Transformer fault diagnosis method based on dynamic characteristics of dissolved gas in insulating oil

Abstract

According to the transformer fault diagnosis method based on the dynamic characteristics of the dissolved gas in the insulating oil, the theoretical limitation of a traditional static ratio method is broken through, and multi-dimensional dynamic characteristic extraction, a nonlinear quantitative scoring model and a multi-factor decision fusion mechanism are introduced; according to the method, a complete fault early warning and severity quantitative evaluation system is constructed, ultra-early perception, accurate identification and risk grading of latent faults in the transformer are realized, and the intelligent level of state maintenance of power equipment and the active security defense capability of power grid operation are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of power equipment fault diagnosis technology, and specifically relates to a transformer fault diagnosis method based on the dynamic characteristics of dissolved gases in insulating oil. Background Technology

[0002] Transformers are core hub equipment in power grids, and their operating status directly affects the safety and stability of the entire power system. Insulating oil in transformers plays a dual role of insulation and cooling. During operation, due to factors such as electrical, thermal, and mechanical stress, the insulating oil and solid insulating materials age and decompose, producing various characteristic gases, including hydrogen (H2), methane (CH4), ethane (C2H6), ethylene (C2H4), acetylene (C2H2), carbon monoxide (CO), and carbon dioxide (CO2). The types, concentrations, and variation patterns of these gases dissolved in the oil are crucial information reflecting latent faults inside the transformer (such as partial discharge, overheating, and arcing). Traditional transformer fault diagnosis methods heavily rely on static analysis of DGA data, such as the three-ratio method (IEC 60599 standard). These methods calculate the ratio of specific gas pairs and map them to preset fault type codes. However, the static ratio method has obvious limitations: First, it ignores the dynamic changes in gas concentration and cannot capture the evolution trend of faults, resulting in a slow response to early and rapidly developing faults; second, the ratio interval boundaries are rigid and there is a "coding blind zone," so the diagnostic results are unreliable when the gas ratio falls into the critical region; finally, it is difficult to quantify the severity of the fault, which is not conducive to refined maintenance decisions.

[0003] With the advancement of smart grids and condition-based maintenance systems, higher demands are placed on transformer fault diagnosis: not only must fault types be identified, but also the severity of faults must be quantitatively assessed and early warnings provided. Therefore, a new method is urgently needed that can fully utilize the time-series characteristics of DGA data to achieve dynamic, accurate, and quantitative diagnosis. Summary of the Invention

[0004] The present invention aims to at least partially solve one of the technical problems in the related art.

[0005] Therefore, the first objective of this invention is to propose a method for diagnosing transformer faults based on the dynamic characteristics of dissolved gases in insulating oil.

[0006] The second objective of this invention is to provide a transformer fault diagnosis device based on the dynamic characteristics of dissolved gases in insulating oil.

[0007] To achieve the above objectives, a first aspect of the present invention provides a transformer fault diagnosis method based on the dynamic characteristics of dissolved gases in insulating oil, comprising: S1. Periodically acquire concentration data of various characteristic dissolved gases in transformer insulating oil through online monitoring devices or offline experiments, and construct a high-resolution time series; based on time series analysis, extract the multi-scale dynamic change characteristics of each gas concentration, including short-term mutation rate, long-term evolution trend slope and concentration range fluctuation entropy. S2. Based on the comparison between the dynamic change characteristics and the preset multi-level adaptive thresholds, and by integrating the absolute concentration exceeding the standard criterion, the initial fault level corresponding to each gas is comprehensively determined. S3. Based on the pre-defined differential influence factors of the physical sensitivity, chemical specificity and insulation damage severity of different gases to transformer faults, the individual fault score of each gas under each detection item is accurately calculated by querying the non-linearly growing multidimensional fault scoring table. S4. A weighted aggregation algorithm is used to combine the individual fault scores of all gases to obtain a comprehensive fault score that characterizes the overall insulation health of the transformer. S5. Based on the individual fault score and the comprehensive fault score, and by introducing the key gas combination effect rule and the veto mechanism, accurately determine the fault diagnosis status of the transformer and automatically generate differentiated operation and maintenance strategies.

