical structure module of the current digital twin model, and determine the standard geometric dimensions and standard stress distribution of the mechanical structure module.

[0158] Step S393: Perform dimensional correction on the standard geometric dimensions based on the model deformation parameters, and update the standard maintenance plan in response to the maintenance priority being higher than the preset normal priority.

[0159] Step S394: Based on the determination of the evaluation terminal that there is a three-dimensional model of the target transformer on the evaluation display interface, the evaluation terminal is controlled to highlight the deformation based on the updated digital twin model, and generate a deformation area identification box to identify the winding area corresponding to the low correlation coefficient interval.

[0160] Specifically, first determine whether the winding deformation degree output by step S385 is lower than the preset emergency threshold (0.7 is selected as an example here). If the current comprehensive evaluation value is 0.62, it can be determined that the target transformer does not meet the emergency maintenance conditions and does not need to be shut down immediately, but it still needs to be included in the tracking maintenance sequence.

[0161] Furthermore, mechanical structure modules are extracted from the current digital twin model. These modules contain the initial geometric parameters and material mechanical properties of components such as windings, cores, and clamps, from which standard geometric dimensions and standard stress distributions can be read.

[0162] Furthermore, based on the model deformation parameters generated in step S38 (for example, using an axial compression of 12 mm and a radial offset of 8 mm as the model deformation parameters here), the standard geometric dimensions are corrected item by item, finally forming an updated geometric model that reflects the actual state. At the same time, the maintenance priority is mapped according to the degree of winding deformation, and the maintenance priority is written into the annual maintenance plan database, triggering the plan adjustment process.

[0163] Furthermore, when the evaluation terminal detects that the target transformer's 3D model has been loaded, the driving graphics engine highlights the corrected winding geometry deformation through color gradient or mesh distortion, and overlays a semi-transparent rectangle at the corresponding physical location as a deformation area marker. The text displayed inside the rectangle can include key information such as "Winding Deformation: Middle of Phase A High Voltage", "Deformation Level: Slight", and "Suggested Processing Period: Within 3 Months".

[0164] The preset emergency threshold refers to the critical value of the winding deformation degree used to determine whether immediate shutdown and maintenance are required; the maintenance priority refers to the urgency level of the maintenance task based on the degree of winding deformation, which is divided into three levels: low, medium and high.

[0165] In some embodiments, in step S310, in response to determining that the target transformer meets the emergency maintenance conditions based on the degree of winding deformation, the emergency handling steps established based on model deformation parameters and maintenance priorities are added to the preset health assessment process. The specific steps for controlling the assessment terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on standard assessment steps and emergency handling steps may include:

[0166] Step S3101: If the degree of deformation of the response winding is not lower than the preset emergency threshold, it is determined that the target transformer meets the emergency maintenance conditions.

[0167] Step S3102: Based on the center frequency point of the corresponding low correlation coefficient interval, generate a treatment plan library that includes low correlation coefficient intervals and is associated with preset deformation types, and establish a preset treatment template;

[0168] Step S3103: Fill the model deformation parameters and maintenance priorities into the parameter text slots and priority text slots located in the preset treatment template, and add the obtained emergency treatment steps to the preset health assessment process.

[0169] Step S3104: In response to the determination by the evaluation terminal that a three-dimensional model of the target transformer exists on the evaluation display interface, the evaluation terminal is controlled to issue alarm prompts based on the standard evaluation steps and emergency handling steps, and generate equipment alarm boxes that identify the target transformer and emergency handling boxes that identify the deformed areas.

[0170] Specifically, first determine whether the winding deformation degree output in step S385 has reached or exceeded the preset emergency threshold. If the current comprehensive evaluation value is 0.82, then the target transformer is determined to meet the emergency maintenance conditions and the emergency response mechanism should be activated immediately.

[0171] Furthermore, taking the center frequency point of the low correlation coefficient range as the core, the winding deformation type strongly correlated with the corresponding frequency band is retrieved in the pre-set power equipment fault knowledge system, and the corresponding contingency plan library is retrieved. This contingency plan library can include typical contingency measures, required tools and equipment, estimated downtime and risk control points. The winding deformation type mentioned above can include types such as "axial compression type severe displacement" or "local inter-turn collapse".

