An AI early warning system based on dissolved gas monitoring in transformer oil

Abstract

The invention relates to the technical field of transformers, and discloses an AI early warning system based on monitoring of dissolved gas in transformer oil. The gas precipitation module is used for clustering the solubility of dissolved gas in transformer oil to obtain precipitation temperatures of multiple dissolved gases, and specifically comprises: setting multiple precipitation points in transformer oil, setting precipitation temperatures for each precipitation point, and respectively precipitating dissolved gas in transformer oil; a data acquisition module is used for setting capacitance monitoring devices at multiple precipitation points in transformer oil, and collecting capacitance parameters of each precipitation point in real time; a data processing module is used for comparing capacitance parameters of multiple precipitation points at the same time, determining capacitance differences of precipitation points at adjacent precipitation temperatures, and the capacitance differences are used for approximately reflecting the concentration of dissolved gas in transformer oil.

Description

Technical Field

[0001] The present invention relates to the field of transformers, and more specifically, to an AI early warning system based on monitoring of dissolved gas in transformer oil. Background Art

[0002] Transformer oil is an important insulating medium between the transformer's internal windings and the iron core, and also serves as a heat dissipation and cooling function. During the operation of the transformer, gases such as hydrogen, methane and acetylene are generated. When a transformer fails, the amount of gas generated increases significantly. Transformer fault assessment mainly relies on gas chromatographs to analyze the gas composition and concentration in transformer oil.

[0003] However, continuous operation of the gas chromatograph will produce problems such as temperature drift, carrier gas stability and detector sensitivity fluctuation, which will affect the reliability of the final detection results. Summary of the invention

[0004] The present invention provides an AI early warning system based on monitoring of dissolved gas in transformer oil, which solves the technical problems raised in the background technology.

[0005] The present invention provides an AI early warning system based on monitoring of dissolved gas in transformer oil, comprising:

[0006] The gas precipitation module is used to cluster the solubility of dissolved gases in transformer oil and obtain the precipitation temperatures of multiple dissolved gases, including:

[0007] A plurality of precipitation points are respectively set in the transformer oil, and a precipitation temperature is respectively set for each precipitation point, so that the dissolved gas in the transformer oil is precipitated respectively;

[0008] A data acquisition module is used to set capacitance monitoring devices at multiple precipitation points in transformer oil and collect capacitance parameters of each precipitation point in real time;

[0009] The data processing module is used to compare the capacitance parameters of multiple precipitation points at the same time and determine the capacitance difference of precipitation points at adjacent precipitation temperatures, as follows:

[0010] Constructing a capacitance curve of each precipitation point according to the capacitance parameter of each precipitation point;

[0011] Perform slope analysis on the capacitance curve to determine the plateau of each capacitance curve;

[0012] Determine the intersection of the time sets corresponding to the plateau segments of each capacitance curve;

[0013] Select any moment in the intersection of the time set corresponding to the stable segment, compare the capacitance parameters of multiple precipitation points at the moment, and obtain the capacitance difference of the precipitation points at adjacent precipitation temperatures;

[0014] The capacitance difference is used to approximate the concentration of dissolved gas in a transformer oil;

[0015] The risk prediction module is used to analyze the transformer faults based on the capacitance difference of the precipitation points at adjacent temperatures and determine the risk of each type of transformer fault, as follows:

[0016] Conduct simulation experiments, including:

[0017] Take N transformer oils of the same amount as N samples;

[0018] For each fault-indicating gas, a different amount is dissolved in each sample;

[0019] A mapping model of the capacitance difference of precipitation points at adjacent temperatures and the approximate concentration of the fault gas is established. The formula of the mapping model is as follows:

[0020] ;

[0021] in, The first one represents the capacitance difference mapping of precipitation points at adjacent temperatures The predicted value of the concentration of the fault indicating gas; represents the exponential term, represents the first proportionality coefficient, represents a natural constant, represents the growth rate coefficient of the exponential term, Indicates precipitation temperature The precipitation point and precipitation temperature The capacitance difference of the precipitation point; represents the quadratic term, represents the second proportionality coefficient; represents a constant term;

[0022] The error function of the mapping model is as follows:

[0023] ;

[0024] in, Represents the sum of the errors between the predicted values ​​and sample values ​​of all samples; represents the number of samples, The sample size index representing the sample; Indicates The sample value of samples, Indicates The predicted value of samples;

[0025] The concentration of each fault-indicating gas is divided into a maximum warning concentration by an expert method, and the predicted value of the mapping model is normalized to obtain a normalized value, which represents the risk probability of the corresponding transformer fault;

[0026] The sample value represents the concentration of the fault-indicating gas dissolved in the actual sample; the predicted value represents the concentration of the fault-indicating gas dissolved in the sample predicted by the mapping model; the custom coefficient in the formula of the mapping model The values ​​are determined by optimization algorithms;

[0027] The fault chain module is used to perform spatiotemporal analysis on the risk of each type of transformer fault, verify the judgment of the risk prediction module, and obtain the transformer verification judgment, as follows:

