Transformer online monitoring and remote diagnosis system and method based on multi-parameter fusion

Abstract

The present invention discloses a transformer online monitoring and remote diagnosis system and method based on multi-parameter fusion, and relates to the technical field of power equipment monitoring. The system includes one or more status detection devices installed on the transformer and a cloud platform. The status detection device is used to perform real-time online monitoring of the transformer status, including a data acquisition module, a feature matching module, a probability calculation module, a data upload module and a feature library update module. The cloud platform receives the data uploaded by the data upload module, which is used to perform remote diagnosis of the transformer status, including a feature library, an abnormal secondary analysis module, an abnormal confirmation module and a feature library distribution module. By integrating multiple sensors and data processing technologies to monitor the transformer status in real time, the reliability and safety of the transformer operation are improved, online monitoring and remote diagnosis are realized, the dependence on manual regular inspections and maintenance is reduced, and the operation and maintenance costs and workload are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment monitoring, and in particular to a transformer online monitoring and remote diagnosis system and method based on multi-parameter fusion. Background Art

[0002] The power system is one of the indispensable infrastructures in modern society, and the transformer, as the core component of the power system, undertakes the important task of converting electric energy from one voltage level to another. However, due to changes in grid load and environmental conditions, transformers may produce some faults during operation, affecting the stable operation of the power system.

[0003] Distribution transformers refer to transformers with voltage levels less than 35KV. There are many distribution transformers and their installation locations are scattered, which is extremely unfavorable for monitoring. In addition, the causes of failures are relatively complex. Structure, operating conditions, climate, and even strong currents and electromagnetic fields can affect the operation of distribution transformers. Fault manifestations are diverse, making it very difficult to obtain fault characteristics. Therefore, how to efficiently monitor the status of distribution transformers and improve fault diagnosis efficiency is particularly important.

[0004] Existing transformer monitoring technologies mainly include temperature monitoring, current monitoring, oil chromatography analysis and partial discharge monitoring, etc. Most of these methods are single parameter monitoring, which makes it difficult to fully reflect the operating status of the transformer. Some monitoring methods (such as oil chromatography analysis) require regular sampling, which makes it difficult to achieve real-time monitoring and cannot timely reflect the operating status and fault conditions of the transformer. In addition, the existing monitoring system has limited capabilities in data fusion and analysis, making it difficult to conduct comprehensive analysis of multiple monitoring data, and unable to provide comprehensive and accurate fault diagnosis information. At the same time, it lacks remote diagnosis capabilities and cannot achieve remote real-time monitoring and diagnosis of the transformer's operating status, increasing the cost and difficulty of operation and maintenance. Summary of the invention

[0005] In view of the problems existing in the existing distribution transformer monitoring, such as single parameter monitoring, insufficient real-time performance, limited data fusion and analysis capabilities, and lack of remote diagnosis capabilities, the present invention proposes a transformer online monitoring and remote diagnosis system and method based on multi-parameter fusion. By integrating multiple sensors and data processing technologies, real-time monitoring of the transformer status is achieved, thereby improving the reliability and safety of the transformer operation.

[0006] In order to achieve the above object, the present invention is implemented by the following technical solutions:

[0007] A transformer online monitoring and remote diagnosis system based on multi-parameter fusion, the system includes one or more status detection devices installed on the transformer and a cloud platform, the status detection device and the data concentrator are connected via RS485, Bluetooth or other micro-power wireless methods, and the data concentrator and the cloud platform use the transformer owner's private network;

[0008] The state detection device is used to perform real-time online monitoring of the transformer state, including:

[0009] A data acquisition module is used to integrate multiple sensors to collect the status parameters of the transformer in real time, wherein the status parameters include the temperature, pressure, oil level and vibration data in the transformer, perform Fourier transform on the vibration data, analyze the vibration energy of different frequency components, and perform temperature calibration on the vibration data after Fourier transform;

[0010] A feature matching module is used to extract vibration feature data from the calibrated vibration data, wherein the vibration feature data includes 100 Hz proportion, total harmonic distortion, frequency point vibration energy and odd-even harmonic ratio, and compare the vibration feature data with the local feature library to determine whether it exceeds a threshold;

[0011] The probability calculation module is used to calculate the matching probability between the extracted vibration feature data and the vibration data features in the feature library, and to judge the transformer state based on the matching probability. If an abnormal state is matched, the data is collected three times and matched. If an abnormal state is still matched after repeated data collection three times, it is marked as abnormal and uploaded to the cloud platform.

[0012] The data upload module is used to upload the collected state parameters, vibration characteristic data and calculated matching probability to the cloud platform;

[0013] The feature library update module is used to receive the latest vibration data features sent by the cloud platform to update the local feature library;

[0014] The cloud platform is used to remotely diagnose the transformer status, including:

[0015] A feature library is used to store the reference curve of the transformer under normal conditions and the vibration data characteristics under abnormal conditions;

[0016] The abnormal secondary analysis module is used to perform secondary analysis based on the data uploaded by the data upload module and the historical load information to further confirm whether the transformer is in an abnormal state. If the abnormality is still confirmed after the secondary analysis, an abnormal alarm will be issued to notify manual processing;

[0017] The abnormality confirmation module is used to manually confirm whether the transformer has an abnormality. If the abnormality is manually confirmed, an abnormality confirmation message is sent;

[0018] The feature library sending module is used to send the latest vibration data features to the state detection device.

