Transformer oil quality comprehensive monitoring method and system and storage medium

Abstract

The invention discloses a transformer oil quality comprehensive monitoring method and system and a storage medium, relates to the field of transformer oil quality monitoring, and solves the technical problem that the oil quality change is difficult to find in time due to the fact that the transformer oil quality cannot be monitored in real time in the prior art. The method comprises the following steps: acquiring oil quality parameters of transformer oil in real time, and preprocessing; performing temperature compensation on the preprocessed refractive index based on the temperature to obtain an actual refractive index, and constructing an oil quality feature vector; a plurality of oil quality feature vectors are extracted from historical data to train an LSTM-Attention model, and an oil quality evaluation model is obtained through training; the oil quality evaluation model evaluates the early warning grade of the oil quality of the to-be-detected transformer oil; the physical state of the transformer oil can be monitored in real time, oil quality deterioration early warning can be given out in time, and safe operation of an electric power system is ensured.

Description

Technical Field

[0001] The present invention belongs to the field of transformer oil quality monitoring, and specifically relates to a comprehensive transformer oil quality monitoring method, system and storage medium. Background Art

[0002] Transformer oil, as the core medium in the power system, plays important roles in insulation, cooling and arc resistance. Its quality is directly related to the normal operation of the transformer and the safety of the power system. Physical properties of transformer oil, such as refractive index, water content, acid value, viscosity, etc., can reflect the degree of oil aging, pollutant content and its potential failure risks. Therefore, real-time and accurate monitoring of transformer oil is of great significance for ensuring the safe operation of power equipment and extending the equipment life.

[0003] Traditional oil quality detection usually requires shutting down the transformer to collect oil samples, and then sending them to the laboratory for chemical analysis. This off-line detection method not only causes the transformer to shut down, increasing the outage cost, but also cannot achieve real-time monitoring, making it difficult to timely detect changes in oil quality and potential equipment problems. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; for this purpose, the present invention provides a comprehensive transformer oil quality monitoring method, system and storage medium, which are used to solve the technical problem that the prior art cannot perform real-time monitoring on transformer oil quality, resulting in difficulty in timely detecting changes in oil quality.

[0005] To achieve the above object, the first aspect of the present invention provides a comprehensive transformer oil quality monitoring method, including: Real-time obtain the oil quality parameters of transformer oil and perform preprocessing; wherein, the oil quality parameters include refractive index, temperature and oil flow pressure; Perform temperature compensation on the preprocessed refractive index based on the temperature to obtain the actual refractive index; Construct an oil quality feature vector based on the actual refractive index, the preprocessed temperature and the oil flow pressure; Extract a number of oil quality feature vector pairs from historical data to train the LSTM-Attention model, and obtain an oil quality evaluation model through training; the LSTM-Attention model is used to predict the warning level of transformer oil quality based on the oil quality feature vector of transformer oil; Input the oil quality feature vector of the transformer oil to be detected into the oil quality evaluation model, and output the warning level of the transformer oil quality.

[0006] Preferably, the oil quality parameters are obtained in real time through an optical fiber refractive index sensor; The optical fiber refractive index sensor is a dual-parameter composite probe including a fiber Bragg grating and a fiber grating tilt grating; The fiber Bragg grating is used to detect the refractive index and temperature, and the fiber grating tilted grating is used to monitor the oil flow pressure.

[0007] Preferably, the preprocessing process of the oil quality parameters includes: Fitting each of the oil quality parameters collected within the analysis period to obtain the curves of each oil quality parameter; placing the curves of each oil quality parameter and the corresponding preset standard parameters in the same coordinate system; Obtaining the abnormal regions from the regions enclosed by the oil quality curve and the preset standard curve, calculating the area of the abnormal regions, and determining whether the area of the abnormal regions is less than the preset area; if yes, removing each of the oil quality parameters in the abnormal regions; if no, retaining each of the oil quality parameters in the abnormal regions; The calculation formula for the area of the abnormal regions is as follows: ; where i is the serial number of each oil quality parameter, i = 0, 1, 2, representing temperature, actual refractive index, and oil flow pressure respectively; is the change curve of the oil quality parameter i within the analysis period, is the preset standard curve of the oil quality parameter i; t1 and t2 are the corresponding times at the intersection of the two curves respectively.

