Smart transformer for power systems

Abstract

The application relates to the technical field of transformers, and discloses an intelligent transformer for a power system, which comprises a data receiving module, a parameter analysis module, a cooperative analysis module, an abnormal positioning module and an early warning and maintenance module; the data receiving module is used for collecting transformer data of each transformer; the parameter analysis module is used for calculating a comprehensive health index of each transformer; the cooperative analysis module is used for correcting the comprehensive health index of each transformer to obtain a health index; the abnormal positioning module is used for determining an abnormal transformer and an abnormal degree thereof; and the early warning and maintenance module is used for generating early warning information and a maintenance strategy; the application establishes a health index evaluation system by comprehensively analyzing a plurality of operation parameters, improves monitoring precision and reliability, comprehensively evaluates the health state of equipment under a multi-transformer parallel operation environment by analyzing inter-transformer circulating current and loss, and maintains the intelligent transformer according to the health state.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of transformers, and more particularly to an intelligent transformer for a power system. BACKGROUND

[0002] With the rapid development of power systems, the safe and reliable operation of transformers, as key power conversion equipment in power systems, has an important influence on the stability of the entire power grid. Traditional transformers mainly realize voltage transformation function, and lack real-time monitoring, diagnosis and early warning capabilities.

[0003] The existing Chinese patent with the publication number CN116937815A discloses an intelligent maintenance system for a power transformer, which comprises: a control terminal, which is the main control end of the system and is used to issue execution commands; a receiving module, which is used to receive real-time operation data of the transformer; a loading module, which is used to load safety parameters of the running state of the transformer service equipment; and an analysis module, which is used to receive the real-time operation data of the transformer received by the receiving module and the safety parameters of the running state of the transformer service equipment loaded in the loading module. The application can analyze and determine whether the running state of the power transformer is safe by combining the received real-time operation data of the transformer with the loaded safety parameters of the running state of the transformer service equipment, thereby achieving the purpose of maintaining the safe operation of the power transformer.

[0004] However, large data centers need to use N+X redundant transformer architecture, which may result in multiple transformers in parallel. A single transformer failure may cause a cascading downtime risk. The existing technology only evaluates single machines and does not consider the electrical coupling effect between multiple transformers. SUMMARY

[0005] The purpose of the present application is to provide an intelligent transformer for a power system to solve the above problems.

[0006] The present application provides an intelligent transformer for a power system, which comprises:

[0007] a data receiving module, configured to collect transformer data of each transformer;

[0008] a parameter analysis module, configured to calculate a load rate of each transformer according to the transformer data, identify a current working condition type according to the load rate and the transformer data, select a corresponding safety threshold according to the working condition type, and calculate a comprehensive health index of each transformer;

[0009] a cooperative analysis module, configured to analyze the coupling effect between parallel transformers and correct the comprehensive health index of each transformer to obtain a health index;

[0010] an abnormal positioning module, configured to perform group data analysis on the health indexes of multiple transformers and compare the health indexes with safety thresholds to determine an abnormal transformer and an abnormal degree thereof.

[0011] The early warning maintenance module generates early warning information and maintenance strategies based on the abnormal transformer and the abnormal degree thereof.

[0012] Further, the specific steps of selecting the corresponding safety threshold value include:

[0013] In step 201, a load rate is obtained by calculating a ratio of an actual apparent power to a rated capacity of the transformer in the transformer data, the actual apparent power being a square root of a sum of squares of an active power in the transformer data and a reactive power in the transformer data;

[0014] In step 202, a working condition type of the current transformer is identified based on the load rate and a harmonic content in the transformer data.

[0015] In step 203, a corresponding safety threshold value is selected according to the identified working condition type.

[0016] Further, the calculation method of the comprehensive health index of the transformer includes:

[0017] A hot spot temperature rise coefficient is calculated, the hot spot temperature rise coefficient being a ratio of a winding temperature in the transformer data to a rated winding temperature;

[0018] A life consumption rate is calculated using the Arrhenius equation, a default value of an activation energy related parameter and a reference temperature in the Arrhenius equation being set according to an execution standard of the transformer; the winding temperature is converted from Celsius to Kelvin, input into the Arrhenius equation, and the life consumption rate is output; wherein the activation energy related parameter is a ratio of an activation energy to a Boltzmann constant;

[0019] A weighted sum of the hot spot temperature rise coefficient, the life consumption rate, and a ratio of a current in the transformer data to a rated current of each transformer is calculated, the weighted sum representing the comprehensive health index of each transformer.

[0020] Further, the specific steps of the collaborative analysis module include:

[0021] In step 301, an absolute value of a current difference between two adjacent transformers is calculated to obtain a circulating current value;

[0022] In step 302, a circulating current loss is evaluated by a heat loss generated by the circulating current value on an equivalent resistance of the transformer, wherein the circulating current loss is a product of a square of the circulating current value multiplied by the equivalent resistance in the transformer data by a phase number multiple, the phase number multiple being related to a number of phase currents of the transformer;

[0023] In step 303, the health index is equal to the comprehensive health index plus a weighted value of the circulating current loss to a rated power ratio.

[0024] Further, the steps of the abnormal positioning module include:

[0025] Step 401, calculate the average value and standard deviation of all transformer health indicators, obtain the population mean and population standard deviation;

[0026] Step 402, determine whether each transformer is abnormal, and determine it as an abnormal transformer if any condition is met, including:

[0027] Calculate the deviation of the health indicator of the transformer relative to the population mean, when the deviation exceeds the preset multiple of the population standard deviation;

[0028] When the health indicator of the transformer exceeds the safety threshold under the current operating condition;

[0029] Step 403, for the determined abnormal transformer, use the Mahalanobis distance method to quantify its abnormality.