[0019] A further improvement of the present invention is that the data acquisition module comprises:

[0020] The PT100 temperature measuring unit is used to collect the voltage of the PT100 in real time through the temperature measuring circuit, calculate the resistance value of the PT100 according to the voltage, and thus map the temperature data in the transformer. When the temperature data exceeds the threshold value, the temperature alarm is triggered; the PT100 is installed at the oil thermometer position of the transformer and is connected in a two-wire system;

[0021] The air pressure measuring unit is used to collect the pressure data of the transformer in real time through the pressure sensor, and trigger the pressure alarm when the pressure data exceeds the threshold value;

[0022] The laser ranging unit is used to measure the oil level data of the transformer in real time through the oil level sensor, and trigger the oil level alarm when the oil level data exceeds the threshold value;

[0023] A vibration detection unit is used to collect vibration data in the transformer through a vibration sensor at a sampling frequency of 50kHz, calculate the vibration spectrum through short-time Fourier transform, extract the fundamental frequency and multiple frequencies, perform temperature calibration on the transformed vibration data, and transform the vibration characteristic data into data at 20°C;

[0024] Vibration signal Frequency components obtained by short-time Fourier transform Defined as:

[0025] ;

[0026] In the formula, is the window function, represents angular frequency, * represents conjugation; is the value of the original vibration signal at time t′; Indicates at a fixed time t Based on the t′ The angular frequency is adjusted by the change of ω ;

[0027] The PT100 temperature measuring unit, the laser distance measuring unit, the air pressure measuring unit and the vibration detecting unit are centrally mounted on the same PCB board.

[0028] A further improvement of the present invention is that, in the data acquisition module, the method for temperature calibration of the vibration characteristic data comprises:

[0029] According to the influence of the temperature of the distribution transformer on the vibration, the temperature influence curve is drawn, and a temperature acceleration calibration mapping table is established according to the temperature influence curve. The acceleration of the real-time collected temperature data is calibrated by table lookup, so that all vibration data are converted into data at 20°C.

[0030] A further improvement of the present invention is that in the feature matching module, 100 Hz accounts for The calculation formula is:

[0031] ;

[0032] In the formula, is the amplitude at frequency point f in the spectrum, is the amplitude when f=100Hz; f max Indicates the maximum frequency, the value is 1000Hz;

[0033] Total Harmonic Distortion The calculation formula is:

[0034] ;

[0035] In the formula, is the i-th harmonic amplitude, N is the harmonic number, is the i-th harmonic weight;

[0036] Odd-even harmonic ratio The calculation formula is:

[0037] ;

[0038] In the formula, and are the amplitudes of odd and even harmonics respectively; is the upper limit of odd frequency, is the upper limit of even frequency;

[0039] The vibration energy is calculated using the root mean square value of the vibration velocity. The formula is:

[0040] ;

[0041] Where R is the root mean square value of the vibration velocity, T is the vibration period, is the vibration speed at time t;

[0042] The proportion of vibration energy at frequency point f for:

[0043] .

[0044] A further improvement of the present invention is that in the probability calculation module, a frequency-amplitude curve is drawn with the multiple frequency of 50 Hz as the x-axis, the maximum value is 1000 Hz, and the amplitude as the y-axis, and the curve is compared with the reference curve of the transformer in the feature library to calculate the similarity between the curves. , the calculation formula is:

[0045] ;

[0046] In the formula, , are the nth eigenvalues ​​in the data sequences X and Y, respectively. , are the average values ​​in the data series X and Y respectively, and N is the total number of eigenvalues;

[0047] When the similarity is lower than the preset threshold, the characteristic curve generated by each feature in the vibration feature data is compared with the reference curve in the feature library for similarity, and the characteristic curve with the highest similarity is selected and reported to the cloud platform.

[0048] A further improvement of the present invention is that the implementation method of the abnormal secondary analysis module includes:

[0049] The IQR method is used to automatically detect and process outliers in state parameters. The formula is: , in is the interquartile range, and are the first and third quartiles of the data, respectively. Outliers are defined as values ​​less than or greater than The value of

[0050] The Z-score standardization process state parameters are applied, and the formula is:

[0051] ;

[0052] In the formula, is the state parameter, , are the mean and standard deviation of the state parameters, is the standardized state parameter;

[0053] Apply wavelet transform to vibration data to capture local features and anomalies of vibration signals in the frequency domain;

[0054] The random forest method is used to build a classification model to classify the data uploaded by the status detection device to determine whether there is anomaly. The online gradient descent algorithm is introduced to enable the classification model to continuously learn from new data and update in real time.

