Transformer insulation health state comprehensive evaluation method based on type-2 hyper-fusion cloud spectrum

Abstract

The invention relates to a transformer insulation health state comprehensive evaluation method based on a type-2 hyper-fusion cloud spectrum, and belongs to the technical field of power equipment state monitoring and intelligent diagnosis. The method takes electrical, thermal, chemical and gas operation indexes as unified input, and faces an engineering reality of'multi-source acquisition error, sample fluctuation and state boundary fuzziness'. And outputting a health state score, a membership degree vector and a corresponding state grade judgment result of the transformer through a closed-loop structure of global mixed weighting-single index types-2 cloud spectrum uncertainty expression-multi-source evidence fusion and type dimensionality reduction-cross-index nonlinear aggregation-comprehensive cloud spectrum reconstruction and similarity decision-making. According to the method, various operating state parameters of different dimensions can be fused, and accurate classification evaluation of the operating state of the transformer is realized while the data fuzziness and uncertainty expression ability is maintained.

Description

Technical Field

[0001] This invention belongs to the field of power equipment condition monitoring and intelligent diagnosis technology, and relates to a comprehensive assessment method for transformer insulation health status based on Type-2 hyper-fusion cloud spectrum. It can be widely applied to scenarios such as transformer in-service condition assessment, service life prediction, fault early warning and operation and maintenance decision optimization, providing technical support for transformer operation status management and health assessment. Background Technology

[0002] Power transformers are core equipment in power transmission and transformation, and their insulation health directly affects the safety of the power grid. Long-term electro-thermal-mechanical coupling stress gradually degrades the insulation, easily leading to breakdown, increased partial discharge, and accelerated aging. Failure can cause power outages and significant losses. As the power grid develops towards ultra-high voltage, large capacity, and intelligence, the operating conditions are becoming more complex, placing higher demands on the accuracy, robustness, and interpretability of condition assessments.

[0003] Tang Huiling and Wu Jiekang of Guangdong University of Technology achieved fault state assessment by constructing feature libraries for oil, paper, gas, and temperature, combined with probabilistic fuzzy sets and clustering. However, their shortcomings lie in the strong dependence of data on the feature libraries and biased fault discrimination. Gu Cailian et al. of Shenyang Institute of Technology completed anomaly monitoring, fault identification, and visual classification based on online data of multiple parameters such as current, voltage, temperature, humidity, and pressure. However, their shortcomings lie in the insufficient characterization of insulation properties such as oil chemistry. Wu Junjian et al. of State Grid Zhejiang Electric Power Co., Ltd. achieved comprehensive state determination by fusing the threshold membership of four signals: partial discharge, core grounding current, oil chromatography, and winding temperature. However, their shortcomings lie in the static nature of the threshold rules and their difficulty in adapting to uncertain fluctuations. Chen Feiyang et al. of Shanghai Dianji University completed high-conflict evidence fusion assessment based on Lance distance and Deng entropy-improved evidence theory. However, their shortcomings lie in the need for reliable weights and probability modeling and the computational complexity.

