Detection device for paper oil insulation discharge fault of power transformer

Abstract

The embodiment of the invention provides a power transformer oil paper insulation discharge fault detection device, which comprises a characteristic parameter acquisition unit and a calculation unit, and is characterized in that the characteristic parameter acquisition unit is used for acquiring target electric energy data and oil sample dissolved gas component data of an oil product in a target transformer; the electric energy data and oil sample dissolved gas component data are transmitted to a calculation unit, and the oil sample dissolved gas component data comprise component data of hydrogen, acetylene, total hydrocarbon, carbon monoxide and carbon dioxide; the calculation unit is used for importing the target electric energy data and the oil sample dissolved gas component data into a trained fault recognition model to generate a fault type result, and the fault recognition model is obtained by training a preset power transformer oil paper insulation discharge fault experiment system. According to the embodiment of the invention, the reliability and real-time performance of fault diagnosis are remarkably improved, the health state of the power transformer can be monitored in time, and early warning and processing can be carried out in time.

Description

Technical Field

[0001] The embodiments of this specification relate to the technical field of power transformer discharge fault detection, and in particular to a detection device for oil-paper insulation discharge fault of a power transformer. Background Art

[0002] As a key equipment in the power transmission and distribution system, the stability and reliability of power transformers are directly related to the safe operation of the entire power grid. The oil-paper insulation system is one of the most widely used insulation structures inside power transformers, which can provide electrical insulation, thermal conduction and mechanical support. However, during long-term operation, due to the combined effects of electrical, thermal, chemical and mechanical stresses, the oil-paper insulation material will gradually age, which may lead to phenomena such as partial discharge, spark discharge and even arc discharge, thus affecting the normal operation of the transformer and even causing serious failures, resulting in major safety hazards.

[0003] Traditional methods for evaluating the health status of oil-paper insulation systems mainly include dissolved gas analysis in oil, electrical testing, and physical inspections. Although these methods can reflect the insulation status to a certain extent, they are usually inefficient and cannot achieve real-time monitoring. Moreover, for early weak partial discharge signals, the sensitivity and accuracy of traditional detection methods are often insufficient, making it difficult to accurately capture fault signs, determine the fault type, and issue early warnings. Summary of the invention

[0004] In view of this, the embodiments of this specification provide a detection device for oil-paper insulation discharge fault of a power transformer to solve the technical defects existing in the prior art.

[0005] According to an embodiment of the present specification, a detection device for oil-paper insulation discharge fault of a power transformer is provided, comprising: a characteristic parameter acquisition unit and a calculation unit, wherein: The characteristic parameter acquisition unit is used to collect target power data and oil sample dissolved gas component data of the oil in the target transformer, and transmit the power data and oil sample dissolved gas component data to the calculation unit, wherein the oil sample dissolved gas component data includes component data of hydrogen, acetylene, total hydrocarbons, carbon monoxide and carbon dioxide; The calculation unit is used to import the target electric energy data and the component data of hydrogen, acetylene, total hydrocarbons, carbon monoxide and carbon dioxide into a trained fault identification model to generate a fault type result, wherein the fault identification model is trained by a preset power transformer oil-paper insulation discharge fault experimental system.

[0006] In some embodiments, the detection device for oil-paper insulation discharge fault of a power transformer also includes a preprocessing unit, which is configured to obtain a real-time voltage signal and a real-time current signal of the oil in the target transformer; clean the real-time voltage signal and the real-time current signal; and calculate the target electric energy data based on the cleaned voltage signal and the current signal.