[0028] According to the risk probability of each type of transformer fault, the transformer faults are sorted, and the transformer faults higher than a preset probability threshold are determined to obtain a first candidate fault set;

[0029] Based on prior knowledge, a fault relationship chain of the transformer is constructed, and the fault relationship chain represents the logical sequence of the transformer fault occurrence;

[0030] Fitting the faults in the first candidate fault set with the fault relationship chain;

[0031] If there is at least one fault relationship chain in the first candidate fault set, the first candidate fault set obtains a first verification judgment;

[0032] According to the fault relationship chains existing in the first candidate fault set, determine the starting fault and the ending fault of each fault relationship chain;

[0033] If the risk probability from the starting fault to the ending fault in each fault relationship chain remains in a decreasing state, the first candidate fault set obtains a second verification judgment;

[0034] The planning module schedules the maintenance and management of transformers based on the transformer verification judgment.

[0035] Furthermore, the solubility of dissolved gas in transformer oil is clustered to obtain the precipitation temperatures of multiple dissolved gases, including:

[0036] Set the precipitation temperature range of the precipitation point;

[0037] Obtaining the solubility of each dissolved gas in transformer oil within the precipitation temperature range;

[0038] Classify dissolved gases into fault indicating gases and irrelevant gases;

[0039] According to the solubility, determine the dissolution fluctuation temperature of each dissolved gas in the precipitation temperature range;

[0040] The dissolution fluctuation temperatures are sorted to determine the dissolution fluctuation temperature corresponding to the fault indicating gas; the dissolution fluctuation temperatures of irrelevant gases are screened by the dissolution fluctuation temperature corresponding to the fault indicating gas to obtain the corresponding dissolution fluctuation temperature in the irrelevant gas, and the corresponding dissolution fluctuation temperature is used as the first precipitation temperature;

[0041] The dissolution fluctuation temperatures of each fault indicating gas are sorted, and the dissolution fluctuation temperatures of each fault indicating gas within the precipitation temperature range are respectively used as the second precipitation temperature to the Nth precipitation temperature.

[0042] Furthermore, a capacitance monitoring device is set at multiple precipitation points in the transformer oil, and the capacitance parameters of each precipitation point are collected in real time, including:

[0043] The precipitation point includes a needle body, and a plurality of needle bodies for heating the transformer oil are arranged at fixed intervals at the bottom of the transformer oil, and each needle body is respectively set with a different precipitation temperature;

[0044] A capacitance monitoring device is fixedly installed above the needle body, and the capacitance monitoring device collects capacitance parameters of the precipitation point at fixed time intervals.

[0045] Furthermore, the custom coefficients in the mapping model formula are all determined by an optimization algorithm, including:

[0046] Randomly generate an initial population with R individuals that meet the constraints, and initialize the current number of iterations to 1;

[0047] Encode the individuals in the population, and the encoding vector is expressed as: ;

[0048] The error function of the mapping model is used as the objective function to obtain the objective function value of each individual;

[0049] According to the objective function value of the individual, the individuals are sorted from small to large, and a fixed proportion of individuals with smaller objective function values ​​in the population are retained. The remaining individuals are updated as parents until the preset number of iterations is reached to obtain the corresponding individuals, and the corresponding individuals are used as the values ​​of the custom coefficients in the formula of the mapping model.

[0050] Further, the remaining individuals are updated as parents, including:

[0051] Pause Update Factor , is a random number used to obtain a random value for each parent before iteratively updating the parent. ;when When the preset pause value is less than the default pause value, the iterative update of the corresponding parent generation will be suspended;

[0052] Updates to the parent include global updates, local updates, and convergence updates;

[0053] The formula for global update is as follows:

[0054] ;

[0055] in, Indicates The update vector of the individual in the iteration, Indicates The update vector of the individual in the iteration, Indicates The individual with the lowest global objective function value in the iteration, represents a natural constant, represents the exponential coefficient, Represents a random number, , represents the preset number of iterations, t represents the current number of iterations, represents the proportionality coefficient;

[0056] The formula for local update is as follows:

[0057] ;in, represents the scaling factor, Represents random numbers that conform to the standard normal distribution;

[0058] The formula for convergence update is as follows:

[0059] ; Represents the step size coefficient.

[0060] Furthermore, the maintenance and management of the transformer is scheduled based on the transformer verification judgment, including:

[0061] In response to the transformer having a risk probability of transformer failure, the transformer oil is tested by starting a gas chromatograph;

[0062] In response to the transformer of the first verification judgment, the transformer oil is tested by starting the gas chromatograph and the person in charge of the transformer is notified immediately;

[0063] In response to the transformer judged by the second verification, the transformer oil is measured by starting a gas chromatograph and the transformer is immediately shut down for maintenance.

