Power transformer fault early warning method and system based on big data processing

Abstract

The invention discloses a power transformer fault early warning method and system based on big data processing. The method comprises the steps of obtaining historical carbon emission data; the power transformer fault early warning model is used for power transformer fault prediction; inputting the real-time carbon emission data into the power transformer fault early warning model for prediction; if the fault exists, determining possible non-fault factors of the fault type according to the predicted fault type; whether the possible non-fault factors are true or not is judged according to the corresponding transformer operation parameters; if the corresponding possible non-fault factor is true, sending the possible non-fault factor and the predicted fault type to a target client; and otherwise, sending the predicted fault type to the target client. The invention belongs to the field of transformer fault prediction. According to the invention, accurate prediction of the transformer fault can be realized based on carbon emission.

Description

Technical Field

[0001] This invention relates to the field of transformer fault prediction, and in particular to a method and system for early warning of power transformer faults based on big data processing. Background Technology

[0002] With the continuous expansion of power systems and the increasing demands for power supply reliability, the safe and stable operation of power transformers, as key equipment in the power grid, is of paramount importance. Traditional transformer fault early warning methods mainly rely on dissolved gas analysis (DGA), partial discharge detection, and electrical testing. While these methods are effective, they often focus on monitoring internal physicochemical changes or specific electrical parameters, making it difficult to comprehensively reflect the combined effects of faults on the efficiency and environmental impact of the entire energy system.

[0003] In recent years, fault diagnosis technologies based on big data and artificial intelligence have gradually emerged. By collecting and analyzing massive amounts of operational data and utilizing machine learning models to identify and predict fault modes, they have significantly improved the intelligence level of early warning systems. However, most existing technologies focus on direct operating parameters such as electrical quantities, temperature, and vibration, and rarely conduct correlation analysis between equipment faults and environmental performance indicators such as carbon emissions. Under the "dual carbon" goal, the green and low-carbon operation of power systems has become an important development direction, and the intrinsic link between equipment energy efficiency losses and carbon emissions urgently needs to be explored and utilized. Therefore, there is an urgent need for a transformer fault prediction method that can integrate equipment operating status and environmental impact data. Summary of the Invention

[0004] This invention provides a power transformer fault early warning method and system based on big data processing, which solves the technical problem of how to use environmental indicators such as carbon emissions to predict transformer faults in the prior art, and achieves the technical effect of using environmental indicators such as carbon emissions to predict transformer faults.

[0005] In a first aspect, the present invention provides a power transformer fault early warning method based on big data processing, including:

[0006] The system acquires the fault types of several power transformers within a historical time period, as well as the historical carbon emission data for the historical time period in which the faults of that type occurred. All power transformers belong to the same model, and the carbon emission data includes both direct and indirect carbon emission data. The fault type and the corresponding historical carbon emission data are bound into data groups, resulting in a total of several data groups. These data groups are then used to train the neural network to be trained. Once the training is complete, a power transformer fault early warning model is obtained. This model is then used to predict power transformer faults. The output of the power transformer fault early warning model includes either the fault type or no fault. The real-time carbon emission data of the target power transformer to be predicted is obtained and input into the power transformer fault early warning model for prediction. If a fault exists, then based on the predicted fault type, determine the possible non-fault factors of that fault type; Obtain the transformer operating parameters corresponding to the possible non-fault factor, and determine whether the possible non-fault factor is true based on the corresponding transformer operating parameters; If the corresponding possible non-fault factor is true, then both the possible non-fault factor and the predicted fault type are sent to the target client; otherwise, the predicted fault type is sent to the target client.

[0007] Furthermore, the types of faults include: Local overheating of the iron core, partial discharge of the winding, overheating of the oil-impregnated paper, short circuit fault in the winding, poor contact of the tap changer, multiple grounding of the iron core, and short circuit between winding turns.