[0030] Further, the method of using the Mahalanobis distance method to quantify the abnormality includes:

[0031] Construct a parameter vector for each transformer, including: health indicators, oil temperature, winding temperature, load rate, and ring current percentage in transformer data, wherein the ring current percentage is the ratio of ring current value to rated current; standardize all parameters of the parameter vector to obtain a standardized parameter vector;

[0032] For the normal transformer population, calculate the mean vector and covariance matrix of the normal transformer population according to the standardized parameter vector, and calculate the Mahalanobis distance of each abnormal transformer according to the mean vector and the covariance matrix; that is, the difference between the standardized parameter vector of the abnormal transformer and the mean vector of the normal transformer population is weighted by the inverse matrix of the covariance matrix, and the square root is taken to obtain the final Mahalanobis distance value, which represents the abnormality, wherein the normal transformer population is a set of transformers that are not determined to be abnormal.

[0033] Further, the elements in the covariance matrix include the covariances between parameters, including: health indicators and oil temperature, health indicators and winding temperature, health indicators and load rate, health indicators and ring current percentage, oil temperature and winding temperature, oil temperature and load rate, oil temperature and ring current percentage, winding temperature and load rate, winding temperature and ring current percentage, and load rate and ring current percentage.

[0034] Further, the specific steps of the early warning maintenance module include:

[0035] According to the abnormality and the number of abnormal transformers, combined with the number of redundant transformers, the early warning level is divided into four levels, including: normal, low-level early warning, medium-level early warning, and high-level early warning;

[0036] For different early warning levels, maintenance strategies are generated in combination with the Mahalanobis distance value and the early warning level. The maintenance strategies include:

[0037] When the early warning level is normal, no processing is performed.

[0038] When the early warning level is low-level early warning, the load of the abnormal transformer is reduced until the early warning level is reduced to normal.

[0039] When the early warning level is medium-level early warning, load transfer is implemented to distribute the load of the abnormal transformer to the normal transformer until the early warning level is reduced to low-level early warning.

[0040] When the early warning level is high-level early warning, load switching is performed to transfer the load from the abnormal transformer to the normal transformer or the redundant transformer, and the abnormal transformer is isolated.

[0041] Further, when all transformers are normal or the Mahalanobis distance value of all abnormal transformers is less than the minimum value of low-level early warning, and the transformer group maintains complete redundancy, it is determined that the early warning level is normal. The complete redundancy is that X in the N+X architecture is unchanged, N in the N+X architecture represents the minimum number of transformers required to meet normal operation, and X represents the number of redundant transformers.

[0042] When the Mahalanobis distance value of any abnormal transformer is between the minimum value of low-level early warning and the maximum value of low-level early warning, the early warning level is low-level early warning.

[0043] When the Mahalanobis distance value of any abnormal transformer is between the minimum value of medium-level early warning and the maximum value of medium-level early warning, or the Mahalanobis distance value of two or more abnormal transformers simultaneously exceeds the minimum value of low-level early warning, or X=1, the early warning level is medium-level early warning.

[0044] When the Mahalanobis distance value of any abnormal transformer exceeds the maximum value of medium-level early warning, or the Mahalanobis distance value of two or more abnormal transformers simultaneously exceeds the minimum value of medium-level early warning, or X=0, the early warning level is high-level early warning.

[0045] Wherein the minimum value of low-level early warning is less than the maximum value of low-level early warning, the maximum value of low-level early warning is equal to the minimum value of medium-level early warning, and the minimum value of medium-level early warning is less than the maximum value of medium-level early warning.

[0046] If multiple early warning levels are met simultaneously, the highest early warning level is output.

[0047] Further, the transformer data includes real-time operating parameters, rated parameters, and safety thresholds, including light-load operating threshold, full-load operating threshold, harmonic environment threshold, and normal operating threshold.

[0048] The beneficial effects of the present application are: by comprehensively analyzing multiple operating parameters, establishing a health index evaluation system, comprehensively evaluating the operating state of the transformer, improving the monitoring accuracy and reliability; by analyzing the inter-loop current and its loss of the transformer, comprehensively evaluating the health state of the equipment under the parallel operation environment of multiple transformers.

[0049] The transformer operating condition can be intelligently identified according to the load rate and harmonic content, the corresponding safety threshold can be selected, the accuracy of monitoring can be improved; the statistical analysis method such as Mahalanobis distance is adopted, the abnormal transformer is locked based on group data analysis, the accuracy of fault positioning is improved; according to the abnormal degree and the characteristics of the redundant architecture, multi-level early warning and differentiated maintenance strategy are provided, and the maintenance resource configuration is optimized; for N+X redundant transformer architecture, load optimization suggestions are provided, the reliability is ensured while the energy efficiency is improved; through the deep learning method, complex features are automatically extracted, high-precision fault diagnosis and predictive maintenance are realized. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 is an example of a module for the intelligent transformer for the power system of the present application Figure One ;

[0051] Figure 2 is an example of a parameter analysis module for the intelligent transformer for the power system of the present application

[0052] Figure 3 is an example of an abnormal positioning module for the intelligent transformer for the power system of the present application

[0053] Figure 4 is an example of a module for the intelligent transformer for the power system of the present application Figure Two . DETAILED DESCRIPTION

[0054] 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 merely meant to provide a better understanding of the subject matter described herein and can be changed in function and arrangement without departing from the scope of the present description. Various processes or components can be omitted, substituted, or added according to desired implementations. Additionally, features described with respect to some examples can be combined in other examples.

[0055] Example 1:

[0056] The intelligent transformer for the power system, as shown in Figure 1 includes:

[0057] The data receiving module 101 is configured to collect transformer data of each transformer, wherein the transformer data includes real-time operating parameters, rated parameters, and safety thresholds.

[0058] wherein the operating parameters specifically include:

[0059] voltage, the voltage value of the primary side and the secondary side of the transformer; current, the current value of the primary side and the secondary side of the transformer; active power; reactive power; winding temperature, the actual temperature of the winding of the transformer; oil temperature, the actual temperature of the oil of the transformer; harmonic content, the harmonic distortion rate of the power grid;

[0060] The rated parameters of the transformer include: rated capacity, rated voltage, rated current, rated frequency, rated winding temperature, rated power equivalent resistance, etc.

[0061] and the safety threshold values under each working condition: light load working condition threshold, full load working condition threshold, harmonic environment threshold and normal working condition threshold.