[0172] Furthermore, a matching preset handling template is loaded on this basis. The preset handling template is a structured text frame containing multiple fillable slots, which can include "{model deformation parameters}", "{maintenance priority}", "{suggested shutdown window}", etc.

[0173] Furthermore, the model deformation parameters output in step S38 are filled into the parameter text slot, and "emergency" is filled into the priority text slot, thus forming a complete emergency response procedure. The obtained emergency response procedure is then inserted into the starting position of the preset health assessment process to ensure priority execution.

[0174] Furthermore, when the evaluation terminal detects that the 3D model of the target transformer has been loaded, two types of visual prompts are triggered simultaneously: First, based on the standard evaluation procedure, a flashing red equipment alarm box is displayed around the entire equipment, indicating "Severe winding deformation, shutdown required within 72 hours"; second, a semi-transparent red highlight layer is overlaid on the physical area corresponding to the low correlation coefficient range in the 3D model (for example, the middle of the A-phase high-voltage winding is selected here), and an emergency handling box pops up, displaying "Deformation type: Axial compression", "Deformation amount: 18mm", and "Handling suggestion: Arrange for immediate inspection with the cover". The parameter text slots refer to the positions in the preset handling template used to fill in the specific model deformation parameters; the priority text slots refer to the specified fields used to fill in the maintenance priority.

[0175] In some embodiments, the method further includes: step S41, responding to the load current in the real-time electrical characteristic data exceeding a preset overload threshold, generating a thermal accumulation evaluation factor based on the duration and peak value of the corresponding load current;

[0176] Specifically, the load current sequence can first be extracted from the real-time electrical characteristic data stream, and it can be determined whether it exceeds the preset overload threshold (this invention sets this threshold to 1.2 times the rated current); when an overload event is detected, the start and end times of the overload are recorded to calculate the duration, and the maximum current value during the period is obtained synchronously as the peak value.

[0177] Furthermore, the heat accumulation calculation model recommended by the IEC 60076-7 standard is invoked to normalize the peak value to a multiple relative to the rated current. And substitute into the formula Numerical integration is performed, where H represents the heat accumulation assessment factor; dt represents the normalized load current, which is the ratio of the real-time load current to the rated current; dt represents the derivative with respect to the time variable t. The thermal time constant of the winding is represented by T; the overload duration is represented by T; the integration interval is represented by T. Covering the entire overload event period, the thermal accumulation assessment factor is used to characterize the equivalent loss to insulation life caused by this overload.

[0178] Step S42: Dynamically correct the thermal path model of the current digital twin model based on the thermal accumulation assessment factor, and generate a temperature rise warning area corresponding to the thermal accumulation assessment factor on the assessment display interface of the assessment terminal.

[0179] In some embodiments, the specific steps of dynamically correcting the thermal circuit model of the current digital twin model based on the thermal accumulation evaluation factor in step S42 may include: step S421, obtaining the top oil temperature calculation node and winding hot spot calculation node in the thermal circuit model of the current digital twin model;

[0180] Step S422: Based on the heat accumulation evaluation factor, the heat dissipation coefficient of the top oil temperature calculation node is attenuated and corrected, and the thermal resistance parameter of the winding hot spot calculation node is increased and corrected.

[0181] Step S423: Based on the corrected top oil temperature and winding hot spot temperature, recalculate the remaining life assessment value of the target transformer and push the remaining life assessment value to the assessment terminal.

[0182] Specifically, the thermal circuit model is first extracted from the current digital twin model. The thermal circuit model can be a multi-node equivalent thermal network constructed using the IEC60076-7 standard. The top oil temperature calculation node is used to simulate the dynamic response of the upper oil temperature in the oil tank, and the winding hot spot calculation node is used to characterize the temperature of the hottest spot in the winding.

