Main transformer state monitoring method and device based on multi-source fusion

Abstract

The invention discloses a main transformer state monitoring method and device based on multi-source fusion, and belongs to the technical field of power equipment state monitoring, and the method comprises the steps: collecting the multi-source data of a to-be-detected main transformer in a current time period; performing feature extraction on the multi-source data to form current feature data; acquiring historical feature data and reference feature data from a historical database; calculating a noise standard deviation of the current feature data based on the historical feature data, and generating a reliability weight; calculating the similarity between the current feature data and the reference feature data, and generating an importance weight; generating feature fusion data based on the current feature data, the reliability weight and the importance weight; and obtaining a state monitoring result of the to-be-detected main transformer based on the feature fusion data by using a machine learning model, and adjusting the operation state of the to-be-detected main transformer based on the state monitoring result. Therefore, by implementing the invention, the problem of low monitoring accuracy of the state of the main transformer in the prior art can be solved.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring technology, and in particular to a method and device for monitoring the condition of a main transformer based on multi-source fusion. Background Technology

[0002] The main transformer is a core piece of equipment in a power system, undertaking the important functions of voltage transformation, power distribution, and transmission. Its operating status directly affects the stability, security, and power supply quality of the power grid. Therefore, real-time status monitoring and fault diagnosis of the main transformer are of great significance.

[0003] Traditional methods for monitoring the condition of main transformers typically rely on a single data source (such as oil chromatography data or temperature data), which makes it difficult to comprehensively reflect the operating status of the equipment. Furthermore, current methods for fusing multi-source data often employ fixed weights, failing to correlate actual data with baseline data. Consequently, the fused data struggles to cope with complex real-world situations such as differences in data quality, varying indicator importance, and dynamic changes in the environment. Summary of the Invention

[0004] This invention provides a method and device for monitoring the condition of a main transformer based on multi-source fusion, which can solve the problem of low accuracy in the condition monitoring of main transformers in the prior art.

[0005] To address the aforementioned technical problems, this invention provides a method for monitoring the condition of a main transformer based on multi-source fusion, comprising:

[0006] Collect multi-source data of the main transformer under test in the current time period; wherein, the multi-source data includes equipment temperature data, oil chromatography data, vibration data and ambient temperature and humidity data;

[0007] Feature extraction is performed on the multi-source data to form current feature data; wherein, the current feature data includes temperature features, gas features, amplitude features, and temperature-humidity coupled features;

[0008] Obtain historical characteristic data and baseline characteristic data of the main transformer under test from the historical database;

[0009] Based on the historical feature data, calculate the noise standard deviation of the current feature data, and generate a reliability weight based on the noise standard deviation;

[0010] Calculate the similarity between the current feature data and the baseline feature data, and generate an importance weight based on the similarity.

[0011] Based on the current feature data, the reliability weight, and the importance weight, feature fusion data is generated;

[0012] Using a machine learning model, the status monitoring results of the main transformer under test are obtained based on the feature fusion data, and the operating status of the main transformer under test is adjusted based on the status monitoring results.

[0013] As a preferred embodiment, the step of extracting features from the multi-source data to form the current feature data includes:

[0014] The dynamic characteristics of temperature change are extracted from the equipment temperature data to form temperature features; wherein, the temperature features include the rate of temperature change and the standard deviation of temperature fluctuation;

[0015] Gas concentration change characteristics are extracted from the oil chromatography data to form gas features, wherein the gas features include gas concentration change rate and gas ratio;

[0016] Amplitude features are extracted from the vibration data; wherein, the amplitude features include the mean vibration amplitude, the peak-to-peak vibration amplitude, and the amplitude of the dominant frequency.

[0017] Extract temperature and humidity coupling features from the environmental temperature and humidity data; wherein, the temperature and humidity coupling features include a temperature and humidity coupling index;

[0018] The temperature change rate, the standard deviation of temperature fluctuation, the gas concentration change rate, the gas ratio, the mean vibration amplitude, the peak-to-peak vibration amplitude, the dominant frequency amplitude, and the temperature-humidity coupling index are determined as the current characteristic data.

[0019] As a preferred embodiment, the step of extracting dynamic characteristics of temperature changes from the device temperature data to form temperature features includes:

[0020] The rate of temperature change is calculated using the following formula:

[0021]

[0022] In the formula, f T Let T(t) be the rate of temperature change; Δt be the time interval; T(t) be the equipment temperature at time t; and T(t+Δt) be the equipment temperature at time t+Δt.

[0023] The standard deviation of temperature fluctuation is calculated using the following formula:

[0024]

[0025] In the formula, f T_std N1 represents the standard deviation of temperature fluctuation; N1 represents the number of equipment temperatures; T i Let i be the temperature of the i-th device; This represents the average temperature of the equipment.

[0026] As a preferred embodiment, the step of extracting gas concentration change characteristics from the oil chromatography data to form gas characteristics includes:

[0027] The gas concentration data of dissolved gases in several types of oil were determined from the oil chromatography data;

[0028] The gas concentration change rate and gas ratio were calculated based on the gas concentration data of dissolved gases in several types of oils.

[0029] The following formula is used to calculate the rate of change of gas concentration:

[0030]

[0031] In the formula, f G ΔG is the rate of change of gas concentration; Δt is the amount of change of gas concentration; Δt is the time interval.