[0062] After the collected data is preliminarily verified, it is transmitted to the parameter analysis module 102 for further processing.

[0063] The parameter analysis module 102 calculates the load rate of each transformer according to the collected transformer data, identifies the current working condition according to the load rate and the transformer data, selects the corresponding safety threshold value according to the working condition, and finally calculates the comprehensive health index of each transformer.

[0064] Specifically as shown in Figure 2 , comprising:

[0065] Step 201, the load rate is a key indicator to measure the load level of the transformer. The load rate is obtained by calculating the ratio of the actual apparent power to the rated capacity of the transformer. Specifically, first, the active power and the reactive power of the transformer are obtained from the data receiving module 101, then the actual apparent power is calculated, wherein the actual apparent power is the square root of the sum of the squares of the active power and the reactive power, and finally the actual apparent power is divided by the rated capacity of the transformer to obtain the load rate.

[0066] The load rate is a dimensionless ratio, which reflects the current load level of the transformer. The higher the load rate, the closer the load of the transformer to its rated capacity; the lower the load rate, the lighter the load level of the transformer.

[0067] Step 202, based on the calculated load rate and the harmonic content in the transformer data, the working condition type of the current transformer is identified:

[0068] When the load rate is less than the light load threshold, it is determined as a light load working condition;

[0069] When the load rate is greater than the full load threshold, it is determined as a full load working condition;

[0070] When the total harmonic distortion rate is greater than the harmonic threshold, it is determined as a harmonic environment working condition;

[0071] When the load ratio is greater than the light load threshold and less than the full load threshold, and the total harmonic distortion rate is less than the harmonic threshold, it is determined to be a normal working condition.

[0072] The initial setting of these thresholds is based on the power industry standards and transformer characteristics. Specifically, the default value of the light load threshold is 0.3, which refers to engineering practice and traditional standards; the default value of the full load threshold is 0.7, which considers the upper limit of the load for long-term safe operation of the transformer; the default value of the harmonic threshold is 5%, which refers to the industry standard for power system harmonic limit. In practical applications, these thresholds will be fine-tuned according to specific transformer models and application scenarios.

[0073] The working condition recognition result will be passed to the subsequent module for dynamic safety threshold selection and health index calculation, ensuring accurate assessment and early warning according to the actual operating state of the transformer.

[0074] Step 203, according to the identified working condition, select the corresponding safety threshold: in light load working condition, the safety threshold is the light load working condition threshold; in full load working condition, the safety threshold is the full load working condition threshold; in harmonic environment working condition, the safety threshold is the harmonic environment threshold; in normal working condition, the safety threshold is the normal working condition threshold; for composite working conditions, such as meeting the full load and harmonic environment conditions at the same time, the most stringent safety threshold is adopted, and the safety threshold is the minimum value of each threshold.

[0075] Dynamic selection of safety threshold can adapt to the operating characteristics of transformer under different working conditions, improving the accuracy of monitoring.

[0076] Step 204, the transformer comprehensive health index is a comprehensive index to evaluate the operating state of each transformer, involving the weighted calculation of multiple parameters. First, calculate the hot spot temperature rise coefficient, which is the ratio of winding temperature to rated winding temperature. The hot spot temperature rise coefficient is a dimensionless value, when the value approaches or exceeds 1, it indicates that the transformer temperature approaches or exceeds the design limit, and there may be overheating risk.

[0077] Then, the life consumption rate is calculated, which is used to evaluate the aging rate of the transformer insulation material at the current temperature. The calculation is based on the Arrhenius equation, which describes the effect of temperature on the rate of a chemical reaction. In the calculation process, the activation energy related parameter in the Arrhenius equation, which is the ratio of activation energy to Boltzmann constant, is set according to the default value of the reference temperature based on the execution standard of the transformer, and the winding temperature is converted from Celsius to Kelvin. The life consumption rate is a dimensionless value, when the value is 1, it means that the aging rate of the insulation material at the current winding temperature is the same as at the reference temperature; when the value is greater than 1, it means that the aging is accelerated; when the value is less than 1, it means that the aging is slowed down. The activation energy related parameter and the reference temperature are set based on the recommended values in the IEEE C57.91-2011 standard for the relationship between the transformer hot spot temperature and the life, and the default value of the activation energy related parameter is 15000K, and the default value of the reference temperature is 383K (110℃).

[0078] The comprehensive health index of each transformer is calculated, which takes into account the weighted sum of the hot spot temperature rise coefficient, the life consumption rate and the ratio of current to rated current. These three factors are given different weight coefficients to reflect their influence on the health status of the transformer. The initial setting of the weight coefficients is based on historical operation data analysis, usually the hot spot temperature rise coefficient weight is set to 0.4, the life consumption rate weight is set to 0.3, and the current ratio weight is set to 0.3. These weights can be dynamically adjusted according to actual operation data.

[0079] The comprehensive health index is a dimensionless value, usually between 0 and 10, the larger the value, the more the transformer state deviates from the normal range. The calculated comprehensive health index of each transformer is passed to the collaborative analysis module 103 for further correction.

[0080] The collaborative analysis module 103 mainly analyzes the coupling effect between parallel transformers and corrects the comprehensive health index of each transformer to obtain the health index. In the transformer redundancy architecture, the mutual influence between transformers cannot be ignored.

[0081] Step 301, when the transformers are running in parallel, due to the small differences in impedance, voltage and other parameters of each transformer, circulating current will be generated between the transformers. The circulating current value is obtained by calculating the absolute value of the current difference between two adjacent transformers. The existence of circulating current will cause additional loss and temperature rise of the transformer, which is an important parameter for evaluating the operating state of parallel transformers.

[0082] At step 302, the circulating current loss is evaluated by calculating the heat loss generated by the circulating current on the equivalent resistance of the transformer. The circulating current loss is the square of the circulating current multiplied by the equivalent resistance in the transformer data, and the phase multiple is determined according to the number of phase currents of the transformer, for example, three-phase current, and the phase multiple is 3 times. The equivalent resistance value is usually between 0.01Ω and 0.1Ω, depending on the capacity and design characteristics of the transformer. The calculation result of the circulating current loss is used to evaluate the influence of the circulating current on the operating efficiency and service life of the transformer.