[0183] Furthermore, based on the heat accumulation evaluation factor generated in step S41, the thermal path parameters are physically consistent according to a preset mapping relationship, and the heat dissipation coefficient associated with the top oil temperature calculation node is subject to exponential decay, for example, the heat dissipation coefficient... ,in, This indicates the corrected heat dissipation coefficient of the top layer oil; This represents the heat dissipation coefficient in the initial (unaged) state. This represents the aging sensitivity coefficient for heat dissipation performance and is a positive real number. Indicates the heat accumulation assessment factor; This indicates the decrease in heat dissipation capacity due to insulation aging and oil channel deterioration.

[0184] Simultaneously, a linear increase correction is applied to the thermal resistance of the winding hot spot calculation node. The thermal resistance can be expressed as: , This indicates the corrected thermal resistance between the winding hot spot and the top layer oil; This represents the reference thermal resistance in the initial (unaged) state. This represents the thermal resistance degradation coefficient, which is a non-negative real number. Indicates the heat accumulation assessment factor; This represents the linear increase in thermal resistance caused by factors such as deterioration of insulation materials and obstruction of oil flow.

[0185] Furthermore, after completing the parameter update, the system uses the current load current and ambient temperature as boundary conditions to resolve the thermal circuit differential equation, and obtains the corrected top oil temperature and winding hot spot temperature.

[0186] Furthermore, the hotspot temperature is substituted into the Monzinger aging model, combined with the current degree of polymerization of the insulating paper, to calculate the updated remaining life assessment value; finally, the value is pushed to the assessment terminal in real time through the message middleware for maintenance personnel to view.

[0187] In some embodiments, the method further includes: step S51, periodically comparing the output evaluation result of the current digital twin model with the offline test data of the target transformer to obtain a comparison deviation value.

[0188] Specifically, a fixed verification cycle can be set (in this invention, a fixed verification cycle is set every quarter or after each preventive test). The latest offline test data of the target transformer can be extracted from the power equipment test management system. This offline test data includes key indicators such as furfural content in oil, degree of polymerization of insulating paper, correlation coefficient of winding deformation frequency response analysis, and DC resistance imbalance rate.

[0189] Furthermore, the simulation output values ​​of the corresponding state variables are read from the current digital twin model. The simulation output values ​​may include the degree of aggregation derived from the thermal-aging coupling module and the correlation coefficient of the frequency response analysis inverted from the mechanical structure module.

[0190] Furthermore, the absolute or relative deviation is calculated for each group of similar indicators. For example, if the measured value of the degree of aggregation is 480 and the model output is 510, then the comparison deviation value is -30.

[0191] Step S52: If the response comparison deviation value is greater than the preset allowable deviation, a global calibration process for the current digital twin model is triggered. The global calibration process includes inverting and optimizing the core physical parameters of the model using offline experimental data.

[0192] Specifically, the system first determines whether the comparison deviation value output in step S51 exceeds the preset allowable deviation of the corresponding indicator. In this invention, the allowable deviation is set to a convergence degree deviation tolerance of ±25 and a frequency response analysis correlation coefficient tolerance of ±0.02. Technical personnel can reset these settings according to actual needs. When any key indicator exceeds the limit, the system activates the global calibration process, loading the current digital twin model of the target transformer and its associated multiphysics submodules. These multiphysics submodules include the thermal circuit model, aging model, and mechanical structure model.

[0193] Furthermore, an objective function is then constructed using offline experimental data as observations and model outputs as predictions. Nonlinear least squares or Bayesian inversion methods are then used to jointly optimize the core physical parameters, such as inverting the insulation aging activation energy, the winding thermal resistance reference value, and the thermal conductivity coefficient of the oil paper.

[0194] Furthermore, physical rationality constraints can be imposed during the optimization process. These constraints may include that the aging rate must not be negative and the thermal resistance must not be lower than the theoretical lower limit of the material.

[0195] Furthermore, the optimized parameter set is written back into the current digital twin model to generate a calibrated new version of the digital twin model, while a calibration log is recorded for easy auditing and traceability later. Among them, the core physical parameters refer to the basic material or structural properties that directly affect the accuracy of the model output, such as the aging activation energy of insulating paper, winding thermal resistance, and thermal conductivity of oil paper.