[0032] The gas ratio is calculated using the following formula:

[0033]

[0034] In the formula, f ratio G represents the gas ratio. i G represents the gas concentration of the i-th gas; j Let be the gas concentration of the j-th gas.

[0035] As a preferred embodiment, the extraction of amplitude features from the vibration data includes:

[0036] The mean vibration amplitude is calculated using the following formula:

[0037]

[0038] In the formula, f V_mean N1 represents the average vibration amplitude; N2 represents the number of vibration frequencies. Let i be the i-th vibration frequency;

[0039] The peak-to-peak value of the vibration amplitude is calculated using the following formula:

[0040] f V_pp =max(V amp )-min(V amp )

[0041] In the formula, f V_pp The peak-to-peak value of the vibration amplitude; max(V amp ) is the maximum vibration frequency; min(V) amp () is the minimum vibration frequency;

[0042] The main frequency amplitude is calculated using the following formula:

[0043] fV_freq =max(|FFT(V) amp )|)

[0044] In the formula, f V_freq Main frequency amplitude; V amp is the vibration frequency; FFT(·) is the Fast Fourier Transform function.

[0045] As a preferred embodiment, the extraction of temperature and humidity coupling features from the environmental temperature and humidity data includes:

[0046] Extract the ambient temperature and ambient humidity from the ambient temperature and humidity data;

[0047] Based on the ambient temperature and humidity, the temperature-humidity coupling index is calculated using the following formula:

[0048] f THI =T env ×(1-k×H env )

[0049] In the formula, f THI T is the temperature and humidity coupling index; env For ambient temperature; H env is the ambient humidity; k is a preset parameter.

[0050] As a preferred embodiment, the step of calculating the noise standard deviation of the current feature data based on the historical feature data, and generating a reliability weight based on the noise standard deviation, includes:

[0051] For each feature value in the current feature data, several historical features are obtained from the historical feature data;

[0052] Calculate the mean of the feature values ​​based on several historical characteristics;

[0053] The eigenvalues ​​and the noise standard deviation calculated based on the mean of the eigenvalues;

[0054] After determining the noise standard deviation of each feature value in the current feature data, the reliability weight of each feature value is calculated according to the following formula:

[0055]

[0056] In the formula, Reliability i The reliability weight for the i-th feature value; NoiseLevel i is the noise standard deviation of the i-th eigenvalue; max(NoiseLevel) is the maximum noise standard deviation.

[0057] As a preferred embodiment, the step of calculating the similarity between the current feature data and the benchmark feature data, and generating an importance weight based on the similarity, includes:

[0058] For each feature value in the current feature data, the corresponding benchmark feature value is obtained from the benchmark feature data;

[0059] The similarity between the feature value and the benchmark feature value is calculated using the normalized difference method.

[0060] The similarity is determined as the importance weight of the feature value;

[0061] The importance weights are calculated using the following formula:

[0062]

[0063] In the formula, Importance i f is the importance weight of the i-th feature value; i f is the i-th eigenvalue; i0 Let be the baseline eigenvalue of the i-th eigenvalue.

[0064] As a preferred embodiment, generating feature fusion data based on the current feature data, the reliability weight, and the importance weight includes:

[0065] Based on the reliability weight and the importance weight, generate the feature weights for each feature value in the current feature data;

[0066] Based on each feature value and its feature weight in the current feature data, linear weighting is performed to generate feature fusion data;

[0067] The feature weights are generated using the following formula:

[0068]

[0069] In the formula, ω i The feature weights for the i-th feature value; Reliability i The reliability weight for the i-th feature value; Importance i The importance weight for the i-th feature value; Reliability j The reliability weight for the j-th eigenvalue; Importance j The importance weight of the j-th feature value; M α is the number of eigenvalues; α is the first harmonic coefficient; β is the second harmonic coefficient;

[0070] The feature fusion data is calculated using the following formula:

[0071]

[0072] In the formula, F represents the feature fusion data; ω i f is the feature weight of the i-th feature value; i Let i be the i-th eigenvalue; M The number of eigenvalues.

[0073] Accordingly, the present invention provides a main transformer condition monitoring device based on multi-source fusion, comprising: a data acquisition module, a feature extraction module, a data acquisition module, a reliability weight calculation module, an importance weight calculation module, a feature fusion module, and a condition monitoring module;

[0074] The data acquisition module is used to collect multi-source data of the main transformer under test in the current time period; wherein, the multi-source data includes equipment temperature data, oil chromatography data, vibration data and ambient temperature and humidity data;

[0075] The feature extraction module is used to extract features from the multi-source data to form current feature data; wherein, the current feature data includes temperature features, gas features, amplitude features, and temperature-humidity coupling features;

[0076] The data acquisition module is used to acquire historical characteristic data and baseline characteristic data of the main transformer under test from the historical database;

[0077] The reliability weight calculation module is used to calculate the noise standard deviation of the current feature data based on the historical feature data, and generate a reliability weight based on the noise standard deviation.

[0078] The importance weight calculation module is used to calculate the similarity between the current feature data and the benchmark feature data, and generate importance weights based on the similarity.

[0079] The feature fusion module is used to generate feature fusion data based on the current feature data, the reliability weight, and the importance weight;

[0080] The status monitoring module is used to use a machine learning model to obtain the status monitoring results of the main transformer under test based on the feature fusion data, and to adjust the operating status of the main transformer under test based on the status monitoring results.