[0008] In one embodiment of the present invention, S1 includes: Calculate the short-term rate of change of gas concentration within adjacent sampling periods; Based on concentration data of a preset time length, the long-term trend slope is calculated through linear fitting; the fluctuation range of gas concentration within the sliding time window is statistically analyzed, and the fluctuation range is the difference between the maximum and minimum values ​​or the standard deviation.

[0009] In one embodiment of the present invention, S2 includes: The short-term rate of change, long-term trend slope, and volatility are compared with the corresponding first-level, second-level, and third-level thresholds, respectively. The highest level corresponding to each characteristic is determined as the final initial fault level of the gas; wherein, the initial fault level is divided into Level I Normal, Level II Caution, Level III Abnormal, and Level IV Severe; If the current absolute concentration of any gas exceeds the warning value specified in the industry standard, the initial fault level of that gas will be directly determined as Level IV, which is severe.

[0010] In one embodiment of the present invention, S3 includes: The influence factor is set according to the criticality of the gas in diagnosing discharge faults and overheating faults and its gas production rate. The gas includes hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, and carbon dioxide, with acetylene being assigned the highest influence factor. Obtain a predefined fault detection scoring table, which is a two-dimensional matrix. The row headers correspond to the levels of the influencing factors, the column headers correspond to the initial fault levels, and the matrix element values ​​are the corresponding individual fault scores. Based on the gas influencing factor levels and the determined initial fault levels, retrieve the corresponding individual fault scores from the scoring table.

[0011] In one embodiment of the present invention, S5 includes: To determine if a condition is normal, all individual fault scores for all gases must be less than the first threshold; and the overall fault score must also be less than the first threshold. To determine a state of alert, the individual fault score of any gas must be within the range of the first and second thresholds, and the overall fault score must be greater than the first threshold; or there must be a specific combination of key gases whose concentration change trends simultaneously exhibit the characteristics of the alert level. An abnormal state is determined to be one where the individual fault score of any gas falls within the range of the second and third thresholds; or the overall fault score falls within the range of the second and third thresholds. A severe condition is determined if the individual fault score of any gas is greater than the third threshold; or, the overall fault score is greater than or equal to the third threshold; or, the individual fault score of acetylene gas exceeds an extremely high threshold.

[0012] To achieve the above objectives, a second aspect of the present invention provides a transformer fault diagnosis device based on the dynamic characteristics of dissolved gases in insulating oil, comprising: The data acquisition and dynamic feature extraction module is used to periodically acquire the concentration data of various characteristic dissolved gases in transformer insulating oil through online monitoring devices or offline experiments, and construct a high-resolution time series. Based on the time series analysis, the module extracts the multi-scale dynamic change features of each gas concentration, including short-term mutation rate, long-term evolution trend slope, and concentration range fluctuation entropy. The threshold comparison and level determination module is used to compare the dynamic change characteristics with the preset multi-level adaptive thresholds and integrate the absolute concentration exceedance criteria to comprehensively determine the initial fault level corresponding to each gas. The scoring table query and score calculation module is used to combine the differentiating influence factors preset by different gases on the physical sensitivity, chemical specificity and insulation damage severity of transformer faults, and accurately calculate the single fault score of each gas under each detection item by querying the non-linearly growing multidimensional fault scoring table. The comprehensive scoring calculation module is used to combine the individual fault scores of all gases using a weighted aggregation algorithm to obtain a comprehensive fault score that characterizes the overall insulation health of the transformer. The outlier detection and diagnosis module is used to accurately determine the fault diagnosis status of the transformer and automatically generate differentiated operation and maintenance strategies based on the individual fault scores and the comprehensive fault scores, and by introducing key gas combination effect rules and a veto mechanism.

[0013] This invention discloses a transformer fault diagnosis method and apparatus based on the dynamic characteristics of dissolved gases in insulating oil. By introducing dynamic change characteristics, it enables the tracking of the fault evolution process and improves the sensitivity of early fault detection. Through a multi-level quantitative scoring model, it achieves an objective assessment of the fault severity. By integrating multiple indicators and rules, it improves the accuracy and reliability of the diagnostic results, providing strong support for predictive maintenance of transformers.