[0064] An AI early warning method based on monitoring of dissolved gas in transformer oil, applied to the system, includes:

[0065] Step 1, clustering the solubility of dissolved gas in transformer oil to obtain precipitation temperatures of multiple dissolved gases, specifically including:

[0066] A plurality of precipitation points are respectively set in the transformer oil, and a precipitation temperature is respectively set for each precipitation point, so that the dissolved gas in the transformer oil is precipitated respectively;

[0067] Step 2, setting a capacitance monitoring device at multiple precipitation points in the transformer oil, and collecting capacitance parameters of each precipitation point in real time;

[0068] Step 3, comparing the capacitance parameters of multiple precipitation points at the same time, determining the capacitance difference of the precipitation points at adjacent precipitation temperatures, the capacitance difference being used to approximately reflect the concentration of dissolved gas in a transformer oil;

[0069] Step 4, performing fault analysis on the transformer according to the capacitance difference of precipitation points at adjacent temperatures, and determining the risk of each type of transformer fault;

[0070] Step 5, perform spatiotemporal analysis on the risk of each type of transformer fault, verify the judgment of the risk prediction module, and obtain a verified judgment of the transformer;

[0071] Step 6, scheduling maintenance and management of the transformer based on the transformer verification judgment.

[0072] The beneficial effects of the present invention are: by real-time monitoring of capacitance changes and using a gas concentration mapping model, an approximate gas monitoring method is provided, which reduces the need for continuous operation of the gas chromatograph, and effectively reduces costs and maintenance difficulties. The system achieves highly accurate fault warnings through fault chain analysis constructed by optimizing algorithms and prior knowledge; at the same time, further verification is combined with the interval operation of the gas chromatograph, making the detection more accurate, providing an efficient, economical and stable fault warning solution for the transformer. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 It is an architecture diagram of an AI early warning system based on monitoring of dissolved gas in transformer oil of the present invention;

[0074] Figure 2 It is a flow chart of an AI early warning method based on monitoring of dissolved gas in transformer oil of the present invention. DETAILED DESCRIPTION

[0075] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0076] like Figure 1 As shown, an AI early warning system based on dissolved gas monitoring in transformer oil includes:

[0077] The gas precipitation module is used to cluster the solubility of dissolved gases in transformer oil and obtain the precipitation temperatures of multiple dissolved gases, including:

[0078] A plurality of precipitation points are respectively set in the transformer oil, and a precipitation temperature is respectively set for each precipitation point, so that the dissolved gas in the transformer oil is precipitated respectively;

[0079] A data acquisition module is used to set capacitance monitoring devices at multiple precipitation points in transformer oil and collect capacitance parameters of each precipitation point in real time;

[0080] The data processing module is used to compare the capacitance parameters of multiple precipitation points at the same time and determine the capacitance difference of precipitation points at adjacent precipitation temperatures, as follows:

[0081] Constructing a capacitance curve of each precipitation point according to the capacitance parameter of each precipitation point;

[0082] Perform slope analysis on the capacitance curve to determine the plateau of each capacitance curve;

[0083] Determine the intersection of the time sets corresponding to the plateau segments of each capacitance curve;

[0084] Select any moment in the intersection of the time set corresponding to the stable segment, compare the capacitance parameters of multiple precipitation points at the moment, and obtain the capacitance difference of the precipitation points at adjacent precipitation temperatures;

[0085] The capacitance difference is used to approximate the concentration of dissolved gas in a transformer oil;

[0086] The risk prediction module is used to analyze the transformer faults based on the capacitance difference of the precipitation points at adjacent temperatures and determine the risk of each type of transformer fault, as follows:

[0087] Conduct simulation experiments, including:

[0088] Take N transformer oils of the same amount as N samples;

[0089] For each fault-indicating gas, a different amount is dissolved in each sample;

[0090] A mapping model of the capacitance difference of precipitation points at adjacent temperatures and the approximate concentration of the fault gas is established. The formula of the mapping model is as follows:

[0091] ;

[0092] in, The first one represents the capacitance difference mapping of precipitation points at adjacent temperatures The predicted value of the concentration of the fault indicating gas; represents the exponential term, represents the first proportionality coefficient, represents a natural constant, represents the growth rate coefficient of the exponential term, Indicates precipitation temperature The precipitation point and precipitation temperature The capacitance difference of the precipitation point; represents the quadratic term, represents the second proportionality coefficient; represents a constant term;

[0093] The error function of the mapping model is as follows:

[0094] ;

[0095] in, Represents the sum of the errors between the predicted values ​​and sample values ​​of all samples; represents the number of samples, The sample size index representing the sample; Indicates The sample value of samples, Indicates The predicted value of samples;

[0096] The concentration of each fault-indicating gas is divided into a maximum warning concentration by an expert method, and the predicted value of the mapping model is normalized to obtain a normalized value, which represents the risk probability of the corresponding transformer fault;

[0097] The sample value represents the concentration of the fault-indicating gas dissolved in the actual sample; the predicted value represents the concentration of the fault-indicating gas dissolved in the sample predicted by the mapping model; the custom coefficient in the formula of the mapping model The values ​​are determined by optimization algorithms;