[0081] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0082] This invention provides a method for monitoring the condition of a main transformer based on multi-source fusion. The method involves collecting multi-source data of the main transformer under test within the current time period; extracting features from the multi-source data to form current feature data; acquiring historical feature data and benchmark feature data of the main transformer under test from a historical database; calculating the noise standard deviation of the current feature data based on the historical feature data, and generating a reliability weight based on the noise standard deviation; calculating the similarity between the current feature data and the benchmark feature data, and generating an importance weight based on the similarity; generating feature fusion data based on the current feature data, reliability weight, and importance weight; and using a machine learning model to derive the condition monitoring results of the main transformer under test based on the feature fusion data, and adjusting the operating state of the main transformer under test based on the condition monitoring results. This invention monitors the condition of the main transformer based on multi-source data, comprehensively reflecting the operating status of the equipment. By analyzing historical and current feature data to generate reliability weights, analyzing benchmark and current feature data to generate importance weights, and fusing features based on reliability and importance weights, the feature fusion data is better adapted to the complexities of real-world data, improving the accuracy of main transformer condition monitoring. Attached Figure Description

[0083] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0084] Figure 1 A flowchart illustrating an embodiment of the main transformer condition monitoring method based on multi-source fusion provided by the present invention;

[0085] Figure 2 This is a schematic diagram of an embodiment of the main transformer condition monitoring device based on multi-source fusion provided by the present invention. Detailed Implementation

[0086] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0087] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0088] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0089] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0090] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0091] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0092] In the description of the embodiments of this application, unless otherwise expressly specified and limited, technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0093] See Figure 1To address the problem of low accuracy in main transformer condition monitoring in existing technologies, an embodiment of the present invention provides a main transformer condition monitoring method based on multi-source fusion. This method includes steps 101 to 107, each step of which is detailed below:

[0094] Step 101: Collect multi-source data of the main transformer under test in the current time period; wherein, the multi-source data includes equipment temperature data, oil chromatography data, vibration data and ambient temperature and humidity data.

[0095] In this embodiment of the invention, when monitoring the condition of the main transformer under test, multiple sensors of the main transformer can be used to collect real-time multi-source data. To collect data that represents the state changes of the main transformer under test, when condition monitoring is required, not only instantaneous data at the current time point but also continuous data over a period of time preceding the current time point is collected. The multi-source data includes equipment temperature data, oil chromatography data, vibration data, and ambient temperature and humidity data. Specifically, equipment temperature data includes high-voltage winding temperature, low-voltage winding temperature, core temperature, and transformer oil temperature, which can be collected using temperature sensors. Oil chromatography data includes gas concentration data such as hydrogen concentration, carbon monoxide concentration, methane concentration, and acetylene concentration, which can be collected using oil chromatography sensors. Vibration data includes vibration amplitude and frequency spectrum data, which can be collected using vibration sensors. Ambient temperature and humidity data includes ambient temperature data and ambient humidity data, which can be collected using environmental sensors.

[0096] In this embodiment of the invention, when collecting multi-source data, the sensors on the main transformer under test achieve synchronous sampling through a unified clock signal (such as GPS timing), and the sampling frequency is set according to the data type. For example, equipment temperature and oil chromatography data are sampled once per minute; vibration data is sampled 1000 times per second (high-frequency vibration requires a higher sampling rate); environmental data is sampled once every 5 minutes. The collected data can be transmitted to edge computing nodes or cloud servers through industrial IoT protocols (such as MQTT, OPC UA), and the data storage format is a time-series database.

[0097] In this embodiment of the invention, after collecting multi-source data of the main transformer under test for the current time period, it is necessary to preprocess the collected raw data first, and use the processed multi-source data as the data basis for condition monitoring. Data preprocessing includes noise reduction, normalization, and missing value imputation.

[0098] As an example of an embodiment of the present invention, multi-scale noise separation based on wavelet transform can be used to eliminate random noise and electromagnetic interference in sensor data. First, a suitable wavelet basis function is selected based on the characteristics of the main transformer data (such as the spectral distribution of equipment temperature and vibration signals). For example, the db4 wavelet basis function is selected for equipment temperature / oil chromatography data, as this wavelet basis function is suitable for smooth signals; the sym5 wavelet basis function is selected for vibration data, as this wavelet basis function is suitable for high-frequency non-stationary signals. After selecting the appropriate wavelet decomposition function, multi-level wavelet decomposition is performed on the original data to obtain high-frequency coefficients (detail components) and low-frequency coefficients (approximate components). The expression for wavelet decomposition is:

[0099]

[0100] In the formula, cA j cD represents the low-frequency approximation coefficients of the j-th layer. j Here, cA represents the high-frequency detail coefficients of the j-th layer. j0,k cD represents the coefficients of the k-th basis function in the j-th approximation part. j,k φ represents the coefficients of the k-th wavelet basis function in the j-th layer; j0,k (x) is the scaling function; ψ j,k (x) is the wavelet function; j is the scaling parameter; k is the translation parameter; j0 is the upper limit of the scaling.

[0101] After wavelet decomposition of the original data, soft thresholding is applied to the high-frequency detail coefficients to suppress noise components. The expression for this is:

[0102]

[0103] In the formula, cD j denoised σ represents the high-frequency detail coefficients of the j-th layer after denoising; σ is the noise standard deviation; and L is the signal length.

