Transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition

Abstract

The invention discloses a transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition, and relates to the technical field of power equipment fault diagnosis. The method comprises the following steps: collecting concentration data of various characteristic dissolved gases in transformer oil; receiving the concentration data of the various characteristic dissolved gases and calculating the absolute gas production rate of each gas; according to the concentration data of the gas, establishing a mapping relation between fault features and fault types by adopting a machine learning algorithm; and judging whether fault early warning is triggered or not based on the change trend of the absolute gas production rate of each gas, and performing final fault type judgment by adopting a pentagonal method according to the mapping relationship between fault characteristics and fault types to obtain a fault diagnosis result and output the fault diagnosis result. The method is mainly suitable for sealed and oil-immersed transformers with the voltage class of 110kV and above, and can improve the accuracy and timeliness of transformer fault diagnosis in combination with big data analysis and gas production rate calculation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power equipment fault diagnosis, in particular to a transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition. BACKGROUND

[0002] As the core equipment of the power system, the operation reliability of the power transformer directly determines the safety and stability of the power system. When an electrical or thermal fault occurs in the transformer, the insulating oil and solid insulating materials will decompose under the action of electric and thermal stress, producing characteristic gases such as hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, and carbon dioxide, which are dissolved in the oil. The Dissolved Gas Analysis (DGA) technology detects the types and contents of these gases to identify fault types and assess the state, and has become an important means of transformer online monitoring and fault diagnosis.

[0003] However, the traditional DGA method mainly relies on the absolute value or ratio of gas concentration, such as the Duval triangle method and the IEC three-ratio method. However, these methods have some limitations, such as inability to accurately reflect the development trend of the fault, insufficient sensitivity to early faults, and susceptibility to external factors. With the development of big data technology, how to combine big data analysis and gas production rate calculation to improve the accuracy and timeliness of transformer fault diagnosis has become a problem to be solved.

[0004] Therefore, the present application proposes a transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition. SUMMARY

[0005] The purpose of the present application is to provide a transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides the following technical solution: a transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition, comprising: Collecting concentration data of multiple characteristic dissolved gases in the transformer oil; Receiving the concentration data of multiple characteristic dissolved gases and calculating the absolute gas production rate of each gas; According to the concentration data of the gas, a machine learning algorithm is used to establish a mapping relationship between the fault characteristics and the fault type; Based on the change trend of the absolute gas production rate of each gas, it is determined whether to trigger a fault warning, and based on the mapping relationship between the fault characteristics and the fault type, a pentagon method is used for final fault type judgment to obtain a fault diagnosis result and output.

[0007] Further, the concentration data of a plurality of characteristic dissolved gases in the transformer oil is collected by sensors arranged in the transformer oil tank, which specifically includes the concentration of hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide and carbon dioxide.

[0008] Further, the concentration data of a plurality of characteristic dissolved gases is received and the absolute gas production rate of each gas is calculated, and the specific calculation formula is as follows: R i =(C i2 -C i1 )·G / (Δt*d)(1) Wherein, R i represents the absolute gas production rate; C i1 represents the concentration of a certain gas in the oil measured by the first sampling; C i2 represents the concentration of a certain gas in the oil measured by the second sampling; Δt represents the actual running time in the time interval between the two samplings; G represents the total oil volume of the equipment; d represents the density of the oil.

[0009] Further, the calculated absolute gas production rates of acetylene, methane, ethylene, ethane, hydrogen, carbon monoxide and carbon dioxide are compared with the preset attention value, and when it is preliminarily judged that the gas production rate exceeds the preset attention value, a fault warning signal is triggered.