[0098] The fault chain module is used to perform spatiotemporal analysis on the risk of each type of transformer fault, verify the judgment of the risk prediction module, and obtain the transformer verification judgment, as follows:

[0099] According to the risk probability of each type of transformer fault, the transformer faults are sorted, and the transformer faults higher than a preset probability threshold are determined to obtain a first candidate fault set;

[0100] Based on prior knowledge, a fault relationship chain of the transformer is constructed, and the fault relationship chain represents the logical sequence of the transformer fault occurrence;

[0101] Fitting the faults in the first candidate fault set with the fault relationship chain;

[0102] If there is at least one fault relationship chain in the first candidate fault set, the first candidate fault set obtains a first verification judgment;

[0103] According to the fault relationship chains existing in the first candidate fault set, determine the starting fault and the ending fault of each fault relationship chain;

[0104] If the risk probability from the starting fault to the ending fault in each fault relationship chain remains in a decreasing state, the first candidate fault set obtains a second verification judgment;

[0105] The planning module schedules the maintenance and management of transformers based on the transformer verification judgment.

[0106] In one embodiment of the present invention, the solubility of dissolved gas in transformer oil is clustered to obtain a plurality of precipitation temperatures of dissolved gas, including:

[0107] Set the precipitation temperature range of the precipitation point;

[0108] Obtaining the solubility of each dissolved gas in transformer oil within the precipitation temperature range;

[0109] Classify dissolved gases into fault indicating gases and irrelevant gases;

[0110] According to the solubility, determine the dissolution fluctuation temperature of each dissolved gas in the precipitation temperature range;

[0111] The dissolution fluctuation temperatures are sorted to determine the dissolution fluctuation temperature corresponding to the fault indicating gas; the dissolution fluctuation temperatures of irrelevant gases are screened by the dissolution fluctuation temperature corresponding to the fault indicating gas to obtain the corresponding dissolution fluctuation temperature in the irrelevant gas, and the corresponding dissolution fluctuation temperature is used as the first precipitation temperature;

[0112] The dissolution fluctuation temperatures of each fault indicating gas are sorted, and the dissolution fluctuation temperatures of each fault indicating gas within the precipitation temperature range are respectively used as the second precipitation temperature to the Nth precipitation temperature.

[0113] Specifically, the dissolved gas in transformer oil includes fault-indicating gas and irrelevant gas. The precipitation temperature of irrelevant gas is 30-40 degrees Celsius, while the temperature of fault-indicating gas is 50-300 degrees Celsius.

[0114] Therefore, irrelevant gases can be precipitated by setting different precipitation temperatures, and fault-indicating gases can be precipitated separately by setting step temperatures. The dissolution fluctuation temperature indicates the temperature at which the precipitation rate of the dissolved gas fluctuates dramatically. For example, at 40 degrees Celsius, the solubility is 100, at 50 degrees Celsius, the solubility is 95, ..., at 80 degrees Celsius, the solubility is 10; therefore, 80 degrees is used as the dissolution fluctuation temperature of the dissolved gas; for the smallest dissolution fluctuation temperature of the fault gas, such as 50 degrees; the largest dissolution fluctuation temperature of the irrelevant gas, such as 47 degrees; then 47 degrees is selected as the first precipitation temperature, which is used to precipitate most of the irrelevant gases; then 50 degrees is used as the second precipitation temperature; and so on.

[0115] It should be noted that the precipitation temperature range of the precipitation point shall not be higher than 300 degrees Celsius, because the ignition point of transformer oil is 400 degrees Celsius.

[0116] In one embodiment of the present invention, a capacitance monitoring device is provided at multiple precipitation points in transformer oil, and capacitance parameters of each precipitation point are collected in real time, including:

[0117] The precipitation point includes a needle body, and a plurality of needle bodies for heating the transformer oil are arranged at fixed intervals at the bottom of the transformer oil, and each needle body is respectively set with a different precipitation temperature;

[0118] A capacitance monitoring device is fixedly installed above the needle body, and the capacitance monitoring device collects capacitance parameters of the precipitation point at fixed time intervals.

[0119] Specifically, dissolved gases include nitrogen, oxygen, methane, ethane, ethylene and other gases; wherein irrelevant gases include nitrogen and oxygen; and fault indicating gases include methane, ethane, ethylene and other gases.

[0120] Step 301, when the transformer oil is heated to nearly 40 degrees Celsius, nitrogen and oxygen are precipitated in large quantities. At this time, the precipitation temperature is a precipitation point of 40 degrees Celsius, and floating bubbles will be generated above it. The bubbles will affect the capacitance parameter of the transformer oil above this point, thereby causing the capacitance parameter to decrease.

[0121] Step 302, when the transformer oil is heated to nearly 60 degrees Celsius, nitrogen, oxygen and methane are precipitated in large quantities, and the bubble density is further increased, thereby causing the capacitance parameter above the precipitation point with a precipitation temperature of 60 degrees Celsius to further decrease.