[0104] The high-frequency detail coefficients after denoising are reconstructed into a signal, and its expression is:

[0105] x denoised =Waveletreconstryct(cA n ,cD n denoised cD1 denoised )

[0106] In the formula, x denoised For reconstructed signal; n is the maximum scale level.

[0107] After eliminating random noise and electromagnetic interference in sensor data through multi-scale noise separation based on wavelet transform, the data is normalized to eliminate the influence of different dimensions and map the data to a unified range ([0,1]). The global minimum and global maximum values ​​are calculated for each data dimension (e.g., equipment temperature, vibration amplitude). The normalization expression is as follows:

[0108]

[0109] In the formula, x norm x represents the normalized data; x represents the data before normalization; x max x is the global maximum value; min It is the global minimum value.

[0110] Step 102: Extract features from the multi-source data to form current feature data; wherein, the current feature data includes temperature features, gas features, amplitude features and temperature-humidity coupling features.

[0111] As a preferred embodiment, feature extraction is performed on the multi-source data to form current feature data, including:

[0112] The dynamic characteristics of temperature change are extracted from the equipment temperature data to form temperature features; wherein, the temperature features include the rate of temperature change and the standard deviation of temperature fluctuation;

[0113] Gas concentration change characteristics are extracted from the oil chromatography data to form gas features, wherein the gas features include gas concentration change rate and gas ratio;

[0114] Amplitude features are extracted from the vibration data; wherein, the amplitude features include the mean vibration amplitude, the peak-to-peak vibration amplitude, and the dominant frequency amplitude;

[0115] Extract temperature and humidity coupling features from the environmental temperature and humidity data; wherein, the temperature and humidity coupling features include a temperature and humidity coupling index;

[0116] The temperature change rate, the standard deviation of temperature fluctuation, the gas concentration change rate, the gas ratio, the mean vibration amplitude, the peak-to-peak vibration amplitude, the dominant frequency amplitude, and the temperature-humidity coupling index are determined as the current characteristic data.

[0117] In this embodiment of the invention, key features are extracted from the preprocessed data, which can be used for subsequent multi-source data fusion and status assessment, thereby reducing data dimensionality and highlighting sensitive indicators related to equipment status. Specifically, feature extraction from equipment temperature data can capture the dynamic characteristics of transformer temperature changes. The rate of temperature change reflects the rate of change in main transformer load or internal losses; the standard deviation of temperature fluctuation reflects quantified temperature stability, and abnormal fluctuations may indicate abnormal heat dissipation or local overheating. Feature extraction from oil chromatography data can analyze changes in the concentration of dissolved gases in the oil and identify early faults. A sudden increase in hydrogen concentration may indicate partial discharge, and a sudden increase in acetylene concentration may indicate arcing faults. Feature extraction from vibration data can identify abnormal frequency components in mechanical vibration. Feature extraction from environmental temperature and humidity data can quantify the impact of environmental factors on transformer status. Environmental features are mainly reflected in the temperature-humidity coupling index; a high temperature-humidity coupling index may exacerbate the heat dissipation pressure on the transformer.

[0118] As a preferred embodiment, the dynamic characteristics of temperature change are extracted from the device temperature data to form temperature features, including:

[0119] The rate of temperature change is calculated using the following formula:

[0120]

[0121] In the formula, f T Let T(t) be the rate of temperature change; Δt be the time interval; T(t) be the equipment temperature at time t; and T(t+Δt) be the equipment temperature at time t+Δt.

[0122] The standard deviation of temperature fluctuation is calculated using the following formula:

[0123]

[0124] In the formula, f T_std N1 represents the standard deviation of temperature fluctuation; N1 represents the number of equipment temperatures; T i Let i be the temperature of the i-th device; This represents the average temperature of the equipment.

[0125] In this embodiment of the invention, the equipment temperature data is collected within the current time period, so it includes multiple equipment temperatures. The temperature change rate and temperature fluctuation standard deviation can be calculated separately using the above formula, thereby forming the equipment temperature characteristics.

[0126] As a preferred embodiment, gas concentration change characteristics are extracted from the oil chromatography data to form gas characteristics, including:

[0127] The gas concentration data of dissolved gases in several types of oil were determined from the oil chromatography data;

[0128] The gas concentration change rate and gas ratio were calculated based on the gas concentration data of dissolved gases in several types of oils.

[0129] The following formula is used to calculate the rate of change of gas concentration:

[0130]

[0131] In the formula, f G ΔG is the rate of change of gas concentration; Δt is the amount of change of gas concentration; Δt is the time interval.

[0132] The gas ratio is calculated using the following formula:

[0133]

[0134] In the formula, f ratio G represents the gas ratio. i G represents the gas concentration of the i-th gas; j Let be the gas concentration of the j-th gas.

[0135] In this embodiment of the invention, the dissolved gases in the main transformer oil include a variety of different gases; therefore, the collected oil chromatographic data also includes gas concentration data for various dissolved gases. Since the oil chromatographic data is collected within the current time period, each dissolved gas in the oil has multiple corresponding gas concentration data. Based on the multiple gas concentration data corresponding to each dissolved gas in the oil, the gas concentration change rate of each dissolved gas in the oil can be calculated using the above formula. Based on the gas concentration data of each dissolved gas in the oil at the current moment, multiple gas ratios can be calculated by random combination using the above formula.