[0014] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0015] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a transformer fault diagnosis method based on the dynamic characteristics of dissolved gases in insulating oil according to an embodiment of the present invention; Figure 2 This is a structural diagram of a transformer fault diagnosis device based on the dynamic characteristics of dissolved gas in insulating oil according to an embodiment of the present invention. Detailed Implementation

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0017] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0018] The following describes, with reference to the accompanying drawings, a transformer fault diagnosis method and apparatus based on the dynamic characteristics of dissolved gases in insulating oil, according to an embodiment of the present invention.

[0019] Example 1 Figure 1This is a flowchart of a transformer fault diagnosis method based on the dynamic characteristics of dissolved gases in insulating oil according to an embodiment of the present invention, such as... Figure 1 As shown, it includes: S1. Periodically acquire concentration data of various characteristic dissolved gases in transformer insulating oil through online monitoring devices or offline experiments, and construct a high-resolution time series; based on time series analysis, extract the multi-scale dynamic change characteristics of each gas concentration, including short-term mutation rate, long-term evolution trend slope and concentration range fluctuation entropy. S2. Based on the comparison between the dynamic change characteristics and the preset multi-level adaptive thresholds, and by integrating the absolute concentration exceeding the standard criterion, the initial fault level corresponding to each gas is comprehensively determined. S3. Based on the pre-defined differential influence factors of the physical sensitivity, chemical specificity and insulation damage severity of different gases to transformer faults, the individual fault score of each gas under each detection item is accurately calculated by querying the non-linearly growing multidimensional fault scoring table. S4. A weighted aggregation algorithm is used to combine the individual fault scores of all gases to obtain a comprehensive fault score that characterizes the overall insulation health of the transformer. S5. Based on the individual fault score and the comprehensive fault score, and by introducing the key gas combination effect rule and the veto mechanism, accurately determine the fault diagnosis status of the transformer and automatically generate differentiated operation and maintenance strategies.

[0020] This invention discloses a transformer fault diagnosis method based on the dynamic characteristics of dissolved gases in insulating oil. By introducing dynamic change characteristics, it enables the tracking of the fault evolution process and improves the sensitivity of early fault detection. Through a multi-level quantitative scoring model, it achieves an objective assessment of the fault severity. By integrating multiple indicators and rules, it improves the accuracy and reliability of the diagnostic results, providing strong support for predictive maintenance of transformers.

[0021] Example 2 The following describes in detail, with reference to the accompanying drawings, a transformer fault diagnosis method based on the dynamic characteristics of dissolved gases in insulating oil according to an embodiment of the present invention.

[0022] This invention proposes a transformer fault diagnosis method based on the dynamic characteristics of dissolved gases in insulating oil, the specific steps of which are as follows: S10: Multi-dimensional dynamic feature extraction and fusion. Obtaining concentration data sequences of various dissolved gases in transformer insulating oil; S101: Data Acquisition and Preprocessing Before feature extraction, the raw data is cleaned and organized. Data source: Typically from periodic sampling (e.g., daily) from online monitoring systems or offline laboratory chromatographic analysis. Data structure: Data should contain at least two fields: timestamp and concentration values ​​for various gases. Preprocessing steps: Alignment and resampling to ensure data are at equal time intervals. For irregular data, interpolation (e.g., linear interpolation) or resampling at a fixed frequency is required. Missing value handling: For a small number of missing data points, interpolation or moving averages can be used to fill in the missing values. For a large number of consecutive missing values, consider excluding that time period from the calculation. Outlier handling: Use statistical methods (e.g., the 3σ principle) or domain knowledge (e.g., concentration cannot suddenly drop to zero) to identify, correct, or remove obvious outliers. Data smoothing: If the data is noisy, use a moving average filter for smoothing to reduce the interference of high-frequency fluctuations on feature calculation. After preprocessing, the regularized time series for each gas is obtained: C(t) = [c1, c2, c3, ..., c n ], where c i is the gas concentration at the i-th time point, and n is the total number of data points.

[0023] S20: Based on this time series, extract the dynamic change characteristics of each gas concentration, including short-term change rate, long-term trend slope, and fluctuation amplitude. Let c be the data point in the series. i Calculate the three features.