[0083] At step 303, the comprehensive health index of each transformer is modified considering the circulating current factor. The health index obtained after modifying the comprehensive health index of each transformer is equal to the comprehensive health index plus the weighted value of the ratio of circulating current loss to rated power. The initial value of the weight coefficient of the ratio of circulating current loss to rated power is set to 0.2, which is determined based on historical operation data analysis and reflects the degree of influence of circulating current loss on the health status of the transformer. The modified comprehensive health index is used as the health index.

[0084] The modified comprehensive health index more comprehensively reflects the actual state of the transformer in the parallel operation environment, and is transmitted to the abnormality detection and fault location module 104 for analysis. In this way, those transformers that may have potential risks due to the influence of circulating current can be identified, even if their individual parameters are within the normal range. If the influence of circulating current is not identified, the following consequences may occur:

[0085] The additional loss caused by circulating current is ignored, which accelerates the aging of transformer insulation and significantly shortens the service life of the transformer;

[0086] The heat generated by circulating current may cause local hot spots, which may cause insulation breakdown failure if they exist for a long time;

[0087] When the load changes dynamically, the circulating current may suddenly increase, causing protection devices to malfunction or refuse to act;

[0088] Circulating current loss reduces overall energy efficiency and increases operating costs;

[0089] Due to the failure to discover the circulating current problem in time, the best maintenance opportunity may be missed, and ultimately a catastrophic failure of the transformer may occur, affecting the reliability and safety of the entire power system.

[0090] The abnormality location module 104 performs group data analysis on the health indexes of multiple transformers to determine abnormal transformers and their abnormality degrees. In a multi-transformer system, abnormal equipment can be more accurately identified by comparative analysis.

[0091] As shown in the specific embodiment of FIG. 4, the method comprises the following steps: Figure 3

[0092] ​Step 401, first calculate the average value of all transformer health indicators, get the population mean, that is, the sum of all transformer health indicators divided by the total number of transformers. This average value represents the overall health level of the transformer population under the current operating environment, and is an important reference benchmark for judging whether a single transformer is abnormal.

[0093] Next, calculate the standard deviation of the health indicators, get the population standard deviation, that is, the square root of the average of the square sum of the difference between each transformer health indicator and the population mean. The standard deviation reflects the dispersion of the health status of the transformer population, the smaller the standard deviation, the closer the state of each transformer, the larger the standard deviation, the more obvious the difference between the states of the transformers.

[0094] The calculation of these statistical values is based on the data of all online running transformers, which is usually updated every 5-15 minutes to ensure that the dynamic changes of the transformer population can be reflected in time. The calculation results are stored in the system database, which is used for current abnormality judgment and also as historical data for future analysis.

[0095] Step 402, based on two conditions to determine whether each transformer is abnormal:

[0096] The first condition is a statistical principle, to calculate the deviation of the transformer health indicator from the population mean, when the deviation exceeds the preset multiple of the population standard deviation; the default value of the preset multiple is 2. This judgment is based on the normal distribution theory, in the normal distribution, the probability of data falling outside the range of mean ± 2 times standard deviation is about 5%, so these data points can be considered as statistically abnormal values. Adopt 2 times standard deviation as the initial threshold value, which can be adjusted through system setting, usually between 1.5-3, depending on the user's requirement for abnormal sensitivity.

[0097] The second condition is the principle of engineering practice, when the health indicator of the transformer exceeds the safety threshold value under the current working condition. The safety threshold value is pre-set according to the transformer model, operating environment and industry standard, and is dynamically corrected by the threshold adjustment module. The initial safety threshold value is usually set based on the technical specifications provided by the transformer manufacturer and the power industry standards, the light load threshold value under light load working condition is about 7.5, the full load threshold value under full load working condition is about 8.5, the harmonic environment threshold value under harmonic environment is about 8.0, and the normal working condition threshold value under normal working condition is about 8.0.

[0098] The "or" logic is adopted, that is, any condition is met to determine that it is abnormal, this design enhances the sensitivity of the system, which can capture different types of abnormal conditions. The judgment result is stored in the form of Boolean value, and is transmitted to the abnormal source positioning step, where the Boolean value includes normal and abnormal.

[0099] Step 403, for the transformer determined as abnormal, further quantify its abnormality degree using Mahalanobis distance method.

[0100] Mahalanobis distance is a multi-dimensional distance calculation method considering the correlation between variables. The multiple parameters of each transformer are regarded as a point in a multi-dimensional space, and the distance from this point to the center of the normal transformer group is calculated. Unlike simple Euclidean distance, Mahalanobis distance considers the correlation between parameters through the covariance matrix, which can more accurately identify abnormal points in multi-dimensional space.

[0101] In actual calculation, first construct the parameter vector of each transformer, including health index, oil temperature, winding temperature, load rate, and loop current percentage. Specifically, the following elements of multiple transformers are included:

[0102] Health index reflects the comprehensive health status of the transformer; load rate is the ratio of actual load to rated capacity; loop current percentage is the ratio of loop current value to rated current;

[0103] For example, the parameter vector of a transformer is [7.2, 65°C, 85°C, 0.85, 8%], indicating that the health index is 7.2, the oil temperature is 65°C, the winding temperature is 85°C, the load rate is 85%, and the loop current is 8% of the rated current.

[0104] Since these parameters have different dimensions and numerical ranges, direct calculation will result in larger numerical parameters (such as temperature) dominating the covariance matrix, while smaller numerical parameters (such as load rate) are weakened. To solve this problem, all parameters are standardized before calculating Mahalanobis distance, obtaining the standardized parameter vector:

[0105] Apply Z-score standardization to each parameter, subtract the mean of the parameter in the normal transformer group from the original parameter value, and divide by the standard deviation.

[0106] After standardization, all parameters are converted to dimensionless values with similar numerical ranges, ensuring that each parameter has the same weight in subsequent calculations.