[0008] Furthermore, if a fault exists, based on the predicted fault type, possible non-fault factors for that fault type are determined, including: When the predicted fault type is a winding short-circuit fault, the possible non-fault factors for this fault type include oil-impregnated paper insulation, transformer oil oxidation and aging, and aging of the inter-turn insulation varnish layer of the winding conductors. When the predicted fault type is multi-point grounding of the iron core, the possible non-fault factors for this fault type include aging of the insulating pad, aging of the oxide layer of the insulating varnish between iron cores, and aging of the insulating bushing.

[0009] Furthermore, based on the corresponding transformer operating parameters, it is determined whether the possible non-fault factor is true, including: When the predicted fault type is a winding short-circuit fault, obtain content, content, Content, water content, and media loss factor; according to content, content, Based on the content, water content, and media loss factor, determine whether the corresponding non-fault factors are true, including:

[0010] in, This represents the aging probability corresponding to a winding short-circuit fault. All are weights. For normalized content, For normalized After content normalization, The normalized water content, This is the normalized mass loss factor. For normalized content, like If the probability exceeds a preset probability threshold, then the possible non-fault factors corresponding to the winding short-circuit fault are true; otherwise, they are not true.

[0011] Furthermore, based on the corresponding transformer operating parameters, it is determined whether the possible non-fault factor is true, including: When the predicted fault type is multi-point grounding of the iron core, obtain the iron core grounding current, content, Content and water content; According to the core grounding current, content, The content and water content are used to determine whether the corresponding non-fault factors are true, including:

[0012] in, The aging probability corresponding to multi-point grounding of the iron core. This is the normalized core grounding current. All are weights; when If the probability exceeds the preset probability threshold, then the possible non-fault factors corresponding to the multi-point grounding of the iron core are true; otherwise, they are not true.

[0013] Furthermore, indirect carbon emissions are identified, including: Based on the no-load loss and load loss of the power transformer, the power loss of the power transformer is determined, including:

[0014] in, For power loss, For rated iron loss, For runtime, For rated copper loss, This represents the maximum load loss duration during the operating time. Load rate; Based on power loss, the indirect carbon emissions of power transformers are determined, including:

[0015] in, Indirect carbon emissions from power transformers, This is a carbon emission factor for the power grid.

[0016] Furthermore, direct carbon emissions are identified, including:

[0017] in, For direct carbon emissions, For runtime Carbon dioxide emissions within the region. For runtime The amount of methane emitted within the country.

[0018] Secondly, the present invention provides a power transformer fault early warning system based on big data processing, comprising: The carbon emission acquisition module is used to acquire the fault types of several power transformers in a historical period, as well as the historical carbon emission data of the historical period in which the faults of that type occurred. The power transformers all belong to the same model, and the carbon emission data includes direct carbon emission data and indirect carbon emission data. The model training module is used to bind the fault type and the corresponding historical carbon emission data into a data group, resulting in a total of several data groups, which are then used to train the neural network to be trained. The model output module is used to obtain the power transformer fault early warning model after training is completed, and to use the power transformer fault early warning model for power transformer fault prediction. The output of the power transformer fault early warning model includes fault type or no fault. The prediction module is used to acquire real-time carbon emission data of the target power transformer to be predicted, and input the real-time carbon emission data into the power transformer fault early warning model for prediction. The factor acquisition module is used to determine the possible non-fault factors of the fault type based on the predicted fault type if a fault exists. The judgment module is used to obtain the transformer operating parameters corresponding to the possible non-fault factor, and to determine whether the possible non-fault factor is true based on the corresponding transformer operating parameters; The result output module is used to send both the possible non-fault factors and the predicted fault type to the target client if the corresponding possible non-fault factors are true; otherwise, it sends the predicted fault type to the target client.