[0024] S201: Short-term rate of change characterizes sudden, drastic positive rate of change, indicating the emergence of a new, active fault point within the equipment. This feature reflects the instantaneous rate and direction of gas concentration change within the most recent short period. Algorithm: First-order difference: Short-run rate of change (i) = c i - c i-1 Calculate the absolute change between two adjacent sampling points.

[0025] Relative rate of change: Short-run rate of change (i) = (c i - c i-1 ) / c i-1 * 100%, more sensitive to low concentrations of gas.

[0026] S202: The long-term trend slope characterizes the overall trend of gas concentration over a relatively long period, determining the persistence and severity of the fault. Algorithm: Linear regression fitting. (1) For a given data point c i Take the m points preceding it (including itself) as a time window [c i-m+1 , ..., c i-1 , c i ].

[0027] (2) Perform univariate linear regression on the interval data and their time indices (e.g., [1, 2, ..., m]) to fit a straight line y = kx + b. Extract the slope k as the long-term trend slope of the point. When k>0, the gas concentration continues to increase, which should raise an alarm; when k≈0, the component concentration is stable; when k<0, the component concentration decreases, which may alleviate the fault or slow down the diffusion.

[0028] (3) Set according to sampling frequency and business needs. For example, for daily sampling data, m can be 30 (about one month) or 90 (about one quarter). The larger m is, the smoother the trend, but the slower the response to recent changes.

[0029] S203: Fluctuation amplitude characterizes the stability and volatility of gas concentration in the short term, identifying intermittent faults or measurement noise. Large fluctuation amplitude indicates that the gas production process of the equipment is unstable or there is external interference; small fluctuation amplitude indicates that the gas production process is stable or the equipment is operating smoothly.

[0030] Algorithm: Moving standard deviation. For a given data point c i Take the w points before it (including itself) as a window [c i-w+1 , ..., c i-1 , c i Calculate the standard deviation σ of the data within this window, which is taken as the volatility at that point. Volatility (i) = σ([c i-w+1 , ..., c i ]) S30: Initial Fault Level Mapping Based on Multidimensional Criteria. Each extracted dynamic feature is compared with a preset multi-level threshold. Following the principle of "choosing the highest," the highest level corresponding to each feature is determined as the final initial fault level (Level I to IV) for the gas. If the absolute concentration of the gas exceeds the limit, it is directly determined to be Level IV. The threshold setting scheme for Levels I to IV is as follows: S301 Absolute Concentration Threshold: The absolute concentration threshold is set directly according to GB / T 7252 or IEC 60599. Once exceeded, dynamic characteristics are not required. If the concentration of any gas is ≥ its "hazard value", the component is directly judged as a Level IV fault.

[0031] S302 Short-Term Change Rate Threshold: This feature focuses on rapid changes in the recent period (e.g., between two consecutive detections). Units are μL / L / day (for online monitoring) or μL / L / detection (for offline detection), as shown in Table 1.

[0032] Table 1

[0033] S303: Long-term trend slope threshold. This feature focuses on the stable trend in the medium to long term. The unit is μL / L / day, reflecting the persistence and inertia of the fault, as shown in Table 2.

[0034] Table 2

[0035] S304: Fluctuation amplitude threshold. This feature focuses on instability in the short term (e.g., the past week). The unit is μL / L. Large fluctuations indicate instability in the gas production process or intermittent failures, as shown in Table 3.

[0036] Table 3

[0037] S305: Calculate the individual fault score for each gas based on the initial fault level and influence factors. The influence factors are set according to the gas type and its impact on the transformer insulation performance; for example, acetylene (C2H2) has the highest influence factor. The individual fault score is obtained by querying the preset fault scoring table (as shown in Table 4).

[0038] Table 4

[0039] S40: Quantitative scoring model based on influence factor weighting. Based on the initial fault level and predefined influence factors related to gas type, the individual fault score for each gas is calculated by querying a predefined fault detection scoring table.

[0040] S50: Comprehensive assessment of global health status. Calculate the transformer's overall fault score based on the individual fault scores for all gases. Decision-making based on a rule engine using multiple parameters. Determine the transformer's fault diagnosis status (normal, alert, abnormal, severe) based on the individual fault scores, the overall fault score, and key gas combination rules.