[0007] In some embodiments, the power transformer oil-paper insulation discharge fault experimental system includes an experimental tank, a needle plate electrode model, a measuring device and a detection device, wherein: The needle-plate electrode model is arranged in the experimental tank, an insulating paperboard sample is arranged between the needle electrode and the plate electrode of the needle-plate electrode model, and the needle-plate electrode model is used to simulate a three-level experiment, which includes a partial discharge experiment, a spark discharge experiment and an arc discharge experiment; The measuring device includes a voltage measuring instrument, a current measuring sensor and an oil sample gas analyzer, wherein the voltage measuring instrument is used to measure the voltage signal in the oil sample during the discharge process in the three-level experiment, the current measuring sensor is used to measure the current signal during the discharge process in the three-level experiment, and the oil sample gas analyzer is used to measure the dissolved gas components of the oil sample in the three-level experiment; The detection device is provided with an experiment collection unit and a training unit: The experiment acquisition unit is connected to the measuring device and the computing unit respectively, and is used to receive the voltage signal, current signal and oil sample dissolved gas components collected by the voltage measuring instrument, current measuring sensor and oil sample gas analyzer for the three-level experiment, and transmit the voltage signal, current signal and oil sample dissolved gas components to the training unit; The training unit is used to calculate the electric energy data of the three-level experiment based on the voltage signal and the current signal; import the electric energy data and the dissolved gas components of the oil sample into a preset initial judgment model for iterative training until the loss of the initial judgment model meets the preset requirements, thereby obtaining the trained fault detection model.

[0008] In some embodiments, during the partial discharge experiment, The conditions for collecting voltage signals and current signals include: increasing the needle-plate voltage between the needle electrode and the plate electrode to m times (m>1) the local discharge starting voltage, and starting to collect voltage signals and current signals after the needle-plate electrode model breaks through the insulating paperboard sample set therein; The conditions for collecting oil samples include: when the voltage signal and the current signal are collected, the first collection is started, and the oil sample is collected again after a preset first time period.

[0009] In some embodiments, during the spark discharge experiment, The conditions for collecting voltage signals and current signals include: increasing the needle-plate voltage between the needle electrode and the plate electrode uniformly from zero at a preset speed until the insulating paperboard sample is broken down, and then starting to collect voltage signals and current signals after cutting off the power supply; The conditions for collecting the oil sample include: when the insulating paperboard sample is punctured and the gas in the oil sample reaches equilibrium, the oil sample is collected.

[0010] In some embodiments, during the arc discharge experiment, The conditions for collecting the voltage signal and the current signal include: boosting the needle-plate voltage between the needle electrode and the plate electrode until the insulating paperboard sample is broken down and a stable penetrating discharge channel is formed between the needle electrode and the plate electrode, and then cutting off the power supply after the second time period; The conditions for collecting the oil sample include: when the insulating paperboard sample is punctured and the gas in the oil sample reaches equilibrium, the oil sample is collected.

[0011] In some embodiments, calculating the electric energy data of the partial discharge experiment according to the voltage signal and the current signal includes: Calculating a first voltage peak value according to the voltage signal; Calculating the discharge phase, partial discharge charge and discharge times according to the current signal; The electric energy data of the partial discharge experiment is calculated according to the voltage peak value, the discharge phase, the partial discharge charge and the discharge times, and a preset first calculation formula, wherein the first calculation formula includes:

[0012] in, W PD Represents the electrical energy data of the partial discharge experiment, U PD represents the first voltage peak, q i Indicates the amount of partial discharge charge, n Indicates the number of discharges, Indicates the discharge phase.

[0013] In some embodiments, calculating the electric energy data of the spark discharge experiment / arc discharge experiment according to the voltage signal and the current signal includes: Calculating a second voltage peak value according to the voltage signal; Calculating the current peak value and duration according to the current signal; The electric energy data of the spark discharge experiment / arc discharge experiment is calculated according to the voltage peak value, current peak value and duration, and a preset second calculation formula, wherein the second calculation formula includes:

[0014] in, W bd Indicates the electrical energy data of the spark discharge experiment / arc discharge experiment, U bd represents the second voltage peak, T Indicates duration, i bd Indicates the peak current.

[0015] In some embodiments, the fault detection model is a back propagation neural network.

[0016] In some embodiments, the experimental tank body is also provided with an oil valve for taking out and filling oil, a pressure release valve for relieving pressure, and a pressure gauge for detecting the air pressure in the experimental tank body.