[0136] As a preferred embodiment, the amplitude features are extracted from the vibration data, including:

[0137] The mean vibration amplitude is calculated using the following formula:

[0138]

[0139] In the formula, f V_mean N1 represents the average vibration amplitude; N2 represents the number of vibration frequencies. Let i be the i-th vibration frequency;

[0140] The peak-to-peak value of the vibration amplitude is calculated using the following formula:

[0141] f V_pp =max(V amp )-min(V amp )

[0142] In the formula, f V_ppThe peak-to-peak value of the vibration amplitude; max(V amp ) is the maximum vibration frequency; min(V) amp () is the minimum vibration frequency;

[0143] The main frequency amplitude is calculated using the following formula:

[0144] f V_freq =max(|FFT(V) amp )|)

[0145] In the formula, f V_freq Main frequency amplitude; V amp is the vibration frequency; FFT(·) is the Fast Fourier Transform function.

[0146] In this embodiment of the invention, since the vibration data is collected within the current time period, it includes multiple vibration frequencies. Summing the values ​​of each vibration frequency and averaging the sum yields the average vibration amplitude. The difference between the maximum and minimum vibration frequencies is determined among the multiple vibration frequencies, and this difference is defined as the dominant frequency amplitude. Processing the vibration frequencies using a Fast Fourier Transform function yields the dominant frequency amplitude, representing the frequency domain characteristics.

[0147] A preferred embodiment involves extracting temperature and humidity coupling features from the environmental temperature and humidity data, including:

[0148] Extract the ambient temperature and ambient humidity from the ambient temperature and humidity data;

[0149] Based on the ambient temperature and humidity, the temperature-humidity coupling index is calculated using the following formula:

[0150] f THI =T env ×(1-k×H env )

[0151] In the formula, f THI T is the temperature and humidity coupling index; env For ambient temperature; H env is the ambient humidity; k is a preset parameter.

[0152] In this embodiment of the invention, the environmental temperature and humidity data includes environmental temperature data and environmental humidity data. Based on these two data and preset parameters, a temperature and humidity coupling index can be calculated to characterize environmental features.

[0153] Step 103: Obtain the historical characteristic data and baseline characteristic data of the main transformer under test from the historical database.

[0154] In this embodiment of the invention, after extracting the feature data from various multi-source data, it is necessary to perform data fusion operations by combining weights. Since fixed weights are insufficient to handle complex situations such as differences in data quality, differences in indicator importance, and dynamic changes in scenarios, it is necessary to determine the weights by combining the historical feature data and benchmark feature data of the main transformer under test, thereby correcting data quality deviations and highlighting the impact of core indicators. The historical feature data and benchmark feature data of the main transformer under test are also multi-source data, including equipment temperature data, oil chromatography data, vibration data, and environmental temperature and humidity data.

[0155] Step 104: Based on the historical feature data, calculate the noise standard deviation of the current feature data, and generate a reliability weight based on the noise standard deviation.

[0156] In this embodiment of the invention, the feature weight can be composed of two parts: a reliability weight and an importance weight. The reliability weight is calculated by analyzing historical feature data and current feature data. The importance weight is calculated by analyzing baseline feature data and current feature data.

[0157] As a preferred embodiment, based on the historical feature data, the noise standard deviation of the current feature data is calculated, and a reliability weight is generated based on the noise standard deviation, including:

[0158] For each feature value in the current feature data, several historical features are obtained from the historical feature data;

[0159] Calculate the mean of the feature values ​​based on several historical characteristics;

[0160] The eigenvalues ​​and the noise standard deviation calculated based on the mean of the eigenvalues;

[0161] After determining the noise standard deviation of each feature value in the current feature data, the reliability weight of each feature value is calculated according to the following formula:

[0162]

[0163] In the formula, Reliability i The reliability weight for the i-th feature value; NoiseLevel i is the noise standard deviation of the i-th eigenvalue; max(NoiseLevel) is the maximum noise standard deviation.

[0164] In this embodiment of the invention, the weights formed based on the noise standard deviation of feature data are based on the core logic of using the magnitude of noise to measure the reliability of the feature. Therefore, by analyzing the noise standard deviation of the feature data, the reliability weights can be determined. First, the noise standard deviation of a feature value is a statistical measure of the dispersion of the deviation between the observed value and the true value. It is used to quantify the interference intensity of noise on the feature value estimation results and reflects the stability of the feature observations. Since the true value is data under ideal conditions, this invention determines the mean of historical feature data as the true value and the difference between the measured feature value and the calculated mean as the noise standard deviation. After calculating the noise standard deviations corresponding to all feature values, the maximum noise standard deviation is determined. Then, based on the maximum noise standard deviation and the noise standard deviations calculated for each feature value, the reliability weight of each feature value is calculated according to the above formula. Specifically, when calculating the noise standard deviation of each feature value in the current feature data, multiple historical features of the same type as that feature value are obtained from the historical feature data for calculation. For example, for the feature value of temperature change rate in the current feature data, multiple historical temperature change rates are obtained from the historical feature data for noise standard deviation calculation.

[0165] Step 105: Calculate the similarity between the current feature data and the benchmark feature data, and generate an importance weight based on the similarity.

[0166] As a preferred embodiment, the similarity between the current feature data and the baseline feature data is calculated, and an importance weight is generated based on the similarity, including:

[0167] For each feature value in the current feature data, the corresponding benchmark feature value is obtained from the benchmark feature data;

[0168] The similarity between the feature value and the benchmark feature value is calculated using the normalized difference method.