[0041] Example 4 The technical solution of this disclosure will be described in detail below with reference to specific embodiments. Detailed explanation of the calculation process for fault diagnosis of sudden acetylene component change in a 500kV transformer.

[0042] Data preprocessing (sampled data and automatic processing omitted).

[0043] The dynamic features are calculated as follows: Short-term rate of change (difference between adjacent days): ; .

[0044] Relative rate of change: R short %=0.76.3×100%=11.1%R short% = 6.3 × 100% = 11.1%. Absolute threshold for grade determination: 0.7 μL / L / day < 5 μL / L / day → Grade I; C2H2 sensitive, threshold halved: 0.7 μL / L / day ≥ 0.5 × 5 μL / L / day → Grade I.

[0045] Long-term trend slope (30-day linear regression): time index x=[1,2,...,30], concentration y=[0.3, 0.4,..., 7.0].

[0046] (1) Regression calculation: Slope: k=n∑xy-(∑x)(∑y)n∑x2-(∑x)2 =30×650.1-465×42.330×9455-4652 =0.35μL / L / day; Grade determination: 1μL / L / day≤0.35<3μL / L / day→Grade I.

[0047] (2) Fluctuation range (7-day standard deviation), window data: October 6-12 [5.1, 5.2, 5.8, 6.0, 6.3, 6.5, 7.0]. Calculation: μ=5.1+...+7.07=6.0, σ=∑(yi-μ)27=0.72μL / L, grade determination: 0.72<5μL / L→Grade I 0.72<5μL / L→Grade I

[0048] Fault severity and score calculation includes: Initial fault level: C2H2: absolute concentration 7μL / L ≥ 5μL / L → directly Level IV; H2: absolute concentration 800 μL / L ≥ 150 μL / L → directly Level IV.

[0049] Short-term rate of change threshold: H2 slope calculation (30-day window): kH2=25μL / L / day (>15μL / L / day → Level IV) kH2=25μL / L / day (>15μL / L / day → Level IV); Fluctuation amplitude threshold: C2H2's σ=0.72 μL / L is much lower than the Level I threshold (5μL / L), indicating a stable growth trend and non-intermittent failure.

[0050] Fault status assessment: Acetylene has a long-term slope of 0.35 μL / L / day (Level I), but its absolute concentration is approaching the danger value (reaching 7 μL / L); Hydrogen has a 30-day slope of 25 μL / L / day (Level IV), with a fluctuation range of 50 μL / L (Level IV), indicating that the fault is continuously deteriorating.

[0051] The overall scores are shown in Table 5: Table 5

[0052] 40(C2H2)+30(H2)+16(CH4)+16(C2H4)=102 Diagnostic conclusion: Rule Trigger: Overall score 102 > 60 → "Severe"; Critical gas combination (C2H2 + H2) → Arc discharge (PD). Overall score: 102 points (crossing the full score of 100), exceeding the "Severe" threshold (≥60 points); Critical gas combination: Continuous increase of acetylene (C2H2) + explosive exceedance of hydrogen (H2) → Arc discharge (PD) or high temperature overheating; Immediate shutdown, check for poor contact of tap changer or partial discharge of windings, as shown in Table 6.

[0053] Table 6

[0054] Output processing suggestions: (1) Take immediate action and arrange for a power outage for maintenance; prioritize checking for poor contact of the tap changer, partial discharge of the winding, or multiple grounding faults in the iron core. Further conduct partial discharge detection and winding deformation testing; analyze the metal particles in the oil (to determine whether copper or iron materials are involved in overheating).

[0055] (2) Risk warning: If operation continues, it may cause insulation breakdown or short circuit explosion (acetylene concentration is expected to exceed the dangerous value of 10 μL / L within 3-5 days); historical cases show that it takes only 7-10 days on average for a fault with a similar gas combination to develop into a breakdown.

[0056] Example 4 To achieve the above embodiments, such as Figure 2 As shown, this embodiment also provides a task-oriented visual model construction device 10 based on self-supervised representation. The device 10 includes a data acquisition and dynamic feature extraction module 100, a threshold comparison and level determination module 200, a scoring table query and score calculation module 300, a comprehensive score calculation module 400, and an outlier detection and diagnosis module 500.