[0122] Step 303, calculates the difference between the capacitance parameter of the precipitation point with a precipitation temperature of 60 degrees Celsius and the capacitance parameter of the precipitation point with a precipitation temperature of 40 degrees Celsius, and the difference is used to indicate the approximate concentration of methane in the transformer oil. It should be noted that the precipitation temperature of ethane in transformer oil is close to 70 degrees Celsius, and the precipitation temperature of ethylene in transformer oil is close to 150 degrees.

[0123] Step 304, repeating steps 301, 302 and 303 to obtain capacitance differences of gases such as methane, ethane and ethylene.

[0124] It should be noted that heating the transformer oil by the precipitation needle has a negligible effect on the heat dissipation of the transformer.

[0125] Specifically, the slope of the capacitance curve is analyzed to determine the stable section of each capacitance curve, including:

[0126] Step 401: Constructing a capacitance curve:

[0127] Assume that at each precipitation point i, the reading of the capacitance sensor at time t is The capacitance curve can be expressed as a series of capacitance values ​​versus time. The goal is to find the plateau of the capacitance curve through curve analysis, that is, the period of time when the capacitance value changes little or tends to be stable within a certain period of time.

[0128] Step 402: Calculate the slope of the capacitance curve:

[0129] The slope of the capacitance curve is calculated using the differential method or the derivative method to obtain the rate of change of the capacitance value at each moment.

[0130] For each moment t in the time series, calculate the slope of the capacitance value :

[0131] ; represents the interval between adjacent time points in the capacitance curve, Represents a differential operation.

[0132] Step 403: Determine the threshold of the plateau:

[0133] To determine the stable segment, a threshold can be set , that is, when the absolute value of the slope Less than threshold When , it is considered that the capacitance change is not significant and is in a stable state. The value of depends on the accuracy of the capacitance sensor and the system noise level. For example, if the capacitance sensor accuracy is 0.01uF, then Can be set around 0.01.

[0134] Step 404: Marking a stationary segment:

[0135] By traversing the time series, for each time point ,judge Is it true? If true, mark this time period as a stable period.

[0136] In one embodiment of the present invention, the system performs fault analysis and risk assessment through the following steps:

[0137] Sample preparation: select N transformer oil samples of the same amount, each of which contains a different amount of fault indicating gas (such as methane and ethane). For example, sample 1 contains X1 methane and Y1 ethane, and sample 2 contains X2 methane and Y2 ethane.

[0138] Capacitance difference measurement: By setting multiple different precipitation points and multiple capacitance monitoring devices in each sample, the capacitance parameters of each sample at different temperatures are recorded in real time. Based on these capacitance parameters, the capacitance difference of adjacent temperature precipitation points is calculated and marked as .

[0139] Mapping model establishment: Establish a mapping model between the capacitance difference and the fault gas concentration in the form of: .

[0140] Error function optimization: Define the error function of the mapping model, calculate the error between the predicted values ​​and the actual values ​​of all samples, and minimize the total error.

[0141] Risk assessment: The maximum warning concentration is set for each fault-indicating gas concentration through the expert method, and the predicted value of the mapping model is used for normalization to obtain the fault risk probability.

[0142] By implementing the above embodiments, the following significant effects are obtained:

[0143] Accurate prediction of fault gas concentration: The mapping model in the embodiment can accurately predict the concentration of different fault indicating gases according to the capacitance difference, providing an effective monitoring means for transformer faults.

[0144] Improved model accuracy: By introducing the error function, the prediction accuracy of the model is improved.

[0145] Effective risk differentiation: The normalized predicted values ​​make the risk assessment of transformer faults clearer and can be classified according to the concentration levels of different fault gases, helping to identify and respond to potential risks in a timely manner.

[0146] Improve real-time monitoring capabilities: The capacitance difference used in the implementation is used as a monitoring indicator, providing real-time and continuous fault monitoring capabilities. Compared with traditional offline detection methods, it has a faster response speed and can issue early warnings in a timely manner, reducing the possibility of accidents.

[0147] Significant economic benefits: This method reduces the reliance on complex testing equipment and reduces testing costs. At the same time, through real-time monitoring, it improves the safety and operating efficiency of the transformer and brings economic benefits.

[0148] In one embodiment of the present invention, the custom coefficients in the formula of the mapping model are all determined by an optimization algorithm, including:

[0149] Randomly generate an initial population with R individuals that meet the constraints, and initialize the current number of iterations to 1;

[0150] Encode the individuals in the population, and the encoding vector is expressed as: ;

[0151] The error function of the mapping model is used as the objective function to obtain the objective function value of each individual;

[0152] According to the objective function value of the individual, the individuals are sorted from small to large, and a fixed proportion of individuals with smaller objective function values ​​in the population are retained. The remaining individuals are updated as parents until the preset number of iterations is reached to obtain the corresponding individuals, and the corresponding individuals are used as the values ​​of the custom coefficients in the formula of the mapping model.