[0055] A transformer online monitoring and remote diagnosis method based on multi-parameter fusion, the method comprising:

[0056] Integrate multiple sensors to collect transformer status parameters in real time, including temperature, pressure, oil level and vibration data inside the transformer;

[0057] Perform short-time Fourier transform on the vibration data to calculate the vibration spectrum, analyze the vibration energy of different frequency components, extract the fundamental frequency and the frequency multiples, and perform temperature calibration on the transformed vibration data to convert the vibration characteristic data into data at 20°C; the method for temperature calibration of vibration characteristic data includes: drawing a temperature influence curve according to the influence of the temperature of the distribution transformer on the vibration, establishing a temperature acceleration calibration mapping table according to the temperature influence curve, and calibrating by a table lookup method according to the acceleration of the temperature data collected in real time, so that all vibration data are converted into data at 20°C; extracting vibration characteristic data from the calibrated vibration data, including 100Hz proportion, total harmonic distortion, frequency point vibration energy and odd-even harmonic ratio;

[0058] Upload all data of the status detection device to the cloud platform;

[0059] Device abnormality reporting;

[0060] Perform secondary analysis based on the data uploaded by the status detection device and historical load information to further confirm whether the transformer is in an abnormal state. If the abnormality is still confirmed after the secondary analysis, an abnormal alarm will be issued to notify manual processing;

[0061] Manually confirm whether there is any abnormality in the transformer. If the abnormality is manually confirmed, an abnormality confirmation message is sent;

[0062] Determine whether a new anomaly has occurred. If so, update and send the anomaly data; if not, re-collect the sensor data.

[0063] A further improvement of the present invention is that the vibration signal Frequency components obtained by short-time Fourier transform Defined as:

[0064] ;

[0065] In the formula, is the window function, represents angular frequency, * represents conjugation; is the value of the original vibration signal at time t′; Indicates at a fixed time t Based on the t′ The angular frequency is adjusted by the change of ω ;

[0066] 100Hz ratio The calculation formula is:

[0067] ;

[0068] In the formula, is the amplitude at frequency point f in the spectrum, is the amplitude when f=100Hz; f max Indicates the maximum frequency, the value is 1000Hz;

[0069] Total Harmonic Distortion The calculation formula is:

[0070] ;

[0071] In the formula, is the i-th harmonic amplitude, N is the harmonic number, is the i-th harmonic weight;

[0072] Odd-even harmonic ratio The calculation formula is:

[0073] ;

[0074] In the formula, and are the amplitudes of odd and even harmonics respectively; is the upper limit of odd frequency, is the upper limit of even frequency;

[0075] The vibration energy is calculated using the root mean square value of the vibration velocity. The formula is:

[0076] ;

[0077] Where R is the root mean square value of the vibration velocity, T is the vibration period, is the vibration speed at time t;

[0078] The proportion of vibration energy at frequency point f for:

[0079] .

[0080] A further improvement of the present invention is that the device abnormality reporting includes:

[0081] Whether the temperature, pressure and oil level data exceed the threshold value, if so, trigger the temperature alarm, pressure alarm and / or oil level alarm;

[0082] The vibration feature data is compared with the local feature library to determine whether it exceeds the threshold;

[0083] Calculate the matching probability between the extracted vibration feature data and the vibration data features in the feature library, and judge the transformer status based on the matching probability. If an abnormal state is matched, repeat the collection three times and match. If an abnormal state is still matched after repeating the collection three times, mark it as abnormal and upload it to the cloud platform. Specifically: use the multiple frequency of 50Hz as the x-axis, the maximum value is 1000Hz, and the amplitude as the y-axis to draw a frequency-amplitude curve, compare it with the benchmark curve of the transformer in the feature library, and calculate the similarity between the curves. , the calculation formula is:

[0084] ;

[0085] In the formula, , are the nth eigenvalues ​​in the data sequences X and Y, respectively. , are the average values ​​in the data series X and Y respectively, and N is the total number of eigenvalues;

[0086] When the similarity is lower than the preset threshold, the characteristic curve generated by each feature in the vibration feature data is compared with the reference curve in the feature library for similarity, and the characteristic curve with the highest similarity is selected and reported to the cloud platform;

[0087] A further improvement of the present invention is that the secondary analysis method includes: using the IQR method to automatically detect and process abnormal values ​​in the state parameters, the formula is: , in is the interquartile range, and are the first and third quartiles of the data, respectively. Outliers are defined as values ​​less than or greater than The value of

[0088] The Z-score standardization process state parameters are applied, and the formula is:

[0089] ;

[0090] In the formula, is the state parameter, , are the mean and standard deviation of the state parameters, is the standardized state parameter;

[0091] Apply wavelet transform to vibration data to capture local features and anomalies of vibration signals in the frequency domain;

[0092] The random forest method is used to build a classification model to classify the data uploaded by the status detection device to determine whether there is anomaly. The online gradient descent algorithm is introduced to enable the classification model to continuously learn from new data and update in real time.

[0093] The beneficial effects of the present invention are as follows: by integrating multiple sensors such as temperature, pressure, oil level and vibration, the system can collect the status parameters of the transformer in real time, realize real-time online monitoring of the transformer status, and improve the real-time and accuracy of the data;

[0094] By integrating multiple state parameters and performing Fourier transform and temperature calibration on vibration data, vibration feature data including 100Hz proportion, total harmonic distortion, frequency point vibration energy and odd-even harmonic ratio are extracted, which can comprehensively reflect the operating status of the transformer; by calculating the matching probability between vibration feature data and data features in the feature library, combined with data collected repeatedly, accurate judgment and marking of abnormal states are achieved, and abnormal states are uploaded to the cloud platform for remote diagnosis and processing; the cloud platform receives and analyzes the uploaded data, and uses the abnormal secondary analysis module and abnormal confirmation module to further confirm the abnormal state of the transformer, notify manual processing, and update the feature library to ensure that the system can continue to learn and optimize; it realizes online monitoring and remote diagnosis, reduces dependence on manual regular inspections and maintenance, and reduces operation and maintenance costs and workload. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0096] Figure 1 It is the overall system architecture diagram of the present invention;