[0017] At least one embodiment of the embodiments of this specification determines the fault type based on the electric energy during the discharge process and the analysis data of dissolved gas in the oil, which significantly improves the reliability and real-time performance of fault diagnosis, can timely monitor the health status of the power transformer, and provide timely warning and processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a simple structural diagram of a detection device for oil-paper insulation discharge fault of a power transformer provided by some embodiments of this specification; Figure 2 It is a simple structural diagram of a detection device for oil-paper insulation discharge fault of a power transformer provided by some embodiments of this specification; Figure 3 It is a schematic diagram of the structure of a back-propagation neural network provided in some embodiments of this specification. DETAILED DESCRIPTION

[0019] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0020] The terms used in one or more embodiments of this specification are for the purpose of describing specific embodiments only and are not intended to limit one or more embodiments of this specification. The singular forms of "one" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items. The modifications of "one" and "multiple" mentioned in this disclosure are illustrative and non-restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0021] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0022] As a key equipment in the power transmission and distribution system, the stability and reliability of power transformers are directly related to the safe operation of the entire power grid. The oil-paper insulation system is one of the most widely used insulation structures inside power transformers, which can provide electrical insulation, heat conduction and mechanical support. However, during long-term operation, due to the combined effects of electrical, thermal, chemical and mechanical stresses, the oil-paper insulation material may reduce the mechanical strength and insulation performance of the oil-paper insulation system due to natural aging, oil quality deterioration, environmental abnormalities, etc., and even cause local melting in severe cases, resulting in gradual failure of the oil-paper insulation system and ultimately causing transformer failure. The fault process often evolves gradually. According to the severity of the fault, it can be divided into partial discharge stage, spark discharge stage and arc discharge stage from light to heavy. Different corresponding means can be used to deal with different fault stages. Considering economic factors, the oil-paper insulation system is generally not replaced in the partial discharge stage.

[0023] To solve the above problems, see Figure 1 , Figure 1A power transformer oil-paper insulation discharge fault device provided according to some embodiments of the present specification is shown, specifically comprising: a characteristic parameter acquisition unit and a calculation unit, wherein the characteristic parameter acquisition unit is used to collect target electric energy data of the oil in the target transformer and oil sample dissolved gas component data, and transmit the electric energy data and oil sample dissolved gas component data to the calculation unit, wherein the oil sample dissolved gas component data includes component data of hydrogen, acetylene, total hydrocarbons, carbon monoxide and carbon dioxide; the calculation unit is used to import the target electric energy data and the component data of hydrogen, acetylene, total hydrocarbons, carbon monoxide and carbon dioxide into a trained fault identification model to generate a fault type result, wherein the fault identification model is trained by a preset power transformer oil-paper insulation discharge fault experimental system.

[0024] In some optional implementations, a mapping relationship between electric energy and component data of hydrogen, acetylene, total hydrocarbons, carbon monoxide, and carbon dioxide can be established, and the parameters of the input fault identification model only need the component data of hydrogen, acetylene, total hydrocarbons, carbon monoxide, and carbon dioxide. Although the calculation accuracy of this setting is relatively biased, its input data is relatively simpler (no electric energy input is required), so it can be used as a backup option.

[0025] The target transformer may refer to the transformer to be detected. Due to the different specifications of different transformers, it is necessary to set corresponding different processing parameters. For example, the voltage of the transformer can usually be different specifications such as 110V, 500V, etc. When processing transformers of different specifications, the processing parameters are often different. For another example, even if the specifications of the transformers are the same, due to different brands, materials, operating environments, etc., the data or parameters corresponding to the faults of different transformers are also different. Therefore, it is necessary to first determine the transformer to be detected, and then monitor and process the data. The characteristic parameter acquisition unit may refer to a processing unit that collects the target electric energy data and the dissolved gas component data of the oil sample in the target transformer. In each stage of the partial discharge stage, the spark discharge stage and the arc discharge stage, electrical signals can be detected in the transformer oil due to electric energy leakage, such as the voltage signal and current signal detected from the oil-paper insulation system. The target electric energy data may refer to the discharge energy calculated based on the detected voltage signal and current signal. Insulating paper (usually cellulose material) will undergo chemical decomposition under overheating or partial discharge conditions to generate characteristic gases such as carbon monoxide, carbon dioxide, hydrogen, acetylene, methane, ethylene, etc., which will dissolve in the transformer oil. Different gas components (gas volume components) can partially correspond to different stages of the fault. Therefore, when judging the fault, the gas component data is also a relatively important data indicator. Therefore, it is necessary to collect the component data of hydrogen, acetylene, total hydrocarbons, carbon monoxide and carbon dioxide in the transformer oil sample, that is, the dissolved gas component data of the oil sample.