[0107] The normal transformer group is the set of transformers not determined as abnormal.

[0108] For the normal transformer group, calculate its mean vector and covariance matrix:

[0109] The mean vector is the arithmetic mean of the standardized parameter vectors of all normal transformers. For example, if there are 10 normal transformers, the average of each standardized parameter of the 10 transformers is calculated to obtain the mean vector.

[0110] The covariance matrix is obtained by calculating the covariance between each normalized parameter. The diagonal elements represent the variance of each parameter, and the non-diagonal elements represent the covariance between parameters, reflecting the correlation between parameters.

[0111] In this embodiment, the covariance between parameters contained in the covariance matrix includes: the health index and the oil temperature, the health index and the winding temperature, the health index and the load rate, the health index and the circulating current percentage, the oil temperature and the winding temperature, the oil temperature and the load rate, the oil temperature and the circulating current percentage, the winding temperature and the load rate, the winding temperature and the circulating current percentage, and the load rate and the circulating current percentage.

[0112] If there are 10 normal transformers, each transformer has 5 parameters, including the health index, the oil temperature, the winding temperature, the load rate, and the circulating current percentage, then the covariance matrix is a 5x5 matrix. The (i,j) element of the matrix represents the covariance between the i-th normalized parameter and the j-th normalized parameter, which is calculated based on the data of the 10 transformers. The covariance matrix is updated every 15 minutes to adapt to the dynamic changes of the transformer operating state.

[0113] After obtaining the mean vector and the covariance matrix, the Mahalanobis distance of each abnormal transformer is calculated based on the mean vector and the covariance matrix. That is, the difference between the normalized parameter vector of the abnormal transformer and the mean vector of the normal transformer group is weighted by the inverse matrix of the covariance matrix, and the final Mahalanobis distance value is obtained by taking the square root. The weighted calculation by the inverse matrix of the covariance matrix is to use the inverse matrix of the covariance matrix as the weight to adjust the contribution of different parameters and suppress the interference of parameters with high correlation or small variance.

[0114] The Mahalanobis distance calculation result is a dimensionless value, usually in the range of 0-10, but theoretically there is no upper limit. The larger the value, the higher the abnormality.

[0115] The abnormality detection result is transmitted to the threshold adjustment module and the early warning and maintenance module 105 for subsequent parameter optimization and early warning generation. The abnormality detection result is also recorded in the historical database for long-term trend analysis and model optimization.

[0116] The early warning and maintenance module 105 generates differentiated early warning information and maintenance strategies based on the abnormal transformer and its abnormality degree, according to the characteristics of the redundant transformer architecture, to ensure system reliability while optimizing maintenance resource allocation.

[0117] Specifically, it includes:

[0118] Step 501, according to the Mahalanobis distance value (abnormality degree) and the number of abnormal transformers, combined with the characteristics of the redundant transformer architecture, the early warning level is divided into four levels, including normal, low-level early warning, medium-level early warning and high-level early warning:

[0119] When all transformers are normal or the Mahalanobis distance value of all abnormal transformers is less than the minimum value of the low-level warning, and the transformer group maintains complete redundancy, the complete redundancy is X unchanged in the N+X architecture, and the warning level is determined to be normal.

[0120] When the Mahalanobis distance value of any abnormal transformer is between the minimum value and the maximum value of the low-level warning, the default values of the minimum value and the maximum value of the low-level warning are 3.0 and 5.0 respectively, or the redundancy is slightly reduced, the redundancy is reduced by 1 in the N+X architecture, but X≥2, and the warning level is low-level warning.

[0121] When the Mahalanobis distance value of any abnormal transformer is between the minimum value and the maximum value of the medium-level warning, the default values of the minimum value and the maximum value of the medium-level warning are 5.0 and 7.0 respectively, or the Mahalanobis distance value of two or more abnormal transformers simultaneously exceeds the minimum value of the low-level warning, or the redundancy is significantly reduced, the redundancy is X=1 in the N+X architecture, and the warning level is medium-level warning.

[0122] When the Mahalanobis distance value of any abnormal transformer exceeds the maximum value of the medium-level warning, or the Mahalanobis distance value of two or more abnormal transformers simultaneously exceeds the minimum value of the medium-level warning, or the redundancy is completely lost, the redundancy is X=0 in the N+X architecture, and the warning level is high-level warning.

[0123] If multiple warning levels are met simultaneously, the highest warning level is output. The determination of the warning level takes into account the abnormality degree of a single transformer, the number of abnormal transformers, and the redundancy state of the entire transformer group.

[0124] In the N+X architecture, N represents the minimum number of transformers required for normal operation, and X represents the number of redundant transformers. For example, the N+2 architecture means that N transformers are required for normal operation of the system, and 2 redundant transformers are provided, and the total number of transformers is N+2. This architecture design ensures that the system can still operate normally when some transformers fail.

[0125] In step 502, different maintenance strategies are automatically generated for different warning levels in combination with the Mahalanobis distance value and the characteristics of the redundancy architecture. The maintenance strategies include:

[0126] When the warning level is normal, routine monitoring is performed according to the planned periodic maintenance, and no special treatment is required.

[0127] When the warning level is low, optimize load distribution, reduce the load of abnormal transformers, and reduce the load proportion in direct proportion to the Mahalanobis distance value, which is three times the Mahalanobis distance value divided by 100, usually 10%-20%; arrange non-emergency maintenance, which can be carried out during the planned maintenance window; it is suggested to complete the maintenance within 30 days without affecting the normal operation of the system; prepare spare parts and update the maintenance plan;

[0128] When the warning level is medium, implement load transfer, distribute the load of abnormal transformers to normal transformers or redundant transformers; arrange priority maintenance, which is suggested to be completed within 7 days; start the standby transformer to restore the redundancy of the system.

[0129] When the warning level is high, immediately perform load switching to transfer critical loads from abnormal transformers; arrange emergency maintenance, which is suggested to be handled within 24 hours; isolate abnormal transformers to prevent cascading failures; start the emergency plan to allocate temporary equipment to ensure power supply reliability.