[0097] Figure 2 This is a functional logic diagram of a data acquisition module in an embodiment of the present invention;

[0098] Figure 3 Schematic diagram of a temperature measurement circuit in an embodiment of the present invention;

[0099] Figure 4 A schematic diagram of a pressure detection circuit in an embodiment of the present invention;

[0100] Figure 5 Schematic diagram of a laser ranging circuit in an embodiment of the present invention;

[0101] Figure 6 Schematic diagram of a vibration detection circuit in an embodiment of the present invention;

[0102] Figure 7 A schematic diagram of a temperature influence curve in an embodiment of the present invention;

[0103] Figure 8 is a flow chart of the method of the present invention;

[0104] Fig. 9 It is a logical function block diagram of device abnormality analysis in an embodiment of the present invention. DETAILED DESCRIPTION

[0105] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0106] like Figure 1 As shown, it is an embodiment of the present invention, which provides a transformer online monitoring and remote diagnosis system based on multi-parameter fusion, including one or more status detection devices installed on the transformer and a cloud platform. The status detection device is used to perform real-time online monitoring of the transformer status, including a data acquisition module, a feature matching module, a probability calculation module, a data upload module and a feature library update module. The cloud platform receives the data uploaded by the data upload module for remote diagnosis of the transformer status, including a feature library, an abnormal secondary analysis module, an abnormal confirmation module and a feature library sending module. The status detection device is connected to the data concentrator via RS485, Bluetooth or other micro-power wireless methods. The choice of connection method depends on the actual situation. The data concentrator and the cloud platform use the private network of the transformer owner.

[0107] It should be noted that the transformer referred to in this embodiment is preferably an oil-immersed distribution transformer. Oil is used as a cooling medium, and its temperature change directly affects the thermal expansion and contraction of the internal structure of the oil-immersed transformer, which in turn affects the vibration mode. Therefore, the temperature of the oil-immersed transformer is directly related to the vibration data, which is one of the important indicators of condition monitoring. Calibration of vibration data by temperature can improve the accuracy of transformer online monitoring. The condition monitoring of non-oil-immersed transformers does not involve the judgment of oil level height. The temperature calibration of vibration data is more dependent on ambient temperature and the thermal properties of transformer materials. On the basis of the present invention, the specific impact of temperature on vibration data can be determined through experiments and historical data analysis, so as to formulate targeted calibration strategies, which are also applicable to other monitoring and diagnosis processes. The system modifies the existing components of the transformer and does not require the addition of auxiliary equipment, which greatly improves the installation convenience and operation reliability, and reduces the cost of online monitoring and remote diagnosis.

[0108] Specifically, Figure 2 As shown, the data acquisition module includes a PT100 temperature measurement unit, an air pressure measurement unit, a laser distance measurement unit and a vibration detection unit, which are all centrally installed on a PCB board. It is used to integrate multiple sensors to collect the status parameters of the transformer in real time, including the temperature, pressure, oil level and vibration data in the oil tank, and perform Fourier transform on the vibration data to decompose different frequency components, extract vibration feature data and perform temperature calibration, including 100Hz proportion, total harmonic distortion, frequency point vibration energy and odd-even harmonic ratio.

[0109] The PT100 temperature measurement unit collects the voltage of PT100 in real time through the temperature measurement circuit, calculates the resistance value of PT100 according to the voltage, and maps the temperature data in the transformer. When the temperature data exceeds the threshold value, the temperature alarm is triggered. Figure 3 As shown in the figure, the PT100 is installed at the original oil temperature gauge position of the transformer. Because it is very close to the acquisition equipment, a two-wire connection method is used to meet the high requirements for temperature measurement accuracy.

[0110] The pressure measurement unit collects the pressure data of the transformer in real time through the pressure sensor, and triggers the pressure alarm when the pressure data exceeds the threshold value. This embodiment uses the QMP6989 sensor, and the pressure detection circuit design is as follows Figure 4 shown.

[0111] The laser ranging unit measures the oil level data of the transformer in real time through the oil level sensor. When the oil level data exceeds the threshold value, the oil level alarm is triggered. The laser ranging circuit design is as follows: Figure 5 shown.

[0112] The vibration detection unit collects the vibration data in the transformer through a vibration sensor at a sampling frequency of 50kHz. The vibration detection circuit design is as follows: Figure 6 The vibration spectrum is calculated by short-time Fourier transform, the fundamental frequency and multiple frequencies are extracted, the transformed vibration data is temperature calibrated, the vibration characteristic data is converted to data at 20°C, and uploaded to the cloud platform.

[0113] Vibration signal Frequency components obtained by short-time Fourier transform Defined as:

[0114] ;

[0115] In the formula, is the window function, represents angular frequency, * represents conjugation; is the value of the original vibration signal at time t′; Indicates at a fixed timet Based on the t′ The angular frequency is adjusted by the change of ω ;

[0116] The PT100 temperature measuring unit, the laser distance measuring unit, the air pressure measuring unit and the vibration detecting unit are centrally mounted on the same PCB board.