[0026] The fault identification model may refer to a pre-trained machine learning model. The input parameters of the machine learning model are the target electric energy data collected from the oil sample, as well as the component data of hydrogen, acetylene, total hydrocarbons, carbon monoxide and carbon dioxide. After the model calculation, the fault type result is output to determine different fault types. The fault type result can be various representations representing different results. For example, the fault type result can be a digital type result such as 0 (indicating no fault), 1 (indicating partial discharge), 2 (indicating spark discharge), 3 (indicating arc discharge); it can also be 1 (indicating partial discharge), 2 (indicating spark discharge), 3 (indicating arc discharge). Similarly, the fault type result can also be an English type result, a Chinese type result, a special character type, or a combination of the above-mentioned numbers, English, Chinese, special characters, etc. However, regardless of the type, the result at least includes an output corresponding to any one of the fault types of the partial discharge stage, the spark discharge stage and the arc discharge stage, so as to intuitively and accurately represent the fault type. The fault identification model can adopt various machine learning models that can output identification types.

[0027] The back propagation neural network can automatically calculate the gradient of the loss function to the network weights using the chain rule, avoiding the complexity of manual derivation and significantly improving the training efficiency. Therefore, in some optional implementations, the fault identification model can be a back propagation-based neural network model.

[0028] The power transformer oil-paper insulation discharge fault test system can refer to an experimental system set up according to the characteristics of the transformer oil-paper insulation system. Based on this system, different experiments can be carried out for different fault stages (partial discharge stage, spark discharge stage and arc discharge stage), oil samples can be collected and high-precision calculations can be performed, and training can be performed based on the calculated data to obtain the above-mentioned high-precision fault identification model.

[0029] At least one embodiment of the embodiments of this specification determines the fault type based on the electric energy during the discharge process and the analysis data of dissolved gas in the oil, which significantly improves the reliability and real-time performance of fault diagnosis, can timely monitor the health status of the power transformer, and provide timely warning and processing.

[0030] In some embodiments, Figure 2As shown, the power transformer oil-paper insulation discharge fault device is also provided with a preprocessing unit. It is configured to obtain the real-time voltage signal and real-time current signal of the oil in the target transformer; clean the real-time voltage signal and real-time current signal; calculate the target electric energy data according to the cleaned voltage signal and current signal. Since the voltage signal and current signal in the transformer oil sample acquired / collected in real time have noise signals, it is necessary to clean the noise signal first before performing subsequent calculations. In practical applications, since it is impossible to determine the fault type in advance, it is only necessary to calculate the peak value (voltage value) of the voltage signal after cleaning and the peak value (current value) of the current signal after cleaning, and calculate the target electric energy data according to the voltage value and the current value (that is, the product of the voltage value and the current value).

[0031] In some embodiments, the power transformer oil-paper insulation discharge fault test system includes an experimental tank, a needle-plate electrode model, a measuring device and a detection device, wherein the needle-plate electrode model is arranged in the experimental tank, and an insulating paperboard sample is arranged between the needle electrode and the plate electrode of the needle-plate electrode model, and the needle-plate electrode model is used to simulate a three-level experiment, and the three-level experiment includes a partial discharge experiment, a spark discharge experiment and an arc discharge experiment. For the convenience of observation, the material of the experimental tank can be selected from a transparent material, such as organic glass. The needle-plate electrode model can refer to a laboratory device used to simulate the local discharge phenomenon under an extremely non-uniform electric field. The needle-plate electrode model studies the breakdown, discharge and gas production rules of insulating materials under a strong electric field by combining a high-voltage electrode (needle tip) with a grounding electrode (flat plate). Exemplarily, the plate electrode in the needle-plate electrode model can be set to brass, and the needle electrode can be set to pure tungsten. As a copper-zinc alloy, brass has excellent thermal conductivity while maintaining a high conductivity (second only to pure copper), and is suitable for electrode scenarios that require rapid heat dissipation or uniform heat conduction. In addition, the zinc content of brass (usually 30%-40%) gives it good mechanical strength and wear resistance, making it particularly suitable for plate electrode applications that are subject to repeated contact or friction. Its α solid solution or (α+β) two-phase structure at room temperature further enhances the material's ability to resist deformation. The melting point of pure tungsten is as high as 3422°C, which can maintain structural stability in extremely high temperature environments and avoid melting or deformation caused by local high temperatures generated by discharge or arcing. Therefore, pure tungsten is used as the material for needle electrodes. In addition, brass can control costs as a large-area plate electrode, while the high durability of pure tungsten needle electrodes reduces maintenance frequency.