[0019] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention achieves precise tracing of the root cause of transformer faults by subdividing fault causes into possible non-fault factors (i.e., aging-related) and possible fault factors (i.e., accidents). After identifying the fault type, it analyzes the corresponding component aging probability and performs quantitative verification in conjunction with key operating parameters, effectively distinguishing between natural equipment lifespan degradation and sudden anomalies. This not only avoids misjudging performance degradation caused by aging as a serious fault, reducing unnecessary downtime for maintenance, but also provides early warning of potential aging risks, supporting preventative maintenance. Furthermore, by establishing an aging probability model, the objectivity and scientific rigor of the judgment are enhanced, improving the intelligence level of transformer condition assessment, which helps extend equipment lifespan and ensure the reliability and economy of power system operation.

[0020] This invention uses total carbon emissions as the core indicator, combining six parameters, including direct and indirect carbon emissions, to evaluate transformer fault types, effectively overcoming the limitations of relying on a single parameter. Total carbon emissions are an integrated indicator of direct and indirect carbon emissions, simultaneously capturing the dual impact of faults on losses and gas production. The characteristics of sudden and sustained increases in total carbon emissions visually present the combined impact of faults, avoiding missed diagnoses due to reliance on a single parameter. Furthermore, total carbon emissions have a leading role in fault location: when a transformer fails, it will push total carbon emissions off the normal threshold through direct or indirect paths. Maintenance personnel can quickly identify faulty equipment through abnormal fluctuations in total carbon emissions, and then refine the fault type by combining other parameters, significantly reducing the complexity of analyzing multiple parameters one by one. In addition, total carbon emissions also link fault diagnosis and low-carbon management needs. While identifying faults, the additional impact of faults on carbon emissions can be simultaneously understood, providing a basis for subsequent emission reduction optimization after fault repair, achieving synergy between fault management and low-carbon goals. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart illustrating the power transformer fault early warning method based on big data processing provided by this invention. Detailed Implementation

[0023] This invention provides a power transformer fault early warning method based on big data processing, which solves the technical problem of how to use environmental indicators such as carbon emissions to predict transformer faults in the prior art.

[0024] The technical solution of this invention is to solve the above-mentioned technical problems, and the overall idea is as follows: A power transformer fault early warning method based on big data processing includes: acquiring fault types of several power transformers within a historical time period, and historical carbon emission data within the historical time period in which the faults of that type occurred, wherein all power transformers belong to the same model, and the carbon emission data includes direct carbon emission data and indirect carbon emission data; binding the fault type and the corresponding historical carbon emission data into data groups, resulting in several data groups, and using these data groups to train a neural network; after training, obtaining a power transformer fault early warning model, and using the power transformer fault early warning model for power transformer fault prediction, the output of the power transformer fault early warning model including fault type or no fault; acquiring real-time carbon emission data of the target power transformer to be predicted, and inputting the real-time carbon emission data into the power transformer fault early warning model for prediction; if a fault exists, determining the possible non-fault factors of the fault type based on the predicted fault type; acquiring the transformer operating parameters corresponding to the possible non-fault factors, and determining whether the possible non-fault factors are true based on the corresponding transformer operating parameters; if the corresponding possible non-fault factors are true, sending both the possible non-fault factors and the predicted fault type to the target client; otherwise, sending the predicted fault type to the target client.

[0025] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0026] First, it should be clarified that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0027] Direct carbon emissions during transformer operation refer to the carbon emissions corresponding to greenhouse gases directly generated and emitted by the transformer under fault conditions such as partial discharge or overheating. The emitted greenhouse gases only include two types: carbon dioxide and methane. Indirect carbon emissions during transformer operation refer to the greenhouse gas emissions generated during the production process of transformers due to their own power loss during normal operation.

[0028] The quantitative basis for indirect carbon emissions is the power loss of transformers over a specified period of time, which consists of two types: no-load loss and load loss. No-load loss (iron loss): refers to the energy loss of the iron core due to the change of magnetic field when the transformer is energized but without load (or with extremely low load). Specifically, it includes the hysteresis loss and eddy current loss of the iron core. Load loss refers to the energy loss generated by the current passing through the windings when the transformer is running under load. It mainly includes the resistance loss and leakage flux loss of the windings.