[0057] The data acquisition and dynamic feature extraction module 100 is used to periodically acquire the concentration data of various characteristic dissolved gases in transformer insulating oil through online monitoring devices or offline experiments, and construct a high-resolution time series; based on time series analysis, it extracts the multi-scale dynamic change features of each gas concentration, including short-term mutation rate, long-term evolution trend slope and concentration range fluctuation entropy. The threshold comparison and level determination module 200 is used to comprehensively determine the initial fault level of each gas by comparing the dynamic change characteristics with the preset multi-level adaptive thresholds and integrating the absolute concentration exceedance criterion. The scoring table query and score calculation module 300 is used to combine the preset differentiated influence factors of the physical sensitivity, chemical specificity and insulation damage severity of different gases to transformer faults, and accurately calculate the single fault score of each gas under each detection item by querying the non-linearly growing multidimensional fault scoring table. The comprehensive scoring calculation module 400 is used to combine the individual fault scores of all gases using a weighted aggregation algorithm to obtain a comprehensive fault score that characterizes the overall insulation health of the transformer. The outlier detection and diagnosis module 500 is used to accurately determine the fault diagnosis status of the transformer and automatically generate differentiated operation and maintenance strategies based on the individual fault scores and the comprehensive fault scores, and by introducing key gas combination effect rules and a veto mechanism.

[0058] Furthermore, the aforementioned data acquisition and dynamic feature extraction module 100 is also used for: Calculate the short-term rate of change of gas concentration within adjacent sampling periods; Based on concentration data of a preset time length, the long-term trend slope is calculated through linear fitting; the fluctuation range of gas concentration within the sliding time window is statistically analyzed, and the fluctuation range is the difference between the maximum and minimum values ​​or the standard deviation.

[0059] Furthermore, the threshold comparison and level determination module 200 described above is also used for: The short-term rate of change, long-term trend slope, and volatility are compared with the corresponding first-level, second-level, and third-level thresholds, respectively. The highest level corresponding to each characteristic is determined as the final initial fault level of the gas; wherein, the initial fault level is divided into Level I Normal, Level II Caution, Level III Abnormal, and Level IV Severe; If the current absolute concentration of any gas exceeds the warning value specified in the industry standard, the initial fault level of that gas will be directly determined as Level IV, which is severe.

[0060] Furthermore, the aforementioned scoring table query and score calculation module 300 is also used for: The influence factor is set according to the criticality of the gas in diagnosing discharge faults and overheating faults and its gas production rate. The gas includes hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, and carbon dioxide, with acetylene being assigned the highest influence factor. Obtain a predefined fault detection scoring table, which is a two-dimensional matrix. The row headers correspond to the levels of the influencing factors, the column headers correspond to the initial fault levels, and the matrix element values ​​are the corresponding individual fault scores. Based on the gas influencing factor levels and the determined initial fault levels, retrieve the corresponding individual fault scores from the scoring table.

[0061] Furthermore, the aforementioned outlier detection and diagnosis module 500 is also used for: To determine a normal state, all individual fault scores for all gases must be less than a first threshold, and the overall fault score must also be less than the first threshold. To determine a state of alert, any single fault score for any gas must fall within the range of the first and second thresholds, and the overall fault score must be greater than the first threshold; or there must be a specific combination of key gases whose concentration trends simultaneously exhibit alert level characteristics. To determine an abnormal state, any single fault score for any gas must fall within the range of the second and third thresholds; or the overall fault score must fall within the range of the second and third thresholds. To determine a severe state, any single fault score for any gas must be greater than the third threshold; or the overall fault score must be greater than or equal to the third threshold; or the single fault score for acetylene gas must individually exceed an extremely high threshold. To determine a normal state, all individual fault scores for all gases must be less than the first threshold, and the overall fault score must also be less than the first threshold. To determine a state of alert, the individual fault score of any gas must be within the range of the first and second thresholds, and the overall fault score must be greater than the first threshold; or there must be a specific combination of key gases whose concentration change trends simultaneously exhibit the characteristics of the alert level. An abnormal state is determined to be one where the individual fault score of any gas falls within the range of the second and third thresholds; or the overall fault score falls within the range of the second and third thresholds. A severe condition is determined if the individual fault score of any gas is greater than the third threshold; or, the overall fault score is greater than or equal to the third threshold; or, the individual fault score of acetylene gas exceeds an extremely high threshold.