[0032] In some optional implementations, the experimental tank body is also provided with an oil valve for taking and filling oil, a pressure release valve for pressure relief, and a barometer for detecting the air pressure in the experimental tank body. The oil valve can facilitate the injection and sampling of oil samples. The pressure release valve can relieve pressure to prevent damage to the experimental equipment caused by excessive pressure in the experimental tank body. For the same reason, a pressure gauge is provided in the experimental tank body to observe in real time whether the pressure in the experimental tank body meets the standard, or whether the pressure generated in the experimental tank body is too high, so as to perform corresponding operations.

[0033] The measuring device includes a voltage measuring instrument, a current measuring sensor and an oil sample gas analyzer, wherein the voltage measuring instrument is used to measure the voltage signal in the oil sample during the discharge process in the three-level experiment, the current measuring sensor is used to measure the current signal in the discharge process in the three-level experiment, and the oil sample gas analyzer is used to measure the dissolved gas components of the oil sample in the three-level experiment. As an example, the voltage measuring instrument can be a resistor-capacitor voltage divider. The current measuring sensor can be a high-frequency current coil mutual inductor. In addition, in order to further measure more accurate current during spark discharge and arc discharge, a Hall current sensor can be used to collect / simultaneously collect current signals.

[0034] The detection device is provided with an experimental acquisition unit and a training unit: the experimental acquisition unit is connected to the measuring device and the calculation unit respectively, and the acquisition unit is used to receive the voltage signal, current signal and dissolved gas components of the oil sample collected by the voltage measuring instrument, current measuring sensor and oil sample gas analyzer for the three-level experiment, and transmit the voltage signal, current signal and dissolved gas components of the oil sample to the training unit; the training unit is used to calculate the electric energy data of the three-level experiment according to the voltage signal and current signal; the electric energy data and the dissolved gas components of the oil sample are imported into the preset initial judgment model for iterative training until the loss of the initial judgment model meets the preset requirements, and the trained fault detection model is obtained. By setting up an actual experimental device, the classic characteristics of different stages can be structurally displayed according to the current signal, voltage signal and gas components in the oil sample in different fault stages, and the corresponding data can be generated and trained, so as to obtain a more targeted calculation model and improve the calculation accuracy of the model.

[0035] Since the damage degree of partial discharge, spark discharge and arc discharge faults gradually increases, when conducting spark discharge experiments and arc discharge experiments, appropriately increasing the training weight can increase the prediction accuracy of spark discharge experiments and arc discharge experiments with greater damage degrees.

[0036] Therefore, in some optional implementations, when iteratively training the initial judgment model, the proportions of the first training data, the second training data, and the third training data in the iterative training process are set according to the preset first weight, second weight, and third weight, wherein the first weight / second weight / third weight represent the weights corresponding to the local discharge experiment / spark discharge experiment / arc discharge test, and the first training data / second training data / third training data represent the data volume of the training data corresponding to the local discharge experiment / spark discharge experiment / arc discharge test, the first weight is less than the second weight, the second weight is less than the third weight, and the sum of the first weight, the second weight, and the third weight is 1.

[0037] When dealing with different discharge faults, in order to better respond to different fault situations, different data collection methods should be used for different fault situations.