[0169] The similarity is determined as the importance weight of the feature value;

[0170] The importance weights are calculated using the following formula:

[0171]

[0172] In the formula, Importance i f is the importance weight of the i-th feature value; i f is the i-th eigenvalue; i0 Let be the baseline eigenvalue of the i-th eigenvalue.

[0173] In this embodiment of the invention, to adapt the fused feature data to complex scenarios where the importance of indicators varies in reality, the current feature data and benchmark feature data can be analyzed to consider the influence of core indicators and form importance weights. The importance weight of the current feature data is calculated based on the benchmark feature data, and the core logic is to quantify its importance through the correlation between the current feature and the benchmark feature. Therefore, the importance weight can be determined by analyzing the similarity between the current feature data and the benchmark feature data. Since the benchmark feature data is a quantitative representation of the target standard or ideal state, the more similar a feature is to the benchmark, the closer it is to the target requirement. Therefore, the greater the similarity between the current feature data and the benchmark feature data, the higher the importance weight.

[0174] Step 106: Generate feature fusion data based on the current feature data, the reliability weight, and the importance weight.

[0175] As a preferred embodiment, feature fusion data is generated based on the current feature data, the reliability weight, and the importance weight, including:

[0176] Based on the reliability weight and the importance weight, generate the feature weights for each feature value in the current feature data;

[0177] Based on each feature value and its feature weight in the current feature data, linear weighting is performed to generate feature fusion data;

[0178] The feature weights are generated using the following formula:

[0179]

[0180] In the formula, ω i The feature weights for the i-th feature value; Reliability i The reliability weight for the i-th feature value; Importance i The importance weight for the i-th feature value; Reliability j The reliability weight for the j-th eigenvalue; Importance j The importance weight of the j-th feature value; M α is the number of eigenvalues; α is the first harmonic coefficient; β is the second harmonic coefficient;

[0181] The feature fusion data is calculated using the following formula:

[0182]

[0183] In the formula, F represents the feature fusion data; ω i f is the feature weight of the i-th feature value; iLet be the i-th eigenvalue; M is the number of eigenvalues.

[0184] In this embodiment of the invention, after calculating the reliability weight and importance weight of each feature value, the feature weight of each feature value can be obtained through dynamic adjustment. Then, the feature values ​​are weighted and summed according to their feature weights to obtain feature fusion data. The reliability weight and importance weight can be dynamically adjusted by setting a first harmonic coefficient and a second harmonic coefficient.

[0185] Step 107: Using a machine learning model, obtain the status monitoring results of the main transformer under test based on the feature fusion data, and adjust the operating status of the main transformer under test based on the status monitoring results.

[0186] In this embodiment of the invention, after obtaining the real-time feature fusion data of the main transformer under test, a machine learning model can be used for state assessment to obtain the state monitoring results. Specifically, the machine learning model can define discrete state labels (normal state, warning state, and fault state) from the historical maintenance records and fault data of the main transformer. The historical fused features and corresponding labels are combined to form a training data sample set. The model is then trained, evaluated, and optimized using a machine learning model (such as a support vector machine model) to form a usable state assessment machine learning model. To assess the real-time state of the main transformer under test, the real-time feature fusion data of the main transformer under test is input into the trained state assessment machine learning model, and the state monitoring results are output.

[0187] In this embodiment of the invention, the operating status of the main transformer under test can be adjusted based on the status monitoring results. Specifically, when the status monitoring result is in a normal state, the equipment operating status is continuously monitored and recorded without triggering the fault diagnosis process; when the status monitoring result is in a warning state, a preliminary diagnosis is triggered based on feature fusion data, and a preset combination of warning-level abnormal features is matched through the system rule base to output warning information and corresponding suggestions; when the status monitoring result is in a fault state, a deep diagnosis is triggered, and multi-dimensional cross-validation is performed by combining feature fusion data and the system rule base. Key abnormal data sources are located through feature weight analysis, and condition-conclusion rules containing industry standards and historical cases are matched to achieve accurate location and confidence assessment of fault type and location. At the same time, preset handling strategies are associated to form interpretable diagnostic conclusions and maintenance suggestions.

[0188] Specifically, the rule base needs to integrate industry standards, equipment manual specifications, and historical fault cases to form a set of conditional-conclusion rules that includes multi-source feature thresholds, logical relationships between features, and fault mode mappings. During rule matching, firstly, based on a weight matrix constructed from feature reliability and importance scores, key features that contribute significantly to the current state assessment are selected, and logical matching is performed between real-time feature values ​​and preset feature thresholds in the rule base. Secondly, feature anomalies from at least two independent data sources are required to form cross-validation, avoiding interference from noise from a single sensor. Finally, through forward inference, the fault type, location, and confidence level are output, while simultaneously associating with preset maintenance strategies to achieve interpretable fault diagnosis and decision support.