[0153] In one embodiment of the present invention, the remaining individuals are updated as parents, including:

[0154] Pause Update Factor , is a random number used to obtain a random value for each parent before iteratively updating the parent. ;when When the preset pause value is less than the default pause value, the iterative update of the corresponding parent generation will be suspended;

[0155] Updates to the parent include global updates, local updates, and convergence updates;

[0156] The formula for global update is as follows:

[0157] ;

[0158] in, Indicates The update vector of the individual in the iteration, Indicates The update vector of the individual in the iteration, Indicates The individual with the lowest global objective function value in the iteration, represents a natural constant, represents the exponential coefficient, Represents a random number, , represents the preset number of iterations, t represents the current number of iterations, represents the proportionality coefficient;

[0159] The formula for local update is as follows:

[0160] ;in, represents the scaling factor, Represents random numbers that conform to the standard normal distribution;

[0161] The formula for convergence update is as follows:

[0162] ; Represents the step size coefficient.

[0163] Specific, individual updates include:

[0164] Parent update: When updating, the parent individual will include the following update strategies:

[0165] Pause update factor: This is a random number used to generate a random value for each parent before iterative update. When this value is less than the preset pause value, the update of the parent will be paused to prevent premature convergence.

[0166] Global update: By comparing all individuals, find the individual with the lowest objective function value and use it as the basis for global update. The update formula usually contains natural constants and random factors to introduce a certain degree of randomness.

[0167] Local update: Perform local updates on individuals and adjust their positions to better adapt to the current environment.

[0168] Convergence update: As the iterations proceed, the step size coefficient is gradually reduced to achieve convergence control of the search process.

[0169] Specifically, the first verification judgment includes:

[0170] Logical relationship of the fault chain: First, a fault chain is constructed based on prior knowledge, which describes the cause-effect relationship of different faults and their respective indicator gases. For example, the aging of insulation materials will generate specific fault indicator gases, and the detection of this gas can be used to indicate insulation aging faults.

[0171] Gradual fault development chain: After insulation aging, leakage may occur, which in turn leads to arc discharge, which in turn produces other indicator gases. In this way, different faults will produce different indicator gases in turn, forming a gas indication chain.

[0172] Verify the first candidate fault set: The system fits the fault gas in the first candidate fault set with the fault chain. If the gas type matches the expected fault chain, it can be confirmed that the fault development logic is reasonable and passes the first verification judgment.

[0173] Specifically, the second verification judgment includes:

[0174] Time sequence verification: Since faults occur gradually, the indicator gas of subsequent faults should appear later than that of the previous faults. Therefore, the risk probability corresponding to the early faults should be higher, while the risk probability of subsequent faults should be relatively lower.

[0175] Decreasing risk probability: Assuming that the risk probability of insulation material aging is 0.8, the risk probability of arc discharge as a subsequent fault should be slightly lower, such as 0.7. This can reflect that the indicator gas of the subsequent fault appears for a shorter time and the corresponding risk probability is smaller.

[0176] Verify the decreasing relationship: Verify the risk probability in each fault chain in turn, from the starting fault to the ending fault of the chain, and check whether the risk probability decreases in turn. If the risk probability meets the decreasing trend, the second verification judgment is completed to confirm that the timing logic is reasonable.

[0177] Through these two verification judgments, the system can more accurately determine the fault development path and risk size, thereby improving the reliability and accuracy of transformer fault risk assessment.

[0178] like Figure 2 As shown, in one embodiment of the present invention, scheduling the maintenance and management of the transformer according to the verification judgment of the transformer includes:

[0179] In response to the transformer having a risk probability of transformer failure, the transformer oil is tested by starting a gas chromatograph;

[0180] In response to the transformer of the first verification judgment, the transformer oil is tested by starting the gas chromatograph and the person in charge of the transformer is notified immediately;

[0181] In response to the transformer judged by the second verification, the transformer oil is measured by starting a gas chromatograph and the transformer is immediately shut down for maintenance.

[0182] Specifically, if the output of the risk prediction module does not obtain the first verification judgment and the second verification judgment, only the gas chromatograph will be started for further measurement; the gas chromatograph will conduct a detailed analysis of the dissolved gas composition and concentration in the transformer oil, and analyze the type and concentration of the dissolved gas through the measurement data to help predict the possible fault type of the transformer.

[0183] Specifically, if the output of the risk prediction module only obtains the first verification judgment, the system will immediately notify the person in charge of the transformer to remind them of the potential failure risk and provide preliminary analysis data measured by the gas chromatograph.

[0184] Specifically, if the output of the risk prediction module satisfies both the first verification judgment and the second verification judgment, the shutdown command will be executed immediately to avoid larger failures or accidents; after the shutdown, the transformer will enter the maintenance process, and through detailed inspection and repair, ensure the safety and stability of the transformer before it resumes operation.