[0008] Preferably, the abnormal regions are the regions enclosed above the preset standard curve by the oil quality curve, and the preset standard curve is obtained by fitting a number of historical normal oil quality parameters; among them, the normal oil quality parameters are the oil quality parameters within the preset range.

[0009] It should be noted that the abnormal regions are the oil quality parameters that exceed the normal upper limit.

[0010] In the present invention, each oil quality parameter is respectively fitted into a curve, and the corresponding preset standard parameter is placed in the same coordinate system. This visualization method can intuitively present the change trend of the oil quality parameter over time or other variables. The area enclosed by the oil quality curve and the preset standard curve is extracted, and the area of the region is calculated. Using the region enclosed by the curve as the judgment criterion for the abnormal region can more comprehensively consider the change of the oil quality parameter over a period of time compared with simple numerical threshold judgment. For example, some oil quality parameters may have small fluctuations in a short period of time, but the overall trend is still within the normal range. By using the method of the region enclosed by the curve, such situations can be avoided from being misjudged as abnormal. When the area of the abnormal region is less than the preset area, it is considered that this abnormality may be caused by accidental factors such as measurement errors and environmental interference. Removing the oil quality parameters in this part of the abnormal region can effectively avoid the influence of these irrelevant interference data on subsequent analysis, improve the accuracy and reliability of the data. If the area of the abnormal region is not less than the preset area, the oil quality parameters in the abnormal region are retained. This ensures that the abnormal data with real analysis value will not be omitted, and can provide an important basis for in-depth analysis of the cause of oil quality abnormality and formulation of corresponding treatment measures.

[0011] Preferably, the temperature compensation is performed on the obtained preprocessed refractive index based on temperature to obtain the actual refractive index, including: The actual refractive index of the transformer oil is obtained through a temperature compensation algorithm, and the calculation formula is as follows: ; Wherein, The refractive index collected in real time, RIU / o C, is the difference between the current measured temperature and the standard temperature.

[0012] The refractive index of the transformer oil in the present invention will change with the temperature. If temperature compensation is not performed, the refractive index data collected in real time will contain errors caused by temperature. This algorithm effectively eliminates the influence of temperature on the refractive index measurement by subtracting the refractive index deviation (α×ΔT) caused by the temperature change amount, thereby obtaining a value closer to the true refractive index of the transformer oil and significantly improving the measurement accuracy. It reduces the misjudgment of equipment failures and unnecessary maintenance work that may be caused by inaccurate measurements due to temperature errors, thereby reducing the maintenance cost and operation risk of the transformer.

[0013] Preferably, constructing the oil quality feature vector based on the actual refractive index, the preprocessed temperature, and the oil flow pressure includes: ; Wherein, is the analysis period, is the difference between the highest and lowest actual refractive index within the analysis period, is the actual refractive index change rate, is the difference between the highest and lowest temperatures during the analysis period, f is the oil flow pulsation frequency, is the refractive index-temperature coupling parameter, is the cumulative change in refractive index, is the oil flow-refractive index dynamic coupling, is the refractive index sliding variance, j is the number of historical alarms, and the number of historical alarms is the cumulative number of times that the refractive index change rate of the transformer oil exceeds the preset safety threshold during continuous monitoring.