[0004] Therefore, in order to overcome the problems of insufficient fusion of multidimensional heterogeneous information, subjective or difficult-to-interpret weights, weak robustness under fuzzy boundaries and multi-source fluctuations, and difficulty in characterizing the synergy of indicators by linear weighting, a new comprehensive assessment method for transformer insulation health status is urgently needed to solve these problems. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a comprehensive assessment method for transformer insulation health status based on Type-2 hyper-fusion cloud spectra. Taking electrical, thermal, chemical, and gaseous indicators as inputs, it employs data-driven hybrid weighting and Type-2 cloud uncertainty modeling, and combines nonlinear aggregation and cloud spectra similarity decision-making to output interpretable results, providing technical support for online monitoring and operation and maintenance optimization. This method can integrate multiple operational status parameters from different dimensions, achieving accurate classification and assessment of transformer operating status while maintaining the ability to express data fuzziness and uncertainty.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A comprehensive assessment method for transformer insulation health status based on Type-2 hyper-fusion cloud spectrum, specifically including the following steps: S1: Construct an evaluation index system consisting of four major categories of key operating indicators: electrical, thermal, chemical, and gas. This system covers multiple parameters that can reflect the insulation health status of transformers and clarifies the health meaning and threshold classification source of each type of indicator. S2: The global mixed weight of each indicator is determined by combining the anti-entropy weight method with the association rule analysis method, taking into account both data volatility and the correlation between the indicators and typical faults. S3: For a single indicator, use the Type-2 cloud model to establish uncertainty representations at different state levels to improve the ability to characterize fuzzy and fluctuating data; S4: Evidence fusion is performed on similar indicator data from multiple sources to obtain stable fusion membership degrees, and then transformed into aggregatable single-value results through type dimensionality reduction methods; at the cross-indicator level, the Choquet integral method based on fuzzy measure is used to perform nonlinear weighted aggregation of the fusion membership degrees of each indicator, taking into account the synergistic and inhibitory effects between indicators to reduce information redundancy. S5: The cloud model parameters of each indicator are weighted and combined globally according to the mixed weights, and feedback correction is performed based on the aggregation results. The similarity is compared with the preset standard cloud spectrum set to determine the operating status level of the transformer. This significantly improves the accuracy, stability and interpretability of the assessment in complex operating environments, and is suitable for scenarios such as online monitoring, life prediction and operation and maintenance decision-making.

[0008] Furthermore, in the evaluation index system constructed in step S1, the electrical category is used to characterize operating pressure and electrical defect characteristics, including load rate, core leakage current, partial discharge, winding DC resistance difference, and winding dielectric loss factor. Thermal properties are used to characterize the aging rate and cooling capacity of hot spots, including hot spot temperature rise and top oil temperature. Chemicals are used to characterize the aging of oil paper and the insulation strength of oil products, including furfural content, water content, dielectric loss factor, and breakdown voltage of oil products. Gases are used to characterize the gas production signs and development trends of faults such as discharge or overheating, including acetylene content, total hydrocarbon gas content, absolute gas production rate of total hydrocarbons, and relative gas production rate of total hydrocarbons.

[0009] Furthermore, step S2 specifically includes: using the anti-entropy weight method to calculate the category weights of the four major categories of indicators, reflecting the importance of each category of indicators to the state assessment; then using the association rule analysis method to calculate the correlation degree between each indicator within a category and typical faults, and obtain the weight within the category; combining the category weights and the weights within the category to form the global mixed weight of each indicator.

[0010] Furthermore, step S3 specifically includes: for each indicator, calculating the core parameters of the cloud model based on the threshold range of different state levels, and constructing a type-1 cloud spectrum; on this basis, introducing an uncertainty range to generate a type-2 cloud spectrum, which is used to characterize the fuzziness and uncertainty of the actual measurement data.

[0011] For each indicator (based on the threshold range of each state level), cloud parameters and type-1 membership are calculated as follows:

[0012]

[0013] in, As an indicator The expectation at a certain state level, As an indicator Entropy at a certain state level, , Indicators The upper and lower bounds of the threshold at a certain state level. As an indicator Hyperentropy at a certain state level, The random perturbation factor introduced into the cloud generator; As an indicator The membership degree of a sample value under a certain state level. As an indicator The sample values, The entropy in the cloud parameter triplet. The effective entropy after introducing hyperentropy random perturbation into the cloud parameter triplet; Introducing uncertain parameters The membership interval and membership degree of type-2 are constructed as shown in the following formula:

[0014] in, and Indicators The sample values ​​belong to the type-2 cloud model under a certain state level, which is the lower and upper bound of the membership degree. As an indicator The sample value belongs to the type-1 cloud model membership degree under a certain state level.

[0015] Furthermore, in step S4, evidence fusion is performed on similar indicator data from multiple sources to obtain stable fusion membership degrees, and then converted into aggregatable single-value results through a type dimensionality reduction method. Specifically, when the same indicator has data from different measurement sources or multiple collections, the type-2 membership degree intervals are fused using the evidence fusion method to obtain more stable results; then, the type-2 membership degree in interval form is converted into a fusion membership degree in single-value form through a type dimensionality reduction method.