[0038] In some optional implementations, during the partial discharge experiment, the conditions for collecting voltage signals and current signals include: increasing the needle-plate voltage between the needle electrode and the plate electrode to m times (m>1) the partial discharge starting voltage, until the needle-plate electrode model breaks down the insulating paperboard sample set therein, and then starting to collect voltage signals and current signals. The partial discharge starting voltage can refer to the working voltage of the simulated transformer during normal operation, that is, the voltage across the insulating oil paper. The partial discharge of the insulating oil paper is often a gradual process. Therefore, when conducting a partial discharge experiment, a gradual pressurization method can be used to simulate the breakdown, thereby increasing the fidelity of the data collected during the experiment.

[0039] The conditions for collecting oil samples include: when the voltage signal and the current signal are collected, the first collection is performed, and the oil sample is collected again after a preset first time period. After collecting the oil sample, the collected oil sample can be analyzed for component data (volume components) of hydrogen, acetylene, total hydrocarbons, carbon monoxide and carbon dioxide in the oil sample. The first time period can refer to the preset time period for interval collection of oil samples. As an example, the first time period can be 30 minutes.

[0040] In some optional implementations, during the spark discharge experiment, the conditions for collecting voltage signals and current signals include: increasing the needle-plate voltage between the needle electrode and the plate electrode from zero at a preset speed until the insulating paperboard sample is broken down, and then starting to collect voltage signals and current signals after cutting off the power supply. For example, the preset speed may be 2 kV / s.

[0041] The conditions for collecting oil samples include: when the insulating paperboard sample is broken down and the gas in the oil sample reaches equilibrium, the oil sample is collected. Different from the partial discharge test process, the oil sample collection in the spark discharge test needs to be carried out after the insulating sample is broken down and the gas reaches equilibrium. The judgment of whether the gas reaches equilibrium can be made manually.

[0042] In some optional implementations, during the arc discharge experiment, the conditions for collecting the voltage signal and the current signal include: boosting the needle-plate voltage between the needle electrode and the plate electrode until the insulating paperboard sample is broken down and a stable penetrating discharge channel is formed between the needle electrode and the plate electrode, and then cutting off the power supply after a second time period. As an example, the second time period may be 10 seconds.

[0043] The conditions for collecting oil samples include: when the insulating paperboard sample is broken down and the gas in the oil sample reaches equilibrium, the oil sample is collected. The conditions for collecting oil samples in the arc discharge test and the spark discharge experiment are the same.

[0044] In some optional implementations, calculating the electric energy data of the partial discharge experiment according to the voltage signal and the current signal includes: calculating a first voltage peak according to the voltage signal; calculating a discharge phase, a partial discharge charge, and a discharge number according to the current signal; calculating the electric energy data of the partial discharge experiment according to the voltage peak, the discharge phase, the partial discharge charge, and the discharge number, and a preset first calculation formula, wherein the first calculation formula includes:

[0045] in, W PD Represents the electrical energy data of the partial discharge experiment, U PD represents the first voltage peak, q i Indicates the amount of partial discharge charge, n Indicates the number of discharges, Indicates the discharge phase.

[0046] In actual execution, since the fault type cannot be known in advance, the voltage and current values ​​of the oil sample in the transformer can only be simply calculated, and the corresponding discharge energy can be calculated based on this. However, when conducting experiments, since the type of fault can be known in advance (man-made), a more accurate calculation method corresponding to the fault type can be adopted for different discharge processes (such as the first calculation formula for the partial discharge process mentioned above), so as to more accurately calculate the electric energy data, so the accuracy of the trained model is also higher. In addition, according to the voltage signal, calculating the first voltage peak; according to the current signal, calculating the discharge phase, partial discharge charge and discharge times are conventional technologies in the field, so the detailed calculation steps are no longer described in detail.

[0047] In some optional implementations, calculating the electric energy data of the spark discharge experiment / arc discharge experiment according to the voltage signal and the current signal includes: calculating a second voltage peak according to the voltage signal; calculating a current peak and duration according to the current signal; calculating the electric energy data of the spark discharge experiment / arc discharge experiment according to the voltage peak, current peak and duration, and a preset second calculation formula, wherein the second calculation formula includes:

[0048] in, W bd Indicates the electrical energy data of the spark discharge experiment / arc discharge experiment, U bd represents the second voltage peak, T Indicates duration, i bd Indicates the peak current.