[0185] An AI early warning method based on monitoring of dissolved gas in transformer oil, applied to the system, includes:

[0186] Step 1, clustering the solubility of dissolved gas in transformer oil to obtain precipitation temperatures of multiple dissolved gases, specifically including:

[0187] A plurality of precipitation points are respectively set in the transformer oil, and a precipitation temperature is respectively set for each precipitation point, so that the dissolved gas in the transformer oil is precipitated respectively;

[0188] Step 2, setting a capacitance monitoring device at multiple precipitation points in the transformer oil, and collecting capacitance parameters of each precipitation point in real time;

[0189] Step 3, comparing the capacitance parameters of multiple precipitation points at the same time, determining the capacitance difference of the precipitation points at adjacent precipitation temperatures, the capacitance difference being used to approximately reflect the concentration of dissolved gas in a transformer oil;

[0190] Step 4, performing fault analysis on the transformer according to the capacitance difference of precipitation points at adjacent temperatures, and determining the risk of each type of transformer fault;

[0191] Step 5, perform spatiotemporal analysis on the risk of each type of transformer fault, verify the judgment of the risk prediction module, and obtain a verified judgment of the transformer;

[0192] Step 6, scheduling maintenance and management of the transformer based on the transformer verification judgment.

[0193] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. An AI early warning system based on monitoring of dissolved gas in transformer oil, characterized in that: include: The gas precipitation module is used to cluster the solubility of dissolved gases in transformer oil and obtain the precipitation temperatures of multiple dissolved gases, including: A plurality of precipitation points are respectively set in the transformer oil, and a precipitation temperature is respectively set for each precipitation point, so that the dissolved gas in the transformer oil is precipitated respectively; A data acquisition module is used to set capacitance monitoring devices at multiple precipitation points in transformer oil and collect capacitance parameters of each precipitation point in real time; The data processing module is used to compare the capacitance parameters of multiple precipitation points at the same time and determine the capacitance difference of precipitation points at adjacent precipitation temperatures, as follows: Constructing a capacitance curve of each precipitation point according to the capacitance parameter of each precipitation point; Perform slope analysis on the capacitance curve to determine the plateau of each capacitance curve; Determine the intersection of the time sets corresponding to the plateau segments of each capacitance curve; Select any moment in the intersection of the time set corresponding to the stable segment, compare the capacitance parameters of multiple precipitation points at the moment, and obtain the capacitance difference of the precipitation points at adjacent precipitation temperatures; The capacitance difference is used to approximate the concentration of dissolved gas in a transformer oil; The risk prediction module is used to analyze the transformer faults based on the capacitance difference of the precipitation points at adjacent temperatures and determine the risk of each type of transformer fault, as follows: Conduct simulation experiments, including: Take N transformer oils of the same amount as N samples; For each fault-indicating gas, a different amount is dissolved in each sample; A mapping model of the capacitance difference of precipitation points at adjacent temperatures and the approximate concentration of the fault indicating gas is established. The formula of the mapping model is as follows: ； in, The first one represents the capacitance difference mapping of precipitation points at adjacent temperatures The predicted value of the concentration of the fault indicating gas; represents the exponential term, represents the first proportionality coefficient, represents a natural constant, represents the growth rate coefficient of the exponential term, Indicates precipitation temperature The precipitation point and precipitation temperature The capacitance difference of the precipitation point; represents the quadratic term, represents the second proportionality coefficient; represents a constant term; The error function of the mapping model is as follows: ； in, Represents the sum of the errors between the predicted values ​​and sample values ​​of all samples; represents the number of samples, The sample size index representing the sample; Indicates The sample value of samples, Indicates The predicted value of samples; The concentration of each fault-indicating gas is divided into a maximum warning concentration by an expert method, and the predicted value of the mapping model is normalized to obtain a normalized value, which represents the risk probability of the corresponding transformer fault; The sample value represents the concentration of the fault-indicating gas dissolved in the actual sample; the predicted value represents the concentration of the fault-indicating gas dissolved in the sample predicted by the mapping model; the custom coefficient in the formula of the mapping model The values ​​are determined by optimization algorithms; The fault chain module is used to perform spatiotemporal analysis on the risk of each type of transformer fault, verify the judgment of the risk prediction module, and obtain the transformer verification judgment, as follows: According to the risk probability of each type of transformer fault, the transformer faults are sorted, and the transformer faults higher than a preset probability threshold are determined to obtain a first candidate fault set; Based on prior knowledge, a fault relationship chain of the transformer is constructed, and the fault relationship chain represents the logical sequence of the transformer fault occurrence; Fitting the faults in the first candidate fault set with the fault relationship chain; If there is at least one fault relationship chain in the first candidate fault set, the first candidate fault set obtains a first verification judgment; According to the fault relationship chains existing in the first candidate fault set, determine the starting fault and the ending fault of each fault relationship chain; If the risk probability from the starting fault to the ending fault in each fault relationship chain remains in a decreasing state, the first candidate fault set obtains a second verification judgment; The planning module schedules the maintenance and management of transformers based on the transformer verification judgment.