[0196] In one embodiment, a transformer health assessment system based on digital twins and electrical characteristics is also provided, including:

[0197] The model retrieval module is configured to respond to the data acquisition terminal collecting the operating data of the target transformer, obtain the real-time electrical characteristic data and basic ledger information corresponding to the target transformer, and retrieve the initial digital twin model.

[0198] The model update module is configured to update the model state of the initial digital twin model based on the model correction parameters generated from the corresponding real-time electrical feature data in response to the real-time electrical feature data meeting the model update conditions, so as to obtain the current digital twin model.

[0199] The assessment guidance module is configured to update the standard assessment steps included in the preset health assessment process based on the assessment guidance data generated from the corresponding basic ledger information, and control the assessment terminal to guide the operation and maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps.

[0200] The above-mentioned unit modules can be embedded in the processor of the electronic device in hardware form or independent of it, or they can be stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0201] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-described technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A transformer health assessment method based on digital twins and electrical characteristics, characterized in that, Includes the following steps: The response data acquisition terminal collects the operating data of the target transformer, obtains the real-time electrical characteristic data and basic ledger information corresponding to the target transformer, and retrieves the initial digital twin model; When the real-time electrical feature data meets the model update conditions, the model state of the initial digital twin model is updated based on the model correction parameters generated from the corresponding real-time electrical feature data to obtain the current digital twin model. The assessment guidance data generated based on the corresponding basic ledger information updates the standard assessment steps included in the preset health assessment process, and controls the assessment terminal to guide the operation and maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps.

2. The transformer health assessment method based on digital twins and electrical characteristics according to claim 1, characterized in that, The process of updating the model state of the initial digital twin model using model correction parameters generated based on corresponding real-time electrical feature data to obtain the current digital twin model includes: The control data acquisition terminal performs multi-source data fusion on the target transformer to obtain a fused dataset; The fusion dataset is identified to contain characteristic gas components that indicate anomalies in dissolved gases in oil. Based on the gas concentration status of the corresponding characteristic gas components, the oil chromatography assessment level is determined, and the model correction coefficient and warning threshold corresponding to the oil chromatography assessment level are determined. The response determines that the target transformer does not meet the high-risk criteria based on the oil chromatography assessment level. Based on the model correction coefficient and the early warning threshold, the insulation aging module parameters of the initial digital twin model are updated. The assessment terminal is then controlled to guide the operation and maintenance personnel to perform a health status assessment of the target transformer based on the updated digital twin model. The response determines that the target transformer meets the high-risk criteria based on the oil chromatography assessment level. It then adds a special diagnostic step based on the model correction coefficient and the early warning threshold to the preset health assessment process. The assessment terminal guides the operation and maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment step and the special diagnostic step.

3. The transformer health assessment method based on digital twins and electrical characteristics according to claim 2, characterized in that, The determination of the oil chromatography evaluation level based on the gas concentration state of the corresponding characteristic gas components includes: The concentration values ​​of each gas component constituting the characteristic gas are weighted and summed, and the current gas evaluation value is determined based on the obtained comprehensive gas index. Based on the preset gas ratio method, pattern recognition is performed on the characteristic gas components to obtain the fault feature vector located in the characteristic gas components; Obtain the vector dimension of the corresponding fault feature vector, and normalize the feature value of each dimension and calculate the entropy value of all the feature values ​​to obtain the feature normalization value and feature entropy value. The current mode evaluation value is determined based on the vector dimension, feature normalization value, and feature entropy value. The current rate evaluation value obtained based on the gas production rate of characteristic gas components is weighted and calculated together with the current model evaluation value and the current gas evaluation value, and the oil chromatography evaluation level is determined based on the obtained current comprehensive evaluation value.

4. The transformer health assessment method based on digital twins and electrical characteristics according to claim 3, characterized in that, The method further includes: The current comprehensive evaluation value is input into the pre-trained fault evolution prediction model to obtain the fault evolution duration of the corresponding characteristic gas component; If the fault evolution duration exceeds the preset safety margin duration, the current comprehensive evaluation value is updated based on the retrieved safety margin weights.