[0130] Step 503, for redundant transformer architecture, provide load optimization suggestions based on Mahalanobis distance value, ensure reliability while improving energy efficiency:

[0131] When the Mahalanobis distance value of all abnormal transformers is less than 2.0, it is suggested to implement load balancing strategy to make the load rate of each transformer close, prolonging the overall service life.

[0132] When the Mahalanobis distance value of the abnormal transformer is detected to exceed 3.0, according to the degree of exceeding the threshold value, a gradual load reduction strategy is adopted to ensure that the abnormal transformer operates within a safe range. Specifically, for every 1.0 threshold value exceeded, it is suggested to reduce the load by 10%.

[0133] Real-time assessment of current redundancy capacity, combined with Mahalanobis distance value to predict possible capacity shortage risk.

[0134] During the low load period, it is suggested to appropriately concentrate the load, making some transformers run at full capacity, and only select transformers with a Mahalanobis distance value below 1.5, while the remaining transformers reduce load or standby, improving overall energy efficiency.

[0135] For example, in an N+2 architecture, if a transformer is detected to have a Mahalanobis distance value of 4.5, it is suggested to reduce the load of the transformer by 15% and distribute the load to the normal transformer with the lowest Mahalanobis distance value, while suggesting to arrange maintenance during the next maintenance window.

[0136] This redundancy architecture performance evaluation based on Mahalanobis distance value and historical data can continuously optimize early warning and maintenance strategies, improving the reliability and economy of the entire transformer group.

[0137] Embodiment 2:

[0138] On the basis of embodiment 1, this embodiment introduces a deep learning method to improve the accuracy and prediction ability of transformer health state evaluation, as shown in Figure 4 The embodiment mainly includes:

[0139] The data receiving module 601 is similar to that of embodiment 1, and is used to collect transformer data of each transformer. In addition to the parameters mentioned in embodiment 1, the following real-time operation parameters are added in this embodiment:

[0140] Partial discharge data, which reflects the index of insulation deterioration degree;

[0141] Transformer vibration data, which reflects the index of transformer mechanical state;

[0142] Gas analysis data, which refers to the content of dissolved gas in oil, including H2, CH4, C2H2, etc.

[0143] The data sources include built-in sensors of transformers, online monitoring equipment and periodic manual sampling analysis results. The data is collected in real time through industrial Ethernet, MODBUS protocol or wireless sensor network, and the sampling frequency is set according to the importance of parameters. Key parameters such as temperature are sampled every 5 minutes, while gas analysis data may be updated once a day.

[0144] The data processing module 602 processes the collected transformer data to obtain time window data; the time window data improves the data quality and provides high-quality input for the fault prediction model.

[0145] The embodiment mainly includes:

[0146] The original transformer data is subjected to outlier detection and processing, and abnormal data points are eliminated or corrected to obtain corrected data. The outlier detection uses a method based on mean and standard deviation, and when a data point deviates from the mean by more than 3 times the standard deviation, it is marked as an outlier. These outliers are corrected by linear interpolation or adjacent value replacement method.

[0147] The initial threshold setting is based on historical data statistical analysis. For example, the abnormal judgment threshold of temperature parameter is initially set to 3 times the standard deviation of the mean, while for parameters with large fluctuations such as load current, it may be set to 3.5 times the standard deviation. These thresholds will be automatically adjusted every quarter according to the actual operation.

[0148] The corrected data of different dimensions are unified to the same scale to obtain normalized data, which is convenient for model training. The normalization uses the minimum-maximum standardization method to map each parameter data to the interval of 0 to 1. For example, temperature data may range from -20°C to 150°C, and after normalization processing, it is mapped to the range of 0 to 1.

[0149] The normalization parameters are determined based on the equipment technical specifications, such as the minimum and maximum values of temperature, which are the lowest and highest working temperatures allowed by the equipment. These normalization parameters will be updated periodically to adapt to changes in the data distribution.

[0150] The continuous time series data in the normalized data is divided into time window data with a fixed length, which is used as the input of the fault prediction model. The data in the last period of time (such as 24 hours) is organized into time windows, and the window length is set according to the change characteristics of different parameters. Parameters that change rapidly, such as temperature, may use a 4-hour window, while parameters that change slowly, such as gas content in oil, may use a 7-day window.

[0151] The sliding step length of the time window is usually set to 25% of the window length, ensuring that adjacent windows have enough overlap to capture continuous change trends. The window size and sliding step length can be dynamically adjusted by the parameter optimization module to achieve optimal performance.

[0152] The fault prediction module 603 is the core of the present embodiment; a fault prediction model is constructed, which inputs time window data and outputs health index prediction values and fault type probability distribution.

[0153] The present embodiment adopts a hybrid model combining long short-term memory network (LSTM) and convolutional neural network (CNN). The following layers are included:

[0154] The input layer receives time window data;

[0155] The convolutional layer uses one-dimensional convolution to extract local features;

[0156] The LSTM layer captures time series features, with 128 hidden units;

[0157] The fully connected layer maps the output of the LSTM layer to the final prediction result;

[0158] The output layer outputs health index prediction values and fault type probability distribution, where the health index prediction value is a number between 0 and 1, representing the overall health status of the transformer, and the closer the value is to 1, the better the health status; the fault type probability distribution is the probability value of four main fault types, including insulation aging, overheating fault, mechanical fault and circulating current loss, and the probability value of each fault type is between 0 and 1, and the sum of the four probabilities is 1.

[0159] The model is implemented using the TensorFlow framework and deployed on edge computing devices to ensure real-time response capability. The convolutional layer uses the ReLU activation function to extract local features, the LSTM layer captures long-term dependencies, and the fully connected layer maps the features to the prediction results.

[0160] The model training adopts a composite loss function, considering both the health index prediction error and the fault type classification error. The health index prediction loss adopts the mean square error loss, which is the average of the square of the difference between the predicted value and the true value. The fault type classification loss adopts the cross-entropy loss, which is the cross-entropy between the true label and the predicted probability distribution.