[0029] This invention provides, for example Figure 1 The power transformer fault early warning method based on big data processing shown includes steps S11-S17: Step S11: Obtain the fault types of several power transformers within a historical time period, as well as the historical carbon emission data of the historical time period in which the faults of that type occurred. The power transformers are all of the same model, and the carbon emission data includes direct carbon emission data and indirect carbon emission data.

[0030] Specifically: Collect fault records of several power transformers that occurred in the past time period (including fault type and duration of occurrence). Then, for each fault and for each type of fault, carbon emission data related to the transformer or the power system in which it is located is obtained.

[0031] Carbon emission data includes seven parameters: total carbon emissions, direct carbon emissions, indirect carbon emissions, no-load losses, load losses, carbon dioxide emissions, and methane emissions.

[0032] Fault types can include: Local overheating of the iron core, partial discharge of the winding, overheating of the oil-impregnated paper, short circuit fault in the winding, poor contact of the tap changer, multiple grounding of the iron core, and short circuit between winding turns.

[0033] The following is a correlation between fault types and some carbon emission data obtained by the inventor based on a certain type of power transformer (not all 7 parameters are listed, only some parameters are used as examples). The degree of correlation may vary for different types of power transformers, but the general performance is similar.

[0034] It should be emphasized that both the increase and the magnitude of change are compared with the normal operation period, and the time periods are all equally divided. For example, if the iron core overheats for 1 minute, the iron core overheating can be divided into 12 5-second intervals. Comparing each of the 12 time periods with the normal operation period, a total of 12 data sets can be obtained, as described below. The length of the time period can be determined according to the actual situation, and there is no restriction here.

[0035] Localized overheating of the iron core affects no-load loss, carbon dioxide emissions, and methane emissions. Increased no-load loss (>10% of normal threshold) leads to a 25%-35% increase in carbon dioxide emissions and a 5%-8% increase in methane emissions. Partial discharge in windings: slight fluctuations in load losses (within ±5%), methane emissions increase by 30%-50%, and carbon dioxide emissions increase by <2%; Overheating of oil-impregnated paper: both no-load loss and load loss increase (>8% of normal threshold); carbon dioxide emissions and methane emissions increase by 15%-25%; Winding short circuit fault: Load loss increases sharply (>50% of normal threshold), and the increase in carbon dioxide emissions and methane emissions both exceed 100% of normal threshold; Poor contact of tap changer: abnormally high load loss (>15% of normal threshold), no-load loss change within ±1%, carbon dioxide emission increase of 10%-20%, methane increase of 5%-10%; The no-load loss of the iron core under multi-point grounding increased significantly (>20% of the normal threshold), carbon dioxide emissions increased by 8%-15%, and methane emissions were trace (increase <5%). The short-circuit load loss between winding turns increased slightly (>5% and <15% of the normal threshold), carbon dioxide emissions increased by 5%-12%, and methane emissions increased by 3%-8%.

[0036] Table 1 shows the increase in total carbon emissions: Table 1

[0037] It should be noted that when the above types of failures occur, the changes in no-load loss, load loss, carbon dioxide emissions, and methane emissions are all different. In other words, this directly affects the corresponding direct carbon emissions, indirect carbon emissions, and total carbon emissions.

[0038] This invention uses total carbon emissions as the core indicator, combining six parameters, including direct and indirect carbon emissions, to evaluate transformer fault types, effectively overcoming the limitations of relying on a single parameter. Total carbon emissions are an integrated indicator of direct and indirect carbon emissions, simultaneously capturing the dual impact of faults on losses and gas production. The characteristics of sudden and sustained increases in total carbon emissions visually present the combined impact of faults, avoiding missed diagnoses due to reliance on a single parameter. Furthermore, total carbon emissions have a leading role in fault location: when a transformer fails, it will push total carbon emissions off the normal threshold through direct or indirect paths. Maintenance personnel can quickly identify faulty equipment through abnormal fluctuations in total carbon emissions, and then refine the fault type by combining other parameters, significantly reducing the complexity of analyzing multiple parameters one by one. In addition, total carbon emissions also link fault diagnosis and low-carbon management needs. While identifying faults, the additional impact of faults on carbon emissions can be simultaneously understood, providing a basis for subsequent emission reduction optimization after fault repair, achieving synergy between fault management and low-carbon goals.