2. The AI ​​early warning system based on monitoring of dissolved gas in transformer oil according to claim 1 is characterized in that: The solubility of dissolved gases in transformer oil is clustered to obtain the precipitation temperatures of multiple dissolved gases, including: Set the precipitation temperature range of the precipitation point; Obtaining the solubility of each dissolved gas in transformer oil within the precipitation temperature range; Classify dissolved gases into fault indicating gases and irrelevant gases; According to the solubility, determine the dissolution fluctuation temperature of each dissolved gas in the precipitation temperature range; The dissolution fluctuation temperatures are sorted to determine the dissolution fluctuation temperature corresponding to the fault indicating gas; the dissolution fluctuation temperatures of irrelevant gases are screened by the dissolution fluctuation temperature corresponding to the fault indicating gas to obtain the corresponding dissolution fluctuation temperature in the irrelevant gas, and the corresponding dissolution fluctuation temperature is used as the first precipitation temperature; The dissolution fluctuation temperatures of each fault indicating gas are sorted, and the dissolution fluctuation temperatures of each fault indicating gas within the precipitation temperature range are respectively used as the second precipitation temperature to the Nth precipitation temperature.

3. The AI ​​early warning system based on monitoring of dissolved gas in transformer oil according to claim 1 is characterized in that: Capacitance monitoring devices are set at multiple precipitation points in transformer oil, and the capacitance parameters of each precipitation point are collected in real time, including: The precipitation point includes a needle body, and a plurality of needle bodies for heating the transformer oil are arranged at fixed intervals at the bottom of the transformer oil, and each needle body is respectively set with a different precipitation temperature; A capacitance monitoring device is fixedly installed above the needle body, and the capacitance monitoring device collects capacitance parameters of the precipitation point at fixed time intervals.

4. The AI ​​early warning system based on monitoring of dissolved gas in transformer oil according to claim 1 is characterized in that: The custom coefficients in the mapping model formula are all determined by optimization algorithms, including: Randomly generate an initial population with R individuals that meet the constraints, and initialize the current number of iterations to 1; Encode the individuals in the population, and the encoding vector is expressed as: ; The error function of the mapping model is used as the objective function to obtain the objective function value of each individual; According to the objective function value of the individual, the individuals are sorted from small to large, and a fixed proportion of individuals with smaller objective function values ​​in the population are retained. The remaining individuals are updated as parents until the preset number of iterations is reached to obtain the corresponding individuals, and the corresponding individuals are used as the values ​​of the custom coefficients in the formula of the mapping model.

5. The AI ​​early warning system based on monitoring of dissolved gas in transformer oil according to claim 4 is characterized in that: Update the remaining individuals as parents, including: Pause Update Factor , is a random number used to obtain a random value for each parent before iteratively updating the parent. ;when When the preset pause value is less than the default pause value, the iterative update of the corresponding parent generation will be suspended; Updates to the parent include global updates, local updates, and convergence updates; The formula for global update is as follows: ； in, Indicates The update vector of the individual in the iteration, Indicates The update vector of the individual in the iteration, Indicates The individual with the lowest global objective function value in the iteration, represents a natural constant, represents the exponential coefficient, Represents a random number, , represents the preset number of iterations, t represents the current number of iterations, represents the proportionality coefficient; The formula for local update is as follows: ;in, represents the scaling factor, represents random numbers that conform to the standard normal distribution, Indicates the number of iterations corresponding to the previous iteration of the current iteration number t; The formula for convergence update is as follows: ; Represents the step size coefficient.

6. The AI ​​early warning system based on monitoring of dissolved gas in transformer oil according to claim 1 is characterized in that: Scheduling transformer maintenance and management based on transformer verification judgment, including: In response to the transformer having a risk probability of transformer failure, the transformer oil is tested by starting a gas chromatograph; In response to the transformer of the first verification judgment, the transformer oil is tested by starting the gas chromatograph and the person in charge of the transformer is notified immediately; In response to the transformer judged by the second verification, the transformer oil is measured by starting a gas chromatograph and the transformer is immediately shut down for maintenance.

7. An AI early warning method based on monitoring of dissolved gas in transformer oil, characterized in that: Applicable to the system according to any one of claims 1 to 6, comprising: Step 1, clustering the solubility of dissolved gas in transformer oil to obtain precipitation temperatures of multiple dissolved gases, specifically including: A plurality of precipitation points are respectively set in the transformer oil, and a precipitation temperature is respectively set for each precipitation point, so that the dissolved gas in the transformer oil is precipitated respectively; Step 2, setting a capacitance monitoring device at multiple precipitation points in the transformer oil, and collecting capacitance parameters of each precipitation point in real time; Step 3, comparing the capacitance parameters of multiple precipitation points at the same time, determining the capacitance difference of the precipitation points at adjacent precipitation temperatures, the capacitance difference being used to approximately reflect the concentration of dissolved gas in a transformer oil; Step 4, performing fault analysis on the transformer according to the capacitance difference of precipitation points at adjacent temperatures, and determining the risk of each type of transformer fault; Step 5, perform spatiotemporal analysis on the risk of each type of transformer fault, verify the judgment of the risk prediction module, and obtain a verified judgment of the transformer; Step 6, scheduling maintenance and management of the transformer based on the transformer verification judgment.

Images