[0010] Further, according to the concentration data of the gas, a mapping relationship between the fault feature and the fault type is established by using a machine learning algorithm, and the specific process is as follows: Obtain gas concentration data and gas production rate data from transformers of different substations, different models and different running years; Label the corresponding fault type label for each gas concentration data and gas production rate data, wherein the fault type label includes corona partial discharge, low-energy discharge, high-energy discharge, thermal fault temperature greater than 700℃, thermal fault temperature 300-700℃, thermal fault temperature less than 300℃ and low-temperature overheating of mineral oil stray gas; Preprocess the data after labeling the fault type label, and then use the preprocessed data to form a data set, and divide the preprocessed data set into a training set and a test set; Based on the gas concentration data in the data set, calculate the percentage feature of each gas concentration in the total combustible gas, and form the fault feature with the gas production rate data and the gas concentration data; A mapping model is constructed by using a neural network algorithm, the training set and the fault feature are input into the mapping model for training until the mapping model converges, a trained mapping model is obtained, and the mapping relationship between the fault feature and the fault type is output by using the trained mapping model.

[0011] Further, in the training process of the mapping model, the Adam optimizer is adopted, and the mean square error is used as the loss function.

[0012] Further, according to the mapping relationship between the fault characteristics and the fault type, a pentagon method is used for final fault type judgment, and a fault diagnosis result is obtained and output, and the specific implementation steps are as follows: (71) receiving the gas concentration data of hydrogen, methane, acetylene, ethylene and ethane provided by the data acquisition module; transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition (72) calculate the percentage of each gas in total combustible gas %H2, %CH4, %C2H2, %C2H4, %C2H6 respectively, the specific calculation method is as follows: (2) (3) (4) (5) (6) (73) construct a regular pentagon, the inner part of the regular pentagon is divided into seven regions PD, D1, D2, S, T1, T2 and T3, which correspond to corona partial discharge, low energy discharge, high energy discharge, stray gas generated by mineral oil, thermal fault temperature less than 300 DEG C, thermal fault temperature greater than 700 DEG C and thermal fault temperature 300-700 DEG C seven kinds of fault type; (74) the five vertices of the pentagon correspond to five kinds of gas, namely hydrogen, acetylene, ethylene, methane and ethane, and the connecting lines from the centroid of the regular pentagon to the five vertices form the coordinate axes of the five kinds of gas, and the scales of the coordinate axes from the centroid to the vertices are set to 0-100%; (75) according to the percentage of each gas calculated in step (72), the coordinate axes in step (74) are mapped, five data points P1 to P5 of the five kinds of gas are obtained, P1 to P5 are connected in turn, a new pentagon is obtained, the centroid position of the new pentagon is calculated, and the specific fault type is obtained according to the area position of the centroid of the new pentagon.

[0013] The present application has at least the following advantages: 1.The present application can effectively identify the nonlinear correlation characteristics between gas components by upgrading the static threshold of gas concentration to dynamic trend analysis and multi-dimensional feature fusion diagnosis, and based on machine learning algorithm, using historical fault samples and normal operation data for training, which overcomes the misjudgment and missed judgment problem caused by the rigid boundary of the coding interval of traditional IEC three-ratio method. At the same time, the gas production rate as a direct quantitative indicator of fault activity degree, forms a complement with the gas composition characteristics, so that the diagnosis result has timeliness and comprehensiveness, and greatly reduces the false alarm rate caused by external factors such as load fluctuation and oil temperature change.

[0014] 2.The traditional DGA method is slow to respond to the growth of trace gases in the early stage of failure, and the present application can sensitively capture the trace gas increment by calculating the absolute gas production rate and setting a hierarchical attention value, and when the gas production rate exceeds the threshold, it will trigger an early warning immediately, combined with the regional centroid positioning of the pentagon method, which can accurately identify early defects such as partial discharge and low temperature overheating in the reversible stage of fault development, and the pentagon method converts abstract percentage data into intuitive geometric figures, and realizes visual judgment of fault type through centroid area mapping, making the diagnosis process transparent.

[0015] Of course, implementing any product of the present application does not necessarily require all the advantages described above. BRIEF DESCRIPTION OF DRAWINGS

[0016] Fig. 1 The flowchart of the method described in the present application is shown in the figure; Fig. 2 The area division diagram of the pentagon in the present application is shown in the figure. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present disclosure will be described clearly and completely below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present disclosure.