[0049] Similar to the partial discharge process, during the actual execution, since the fault type cannot be known in advance, the voltage and current values ​​of the oil sample in the transformer can only be simply calculated, and the corresponding discharge energy can be calculated based on this. However, during the experiment, since the type of fault can be known in advance (man-made), a more accurate calculation method corresponding to the fault type can be adopted for different discharge processes (such as the aforementioned second calculation formula for the spark discharge experiment / arc discharge experiment process), so as to more accurately calculate the electric energy data, so the accuracy of the trained model is also higher. In addition, according to the voltage signal, the second voltage peak is calculated; according to the current signal, the current peak and duration are calculated; according to the voltage peak, current peak and duration are also conventional technologies in the field, so the detailed calculation steps are not described in detail.

[0050] As a specific example, the following table shows one of the training data.

[0051]

[0052] Among them, each column from left to right in the table represents the component data of hydrogen, acetylene, total hydrocarbons, carbon monoxide and carbon dioxide, as well as the corresponding different fault types (partial discharge, spark discharge or arc discharge).

[0053] As a specific example, the calculation model used in the power transformer oil-paper insulation discharge fault experimental system provided in this specification is a back propagation neural network, and its specific training conditions include: like Figure 3As shown, the number of input layer nodes of the directional propagation neural network is 6, the number of hidden layer nodes is 8, the number of output layer nodes is 3, the activation function selects the sigmod function, the gradient descent method is used to optimize the weight and bias parameters, the initial maximum number of iterations is set to 10,000 times, the initial learning rate is 0.01, and if the predicted output error rate reaches less than 10% during the training process, the iteration is stopped and the training is completed; if the predicted output error rate is greater than 10% when the initial maximum number of iterations is reached, the learning rate is divided by 10 and the training is restarted until the predicted output error rate is less than 10% during the training process. The iteration is stopped, and the model is considered to have converged at this time, and the neural network training is completed. In this specific example, 450 groups of experimental data with labeled results in the experimental database (a database for storing experimental collection data) are used as training sets to train the back propagation neural network, and 150 groups of experimental data are used as test sets to carry out neural network performance testing, and the diagnostic success rate reaches 94.67%.

[0054] The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to only the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that technicians in the relevant technical field can well understand and use this specification. This specification is only limited by the claims and their full scope and equivalents.

Claims

1. A detection device for oil-paper insulation discharge fault of power transformer, characterized in that: include: Characteristic parameter acquisition unit and calculation unit, wherein, The characteristic parameter acquisition unit is used to collect target power data and oil sample dissolved gas component data of the oil in the target transformer, and transmit the power data and oil sample dissolved gas component data to the calculation unit, wherein the oil sample dissolved gas component data includes component data of hydrogen, acetylene, total hydrocarbons, carbon monoxide and carbon dioxide; The calculation unit is used to import the target electric energy data and the component data of hydrogen, acetylene, total hydrocarbons, carbon monoxide and carbon dioxide into a trained fault identification model to generate a fault type result, wherein the fault identification model is trained by a preset power transformer oil-paper insulation discharge fault experimental system.

2. The device according to claim 1, characterized in that It also includes a preprocessing unit configured to obtain a real-time voltage signal and a real-time current signal of the oil in the target transformer; Cleaning the real-time voltage signal and the real-time current signal; The target electric energy data is calculated according to the voltage signal and the current signal after cleaning.