[0117] Furthermore, the method for temperature calibration of vibration characteristic data includes:

[0118] According to the influence of distribution transformer temperature on vibration, a temperature influence curve is drawn to calibrate the amplitude of vibration data at different frequency points, such as Figure 7 As shown. A temperature acceleration calibration mapping table is established according to the temperature influence curve, and then the acceleration of the real-time collected temperature data is calibrated by table lookup, so that all vibration characteristic data are converted into characteristic curves at 20°C. The processed data can more accurately reflect the state of the transformer and can be used for further analysis and fault diagnosis.

[0119] The feature matching module is used to extract vibration feature data from the calibrated vibration data and compare it with the local feature library to determine whether it exceeds the threshold. The vibration feature data includes low-frequency proportion, total harmonic distortion (THD), vibration energy at each frequency point, and odd-even harmonic ratio;

[0120] The first judgment basis of the device is to use the 100Hz proportion to judge the fault. The vibration characteristics of the transformer fault are often manifested as a decrease in the proportion of the spectrum base frequency, an increase in the proportion of high frequency, and an increase in the complexity of the spectrum. The calculation formula is:

[0121] ;

[0122] In the formula, is the amplitude at frequency point f in the spectrum, is the amplitude when f=100Hz; f max Indicates the maximum frequency, the value is 1000Hz;

[0123] The second judgment of the device is based on the total harmonic distortion of the transformer vibration, the total harmonic distortion The calculation formula is:

[0124] ;

[0125] In the formula, is the i-th harmonic amplitude, N is the harmonic number, is the i-th harmonic weight;

[0126] The third judgment of the device is based on the odd-even harmonic ratio of the vibration frequency (50Hz as the base) of the transformer vibration. The calculation formula is:

[0127] ;

[0128] In the formula, and are the amplitudes of odd and even harmonics respectively; is the upper limit of odd frequency, Upper limit of even frequency.

[0129] The vibration energy is calculated using the root mean square value of the vibration velocity. The formula is:

[0130] ;

[0131] Where R is the root mean square value of the vibration velocity, T is the vibration period, is the vibration speed at time t;

[0132] The proportion of vibration energy at frequency point f for:

[0133] .

[0134] The probability calculation module is used to calculate the matching probability between the extracted vibration feature data and the vibration data features in the feature library, and to judge the transformer state based on the matching probability. If an abnormal state is matched, the data is collected three times and matched. If an abnormal state is still matched after repeated data collection three times, it is marked as abnormal and uploaded to the cloud platform.

[0135] Preferably, a frequency-amplitude curve is drawn with the multiple frequency of 50 Hz as the x-axis, the maximum value is 1000 Hz, and the amplitude as the y-axis. It is compared with the reference curve of the transformer in the feature library and the similarity between the curves is calculated. , the calculation formula is:

[0136] ;

[0137] In the formula, , are the nth eigenvalues ​​in the data sequences X and Y, respectively. , are the average values ​​in the data series X and Y respectively, and N is the total number of eigenvalues;

[0138] When the similarity is lower than the preset threshold, a characteristic curve is generated for each feature in the vibration feature data, and the similarity is compared with the reference curve in the feature library. The characteristic curve with the highest similarity is selected and reported to the cloud platform.

[0139] The data upload module is used to upload the collected state parameters, vibration characteristic data and calculated matching probability to the cloud platform;

[0140] The feature library update module is used to receive the latest vibration data features sent by the cloud platform to update the local feature library.

[0141] The feature library is used to store the reference curve of the transformer in normal state and the vibration data characteristics in abnormal state;

[0142] The abnormal secondary analysis module is used to perform secondary analysis based on the data uploaded by the data upload module and the historical load information to further confirm whether the transformer is in an abnormal state. If the abnormality is still confirmed after the secondary analysis, an abnormal alarm will be issued to notify manual processing;

[0143] Specifically, the implementation method of the abnormal secondary analysis module includes:

[0144] The IQR method is used to automatically detect and process outliers in state parameters. The formula is: , in is the interquartile range, and are the first and third quartiles of the data, respectively. Outliers are defined as values ​​less than or greater than The value of

[0145] The Z-score standardization process state parameters are applied, and the formula is:

[0146] ;

[0147] In the formula, is the state parameter, , are the mean and standard deviation of the state parameters, is the standardized state parameter;

[0148] Apply wavelet transform to vibration data to capture local features and anomalies of vibration signals in the frequency domain;

[0149] The random forest method is used to build a classification model to classify the data uploaded by the status detection device to determine whether there is anomaly. The online gradient descent algorithm is introduced to enable the classification model to continuously learn from new data and update in real time.

[0150] The abnormality confirmation module is used to manually confirm whether the transformer has an abnormality. If the abnormality is manually confirmed, an abnormality confirmation message is sent;

[0151] The feature library sending module is used to send the latest vibration data features to the state detection device.

[0152] like Figure 8 As shown in FIG. 1 , an embodiment of the present invention is provided, which provides a transformer online monitoring and remote diagnosis method based on multi-parameter fusion, comprising the following steps:

[0153] S1: The device collects sensor data;

[0154] Integrate multiple sensors to collect transformer status parameters in real time, including temperature, pressure, oil level and vibration data inside the transformer;

[0155] S2: device data processing;

[0156] Perform short-time Fourier transform on the vibration data to calculate the vibration spectrum, analyze the vibration energy of different frequency components, extract the fundamental frequency and multiple frequencies, and perform temperature calibration on the transformed vibration data to convert the vibration characteristic data into data at 20°C;

[0157] Methods for temperature calibration of vibration characteristic data include:

[0158] According to the influence of distribution transformer temperature on vibration, a temperature influence curve is drawn, and a temperature acceleration calibration mapping table is established according to the temperature influence curve. The acceleration of the real-time collected temperature data is calibrated by table lookup, so that all vibration data are converted to data at 20°C.