[0189] Implementing the above embodiments has the following effects:

[0190] This invention provides a method for monitoring the condition of a main transformer based on multi-source fusion. The method involves collecting multi-source data of the main transformer under test within the current time period; extracting features from the multi-source data to form current feature data; acquiring historical feature data and benchmark feature data of the main transformer under test from a historical database; calculating the noise standard deviation of the current feature data based on the historical feature data, and generating a reliability weight based on the noise standard deviation; calculating the similarity between the current feature data and the benchmark feature data, and generating an importance weight based on the similarity; generating feature fusion data based on the current feature data, reliability weight, and importance weight; and using a machine learning model to derive the condition monitoring results of the main transformer under test based on the feature fusion data, and adjusting the operating state of the main transformer under test based on the condition monitoring results. This invention monitors the condition of the main transformer based on multi-source data, comprehensively reflecting the operating status of the equipment. By analyzing historical and current feature data to generate reliability weights, analyzing benchmark and current feature data to generate importance weights, and fusing features based on reliability and importance weights, the feature fusion data is better adapted to the complexities of real-world data, improving the accuracy of main transformer condition monitoring.

[0191] like Figure 2 As shown, based on the above method embodiments, corresponding apparatus embodiments are provided;

[0192] One embodiment of the present invention provides a main transformer condition monitoring device based on multi-source fusion, comprising: a data acquisition module, a feature extraction module, a data acquisition module, a reliability weight calculation module, an importance weight calculation module, a feature fusion module, and a condition monitoring module;

[0193] The data acquisition module is used to collect multi-source data of the main transformer under test in the current time period; wherein, the multi-source data includes equipment temperature data, oil chromatography data, vibration data and ambient temperature and humidity data;

[0194] The feature extraction module is used to extract features from the multi-source data to form current feature data; wherein, the current feature data includes temperature features, gas features, amplitude features, and temperature-humidity coupling features;

[0195] The data acquisition module is used to acquire historical characteristic data and baseline characteristic data of the main transformer under test from the historical database;

[0196] The reliability weight calculation module is used to calculate the noise standard deviation of the current feature data based on the historical feature data, and generate a reliability weight based on the noise standard deviation.

[0197] The importance weight calculation module is used to calculate the similarity between the current feature data and the benchmark feature data, and generate importance weights based on the similarity.

[0198] The feature fusion module is used to generate feature fusion data based on the current feature data, the reliability weight, and the importance weight;

[0199] The status monitoring module is used to use a machine learning model to obtain the status monitoring results of the main transformer under test based on the feature fusion data, and to adjust the operating status of the main transformer under test based on the status monitoring results.

[0200] It is understood that the above-described device embodiments correspond to the method embodiments of the present invention, and can implement the main transformer condition monitoring method based on multi-source fusion provided by any of the above-described method embodiments of the present invention.

[0201] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can specifically be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0202] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for monitoring the condition of a main transformer based on multi-source fusion, characterized in that, include: Collect multi-source data of the main transformer under test in the current time period; wherein, the multi-source data includes equipment temperature data, oil chromatography data, vibration data and ambient temperature and humidity data; Feature extraction is performed on the multi-source data to form current feature data; wherein, the current feature data includes temperature features, gas features, amplitude features, and temperature-humidity coupled features; Obtain historical characteristic data and baseline characteristic data of the main transformer under test from the historical database; Based on the historical feature data, calculate the noise standard deviation of the current feature data, and generate a reliability weight based on the noise standard deviation; Calculate the similarity between the current feature data and the baseline feature data, and generate an importance weight based on the similarity. Based on the current feature data, the reliability weight, and the importance weight, feature fusion data is generated; Using a machine learning model, the status monitoring results of the main transformer under test are obtained based on the feature fusion data, and the operating status of the main transformer under test is adjusted based on the status monitoring results.

2. The main transformer condition monitoring method based on multi-source fusion according to claim 1, characterized in that, The step of extracting features from the multi-source data to form the current feature data includes: The dynamic characteristics of temperature change are extracted from the equipment temperature data to form temperature features; wherein, the temperature features include the rate of temperature change and the standard deviation of temperature fluctuation; Gas concentration change characteristics are extracted from the oil chromatography data to form gas features, wherein the gas features include gas concentration change rate and gas ratio; Amplitude features are extracted from the vibration data; wherein, the amplitude features include the mean vibration amplitude, the peak-to-peak vibration amplitude, and the dominant frequency amplitude; Extract temperature and humidity coupling features from the environmental temperature and humidity data; wherein, the temperature and humidity coupling features include a temperature and humidity coupling index; The temperature change rate, the standard deviation of temperature fluctuation, the gas concentration change rate, the gas ratio, the mean vibration amplitude, the peak-to-peak vibration amplitude, the dominant frequency amplitude, and the temperature-humidity coupling index are determined as the current characteristic data.

3. The main transformer condition monitoring method based on multi-source fusion according to claim 2, characterized in that, The step of extracting dynamic temperature change characteristics from the device temperature data to form temperature features includes: The rate of temperature change is calculated using the following formula: In the formula, f T Let T(t) be the rate of temperature change; Δt be the time interval; T(t) be the equipment temperature at time t; and T(t+Δt) be the equipment temperature at time t+Δt. The standard deviation of temperature fluctuation is calculated using the following formula: In the formula, f T_std N1 represents the standard deviation of temperature fluctuation; N1 represents the number of equipment temperatures; T i Let i be the temperature of the i-th device; This represents the average temperature of the equipment.