[0018] Please refer to Figs. 1-2 The present application provides a technical solution: a transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition, applied to sealed, oil-immersed transformers with voltage level of 110kV and above, including the following steps: S1.Collect the concentration data of multiple characteristic dissolved gases in the transformer oil; The concentration data of the plurality of characteristic dissolved gases is automatically collected by arranging a sensor in the transformer oil tank, or is collected by taking out the transformer oil from the body for external detection in an automatic or manual manner. The data of the concentration of the characteristic dissolved gases in the transformer oil includes the concentration of hydrogen (H2), methane (CH4), ethane (C2H6), ethylene (C2H4), acetylene (C2H2), carbon monoxide (CO), carbon dioxide (CO2), and the like; S2. Receive the concentration data of the plurality of characteristic dissolved gases and calculate the absolute gas production rate of each gas, specifically as follows: R i =(C i2 -C i1 )·G / (Δt*d) Wherein, R i — absolute gas production rate, mL / d; C i2 — the concentration of a gas in the oil measured by the second sampling, μL / L C i1 — the concentration of a gas in the oil measured by the first sampling, μL / L; Δt— actual running time in the time interval between the two samplings; G— total oil volume of the equipment, t; d— density of the oil, t / m 3 ; It should be further pointed out that the calculated absolute gas production rates of acetylene, methane, ethylene, ethane, hydrogen, carbon monoxide and carbon dioxide are compared with the preset attention values. When it is preliminarily judged that the gas production rate exceeds the preset attention value, a fault early warning trigger signal is output to the fault diagnosis module. The absolute gas production rate preset attention values are shown in Table 1: Table 1. Absolute gas production rate attention values of dissolved gases in the oil of the equipment in operation Unit: mL / d For transformers of 750 kV and above, the typical attention values of the weekly increments of the laboratory offline detection data (see Table 2) can be further judged; Table 2. Absolute increment attention values of the offline oil chromatography gases of the transformers (reactors) of voltage levels of 750 kV and above Unit: mL / d If the absolute gas production rate (absolute increment) does not exceed the above attention values, it can be determined that there is no fault. If the attention values are exceeded, the fault type is further determined by the pentagon method; S3. According to the concentration data of the gases, a mapping relationship between the fault features and the fault types is established by using a machine learning algorithm, specifically as follows: Obtain the gas concentration data and gas production rate data from transformers of different substations, different models and different operating years; annotating corresponding fault type labels for each gas concentration data and gas production rate data, wherein the fault type labels include corona partial discharge, low-energy discharge, high-energy discharge, thermal fault temperature greater than 700 DEG C, thermal fault temperature 300-700 DEG C, thermal fault temperature less than 300 DEG C, and low-temperature overheating corresponding to mineral oil stray gas; performing preprocessing on the data after the fault type labels are annotated, then using the preprocessed data to constitute a data set, and dividing the preprocessed data set into a training set and a test set; based on the gas concentration data in the data set, calculating a percentage feature of each gas concentration accounting for the total combustible gas, and using the percentage feature, the gas production rate data and the gas concentration data to constitute a fault feature; using a neural network algorithm to construct a mapping model, inputting the training set and the fault feature into the mapping model for training until the mapping model converges, obtaining a trained mapping model, and using the trained mapping model to output a mapping relationship between the fault feature and the fault type; It should be noted that after training, the trained mapping model is tested using the test set that did not participate in the training, the fault feature (X_test) of the test set is input, the mapping model outputs the predicted result (Y_pred), and then the real fault type label (Y_test) is compared, and the performance of the mapping model is evaluated by calculating the accuracy, precision, recall, F1 score and other indicators of the mapping model, and the trained mapping model is integrated into the big data analysis module for use; Further, in the training process of the mapping model, the Adam optimizer is used, and the mean square error is used as the loss function; S4. Based on the change trend of the absolute gas production rate of each gas, it is judged whether a fault warning is triggered, and according to the mapping relationship between the fault feature and the fault type, the pentagon method is used for final fault type judgment to obtain a fault diagnosis result and output, as follows: When the absolute gas production rate of each gas exceeds the preset attention value, a fault warning is triggered, for example, when the hydrogen production rate exceeds the set threshold value 1.0 muL / L·h, a fault warning is triggered, at this time the pentagon method is further used for final fault type judgment, and the specific implementation steps are as follows: (S41) receiving the gas concentration data of hydrogen, methane, acetylene, ethylene and ethane provided by the data acquisition module; (S42) calculating the percentage of each gas accounting for the total combustible gas %H2, %CH4, %C2H2, %C2H4, %C2H6 respectively, the specific calculation method is as follows: (2) (3) (4) (5) (6) (S43) Construct a regular pentagon. The interior of the regular pentagon is divided into seven regions: PD, D1, D2, S, T1, T2, and T3, which correspond to seven fault types: corona partial discharge, low-energy discharge, high-energy discharge, stray gas generated by mineral oil, thermal fault temperature less than 300℃, thermal fault temperature greater than 700℃, and thermal fault temperature 300-700℃. (S44) The five vertices of the pentagon correspond to five gases: hydrogen, acetylene, ethylene, methane, and ethane. Lines connecting the centroid of the regular pentagon to the five vertices form the coordinate axes for the five gases, with the scale from the centroid to the vertices set to 0 to 100%. (S45) Based on the percentage of each gas calculated in step (S42), map it onto the coordinate axis in step (S44) to obtain 5 data points P1 to P5 for the five gases. Connect P1 to P5 sequentially to obtain a new pentagon. Calculate the centroid position of the new pentagon and obtain the specific fault type based on the location of the centroid of the new pentagon.