[0159] Extract vibration characteristic data from the calibrated vibration data, including 100Hz proportion, total harmonic distortion (THD), frequency point vibration energy and odd-even harmonic ratio;

[0160] Vibration signal Frequency components obtained by short-time Fourier transform Defined as:

[0161] ;

[0162] In the formula, is the window function, represents angular frequency, * represents conjugation; is the value of the original vibration signal at time t′; Indicates at a fixed time t Based on the t′ The angular frequency is adjusted by the change of ω ;

[0163] 100Hz ratio The calculation formula is:

[0164] ;

[0165] In the formula, is the amplitude at frequency point f in the spectrum, is the amplitude when f=100Hz; f max Indicates the maximum frequency, the value is 1000Hz;

[0166] Total Harmonic Distortion The calculation formula is:

[0167] ;

[0168] In the formula, is the i-th harmonic amplitude, N is the harmonic number, is the i-th harmonic weight;

[0169] Odd-even harmonic ratio The calculation formula is:

[0170] ;

[0171] In the formula, and are the amplitudes of odd and even harmonics respectively; is the upper limit of odd frequency, is the upper limit of even frequency;

[0172] The vibration energy is calculated using the root mean square value of the vibration velocity. The formula is:

[0173] ;

[0174] Where R is the root mean square value of the vibration velocity, T is the vibration period, is the vibration speed at time t;

[0175] The proportion of vibration energy at frequency point f for:

[0176] .

[0177] S3: device data reporting;

[0178] Upload all data of the status detection device to the cloud platform;

[0179] S4: Device abnormality analysis and reporting;

[0180] like Fig. 9 The following are shown: whether the temperature, pressure and oil level data exceed the threshold value;

[0181] The vibration feature data is compared with the local feature library to determine whether it exceeds the threshold;

[0182] Calculate the matching probability between the extracted vibration feature data and the vibration data features in the feature library, and judge the transformer status based on the matching probability. If an abnormal state is matched, repeat the collection three times and match. If an abnormal state is still matched after repeating the collection three times, mark it as abnormal and upload it to the cloud platform. Specifically: use the multiple frequency of 50Hz as the x-axis, the maximum value is 1000Hz, and the amplitude as the y-axis to draw a frequency-amplitude curve, compare it with the benchmark curve of the transformer in the feature library, and calculate the similarity between the curves. , the calculation formula is:

[0183] ;

[0184] In the formula, , are the nth eigenvalues ​​in the data sequences X and Y, respectively. , are the average values ​​in the data series X and Y respectively, and N is the total number of eigenvalues;

[0185] When the similarity is lower than the preset threshold, the characteristic curve generated by each feature in the vibration feature data is compared with the reference curve in the feature library for similarity, and the characteristic curve with the highest similarity is selected and reported to the cloud platform;

[0186] S5: Platform secondary analysis;

[0187] Perform secondary analysis based on the data uploaded by the status detection device and historical load information to further confirm whether the transformer is in an abnormal state. If the abnormality is still confirmed after the secondary analysis, an abnormal alarm will be issued to notify manual processing;

[0188] Secondary analysis methods include:

[0189] The IQR method is used to automatically detect and process outliers in state parameters. The formula is: , in is the interquartile range, and are the first and third quartiles of the data, respectively. Outliers are defined as values ​​less than or greater than The value of

[0190] The Z-score standardization process state parameters are applied, and the formula is:

[0191] ;

[0192] In the formula, is the state parameter, , are the mean and standard deviation of the state parameters, is the standardized state parameter;

[0193] Apply wavelet transform to vibration data to capture local features and anomalies of vibration signals in the frequency domain;

[0194] The random forest method is used to build a classification model to classify the data uploaded by the status detection device to determine whether there is anomaly. The online gradient descent algorithm is introduced to enable the classification model to continuously learn from new data and update in real time.

[0195] S6: Platform abnormality confirmation;

[0196] Manually confirm whether there is any abnormality in the transformer. If the abnormality is manually confirmed, an abnormality confirmation message is sent;

[0197] S7: Determine whether a new abnormality is generated. If so, update and send the abnormal data; if not, re-collect the sensor data.

[0198] In summary, the present invention extracts vibration feature data including 100 Hz proportion, total harmonic distortion, frequency point vibration energy and odd-even harmonic ratio by Fourier transform and temperature calibration of vibration data, which can comprehensively reflect the operating status of the transformer; by calculating the matching probability between vibration feature data and data features in the feature library, combined with data collected repeatedly, accurate judgment and marking of abnormal states can be achieved, and the abnormal states can be uploaded to the cloud platform for remote diagnosis and processing; the cloud platform receives and analyzes the uploaded data, and uses the abnormal secondary analysis module and the abnormal confirmation module to further confirm the abnormal state of the transformer, notify manual processing, and update the feature library to ensure that the system can continue to learn and optimize; online monitoring and remote diagnosis are realized, which reduces the dependence on manual regular inspections and maintenance, and reduces the operation and maintenance costs and workload.