5. The transformer health assessment method based on digital twins and electrical characteristics according to claim 4, characterized in that, The response determines that the target transformer does not meet the high-risk criteria based on the oil chromatography assessment level. It then updates the insulation aging module parameters of the initial digital twin model based on model correction coefficients and early warning thresholds. Finally, it controls the assessment terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on the updated digital twin model, including: If the response oil chromatography assessment level is lower than the preset high-risk level, it is determined that the target transformer does not meet the high-risk determination criteria. Obtain the insulation aging module parameters of the initial digital twin model, and determine the standard aging rate and standard warning threshold of the insulation aging module parameters; The standard aging rate is corrected based on the model correction coefficient, and in response to the warning threshold being less than the standard warning threshold, the warning threshold is replaced with the standard warning threshold. The response is based on the assessment terminal determining that a three-dimensional model of the target transformer exists on the assessment display interface. The assessment terminal is then controlled to perform visualization rendering based on the updated digital twin model and generate a health status identification box to identify the target transformer.

6. The transformer health assessment method based on digital twins and electrical characteristics according to claim 5, characterized in that, The response determines that the target transformer meets the high-risk criteria based on the oil chromatography assessment level. It then adds a specific diagnostic step, established based on the model correction coefficient and early warning threshold, to the preset health assessment process. The assessment terminal guides maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps and the specific diagnostic steps, including: If the response oil chromatography assessment level is not lower than the preset high-risk level, it is determined that the target transformer meets the high-risk determination criteria. A diagnostic knowledge graph is generated, centered on the main characteristic gas of the corresponding characteristic gas component and associated with a preset fault type, and a preset diagnostic template is established. Fill the model correction coefficients and warning thresholds into the coefficient text slots and threshold text slots in the preset diagnostic template, and add the obtained special diagnostic steps to the preset health assessment process. The response is based on the assessment terminal determining that a three-dimensional model of the target transformer exists on the assessment display interface. The assessment terminal is then controlled to provide voice prompts based on the standard assessment steps and the special diagnostic steps, and to generate equipment identification boxes to identify the target transformer and risk area identification boxes to identify the fault feature vector.

7. The transformer health assessment method based on digital twins and electrical characteristics according to claim 6, characterized in that, The step of adding the obtained specialized diagnostic steps to the preset health assessment process includes: Obtain the standard assessment steps included in the preset health assessment process, and determine the standard diagnostic depth located in the standard assessment steps; The specific diagnostic step is added to the preset health assessment process. If the severity of the fault corresponding to the warning threshold is greater than the baseline severity corresponding to the standard diagnostic depth, the specific diagnostic step is adjusted to be placed before the standard assessment step. Otherwise, it is adjusted to be placed after the standard assessment step.

8. The transformer health assessment method based on digital twins and electrical characteristics according to claim 1, characterized in that, The assessment guidance data generated based on the corresponding basic ledger information updates the standard assessment steps included in the preset health assessment process, and controls the assessment terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps, including: The sensor network located on the target transformer is controlled to collect the partial discharge pulse signal of the target transformer to obtain a partial discharge signal sequence; Time-frequency domain analysis was performed on the partial discharge signal sequence to obtain the phase distribution map of the corresponding discharge phase and the amplitude distribution map of the corresponding discharge amplitude. Discharge pulse points in adjacent phase intervals are clustered to obtain discharge clusters, and discharge clusters that are allowed to match the target transformer are determined as valid discharge clusters based on the basic ledger information. Obtain the standard assessment steps included in the preset health assessment process, and fill the text slots at the positions of the effective discharge clusters in the standard assessment steps with the winding positions corresponding to the winding positions of the effective discharge clusters. The response is based on the evaluation terminal determining that there is a three-dimensional model of the target transformer on the evaluation display interface. The evaluation terminal is then controlled to provide visual guidance based on standard evaluation steps and generate a partial discharge identification box to mark the winding area corresponding to the effective discharge cluster.