[0161] The total loss function is a weighted combination of the health index prediction loss and the fault type classification loss, where the weight coefficient of the health index prediction loss is initially set to 0.7, and the weight coefficient of the fault type classification loss is initially set to 0.3. The weight ratio of the two losses is initially set to 7:3, which can be dynamically adjusted according to the actual prediction effect.

[0162] The weight coefficients of the loss function are initially set based on expert experience and optimized on the validation set through grid search method. These weights are re-evaluated every quarter using the latest data to ensure optimal model performance.

[0163] The model training adopts the mini-batch stochastic gradient descent method, with a batch size of 64, an initial learning rate of 0.001, and a training round of 100. To prevent overfitting, L2 regularization method and early stopping strategy are adopted.

[0164] The training data sources include historical operation data, simulation data and industry public data sets. The initial model is pre-trained on a comprehensive data set containing multiple fault scenarios, and then fine-tuned using the historical data of a specific transformer. The model is updated every month using newly collected data to ensure that the model adapts to equipment aging and environmental changes.

[0165] The fault diagnosis module 604, based on the output of the fault prediction model, combines domain knowledge to comprehensively diagnose the health status of the transformer and obtain the fault severity.

[0166] The fault diagnosis module 604 can identify the following several typical fault types:

[0167] Insulation aging, mainly manifested as increased partial discharge, abnormal dissolved gas in oil;

[0168] Overheating fault, mainly manifested as abnormal hot spot temperature, oil temperature rise;

[0169] Mechanical fault, mainly manifested as vibration anomaly, noise increase;

[0170] Loop current loss, mainly manifested as transformer current imbalance, abnormal temperature rise.

[0171] The probability distribution of fault types is calculated by the Softmax function of the model output layer, and the fault type with the highest probability is taken as the diagnosis result. The initial fault recognition threshold is set to 0.6, that is, when the probability of a certain fault type exceeds 60%, it is determined as this type of fault. This threshold can be adjusted according to actual operation experience, usually between 0.5-0.7.

[0172] The fault feature library is established based on power industry standards and historical fault cases, and contains typical feature patterns of various faults. The current state is compared with the fault feature library through a fuzzy matching algorithm to improve the accuracy of diagnosis.

[0173] The fault severity assessment is based on health indicators and fault probability to calculate the fault severity index. The assessment process considers health indicators and maximum fault type probability.

[0174] The fault severity is divided into four levels, including minor fault, moderate fault, severe fault, and critical fault, wherein:

[0175] The fault severity index interval of minor fault is [0, 0.3); the fault severity index interval of moderate fault is [0.3, 0.6); the fault severity index interval of severe fault is [0.6, 0.8); and the fault severity index interval of critical fault is [0.8, 1.0].

[0176] The severity threshold is set based on power industry standards and equipment manufacturer recommendations, and is adjusted according to the importance of the equipment. The threshold for critical transformers may be more stringent, for example, adjusting the lower limit of the "severe fault" level to 0.55.

[0177] The warning control module 605 generates different levels of warning information and control suggestions based on the fault severity.

[0178] Based on the fault severity, the warning levels are divided into low-level warning, medium-level warning, high-level warning, and emergency warning; wherein, the low-level warning corresponds to minor fault, prompting the operation and maintenance personnel to pay attention; the medium-level warning corresponds to moderate fault, suggesting to arrange maintenance; the high-level warning corresponds to severe fault, suggesting immediate maintenance; and the emergency warning corresponds to critical fault, suggesting immediate shutdown for processing.

[0179] The warning threshold is initially set based on industry standards, and is adjusted according to the importance of the equipment and the operating environment. For example, for transformers supporting critical business, the trigger threshold of medium-level warning may be reduced from 0.3 to 0.25 to issue an early warning. These thresholds will be periodically evaluated and adjusted according to the historical warning accuracy.

[0180] For different fault types and severity, specific control recommendations are generated. For example: for insulation aging, recommend checking insulation condition and replacing insulation material if necessary; for overheating fault, recommend reducing load and strengthening cooling; for mechanical fault, recommend checking fasteners and checking vibration sources; for circulating current loss, recommend checking and adjusting transformer impedance matching condition.

[0181] Control recommendations are generated based on expert knowledge base, which contains various types of fault handling methods and best practices. Through case-based reasoning method, the handling scheme suitable for the current situation is extracted from historical successful cases. The expert knowledge base is updated every quarter, incorporating new handling experience and technological progress.

[0182] This embodiment significantly improves the accuracy of transformer health status assessment and fault prediction ability by introducing deep learning method. Compared with embodiment 1, this embodiment has the following advantages:

[0183] Strong feature extraction ability, fault prediction model can automatically extract complex time series features and the correlation between parameters; high precision fault diagnosis, can identify multiple fault types and evaluate fault severity; predictive maintenance, can predict potential faults in advance to support preventive maintenance.

[0184] This embodiment is particularly suitable for key places with extremely high requirements for transformer operating state, such as large data centers, financial institutions, medical facilities, etc.

[0185] The above describes the embodiments of the present application, but the embodiments are not limited to the specific implementation described above, which is only illustrative and not limiting. Those skilled in the art can make more forms of equivalent embodiments under the inspiration of the embodiments, which are all within the protection scope of the embodiments.

Claims

1. An intelligent transformer for power systems, characterized in that, include: The data receiving module is used to collect transformer data from each transformer. The parameter analysis module calculates the load rate of each transformer based on transformer data, identifies the current operating condition type based on the load rate and transformer data, selects the corresponding safety threshold based on the operating condition type, and calculates the comprehensive health index of each transformer. The collaborative analysis module analyzes the coupling effects between parallel transformers and corrects the comprehensive health index of each transformer to obtain the health index; the specific steps include: Step 301: Calculate the absolute value of the current difference between two adjacent transformers to obtain the circulating current value; Step 302: Evaluate the circulating current loss by the heat loss generated on the equivalent resistance of the transformer by the circulating current value, where the circulating current loss is the square of the circulating current value multiplied by the equivalent resistance in the transformer data, and the phase multiple is related to the number of phases of the transformer. Step 303: The health index equals the comprehensive health index plus the weighted value of the ratio of circulating current loss to rated power. The anomaly location module performs group data analysis and safety threshold comparison on the health indicators of multiple transformers to determine the abnormal transformers and the degree of their anomalies. The early warning and maintenance module generates early warning information and maintenance strategies based on abnormal transformers and the degree of their abnormality.