[0199] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. Transformer online monitoring and remote diagnosis system based on multi-parameter fusion, characterized in that: The system includes one or more status detection devices installed on the transformer and a cloud platform. The status detection device is connected to the data concentrator via RS485, Bluetooth or other micro-power wireless methods, and the data concentrator is connected to the cloud platform via a private network of the device. The state detection device is used to perform real-time online monitoring of the transformer state, including: A data acquisition module is used to integrate multiple sensors to collect the status parameters of the transformer in real time, wherein the status parameters include the temperature, pressure, oil level and vibration data in the transformer, perform Fourier transform on the vibration data, analyze the vibration energy of different frequency components, and perform temperature calibration on the vibration data after Fourier transform; A feature matching module is used to extract vibration feature data from the calibrated vibration data, wherein the vibration feature data includes 100 Hz proportion, total harmonic distortion, frequency point vibration energy and odd-even harmonic ratio, and compare the vibration feature data with the local feature library to determine whether it exceeds a threshold; The probability calculation module is used to calculate the matching probability between the extracted vibration feature data and the vibration data features in the feature library, and to judge the transformer state based on the matching probability. If an abnormal state is matched, the data is collected three times and matched. If an abnormal state is still matched after repeated data collection three times, it is marked as abnormal and uploaded to the cloud platform. The data upload module is used to upload the collected state parameters, vibration characteristic data and calculated matching probability to the cloud platform; The feature library update module is used to receive the latest vibration data features sent by the cloud platform to update the local feature library; The cloud platform is used to remotely diagnose the transformer status, including: A feature library is used to store the reference curve of the transformer under normal conditions and the vibration data characteristics under abnormal conditions; The abnormal secondary analysis module is used to perform secondary analysis based on the data uploaded by the data upload module and the historical load information to further confirm whether the transformer is in an abnormal state. If the abnormality is still confirmed after the secondary analysis, an abnormal alarm will be issued to notify manual processing; The abnormality confirmation module is used to manually confirm whether the transformer has an abnormality. If the abnormality is manually confirmed, an abnormality confirmation message is sent; The feature library sending module is used to send the latest vibration data features to the state detection device; The data acquisition module comprises: The PT100 temperature measuring unit is used to collect the voltage of the PT100 in real time through the temperature measuring circuit, calculate the resistance value of the PT100 according to the voltage, and thus map the temperature data in the transformer. When the temperature data exceeds the threshold value, the temperature alarm is triggered; the PT100 is installed at the oil thermometer position of the transformer and is connected in a two-wire system; An air pressure measuring unit, used to collect pressure data of the transformer in real time through a pressure sensor, and trigger a pressure alarm when the pressure data exceeds a threshold value; A laser ranging unit is used to measure the oil level data of the transformer in real time through an oil level sensor, and trigger an oil level alarm when the oil level data exceeds a threshold value; A vibration detection unit is used to collect vibration data in the transformer through a vibration sensor at a sampling frequency of 50kHz, calculate the vibration spectrum through short-time Fourier transform, extract the fundamental frequency and multiple frequencies, perform temperature calibration on the transformed vibration data, and transform the vibration data into data at 20°C; Vibration signal Frequency components obtained by short-time Fourier transform Defined as: ； In the formula, is the window function, represents angular frequency, * represents conjugation; is the value of the original vibration signal at time t′; Indicates at a fixed time t Based on the t′ The angular frequency is adjusted by the change of ω ; The PT100 temperature measuring unit, laser distance measuring unit, air pressure measuring unit and vibration detection unit are centrally installed on the same PCB board; In the data acquisition module, the method for temperature calibration of vibration data includes: According to the influence of distribution transformer temperature on vibration, a temperature influence curve is drawn, and a temperature acceleration calibration mapping table is established according to the temperature influence curve. The acceleration of the real-time collected temperature data is calibrated by table lookup, so that all vibration data are converted to data at 20°C. In the feature matching module, 100Hz accounts for The calculation formula is: ； In the formula, is the amplitude at frequency point f in the spectrum, is the amplitude when f=100Hz; f max Indicates the maximum frequency, the value is 1000Hz; Total Harmonic Distortion The calculation formula is: ； In the formula, is the i-th harmonic amplitude, N is the harmonic number, is the i-th harmonic weight; Odd-even harmonic ratio The calculation formula is: ； In the formula, and are the amplitudes of odd and even harmonics respectively; is the upper limit of odd frequency, is the upper limit of even frequency; The vibration energy is calculated using the root mean square value of the vibration velocity. The formula is: ； Where R is the root mean square value of the vibration velocity, T is the vibration period, is the vibration speed at time t; The proportion of vibration energy at frequency point f for: ； In the probability calculation module, a frequency-amplitude curve is drawn with the multiple frequency of 50 Hz as the x-axis, the maximum value is 1000 Hz, and the amplitude as the y-axis. The curve is compared with the reference curve of the transformer in the feature library to calculate the similarity between the curves. , the calculation formula is: ； In the formula, , are the nth eigenvalues ​​in the data sequences X and Y, respectively. , are the average values ​​in the data series X and Y respectively, and N is the total number of eigenvalues; When the similarity is lower than the preset threshold, the characteristic curve generated by each feature in the vibration feature data is compared with the reference curve in the feature library for similarity, and the characteristic curve with the highest similarity is selected and reported to the cloud platform; The implementation method of the abnormal secondary analysis module includes: The IQR method is used to automatically detect and process outliers in state parameters. The formula is: , in is the interquartile range, and are the first and third quartiles of the data, respectively. Outliers are defined as values ​​less than or greater than The value of The Z-score standardization process state parameters are applied, and the formula is: ； In the formula, is the state parameter, , are the mean and standard deviation of the state parameters, is the standardized state parameter; Apply wavelet transform to vibration data to capture local features and anomalies of vibration signals in the frequency domain; The random forest method is used to build a classification model to classify the data uploaded by the status detection device to determine whether there is anomaly. The online gradient descent algorithm is introduced to enable the classification model to continuously learn from new data and update in real time.