9. The transformer health assessment method based on digital twins and electrical characteristics according to claim 8, characterized in that, The step of determining the discharge clusters that match the target transformer as valid discharge clusters based on the basic ledger information includes: Obtain the equivalent frequency center for each discharge cluster, and determine all equivalent frequency centers that are allowed to match the target transformer as a primary screening group based on the rated voltage in the basic ledger information; Each frequency band range with a corresponding preset bandwidth is formed, which surrounds each equivalent frequency center located in the first screening group, and a preset discharge fingerprint database is retrieved based on the device type corresponding to the target transformer. Features of discharge pulse points within the frequency band are extracted to obtain each discharge feature subset; Each discharge feature subset is matched with a preset discharge fingerprint database for similarity, and based on the matching results, the discharge feature subsets with matching similarity greater than the preset fingerprint similarity are determined as typical discharge electron sets; Obtain the discharge energy of all typical discharge clusters located in the same frequency band, and determine the effective discharge cluster as the center of the equivalent frequency with the largest corresponding discharge energy.

10. The transformer health assessment method based on digital twins and electrical characteristics according to claim 8, characterized in that, The assessment guidance data generated based on the corresponding basic ledger information updates the standard assessment steps included in the preset health assessment process, and controls the assessment terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps, further including: The response data acquisition terminal applies a frequency sweep signal to the target transformer, obtains the frequency response characteristic curve corresponding to the target transformer, and retrieves historical frequency response characteristic curves. Correlation analysis was performed between the frequency response characteristic curve and the historical frequency response characteristic curve to obtain a correlation coefficient sequence; The correlation coefficient sequence is determined to have a low correlation coefficient interval that indicates winding deformation. The degree of winding deformation is determined based on the frequency band offset of the corresponding low correlation coefficient interval. The model deformation parameters and maintenance priority corresponding to the degree of winding deformation are also determined. The response determines that the target transformer does not meet the emergency maintenance conditions based on the degree of winding deformation. Based on the model deformation parameters and maintenance priority, the mechanical structure module of the current digital twin model is updated, and the evaluation terminal is controlled to guide the operation and maintenance personnel to perform a health status assessment of the target transformer based on the updated digital twin model. The response determines that the target transformer meets the emergency maintenance conditions based on the degree of winding deformation. The emergency handling steps established based on the model deformation parameters and maintenance priority are added to the preset health assessment process. The assessment terminal guides the operation and maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps and the emergency handling steps.

11. The transformer health assessment method based on digital twins and electrical characteristics according to claim 10, characterized in that, The determination of the winding deformation degree based on the frequency band offset of the corresponding low correlation coefficient interval includes: Gradient calculation is performed on the correlation coefficients of each frequency point that constitute the low correlation coefficient interval, and the current gradient evaluation value is determined based on the obtained frequency response gradient value. Energy spectrum analysis is performed on the low correlation coefficient interval based on wavelet packet decomposition to obtain the energy concentration frequency bands located in the low correlation coefficient interval; Obtain the number of frequency bands corresponding to each energy concentration frequency band, and calculate the variance of the center frequency of each energy concentration frequency band and sum the energy proportion of each energy concentration frequency band to obtain the frequency band variance and the total energy proportion. The current energy evaluation value is determined based on the number of frequency bands, frequency band variance, and total energy percentage. The current width evaluation value obtained based on the interval width of the low correlation coefficient interval is weighted and summed with the current energy evaluation value and the current gradient evaluation value, and the winding deformation degree is determined based on the obtained current comprehensive evaluation value.

12. The transformer health assessment method based on digital twins and electrical characteristics according to claim 10, characterized in that, The response determines that the target transformer does not meet the emergency maintenance conditions based on the degree of winding deformation. Based on the model deformation parameters and maintenance priority, it updates the mechanical structure module of the current digital twin model and controls the evaluation terminal to guide maintenance personnel to perform a health status assessment of the target transformer based on the updated digital twin model, including: If the degree of winding deformation is lower than a preset emergency threshold, it is determined that the target transformer does not meet the emergency maintenance conditions. Obtain the mechanical structure module of the current digital twin model, and determine the standard geometric dimensions and standard stress distribution of the mechanical structure module; The standard geometric dimensions are corrected based on the model deformation parameters, and the maintenance priority is updated to update the standard maintenance plan in response to the maintenance priority being higher than the preset normal priority. The response is based on the evaluation terminal determining that there is a three-dimensional model of the target transformer on the evaluation display interface. The evaluation terminal is then controlled to highlight the deformation based on the updated digital twin model, and a deformation area identification box is generated to identify the winding area corresponding to the low correlation coefficient interval.