[0019] In summary, this invention upgrades the static threshold judgment of gas concentration to dynamic trend analysis and multi-dimensional feature fusion diagnosis. It adopts machine learning algorithms and is trained based on historical fault samples and normal operation data. It can effectively identify the nonlinear correlation characteristics between gas components. Furthermore, by calculating the absolute gas production rate and setting graded attention values, this invention can sensitively capture trace gas increments. When the gas production rate exceeds the threshold, an early warning is immediately triggered. Combined with the pentagonal method for regional centroid localization, it can accurately identify early defects such as partial discharge and low-temperature overheating in the reversible stage of fault development.

[0020] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0021] Those of ordinary skill in the art, with the benefit of this disclosure, would understand the specific meaning of the above terms in the context of the present application. When an element is referred to as being "on", "connected to", "mounted to", or "disposed to" another element, it can be directly on, connected to, mounted to, or disposed to the other element or intervening elements can also be present. When an element is referred to as being "connected" to another element, it can be directly connected to the other element or intervening elements can be present. As used herein, the terms "vertical", "horizontal", "upper", "lower", "left", "right", and similar expressions are intended for purposes of illustration only and are not intended to be limiting.

[0022] While embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, combinations, and alterations can be made to the embodiments without departing from the principles and spirit of the application, which is defined by the appended claims and their equivalents.

[0023] In the description of the specification, reference to "one embodiment", "an example", "a specific example", or the like means that a particular feature, structure, material, or characteristic being described is included in at least one embodiment or example of the disclosure. The appearances of the above-described terms in various places in the specification are not necessarily referring to the same embodiment or example. Furthermore, the described specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

Claims

1. A method for diagnosing transformer oil dissolved gas faults based on gas production rate and gas composition, characterized in that, Includes the following steps: Collect concentration data of various characteristic dissolved gases in transformer oil; Receive concentration data of various characteristic dissolved gases and calculate the absolute gas production rate of each gas; Based on gas concentration data, machine learning algorithms are used to establish a mapping relationship between fault characteristics and fault types; Based on the changing trend of the absolute gas production rate of each gas, it is determined whether a fault warning is triggered. Based on the mapping relationship between fault characteristics and fault types, the pentagon method is used to make the final fault type judgment, obtain the fault diagnosis result, and output it.