[0016] Furthermore, in step S4, the Choquet integral method based on fuzzy measure is used to perform nonlinear weighted aggregation of the fusion membership of each index. Specifically, this includes: using additive fuzzy measure and performing Choquet integral calculation to obtain the global fusion value. As shown in the following formula:

[0017]

[0018] in, For the fuzzy measure function value defined on the index set, For the sorted subset of indicators The fuzzy measure value, For the first Global mixed weights for each evaluation indicator A For any subset of the index set, m The total number of indicators participating in the comprehensive evaluation. This is a subset of upper-level indicators sorted from largest to smallest by fusion membership degree. To integrate the membership degrees of various indicators The result after sorting in descending order is the first Individual statistical value, For the first The single-value fusion membership degree is obtained by dimensionality reduction of each indicator by type.

[0019] Furthermore, in step S5, the formula for calculating similarity is:

[0020]

[0021]

[0022] in, This is the triplet of comprehensive cloud parameters obtained by linearly weighting the cloud parameters of each indicator according to the global mixed weight. This is the triplet of integrated cloud parameters after structural feedback correction. The correction factor used for feedback correction. For structural feedback correction coefficients, This is the global fusion value. For the corrected integrated cloud spectrum and the first j Similarity of standard cloud spectra between files and The first j The expected and entropy parameters of the standard state level cloud spectrum.

[0023] The beneficial effects of this invention are as follows: Addressing key pain points in transformer condition assessment, such as the difficulty in unifying and integrating multi-source heterogeneous features, the subjective and uninterpretable allocation of index weights, and the difficulty in effectively characterizing sample fluctuations and ambiguous state boundaries, this invention proposes a comprehensive assessment model based on Type-2 hyper-fusion cloud spectrum. This model integrates a complete technical chain encompassing multi-dimensional index system construction, global hybrid weighting, single-index uncertainty modeling, multi-source fusion, cross-index nonlinear aggregation, and comprehensive cloud spectrum decision-making. Under complex operating environments, it can suppress noise interference, explicitly express ambiguity and uncertainty, and utilize the synergistic and complementary relationships between indicators, thereby significantly improving the accuracy, stability, and interpretability of the assessment results. Simultaneously, this method enriches the engineering application paradigm of Type-2 cloud models in the field of power equipment health assessment, providing reliable support for transformer online monitoring, life prediction, defect early warning, and operation and maintenance decision optimization. It enables refined management of the health status throughout the entire life cycle, thereby improving the safety and reliability of power grid operation and reducing fault risks and operation and maintenance costs, demonstrating clear engineering application value and promising prospects for widespread adoption.

[0024] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0025] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a flowchart of the comprehensive assessment method for transformer insulation health status based on Type-2 hyper-fusion cloud spectrum according to the present invention. Figure 2 To verify the cloud model of the cases in the experiment. Specific Implementation The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0027] This invention provides a fusion model for constructing a comprehensive assessment of transformer insulation health status based on Type-2 hyper-fusion cloud spectra. For example... Figure 1 As shown, this fusion model uses four major categories of operating indicators—electrical, thermal, chemical, and gas—as unified inputs. Addressing the engineering reality of "multi-source acquisition errors, sample fluctuations, and ambiguous state boundaries," it employs a closed-loop structure of "global hybrid weighting—single-indicator type-2 cloud spectrum uncertainty expression—multi-source evidence fusion and type dimensionality reduction—cross-indicator nonlinear aggregation—comprehensive cloud spectrum reconstruction and similarity decision-making" to output the transformer's health status score, membership vector, and corresponding state level determination results. Specifically, global hybrid weighting characterizes the relative importance and fault correlation contribution of indicators in the four categories; type-2 cloud spectrum explicitly expresses measurement fluctuations and boundary ambiguity using "dual membership boundaries"; evidence fusion and type dimensionality reduction stabilize multi-source information and form aggregable single-value memberships; Choquet integral characterizes the synergistic and redundant relationships between indicators; and finally, state decision-making is completed through similarity comparison between the comprehensive cloud spectrum and the standard cloud spectrum set, yielding the health status score and level output.

[0028] To realize the above-mentioned fusion model and complete the calculation of health status score / level, this embodiment constructs and parameterizes each module of the model according to the following steps.

[0029] Step 1: Construct a transformer insulation status assessment model based on a Type-2 hyperfusion cloud spectrum of four categories of indicators: electrical, thermal, chemical, and gas.