[0039] Determine indirect carbon emissions, including: Based on the no-load loss and load loss of the power transformer, the power loss of the power transformer is determined, including:

[0040] in, For power loss, For rated iron loss, For runtime, For rated copper loss, This represents the maximum load loss duration during the operating time. Load rate; Based on power loss, the indirect carbon emissions of power transformers are determined, including:

[0041] in, Indirect carbon emissions from power transformers, This is a carbon emission factor for the power grid.

[0042] Identify direct carbon emissions, including:

[0043] in, For direct carbon emissions, For runtime Carbon dioxide emissions within the region. For runtime The amount of methane emitted within the country.

[0044] In addition to the direct carbon emission method provided by this invention, patent document 202411892177.7 also provides a method for calculating direct carbon emissions: for direct carbon emissions, determining the amount of gas generated by a transformer fault within a specified time period, including: establishing a multiple linear regression model for direct carbon emissions, simplifying the multiple linear regression model, using the least squares method to estimate the simplified multiple linear regression model, and predicting the CO2 and CH4 emissions of the transformer within the specified time period. This invention does not limit the calculation method.

[0045] Total carbon emissions are and sum.

[0046] Step S12: Bind the fault type and the corresponding historical carbon emission data into a data group, resulting in a total of several data groups, and use these data groups to train the neural network to be trained.

[0047] The historical carbon emission data (including total carbon emissions, direct carbon emissions, indirect carbon emissions, no-load losses, load losses, carbon dioxide emissions, and methane emissions) corresponding to each transformer fault type and its occurrence period are spatiotemporally aligned and bound into data groups to form a training dataset containing several such data groups. The data groups are then input into the neural network to be trained, and the network weights are continuously adjusted through algorithms such as backpropagation, so that the model learns the nonlinear mapping relationship between different fault types and carbon emission characteristics, thereby completing the training of the fault type identification model.

[0048] Step S13: After training is completed, a power transformer fault early warning model is obtained, and the power transformer fault early warning model is used for power transformer fault prediction. The output of the power transformer fault early warning model includes fault type or no fault.

[0049] If an output fault type exists, the target power transformer is considered to be faulty; otherwise, there is no fault.

[0050] Step S14: Obtain real-time carbon emission data of the target power transformer to be predicted, and input the real-time carbon emission data into the power transformer fault early warning model for prediction.

[0051] The real-time carbon emission data of the target power transformer to be predicted is input into the power transformer fault early warning model at a preset frequency. For example, the real-time carbon emission data within 5 seconds is input every 5 seconds. The real-time carbon emission data also includes total carbon emissions, direct carbon emissions, indirect carbon emissions, no-load losses, load losses, carbon dioxide emissions, and methane emissions.

[0052] Step S15: If a fault exists, determine the possible non-fault factors of the fault type based on the predicted fault type.

[0053] Specifically, non-failure factors refer to transformer failures caused by the aging of components in the transformer (in essence, identifying the aging components that may cause the corresponding failure type); potential failure factors are transformer failures caused by other unexpected factors besides transformer aging.

[0054] In other words, this application classifies the causes of failure into those caused by aging (which may not be a failure factor) and those caused by other accidents (which may be failure factors).

[0055] Taking winding short-circuit faults and multi-point grounding of the iron core as examples: If a fault exists, based on the predicted fault type, determine the possible non-fault factors for that fault type, including: When the predicted fault type is a winding short-circuit fault, the possible non-fault factors for this fault type include oil-impregnated paper insulation, transformer oil oxidation and aging, and aging of the inter-turn insulation varnish layer of the winding conductors. When the predicted fault type is multi-point grounding of the iron core, the possible non-fault factors for this fault type include aging of the insulating pad, aging of the oxide layer of the insulating varnish between iron cores, and aging of the insulating bushing.