13. The transformer health assessment method based on digital twins and electrical characteristics according to claim 12, characterized in that, The response determines that the target transformer meets the emergency maintenance conditions based on the degree of winding deformation. Emergency handling steps, established based on the model deformation parameters and maintenance priorities, are added to the preset health assessment process. The assessment terminal guides maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps and the emergency handling steps, including: If the degree of winding deformation is not lower than a preset emergency threshold, the target transformer is determined to meet the emergency maintenance conditions. Based on the center frequency point of the corresponding low correlation coefficient interval, generate a treatment plan library that includes the low correlation coefficient interval and is associated with a preset deformation type, and establish a preset treatment template; Fill the model deformation parameters and maintenance priorities into the parameter text slots and priority text slots located in the preset treatment template, and add the obtained emergency treatment steps to the preset health assessment process; The response is based on the assessment terminal determining that a three-dimensional model of the target transformer exists on the assessment display interface. The assessment terminal is then controlled to issue alarm prompts based on the standard assessment steps and emergency handling steps, and to generate an equipment alarm box that identifies the target transformer and an emergency handling box that identifies the deformed area.

14. The transformer health assessment method based on digital twins and electrical characteristics according to claim 13, characterized in that, The method further includes: When the load current in the real-time electrical characteristic data exceeds the preset overload threshold, a thermal accumulation evaluation factor is generated based on the duration and peak value of the corresponding load current. The thermal path model of the current digital twin model is dynamically corrected based on the heat accumulation evaluation factor, and a temperature rise warning area corresponding to the heat accumulation evaluation factor is generated on the evaluation display interface of the evaluation terminal.

15. The transformer health assessment method based on digital twins and electrical characteristics according to claim 14, characterized in that, The dynamic correction of the thermal path model of the current digital twin model based on the heat accumulation evaluation factor includes: Obtain the top oil temperature calculation node and winding hot spot calculation node in the thermal circuit model of the current digital twin model; The heat dissipation coefficient of the top oil temperature calculation node is attenuated and corrected based on the heat accumulation evaluation factor, and the thermal resistance parameter of the winding hot spot calculation node is increased and corrected. Based on the corrected top oil temperature and winding hot spot temperature, the remaining life assessment value of the target transformer is recalculated and pushed to the assessment terminal.

16. The transformer health assessment method based on digital twins and electrical characteristics according to claim 15, characterized in that, The method further includes: The output evaluation results of the current digital twin model are periodically compared with the offline test data of the target transformer to obtain the comparison deviation value; When the comparison deviation value exceeds the preset allowable deviation, a global calibration process for the current digital twin model is triggered. The global calibration process includes inverting and optimizing the core physical parameters of the model using offline experimental data.

17. A transformer health assessment system based on digital twins and electrical characteristics, employing the method described in any one of claims 1 to 16, characterized in that, include: The model retrieval module is configured to respond to the data acquisition terminal collecting the operating data of the target transformer, obtain the real-time electrical characteristic data and basic ledger information corresponding to the target transformer, and retrieve the initial digital twin model. The model update module is configured to update the model state of the initial digital twin model based on the model correction parameters generated from the corresponding real-time electrical feature data in response to the real-time electrical feature data meeting the model update conditions, so as to obtain the current digital twin model. The assessment guidance module is configured to update the standard assessment steps included in the preset health assessment process based on the assessment guidance data generated from the corresponding basic ledger information, and control the assessment terminal to guide the operation and maintenance personnel to perform a health status assessment of the target transformer based on the standard assessment steps.

Images