[0030] Based on available monitoring data, standards, and industry experience in the operation of mixed-oil transformers, a four-category evaluation index system is first established, encompassing electrical, thermal, chemical, and gas-related indicators. The health implications and threshold classifications for each category are clearly defined. The electrical category characterizes operating pressure and electrical defect characteristics, including load factor (reflecting operating pressure or utilization rate; sustained high load exacerbates winding heating and insulation aging); core leakage current (reflecting insulation degradation between the core and ground or between clamps and grounding anomalies); partial discharge quantity (reflecting insulation defects and discharge activity intensity); winding DC resistance difference (reflecting winding circuit consistency and contact hazards); and winding dielectric loss factor (reflecting the level of dielectric loss in winding insulation). The thermal category characterizes hot spot aging rate and cooling capacity, including hot spot temperature rise and top oil temperature. The chemical category characterizes oil paper aging and oil insulation strength, including furfural content (reflecting the level of thermal aging degradation of solid insulation paper); water content (reflecting the risk of decreased insulation performance and accelerated aging due to water content); and the dielectric loss factor and oil breakdown voltage. Gas types are used to characterize the gas production signs and development trends of faults such as discharge or overheating. They include acetylene content, which characterizes high-temperature discharge or high-energy fault characteristics; total hydrocarbon gas content, which reflects the accumulation of hydrocarbon gases; and absolute and relative total hydrocarbon gas production rates, which characterize the evolution of potential faults from the perspective of rate and trend.

[0031] All the aforementioned indicators are uniformly categorized into four levels—"Normal—Attention—Abnormal—Severe"—based on standards and industry experience, serving as the unified input foundation for the subsequent evaluation system. Building upon this, a global hybrid weighting mechanism assigns data-driven weight contributions to each indicator, achieving a unified expression of importance at both the category and indicator levels. Furthermore, a hyper-fusion mechanism is introduced to perform consistent fusion and dimensionality reduction stabilization of multi-source data and multi-indicator information. A cross-indicator nonlinear aggregation strategy comprehensively considers the synergistic and redundant relationships between indicators, forming a comprehensive cloud spectrum and health status score characterizing the overall insulation health. Finally, the comprehensive cloud spectrum is compared with a preset standard state cloud spectrum set to output the transformer's current insulation health status level and corresponding reliable quantitative results, providing unified model support for subsequent online monitoring, early warning, and operation and maintenance decisions.

[0032] Step 2: Construct a two-level weighting mechanism of "category anti-entropy weight - intra-class association rule".

[0033] The inverse entropy weight method is used to calculate the category weights of the four major categories of indicators, reflecting the importance of each category to the state assessment. Then, the association rule analysis method is used to calculate the correlation between each indicator within a category and typical faults, obtaining the intra-category weights. The category weights and intra-category weights are combined to form the global mixed weights for each indicator.

[0034] The data is standardized, and the calculation of positive and negative indicators is shown in equations (1) to (2).

[0035] (1) (2) in, To Standardized value after extreme value normalization (dimensionless). For the first The sample at the th The original observed values ​​for each evaluation indicator For the first The set of values ​​for each evaluation indicator in the entire sample. and These are the maximum and minimum values ​​of the indicator in the sample set, respectively.

[0036] Four categories are derived from anti-entropy weights ω c As shown in equations (3) to (5).

[0037] (3) (4) (5) in, For the first The first indicator in the Normalized percentage at each sample For the first The inverse entropy value of each indicator, For the first The amount of anti-entropy information corresponding to each indicator, where n is the number of samples.

[0038] Association rules are introduced to calculate the support and confidence of a single indicator for a failure / deterioration event. The final normalized within-class weights are then calculated. As shown in equations (6) to (9).

[0039] (6) (7) (8) (9) in, For association rules Support In the dataset Simultaneously satisfy and Sample count, The sample set used for association rule mining. For association rules Confidence level, In the dataset The preceding condition is satisfied Sample count, The first one obtained from the association rule The association rule weights of each indicator (the basic weights of the intra-class weights). m The number of metrics participating in the association rule calculation. As an indicator Pointing to a fault or degradation event The correlation strength index This refers to the set of in-class indicators corresponding to the indicator set (such as electrical, thermal, chemical, or gas categories).