2. The intelligent transformer for power systems according to claim 1, characterized in that, The specific steps for selecting the appropriate security threshold include: Step 201: The load factor is obtained by calculating the ratio of the actual apparent power to the rated capacity of the transformer in the transformer data. The actual apparent power is the square root of the sum of the squares of the active power and the reactive power in the transformer data. Step 202: Identify the current operating condition type of the transformer based on the load rate and harmonic content in the transformer data; Step 203: Select the corresponding safety threshold based on the identified working condition type.

3. The intelligent transformer for power systems according to claim 1, characterized in that, The calculation methods for the comprehensive health indicators of a transformer include: Calculate the hot spot temperature rise coefficient, which is the ratio of the winding temperature to the rated winding temperature in the transformer data; The lifetime attrition rate is calculated using the Arrhenius equation. Default values ​​for the activation energy-related parameters and reference temperature in the Arrhenius equation are set according to the transformer's performance standard. The winding temperature is converted from Celsius to Kelvin, input into the Arrhenius equation, and the lifetime attrition rate is output. The activation energy-related parameters are the ratio of activation energy to the Boltzmann constant. The hot spot temperature rise coefficient, life consumption rate, and the weighted sum of the ratio of current to rated current in the transformer data are calculated for each transformer. The weighted sum is expressed as the comprehensive health index of each transformer.

4. The intelligent transformer for power systems according to claim 1, characterized in that, The steps of the anomaly localization module include: Step 401: Calculate the average and standard deviation of all transformer health indicators to obtain the population mean and population standard deviation; Step 402: Determine whether each transformer is abnormal. A transformer is determined to be abnormal if it meets any of the following conditions: Calculate the degree of deviation of the transformer's health index from the group mean. When the deviation exceeds a preset multiple of the group standard deviation; When the transformer's health indicators exceed the safety threshold under the current operating conditions; Step 403: For the identified abnormal transformers, use the Mahalanobis distance method to quantify their degree of abnormality.

5. The intelligent transformer for power systems according to claim 4, characterized in that, Methods for quantifying the degree of anomaly using Mahalanobis distance include: Construct a parameter vector for each transformer. The parameter vector includes: health indicators, oil temperature, winding temperature, load rate, and circulating current percentage from the transformer data, where the circulating current percentage is the ratio of the circulating current value to the rated current. Standardize all parameters in the parameter vector to obtain a standardized parameter vector. For a group of normal transformers, the mean vector and covariance matrix of the normal transformer group are calculated based on the standardized parameter vector. Then, the Mahalanobis distance of each abnormal transformer is calculated based on the mean vector and covariance matrix. Specifically, the difference between the standardized parameter vector of the abnormal transformer and the mean vector of the normal transformer group is weighted by the inverse of the covariance matrix, and the square root is taken to obtain the final Mahalanobis distance value. The Mahalanobis distance value represents the degree of abnormality, where the normal transformer group is the set of transformers that have not been identified as abnormal.

6. The intelligent transformer for power systems according to claim 5, characterized in that, The elements in the covariance matrix contain the covariance between each parameter, which includes: health index and oil temperature, health index and winding temperature, health index and load rate, health index and circulating current percentage, oil temperature and winding temperature, oil temperature and load rate, oil temperature and circulating current percentage, winding temperature and load rate, winding temperature and circulating current percentage, and load rate and circulating current percentage.

7. The intelligent transformer for power systems according to claim 1, characterized in that, The specific steps of the early warning and maintenance module include: Based on the degree of abnormality and the number of abnormal transformers, combined with the number of redundant transformers, the early warning level is divided into four levels, including: normal, low-level early warning, medium-level early warning, and high-level early warning. For different warning levels, maintenance strategies are generated by combining Mahalanobis distance values ​​and warning levels. These maintenance strategies include: No action is taken when the warning level is normal. When the warning level is low, reduce the load on the abnormal transformer until the warning level is reduced to normal. When the warning level is medium, load transfer is implemented, and the load of the abnormal transformer is distributed to the normal transformer until the warning level is reduced to low level. When the warning level is high, load switching is performed to transfer the load from the abnormal transformer to a normal or redundant transformer; the abnormal transformer is isolated.

8. The intelligent transformer for power systems according to claim 7, characterized in that, When all transformers are normal or the Mahalanobis distance of all abnormal transformers is less than the minimum value for low-level warning, and the transformer group maintains complete redundancy, the warning level is determined to be normal; the complete redundancy is that X remains unchanged in the N+X architecture, where N represents the minimum number of transformers to meet normal operation, and X represents the number of redundant transformers. When the Mahalanobis distance value of any abnormal transformer is between the minimum and maximum values ​​of the low-level warning, the warning level is low-level warning. When the Mahalanobis distance value of any abnormal transformer is between the minimum and maximum values ​​of the intermediate warning, or when the Mahalanobis distance values ​​of two or more abnormal transformers simultaneously exceed the minimum value of the low warning, or when X=1, the warning level is intermediate warning. When the Mahalanobis distance value of any abnormal transformer exceeds the maximum value of the intermediate warning, or when the Mahalanobis distance values ​​of two or more abnormal transformers simultaneously exceed the minimum value of the intermediate warning, or when X=0, the warning level is high-level warning. The minimum value for a low-level warning is less than the maximum value for a low-level warning; the maximum value for a low-level warning is equal to the minimum value for a medium-level warning; and the minimum value for a medium-level warning is less than the maximum value for a medium-level warning. If multiple warning levels are met simultaneously, the highest warning level will be output.

9. The intelligent transformer for power systems according to claim 1, characterized in that, Transformer data includes real-time operating parameters, rated parameters, and safety thresholds. Safety thresholds include light load thresholds, full load thresholds, harmonic environment thresholds, and normal operating thresholds.

Images