[0056] The purpose of step S15 is to determine the possible non-fault factors corresponding to the fault type. It is understood that when a transformer fails, it is not necessarily caused by possible non-fault factors, but may also be caused by other unexpected factors.

[0057] Possible non-fault factors corresponding to localized overheating of the iron core include: aging of the oxide layer of the insulating varnish between iron core pieces, moisture or embrittlement and aging of the insulating pads, abnormal leakage flux caused by aging of the fastener insulation, and oxidation and corrosion of the contact parts of the iron core grounding plate. Possible non-fault factors corresponding to partial discharge in windings include: local deterioration of oil-impregnated paper insulation, aging and cracking of the varnish layer between winding conductor turns, aging of transformer oil, and aging of insulating paperboard or support strips due to moisture. Possible non-fault factors corresponding to overheating of oil-impregnated paper include: long-term thermal stress of oil-impregnated paper, oxidation of transformer oil, aging of oil sludge deposits, carbonization or delamination of insulating paper at high temperatures, etc. Possible non-fault factors for poor contact in tap changers include: carbonized deposits of insulating oil on the contact surface, aging and deformation of the insulating pull rod of the operating mechanism, fatigue of the switch contact spring, and aging of the oil chamber seals of the switching switch. Possible non-fault factors corresponding to inter-turn short circuits in windings include: long-term thermal aging and peeling of the inter-turn insulation varnish layer, decreased mechanical strength of oil-impregnated paper insulation, increased acid value of transformer oil corroding conductor insulation, and insulation wear caused by winding vibration.

[0058] Step S16: Obtain the transformer operating parameters corresponding to the possible non-fault factor, and determine whether the possible non-fault factor is true based on the corresponding transformer operating parameters.

[0059] After identifying potential non-fault factors (i.e., identifying the corresponding potential aging devices), determine the corresponding transformer operating parameters.

[0060] Taking winding short-circuit faults and multiple grounding points of the iron core as examples: When the predicted fault type is a winding short-circuit fault, obtain content, content, The content, water content, and dielectric loss factor (i.e., whether aging phenomena such as oil-impregnated paper insulation, transformer oil oxidation aging, and inter-turn insulation varnish aging of winding conductors have occurred) can be determined by... content, content, (Content, water content, and media loss factor are determined). according to content, content, Based on the content, water content, and media loss factor, determine whether the corresponding non-fault factors are true, including:

[0061] in, This represents the aging probability corresponding to a winding short-circuit fault. All are weights. For normalized content, For normalized After content normalization, The normalized water content, This is the normalized mass loss factor. For normalized content, like If the probability exceeds a preset probability threshold, then the possible non-fault factors corresponding to the winding short-circuit fault are true; otherwise, they are not true.

[0062] When the predicted fault type is multi-point grounding of the iron core, obtain the iron core grounding current, content, The content and water content (i.e., whether aging of insulating pads, aging of the oxide layer of insulating varnish between iron cores, and aging of insulating bushings have occurred, can be determined by the iron core grounding current, content, (Content and water content determined). According to the core grounding current, content, The content and water content are used to determine whether the corresponding non-fault factors are true, including:

[0063] in, The aging probability corresponding to multi-point grounding of the iron core. This is the normalized core grounding current. All are weights; when If the probability exceeds the preset probability threshold, then the possible non-fault factors corresponding to the multi-point grounding of the iron core are true; otherwise, they are not true.

[0064] The weights and preset probability thresholds can be determined according to the actual situation.

[0065] Understandably, issues such as localized overheating of the core, partial discharge of the winding, overheating of the oil-impregnated paper, poor contact of the tap changer, and short circuits between winding turns can all be addressed by using the same logic as for "multiple grounding of the core" and "short circuit faults in the winding," thus determining the necessary transformer operating parameters.