[0040] The category and intra-class weights are multiplied together to obtain the global mixed weight, as shown in equation (10).

[0041] (10) in, For the first The category weight of each indicator to which it belongs (such as electrical, thermal, chemical, gas).

[0042] Step 3: Constructing the type-2 cloud spectrum single index state representation of the double membership boundary.

[0043] For each indicator, the core parameters of the cloud model are calculated based on the threshold range of different state levels to construct a Type-1 cloud spectrum. On this basis, an uncertainty range is introduced to generate a Type-2 cloud spectrum, which is used to characterize the fuzziness and uncertainty of the actual measurement data.

[0044] For each indicator (based on the threshold range of each state level), cloud parameters and type-1 membership are calculated as shown in equations (11) to (12).

[0045] (11) (12) in, As an indicator The expectation at a certain state level, As an indicator Entropy at a certain state level, , Indicators The upper and lower bounds of the threshold at a certain state level. As an indicator Hyperentropy at a certain state level, The random perturbation factor introduced into the cloud generator; As an indicator The membership degree of a sample value under a certain state level. As an indicator The sample values, The entropy in the cloud parameter triplet. The effective entropy is the result of introducing hyperentropy random perturbation into the cloud parameter triplet.

[0046] Introducing uncertain parameters (in empirical evidence) Take 0.05), construct type-2 membership interval and membership degree, as shown in equation (13).

[0047] (13) in, and Indicators The sample values ​​belong to the type-2 cloud model under a certain state level, which is the lower and upper bound of the membership degree. As an indicator The sample value belongs to the type-1 cloud model membership degree under a certain state level.

[0048] Step 4: Construct an integrated module for multi-source evidence fusion and type dimensionality reduction.

[0049] When the same indicator has data from different measurement sources or collected multiple times, the evidence fusion method is used to fuse the type-2 membership intervals to obtain more stable results. Subsequently, the type-2 membership in interval form is transformed into fused membership in single-value form using a type dimensionality reduction method.

[0050] The evidence fusion of simple average is then used to perform type dimensionality reduction (central method midpoint), as shown in equations (14) to (15).

[0051] (14) (15) in, For the same indicator The amount of multi-source data, To integrate the membership degrees of various indicators The result after sorting in descending order is the first Individual statistical values.

[0052] Step 5: Construct fuzzy measures and use Choquet integrals to achieve cross-index collaboration.

[0053] The Choquet integral method based on fuzzy measure is used to perform nonlinear weighted aggregation of the fusion membership of each index. This method considers the synergistic effect between the indexes and suppresses redundant information, ensuring that the aggregation result is more reasonable and accurate.

[0054] The global fusion value is obtained by using additive fuzzy measure and Choquet integral calculation. As shown in equations (16) to (17).

[0055] (16) (17) in, For the fuzzy measure function value defined on the index set, For the sorted subset of indicators The fuzzy measure value, For the first Global mixed weights for each evaluation indicator A For any subset of the index set, m The total number of indicators participating in the comprehensive evaluation. This is a subset of upper-level indicators sorted from largest to smallest by fusion membership degree. For the first The single-value fusion membership degree is obtained by dimensionality reduction of each indicator by type.

[0056] Step 6: Reconstruct the integrated cloud spectrum parameters and implement structural feedback correction and standard cloud spectrum similarity decision.

[0057] The cloud model parameters of each indicator are weighted and combined using a global mixed weighting to form a comprehensive cloud model parameter, and feedback correction is performed based on the aggregation results. The corrected comprehensive cloud spectrum is compared with the preset standard cloud spectra for each state level, and the level with the highest similarity is selected as the current operating state level of the transformer.

[0058] The comprehensive cloud parameters are constructed and the structure is corrected by feedback. Finally, the similarity decision is made, as shown in equations (18) to (20).

[0059] (18) (19) (20) in, This is the triplet of comprehensive cloud parameters obtained by linearly weighting the cloud parameters of each indicator according to the global mixed weight. This is the triplet of integrated cloud parameters after structural feedback correction. The correction factor used for feedback correction. For structural feedback correction coefficients, This is the global fusion value. For the corrected integrated cloud spectrum and the first j Similarity of standard cloud spectra between files and The firstj The expected and entropy parameters of the standard state level cloud spectrum.