2. The transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition according to claim 1, characterized in that: Concentration data of various characteristic dissolved gases in transformer oil are collected by sensors placed inside the transformer oil tank. These include the concentrations of hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, and carbon dioxide.

3. The transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition according to claim 2, characterized in that: It receives concentration data of various characteristic dissolved gases and calculates the absolute gas production rate of each gas. The specific calculation formula is as follows: R i =(C i2 -C i1 )·G / (Δt*d)(1) Among them, R i Indicates the absolute gas production rate; C i1 —The concentration of a certain gas in the oil was measured during the first sampling; C i2 Δt represents the concentration of a certain gas in the oil measured in the second sampling; Δt represents the actual running time during the time interval between the two samplings; G represents the total amount of oil in the equipment; and d represents the density of the oil.

4. The transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition according to claim 3, characterized in that: The calculated absolute gas production rates of acetylene, methane, ethylene, ethane, hydrogen, carbon monoxide, and carbon dioxide are compared with preset warning values. When it is preliminarily determined that the gas production rate exceeds the preset warning value, a fault warning signal will be triggered.

5. The transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition according to claim 4, characterized in that: Based on the gas concentration data, a machine learning algorithm is used to establish a mapping relationship between fault characteristics and fault types, as follows: Obtain gas concentration data and gas production rate data from transformers of different substations, models, and service lives; Each gas concentration data and gas production rate data is labeled with a corresponding fault type label. The fault type labels include corona partial discharge, low energy discharge, high energy discharge, thermal fault temperature greater than 700℃, thermal fault temperature 300-700℃, thermal fault temperature less than 300℃, and low temperature overheating corresponding to mineral oil stray gas. The data labeled with fault type is preprocessed, and then the preprocessed data is used to form a dataset. The preprocessed dataset is divided into a training set and a test set. Based on the gas concentration data in the dataset, the percentage characteristics of each gas concentration in the total combustible gas are calculated, and the percentage characteristics are combined with the gas production rate data and gas concentration data to form fault characteristics. A neural network algorithm is used to construct a mapping model. The training set and fault features are input into the mapping model for training until the mapping model converges, resulting in a trained mapping model. The trained mapping model is then used to output the mapping relationship between fault features and fault types.

6. The transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition according to claim 5, characterized in that: The Adam optimizer is used during the training of the mapping model, and the mean squared error is used as the loss function.

7. The transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition according to claim 6, characterized in that: Based on the mapping relationship between fault characteristics and fault types, the pentagon method is used to determine the final fault type, obtain the fault diagnosis result, and output it. The specific implementation steps are as follows: (71) Receive gas concentration data of hydrogen, methane, acetylene, ethylene, and ethane provided by the data acquisition module; Transformer oil dissolved gas fault diagnosis method based on gas production rate and gas composition. (72) Calculate the percentages of each gas in the total combustible gas: %H2, %CH4, %C2H2, %C2H4, and %C2H6. The specific calculation methods are as follows: (2) (3) (4) (5) (6) (73) Construct a regular pentagon. The interior of the regular pentagon is divided into seven regions: PD, D1, D2, S, T1, T2, and T3, which correspond to seven types of faults: corona partial discharge, low-energy discharge, high-energy discharge, stray gas generated by mineral oil, thermal fault temperature less than 300℃, thermal fault temperature greater than 700℃, and thermal fault temperature 300-700℃. (74) The five vertices of the pentagon correspond to five gases: hydrogen, acetylene, ethylene, methane, and ethane. Lines are drawn from the centroid of the regular pentagon to the five vertices to form the coordinate axes for the five gases. The scale of the coordinate axes from the centroid to the vertices is set to 0 to 100%. (75) Based on the percentage of each gas calculated in step (72), map it onto the coordinate axis in step (74) to obtain five data points P1 to P5 for the five gases. Connect P1 to P5 sequentially to obtain a new pentagon. Calculate the centroid position of the new pentagon and obtain the specific fault type based on the location of the centroid of the new pentagon.

Images