2. A transformer online monitoring and remote diagnosis method based on multi-parameter fusion, characterized in that: The method comprises: Integrate multiple sensors to collect transformer status parameters in real time, including temperature, pressure, oil level and vibration data inside the transformer; Perform short-time Fourier transform on the vibration data to calculate the vibration spectrum, analyze the vibration energy of different frequency components, extract the fundamental frequency and the frequency multiples, and perform temperature calibration on the transformed vibration data to convert the vibration data into data at 20°C; the method for temperature calibration of vibration data includes: drawing a temperature influence curve according to the influence of the temperature of the distribution transformer on the vibration, establishing a temperature acceleration calibration mapping table according to the temperature influence curve, and calibrating by table lookup according to the acceleration of the temperature data collected in real time, so that all vibration data are converted into data at 20°C; extracting vibration characteristic data from the calibrated vibration data, including 100Hz proportion, total harmonic distortion, frequency point vibration energy and odd-even harmonic ratio; Vibration signal Frequency components obtained by short-time Fourier transform Defined as: ； In the formula, is the window function, represents angular frequency, * represents conjugation; is the value of the original vibration signal at time t′; Indicates at a fixed time t Based on the t′ The angular frequency is adjusted by the change of ω ; 100Hz ratio The calculation formula is: ； In the formula, is the amplitude at frequency point f in the spectrum, is the amplitude when f=100Hz; f max Indicates the maximum frequency, the value is 1000Hz; Total Harmonic Distortion The calculation formula is: ； In the formula, is the i-th harmonic amplitude, N is the harmonic number, is the i-th harmonic weight; Odd-even harmonic ratio The calculation formula is: ； In the formula, and are the amplitudes of odd and even harmonics respectively; is the upper limit of odd frequency, is the upper limit of even frequency; The vibration energy is calculated using the root mean square value of the vibration velocity. The formula is: ； Where R is the root mean square value of the vibration velocity, T is the vibration period, is the vibration speed at time t; The proportion of vibration energy at frequency point f for: ； Upload all data of the status detection device to the cloud platform; Upload all data of the status detection device to the cloud platform; Device abnormality reporting, including: Whether the temperature, pressure and oil level data exceed the threshold value, if so, trigger the temperature alarm, pressure alarm and / or oil level alarm; The vibration feature data is compared with the local feature library to determine whether it exceeds the threshold; Calculate the matching probability between the extracted vibration feature data and the vibration data features in the feature library, and judge the transformer status based on the matching probability. If an abnormal state is matched, repeat the collection three times and match. If an abnormal state is still matched after repeating the collection three times, mark it as abnormal and upload it to the cloud platform. Specifically: use the multiple frequency of 50Hz as the x-axis, the maximum value is 1000Hz, and the amplitude as the y-axis to draw a frequency-amplitude curve, compare it with the benchmark curve of the transformer in the feature library, and calculate the similarity between the curves. , the calculation formula is: ； In the formula, , are the nth eigenvalues ​​in the data sequences X and Y, respectively. , are the average values ​​in the data series X and Y respectively, and N is the total number of eigenvalues; When the similarity is lower than the preset threshold, the characteristic curve generated by each feature in the vibration feature data is compared with the reference curve in the feature library for similarity, and the characteristic curve with the highest similarity is selected and reported to the cloud platform; Perform secondary analysis based on the data uploaded by the status detection device and historical load information to further confirm whether the transformer is in an abnormal state. If the abnormality is still confirmed after the secondary analysis, an abnormal alarm will be issued to notify manual processing; The secondary analysis method includes: using the IQR method to automatically detect and process outliers in the state parameters, the formula is: , in is the interquartile range, and are the first and third quartiles of the data, respectively. Outliers are defined as values ​​less than or greater than The value of The Z-score standardization process state parameters are applied, and the formula is: ； In the formula, is the state parameter, , are the mean and standard deviation of the state parameters, is the standardized state parameter; Apply wavelet transform to vibration data to capture local features and anomalies of vibration signals in the frequency domain; A classification model is constructed using the random forest method to classify the data uploaded by the status detection device to determine whether there is an anomaly. An online gradient descent algorithm is introduced to enable the classification model to continuously learn from new data and update in real time. Manually confirm whether there is any abnormality in the transformer. If the abnormality is manually confirmed, an abnormality confirmation message is sent; Determine whether a new abnormality has occurred. If so, update and send the abnormal data; if not, re-collect the sensor data.

Images