[0060] Verification experiment: To further verify the effectiveness of the transformer condition comprehensive evaluation model based on Type-2 hyper-fusion cloud spectrum, the operational data of the selected index parameters were analyzed to obtain the comprehensive evaluation results of the transformer. Then, a comparative analysis was conducted with the fuzzy-evidence theory method and the fuzzy-improved evidence theory fusion method.

[0061] The anti-entropy and anti-entropy weight of each index can be obtained through normalization calculation and entropy weight formula, as shown in Table 1.

[0062] Table 1. Normalization, anti-entropy, and anti-entropy weights of the case data

[0063] Based on the obtained power transformer fault samples, the support and confidence of each indicator of the transformer are calculated by the association rule method, and finally the final global mixed weights are obtained, as shown in Table 2.

[0064] Table 2 Global Mixed Weights of Case Data

[0065] Select δ =0.05, the normalized value is used to calculate the interval of evidence fusion, and then the midpoint is used for dimensionality reduction to obtain... f Value. (For) f The values ​​are sorted in ascending order and then assigned their respective mixed weights in sequence. ω i Selecting additive fuzzy measure yields g The function, when finally used with a cross-index nonlinear aggregation formula, yields the global fusion value. C(f) As shown in Table 3.

[0066] Table 3. Fuzzy measures and global fusion values ​​of the case data

[0067] Next, the weights of each indicator are linearly weighted with the cloud parameters to obtain the comprehensive cloud parameters. Then, cross-indicator adjustments are incorporated, and finally, the similarity is calculated with the standard cloud parameters (the expected parameters for the four standard state levels (normal, attention, abnormal, severe) are (80, 60, 40, 20)). The state assessment results are then obtained, as shown in Tables 4 and 5. Figure 2 As shown.

[0068] Table 4. Comprehensive cloud parameters and corrected cloud parameters of case data

[0069] Table 5 Similarity Decisions for Case Data

[0070] The state with the highest similarity is the second level—note this.

[0071] Table 6 Comparison of Evaluation Methods

[0072] As shown in Table 6, the evaluation results of the fuzzy-evidence theory method and the fuzzy-improved evidence theory fusion method for transformers also reach the attention level, which can achieve the evaluation of transformer status. However, their maximum similarity is lower than that of the type-based method proposed in this invention. The comprehensive evaluation method for transformer condition based on hyper-fusion cloud spectrum is relatively low, resulting in low accuracy. Therefore, the type-based method proposed in this invention... The transformer condition comprehensive assessment method based on 2 hyper-fusion cloud spectrum has extremely high reliability and accuracy.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A comprehensive assessment method for transformer insulation health status based on Type-2 hyper-fusion cloud spectrum, characterized in that, The method specifically includes: First, an evaluation index system consisting of four major categories of operating indicators—electrical, thermal, chemical, and gaseous—is constructed, and the health meaning and threshold classification source of each category of indicators are clarified. Secondly, the global mixed weights of each indicator are determined by combining the anti-entropy weight method with the association rule analysis method; Then, for a single indicator, the uncertainty expression of different state levels is established using the Type-2 cloud model; Next, evidence fusion is performed on similar indicator data from multiple sources to obtain stable fusion membership degrees, and then transformed into aggregatable single-value results through type dimensionality reduction methods. At the cross-indicator level, the Choquet integral method based on fuzzy measure is used to perform nonlinear weighted aggregation of the fusion membership degrees of each indicator. Finally, the cloud model parameters of each indicator are weighted and combined according to the global mixed weight, and feedback correction is performed based on the aggregation results. The similarity is compared with the preset standard cloud spectrum set to determine the operating status level of the transformer.