[0062] This invention discloses a transformer fault diagnosis device based on the dynamic characteristics of dissolved gases in insulating oil. By introducing dynamic change characteristics, it enables the tracking of the fault evolution process and improves the sensitivity of early fault detection. Through a multi-level quantitative scoring model, it achieves an objective assessment of the fault severity. By integrating multiple indicators and rules, it improves the accuracy and reliability of the diagnostic results, providing strong support for predictive maintenance of transformers.

[0063] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0064] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A transformer fault diagnosis method of dynamic characteristics of dissolved gases in insulating oil, characterized by, The method comprises the following steps: S1, periodically obtaining concentration data of multiple characteristic dissolved gases in transformer insulating oil through online monitoring devices or offline experiments, and constructing a high-resolution time series; based on time series analysis, extracting multi-scale dynamic change characteristics of each gas concentration, including short-term mutation rate, long-term evolution trend slope and concentration interval fluctuation entropy; S2, according to the comparison of the dynamic change characteristics with the preset multi-level adaptive threshold, and combining the absolute concentration exceeding criterion, the initial fault grade of each gas is comprehensively determined; S3, combining the preset differentiated influence factors of different gases on the physical sensitivity, chemical specificity and insulating hazard severity of transformer faults, the single fault score of each gas under each detection item is accurately calculated by querying the nonlinearly increasing multi-dimensional fault scoring table; S4, using a weighted aggregation algorithm to comprehensively aggregate the single fault scores of all gases to obtain a comprehensive fault score representing the overall insulation health of the transformer; S5, based on the single fault score and the comprehensive fault score, and introducing the key gas combination effect rule and the one-vote veto mechanism, the fault diagnosis state of the transformer is accurately determined and the differentiated operation and maintenance strategy is automatically generated.

2. The method of claim 1, wherein, The S1 comprises: calculating the short-term change rate of gas concentration in the adjacent sampling period; based on the concentration data of the preset time length, the long-term trend slope is calculated by linear fitting; the fluctuation amplitude of the gas concentration in the sliding time window is calculated, and the fluctuation amplitude is the difference between the maximum value and the minimum value or the standard deviation.

3. The method of claim 1, wherein, The S2 comprises: comparing the short-term change rate, the long-term trend slope and the fluctuation amplitude with the corresponding first-level, second-level and third-level thresholds respectively; the highest grade corresponding to each feature is determined as the final initial fault grade of the gas; wherein the initial fault grade is divided into grade I normal, grade II attention, grade III abnormal and grade IV serious; if the current absolute concentration of any gas exceeds the attention value specified in the industry standard, the initial fault grade of the gas is directly determined as grade IV serious.

4. The method of claim 1, wherein, The S3 further comprises: the influence factor is set according to the key degree of the gas in diagnosing discharge fault and overheat fault and the gas production rate, and the gas includes hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide and carbon dioxide, wherein acetylene is given the highest influence factor; a predefined fault detection scoring table is obtained, the scoring table is a two-dimensional matrix, the row title corresponds to the level of the influence factor, the column title corresponds to the initial fault grade, and the matrix element value is the corresponding single fault score; according to the influence factor level and the determined initial fault grade of the gas, the corresponding single fault score is obtained from the scoring table.

5. The method of claim 1, wherein, The S5 comprises: to determine the normal state, the single fault score of all gases must be less than the first threshold value; and the comprehensive fault score is less than the first threshold value; to determine the attention state, the single fault score of any gas must be in the interval between the first threshold value and the second threshold value, and the comprehensive fault score must be greater than the first threshold value; or there is a specific key gas combination, and the concentration change trend of the combination simultaneously shows the attention level characteristics. The abnormal state is determined by satisfying either the single fault score of any gas being in the second threshold and third threshold interval, or the comprehensive fault score being in the second threshold and third threshold interval; The serious state is determined by satisfying either the single fault score of any gas being greater than the third threshold, or the comprehensive fault score being greater than or equal to the third threshold, or the single fault score of acetylene gas alone exceeding an extremely high threshold.