3. The device according to claim 1, characterized in that The power transformer oil-paper insulation discharge fault experimental system comprises an experimental tank, a needle plate electrode model, a measuring device and a detecting device, wherein: The needle-plate electrode model is arranged in the experimental tank, an insulating paperboard sample is arranged between the needle electrode and the plate electrode of the needle-plate electrode model, and the needle-plate electrode model is used to simulate a three-level experiment, which includes a partial discharge experiment, a spark discharge experiment and an arc discharge experiment; The measuring device includes a voltage measuring instrument, a current measuring sensor and an oil sample gas analyzer, wherein the voltage measuring instrument is used to measure the voltage signal in the oil sample during the discharge process in the three-level experiment, the current measuring sensor is used to measure the current signal during the discharge process in the three-level experiment, and the oil sample gas analyzer is used to measure the dissolved gas components of the oil sample in the three-level experiment; The detection device is provided with an experiment collection unit and a training unit: The experiment acquisition unit is connected to the measuring device and the computing unit respectively, and is used to receive the voltage signal, current signal and oil sample dissolved gas components collected by the voltage measuring instrument, current measuring sensor and oil sample gas analyzer for the three-level experiment, and transmit the voltage signal, current signal and oil sample dissolved gas components to the training unit; The training unit is used to calculate the electric energy data of the three-level experiment based on the voltage signal and the current signal; import the electric energy data and the dissolved gas components of the oil sample into a preset initial judgment model for iterative training until the loss of the initial judgment model meets the preset requirements, thereby obtaining the trained fault detection model.

4. The device according to claim 3, characterized in that During the partial discharge experiment, The conditions for collecting voltage signals and current signals include: increasing the needle-plate voltage between the needle electrode and the plate electrode to m times (m>1) the local discharge starting voltage, and starting to collect voltage signals and current signals after the needle-plate electrode model breaks through the insulating paperboard sample set therein; The conditions for collecting oil samples include: when the voltage signal and the current signal are collected, the first collection is started, and the oil sample is collected again after a preset first time period.

5. The device according to claim 3, characterized in that During the spark discharge experiment, The conditions for collecting voltage signals and current signals include: increasing the needle-plate voltage between the needle electrode and the plate electrode uniformly from zero at a preset speed until the insulating paperboard sample is broken down, and then starting to collect voltage signals and current signals after cutting off the power supply; The conditions for collecting the oil sample include: when the insulating paperboard sample is punctured and the gas in the oil sample reaches equilibrium, the oil sample is collected.

6. The device according to claim 3, characterized in that During the arc discharge experiment, The conditions for collecting the voltage signal and the current signal include: boosting the needle-plate voltage between the needle electrode and the plate electrode until the insulating paperboard sample is broken down and a stable penetrating discharge channel is formed between the needle electrode and the plate electrode, and then cutting off the power supply after the second time period; The conditions for collecting the oil sample include: when the insulating paperboard sample is punctured and the gas in the oil sample reaches equilibrium, the oil sample is collected.

7. The device according to claim 3, characterized in that Calculating the electric energy data of the partial discharge experiment according to the voltage signal and the current signal includes: Calculating a first voltage peak value according to the voltage signal; Calculating the discharge phase, partial discharge charge and discharge times according to the current signal; The electric energy data of the partial discharge experiment is calculated according to the voltage peak value, the discharge phase, the partial discharge charge and the number of discharges, and a preset first calculation formula, wherein the first calculation formula includes: in, W PD Represents the electrical energy data of partial discharge experiment, U PD represents the first voltage peak, q i Indicates the amount of partial discharge charge, n Indicates the number of discharges, Indicates the discharge phase.

8. The device according to claim 3, characterized in that Calculating the electric energy data of the spark discharge experiment / arc discharge experiment according to the voltage signal and the current signal includes: Calculating a second voltage peak value according to the voltage signal; Calculating the current peak value and duration according to the current signal; The electric energy data of the spark discharge experiment / arc discharge experiment is calculated according to the voltage peak value, current peak value and duration, and a preset second calculation formula, wherein the second calculation formula includes: in, W bd Indicates the electrical energy data of the spark discharge experiment / arc discharge experiment, U bd represents the second voltage peak, T Indicates duration, i bd Indicates the peak current.

9. The device according to claim 3, characterized in that The fault detection model is a back propagation neural network.

10. The device according to claim 3, characterized in that The experimental tank body is also provided with an oil valve for taking oil and filling oil, a pressure release valve for relieving pressure, and a barometer for detecting the air pressure in the experimental tank body.

Images