2. The comprehensive assessment method for the insulation health status of a transformer according to claim 1, characterized in that, In the constructed evaluation index system, the electrical category is used to characterize operating pressure and electrical defect characteristics, including load rate, core leakage current, partial discharge, winding DC resistance difference, and winding dielectric loss factor. Thermal properties are used to characterize the aging rate and cooling capacity of hot spots, including hot spot temperature rise and top oil temperature. Chemicals are used to characterize the aging of oil paper and the insulation strength of oil products, including furfural content, water content, dielectric loss factor, and breakdown voltage of oil products. Gases are used to characterize the gas production signs and development trends of faults such as discharge or overheating, including acetylene content, total hydrocarbon gas content, absolute gas production rate of total hydrocarbons, and relative gas production rate of total hydrocarbons.

3. The comprehensive assessment method for the insulation health status of a transformer according to claim 1, characterized in that, The method of determining the global mixed weight of each indicator by combining the anti-entropy weight method and the association rule analysis method includes: using the anti-entropy weight method to calculate the category weights of the four categories of indicators to reflect the importance of each category of indicators to the state assessment; then using the association rule analysis method to calculate the correlation degree between each indicator in the category and the typical fault to obtain the category weight; and combining the category weights and the category weights to form the global mixed weight of each indicator.

4. The comprehensive assessment method for the insulation health status of a transformer according to claim 1, characterized in that, The method of using a Type-2 cloud model to establish uncertainty representations at different state levels specifically includes: for each indicator, calculating the core parameters of the cloud model based on the threshold range of different state levels, and constructing a Type-1 cloud spectrum; on this basis, introducing uncertainty ranges to generate a Type-2 cloud spectrum, which is used to characterize the fuzziness and uncertainty of actual measurement data. Each indicator calculates cloud parameters and type-1 membership degree, as shown in the following formula: in, As an indicator The expectation at a certain state level, As an indicator Entropy at a certain state level, , Indicators The upper and lower bounds of the threshold at a certain state level. As an indicator Hyperentropy at a certain state level, The random perturbation factor introduced into the cloud generator; As an indicator The membership degree of a sample value under a certain state level. As an indicator The sample values, The entropy in the cloud parameter triplet. The effective entropy after introducing hyperentropy random perturbation into the cloud parameter triplet; Introducing uncertain parameters The membership interval and membership degree of type-2 are constructed as shown in the following formula: in, and Indicators The sample values ​​belong to the type-2 cloud model under a certain state level, which is the lower and upper bound of the membership degree. As an indicator The sample value belongs to the type-1 cloud model membership degree under a certain state level.

5. The comprehensive assessment method for the insulation health status of a transformer according to claim 1, characterized in that, The process of fusing evidence from multiple sources of similar indicator data to obtain stable fusion membership degrees, and then converting them into aggregatable single-value results through a type dimensionality reduction method, specifically includes: when the same indicator has data from different measurement sources or multiple collections, fusing the type-2 membership degree intervals using an evidence fusion method; and then converting the interval-form type-2 membership degrees into single-value-form fusion membership degrees through a type dimensionality reduction method.

6. The comprehensive assessment method for the insulation health status of a transformer according to claim 1, characterized in that, The method employing a Choquet integral based on fuzzy measure to perform nonlinear weighted aggregation of the fusion membership degrees of each index specifically includes: using additive fuzzy measure and performing Choquet integral calculation to obtain the global fusion value. As shown in the following formula: in, For the fuzzy measure function value defined on the index set, For the sorted subset of indicators The fuzzy measure value, For the first Global mixed weights for each evaluation indicator A For any subset of the index set, m The total number of indicators participating in the comprehensive evaluation. This is a subset of upper-level indicators sorted from largest to smallest by fusion membership degree. To integrate the membership degrees of various indicators The result after sorting in descending order is the first Individual statistical value, For the first The single-value fusion membership degree is obtained by dimensionality reduction of each indicator by type.

7. The comprehensive assessment method for the insulation health status of a transformer according to claim 4, characterized in that, The formula for calculating the similarity is: in, This is the triplet of comprehensive cloud parameters obtained by linearly weighting the cloud parameters of each indicator according to the global mixed weight. This is the triplet of integrated cloud parameters after structural feedback correction. The correction factor used for feedback correction. For structural feedback correction coefficients, This is the global fusion value. For the corrected integrated cloud spectrum and the first j Similarity of standard cloud spectra between files and The first j The expected and entropy parameters of the standard state level cloud spectrum.

Images