6. A transformer fault diagnosis device of dynamic characteristics of dissolved gas in insulating oil, characterized by, Comprise: The data acquisition and dynamic feature extraction module is used for periodically acquiring concentration data of multiple characteristic dissolved gases in transformer insulating oil through online monitoring devices or offline experiments, and constructing high-resolution time series; based on time series analysis, multi-scale dynamic change features of each gas concentration are extracted, including short-term mutation rate, long-term evolution trend slope and concentration interval fluctuation entropy; The threshold comparison and grade determination module is used for determining the initial fault grade of each gas according to the comparison of the dynamic change features with the preset multi-level adaptive threshold, and combining the absolute concentration exceeding criterion; The scoring table query and score calculation module is used for combining the preset differentiated influence factors of different gases on the physical sensitivity, chemical specificity and insulating hazard severity of the transformer fault, accurately calculating the single fault score of each gas under each detection item by querying the nonlinearly growing multi-dimensional fault scoring table; The comprehensive scoring calculation module is used for adopting a weighted aggregation algorithm to comprehensively aggregate the single fault scores of all gases to obtain a comprehensive fault score representing the overall insulation health of the transformer; The abnormal value detection and diagnosis module is used for accurately determining the fault diagnosis state of the transformer and automatically generating differentiated operation and maintenance strategies based on the single fault score and the comprehensive fault score, and introducing key gas combination effect rules and a veto mechanism.

7. The apparatus of claim 6, wherein, The data acquisition and dynamic feature extraction module is further used for: Calculating the short-term change rate of gas concentration in adjacent sampling periods; Based on the concentration data of a preset time length, the long-term trend slope is calculated by linear fitting; the fluctuation amplitude of the gas concentration in the sliding time window is calculated, and the fluctuation amplitude is the difference between the maximum value and the minimum value or the standard deviation.

8. The apparatus of claim 6, wherein, The threshold comparison and grade determination module is further used for: Comparing the short-term change rate, long-term trend slope and fluctuation amplitude with the corresponding first, second and third thresholds, respectively; The highest grade corresponding to each feature is determined as the final initial fault grade of the gas; wherein the initial fault grade is divided into grade I normal, grade II attention, grade III abnormal and grade IV serious; If the current absolute concentration of any gas exceeds the attention value specified by the industry standard, the initial fault grade of the gas is directly determined as grade IV serious.

9. The apparatus of claim 6, wherein, The scoring table query and score calculation module is further used for: The influence factor is set according to the key degree of the gas in diagnosing discharge fault and overheat fault and the gas production rate, and the gas includes hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide and carbon dioxide, wherein acetylene is given the highest influence factor; A predefined fault detection score table is acquired, the score table being a two-dimensional matrix, row titles corresponding to levels of the influence factors, column titles corresponding to initial fault levels, and matrix element values being corresponding single-item fault scores; a corresponding single-item fault score is obtained from the score table according to the influence factor level of the gas and the determined initial fault level.

10. The apparatus of claim 6, wherein, The abnormal value detection and diagnosis module is further configured to: determine that the normal state is required to simultaneously satisfy that single-item fault scores of all gases are less than a first threshold value; and the comprehensive fault score is less than the first threshold value; determine that the attention state is required to satisfy that a single-item fault score of any gas is in a first threshold value and a second threshold value interval, and the comprehensive fault score is greater than the first threshold value; or there is a specific key gas combination, and concentration variation trends thereof simultaneously present attention level characteristics; determine that the abnormal state is required to satisfy that a single-item fault score of any gas is in a second threshold value and a third threshold value interval; or the comprehensive fault score is in the second threshold value and the third threshold value interval; determine that the serious state is required to satisfy that a single-item fault score of any gas is greater than a third threshold value; or the comprehensive fault score is greater than or equal to the third threshold value; or a single-item fault score of acetylene gas alone exceeds an extremely high threshold value.

Images