[0066] This invention achieves precise tracing of the root cause of transformer faults by subdividing fault causes into possible non-fault factors (i.e., aging-related) and possible fault factors (i.e., accidents). After identifying the fault type, it analyzes the corresponding component aging probability and performs quantitative verification in conjunction with key operating parameters, effectively distinguishing between natural equipment lifespan degradation and sudden anomalies. This not only avoids misjudging performance degradation caused by aging as a serious fault, reducing unnecessary downtime for maintenance, but also provides early warning of potential aging risks, supporting preventative maintenance. Furthermore, by establishing an aging probability model, the objectivity and scientific rigor of the judgment are enhanced, improving the intelligence level of transformer condition assessment, which helps extend equipment lifespan and ensure the reliability and economy of power system operation.

[0067] Step S17: If the corresponding possible non-fault factor is true, then send the possible non-fault factor and the predicted fault type to the target client; otherwise, send the predicted fault type to the target client.

[0068] If the corresponding possible non-fault factors are true, it indicates that the fault is likely due to component aging, and maintenance personnel can replace the components accordingly. If not, then aging-related faults can be ruled out, and the fault should be repaired as an accidental cause.

[0069] Based on the same inventive concept, this invention provides a power transformer fault early warning system based on big data processing, comprising: The carbon emission acquisition module is used to acquire the fault types of several power transformers in a historical period, as well as the historical carbon emission data of the historical period in which the faults of that type occurred. The power transformers all belong to the same model, and the carbon emission data includes direct carbon emission data and indirect carbon emission data. The model training module is used to bind the fault type and the corresponding historical carbon emission data into a data group, resulting in a total of several data groups, which are then used to train the neural network to be trained. The model output module is used to obtain the power transformer fault early warning model after training is completed, and to use the power transformer fault early warning model for power transformer fault prediction. The output of the power transformer fault early warning model includes fault type or no fault. The prediction module is used to acquire real-time carbon emission data of the target power transformer to be predicted, and input the real-time carbon emission data into the power transformer fault early warning model for prediction. The factor acquisition module is used to determine the possible non-fault factors of the fault type based on the predicted fault type if a fault exists. The judgment module is used to obtain the transformer operating parameters corresponding to the possible non-fault factor, and to determine whether the possible non-fault factor is true based on the corresponding transformer operating parameters; The result output module is used to send both the possible non-fault factors and the predicted fault type to the target client if the corresponding possible non-fault factors are true; otherwise, it sends the predicted fault type to the target client.

[0070] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0071] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A power transformer fault early warning method based on big data processing, characterized in that, include: The system acquires the fault types of several power transformers within a historical time period, as well as the historical carbon emission data for the historical time period in which the faults of that type occurred. All power transformers belong to the same model, and the carbon emission data includes both direct and indirect carbon emission data. The fault type and the corresponding historical carbon emission data are bound into data groups, resulting in a total of several data groups. These data groups are then used to train the neural network to be trained. Once the training is complete, a power transformer fault early warning model is obtained, and the power transformer fault early warning model is used for power transformer fault prediction. The output of the power transformer fault early warning model includes fault type or no fault. The real-time carbon emission data of the target power transformer to be predicted is obtained, and the real-time carbon emission data is input into the power transformer fault early warning model for prediction. If a fault exists, then based on the predicted fault type, determine the possible non-fault factors of that fault type; Obtain the transformer operating parameters corresponding to the possible non-fault factor, and determine whether the possible non-fault factor is true based on the corresponding transformer operating parameters; If the corresponding possible non-fault factor is true, then both the possible non-fault factor and the predicted fault type are sent to the target client; otherwise, the predicted fault type is sent to the target client.

2. The power transformer fault early warning method based on big data processing as described in claim 1, characterized in that, Fault types include: Local overheating of the iron core, partial discharge of the winding, overheating of the oil-impregnated paper, short circuit fault in the winding, poor contact of the tap changer, multiple grounding of the iron core, and short circuit between winding turns.