4. The main transformer condition monitoring method based on multi-source fusion according to claim 2, characterized in that, The extraction of gas concentration change characteristics from the oil chromatography data to form gas characteristics includes: The gas concentration data of dissolved gases in several types of oil were determined from the oil chromatography data; The gas concentration change rate and gas ratio were calculated based on the gas concentration data of dissolved gases in several types of oils. The following formula is used to calculate the rate of change of gas concentration: In the formula, f G ΔG is the rate of change of gas concentration; Δt is the amount of change of gas concentration; Δt is the time interval. The gas ratio is calculated using the following formula: In the formula, f ratio G represents the gas ratio. i G represents the gas concentration of the i-th gas; j Let be the gas concentration of the j-th gas.

5. The main transformer condition monitoring method based on multi-source fusion according to claim 2, characterized in that, Extracting amplitude features from the vibration data includes: The mean vibration amplitude is calculated using the following formula: In the formula, f V_mean N1 represents the average vibration amplitude; N2 represents the number of vibration frequencies. Let i be the i-th vibration frequency; The peak-to-peak value of the vibration amplitude is calculated using the following formula: f V_pp =max(V amp )-min(V amp ) In the formula, f V_pp The peak-to-peak value of the vibration amplitude; max(V amp ) is the maximum vibration frequency; min(V) amp () is the minimum vibration frequency; The main frequency amplitude is calculated using the following formula: f V_freq =max(|FFT(V amp )|) In the formula, f V_freq Main frequency amplitude; V amp is the vibration frequency; FFT(·) is the Fast Fourier Transform function.

6. The main transformer condition monitoring method based on multi-source fusion according to claim 2, characterized in that, The extraction of temperature and humidity coupling features from the environmental temperature and humidity data includes: Extract the ambient temperature and ambient humidity from the ambient temperature and humidity data; Based on the ambient temperature and humidity, the temperature-humidity coupling index is calculated using the following formula: f THI =T env ×(1-k×H env ) In the formula, f THI The temperature and humidity coupling index; T env For ambient temperature; H env is the ambient humidity; k is a preset parameter.

7. The main transformer condition monitoring method based on multi-source fusion according to claim 1, characterized in that, The step of calculating the noise standard deviation of the current feature data based on the historical feature data, and generating reliability weights based on the noise standard deviation, includes: For each feature value in the current feature data, several historical features are obtained from the historical feature data; Calculate the mean of the feature values ​​based on several historical characteristics; The eigenvalues ​​and the noise standard deviation calculated based on the mean of the eigenvalues; After determining the noise standard deviation of each feature value in the current feature data, the reliability weight of each feature value is calculated according to the following formula: In the formula, Reliability i The reliability weight for the i-th feature value; NoiseLevel i is the noise standard deviation of the i-th eigenvalue; max(NoiseLevel) is the maximum noise standard deviation.

8. The main transformer condition monitoring method based on multi-source fusion according to claim 1, characterized in that, The step of calculating the similarity between the current feature data and the baseline feature data, and generating an importance weight based on the similarity, includes: For each feature value in the current feature data, the corresponding benchmark feature value is obtained from the benchmark feature data; The similarity between the feature value and the benchmark feature value is calculated using the normalized difference method. The similarity is determined as the importance weight of the feature value; The importance weights are calculated using the following formula: In the formula, Importance i f is the importance weight of the i-th feature value; i f is the i-th eigenvalue; i0 Let be the baseline eigenvalue of the i-th eigenvalue.

9. The main transformer condition monitoring method based on multi-source fusion according to claim 1, characterized in that, The step of generating feature fusion data based on the current feature data, the reliability weight, and the importance weight includes: Based on the reliability weight and the importance weight, generate the feature weights for each feature value in the current feature data; Based on each feature value and its feature weight in the current feature data, linear weighting is performed to generate feature fusion data; The feature weights are generated using the following formula: In the formula, ω i The feature weights for the i-th feature value; Reliability i The reliability weight for the i-th feature value; Importance i The importance weight for the i-th feature value; Reliability j The reliability weight for the j-th eigenvalue; Importance j The importance weight of the j-th feature value; M α is the number of eigenvalues; α is the first harmonic coefficient; β is the second harmonic coefficient; The feature fusion data is calculated using the following formula: In the formula, F represents the feature fusion data; ω i f is the feature weight of the i-th feature value; i Let i be the i-th eigenvalue; M The number of eigenvalues.

10. A main transformer condition monitoring device based on multi-source fusion, characterized in that, include: The system includes a data acquisition module, a feature extraction module, a data acquisition module, a reliability weight calculation module, an importance weight calculation module, a feature fusion module, and a status monitoring module. The data acquisition module is used to collect multi-source data of the main transformer under test in the current time period; wherein, the multi-source data includes equipment temperature data, oil chromatography data, vibration data and ambient temperature and humidity data; The feature extraction module is used to extract features from the multi-source data to form current feature data; wherein, the current feature data includes temperature features, gas features, amplitude features, and temperature-humidity coupling features; The data acquisition module is used to acquire historical characteristic data and baseline characteristic data of the main transformer under test from the historical database; The reliability weight calculation module is used to calculate the noise standard deviation of the current feature data based on the historical feature data, and generate a reliability weight based on the noise standard deviation. The importance weight calculation module is used to calculate the similarity between the current feature data and the benchmark feature data, and generate importance weights based on the similarity. The feature fusion module is used to generate feature fusion data based on the current feature data, the reliability weight, and the importance weight; The status monitoring module is used to use a machine learning model to obtain the status monitoring results of the main transformer under test based on the feature fusion data, and to adjust the operating status of the main transformer under test based on the status monitoring results.

Images