3. The power transformer fault early warning method based on big data processing as described in claim 2, characterized in that, If a fault exists, based on the predicted fault type, determine the possible non-fault factors for that fault type, including: When the predicted fault type is a winding short-circuit fault, the possible non-fault factors for this fault type include oil-impregnated paper insulation, transformer oil oxidation and aging, and aging of the inter-turn insulation varnish layer of the winding conductors. When the predicted fault type is multi-point grounding of the iron core, the possible non-fault factors for this fault type include aging of the insulating pad, aging of the oxide layer of the insulating varnish between iron cores, and aging of the insulating bushing.

4. The power transformer fault early warning method based on big data processing as described in claim 2, characterized in that, Determine whether the possible non-fault factor is true based on the corresponding transformer operating parameters, including: When the predicted fault type is a winding short-circuit fault, obtain content, content, Content, water content, and media loss factor; according to content, content, Based on the content, water content, and media loss factor, determine whether the corresponding non-fault factors are true, including: in, This represents the aging probability corresponding to a winding short-circuit fault. All are weights. For normalized content, For normalized After content normalization, The normalized water content, This is the normalized mass loss factor. For normalized content, like If the probability exceeds a preset probability threshold, then the possible non-fault factors corresponding to the winding short-circuit fault are true; otherwise, they are not true.

5. The power transformer fault early warning method based on big data processing as described in claim 4, characterized in that, Determine whether the possible non-fault factor is true based on the corresponding transformer operating parameters, including: When the predicted fault type is multi-point grounding of the iron core, obtain the iron core grounding current, content, Content and water content; According to the iron core grounding current, content, The content and water content are used to determine whether the corresponding non-fault factors are true, including: in, The aging probability corresponding to multi-point grounding of the iron core. This is the normalized core grounding current. All are weights; when If the probability exceeds the preset probability threshold, then the possible non-fault factors corresponding to the multi-point grounding of the iron core are true; otherwise, they are not true.

6. The power transformer fault early warning method based on big data processing as described in claim 1, characterized in that, Determine indirect carbon emissions, including: Based on the no-load loss and load loss of the power transformer, the power loss of the power transformer is determined, including: in, For power loss, For rated iron loss, For runtime, For rated copper loss, This represents the maximum load loss duration during the operating time. Load rate; Based on power loss, the indirect carbon emissions of power transformers are determined, including: in, Indirect carbon emissions from power transformers, This is a carbon emission factor for the power grid.

7. The power transformer fault early warning method based on big data processing as described in claim 1, characterized in that, Identify direct carbon emissions, including: in, For direct carbon emissions, For runtime Carbon dioxide emissions within the region. For runtime The amount of methane emitted within the area.

8. A power transformer fault early warning system based on big data processing, characterized in that, include: The carbon emission acquisition module is used to acquire the fault types of several power transformers in a historical period, as well as the historical carbon emission data of the historical period in which the faults of that type occurred. The power transformers all belong to the same model, and the carbon emission data includes direct carbon emission data and indirect carbon emission data. The model training module is used to bind the fault type and the corresponding historical carbon emission data into a data group, resulting in a total of several data groups, which are then used to train the neural network to be trained. The model output module is used to obtain a power transformer fault early warning model after training is completed, and to use the power transformer fault early warning model for power transformer fault prediction. The output of the power transformer fault early warning model includes fault type or no fault. The prediction module is used to acquire real-time carbon emission data of the target power transformer to be predicted, and input the real-time carbon emission data into the power transformer fault early warning model for prediction; The factor acquisition module is used to determine the possible non-fault factors of the fault type based on the predicted fault type if a fault exists. The judgment module is used to obtain the transformer operating parameters corresponding to the possible non-fault factor, and to determine whether the possible non-fault factor is true based on the corresponding transformer operating parameters; The result output module is used to send both the possible non-fault factors and the predicted fault type to the target client if the corresponding possible non-fault factors are true; otherwise, it sends the predicted fault type to the target client.

Images