[0014] The present invention analyzes the factors influencing transformer oil quality. By incorporating the relevant features of these factors into vectors, it can fully account for the interactions between various factors. The actual refractive index change rate and the cumulative refractive index change can reflect the dynamic changes in transformer oil refractive index in real time. The former reflects the rate of change of refractive index per unit time, while the latter reflects the overall accumulation of refractive index changes within the analysis period, helping to promptly identify short-term fluctuations and long-term trends in oil quality. The refractive index sliding variance measures the degree of refractive index fluctuation within a certain time window and can reflect the stability of oil quality. A larger variance indicates more severe refractive index fluctuations and less stable oil quality; conversely, a smaller variance indicates relatively stable oil quality. This provides an important basis for evaluating transformer oil operational reliability. The refractive index-temperature coupling parameter and the oil flow-refractive index dynamic coupling reveal the intrinsic relationship between transformer oil refractive index, temperature, and oil flow. By analyzing these coupling relationships, we can gain an in-depth understanding of how oil quality changes under different working conditions, providing clues for further research into the mechanism of oil degradation. The historical alarm count also records the cumulative number of times the refractive index change rate of transformer oil exceeds the preset safety threshold during continuous monitoring. When the number of historical alarms is large, it indicates that the oil quality may be abnormal and requires special attention and further inspection.

[0015] Preferably, the training process of the oil quality assessment model is as follows: Several groups of historical oil quality feature vectors are collected from historical data and labeled with warning labels. The labeled historical oil quality feature vectors are divided into a training dataset and a validation dataset. The LSTM-Attention model is trained using the labeled historical oil quality feature vectors to obtain an oil quality assessment model.

[0016] Preferably, the LSTM-Attention model includes an input layer, an LSTM layer, an Attention layer and an output layer; The input layer is used to receive the oil quality feature vector; The LSTM layer includes a first layer and a second layer; the first layer is used to extract short-term dynamic features in the oil quality feature vector with a span less than N time steps; the second layer is used to extract long-term dynamic features in the oil quality feature vector with a span greater than 3N time steps; where N is a positive integer; The Attention layer is used to determine the weight coefficients of the time steps; Based on the weight coefficients of the time steps determined by the Attention layer, the output layer performs weighted fusion on the short-term dynamic features and long-term dynamic features extracted by the LSTM layer to generate a comprehensive feature vector, maps it to a preset transformer oil quality warning level classification space, and finally outputs the warning level of the transformer oil quality.

[0017] Preferably, the weight coefficients of the time steps are obtained through the following method, including: Through the formula The weight coefficients of the time steps are calculated ; where t is the time step, softmax is the activation function, tanh is the hyperbolic tangent activation function, h t is the hidden state at time step t, W is the weight matrix, b is the bias vector, u is the attention vector, and T is the transpose of the attention vector.

[0018] The LSTM-Attention model of the present invention consists of an input layer, an LSTM layer, an Attention layer, and an output layer, and the functions of each layer are clearly divided. The input layer is responsible for receiving the oil quality feature vector and providing a data basis for subsequent processing; the LSTM layer is used to extract short-term and long-term dynamic features and capture the law of oil quality change over time; the Attention layer determines the weight coefficients of the time steps and enhances the attention to important time information; the output layer then performs feature fusion and warning level output. The entire model structure is compact and the logic is clear.

[0019] Among them, the LSTM layer is divided into a first layer and a second layer, which respectively extract short-term dynamic features with a span less than N time steps and long-term dynamic features with a span greater than 3N time steps. This hierarchical feature extraction method can simultaneously consider the short-term fluctuations and long-term trends of oil quality, enabling the model to have a more comprehensive and in-depth understanding of the changes in oil quality status; The Attention layer can automatically focus on the time step information that is more important for oil quality assessment by calculating the weight coefficients of the time steps, highlighting the influence of key features. This mechanism can avoid the drawback of treating all time step information equally in traditional models and improve the sensitivity and utilization efficiency of the model to important information; The output layer performs weighted fusion on the short-term and long-term dynamic features extracted by the LSTM layer based on the time-step weight coefficients determined by the Attention layer to generate a comprehensive feature vector. The weighted fusion method can make full use of the advantages of different features, enabling the comprehensive feature vector to more accurately reflect the true state of the oil quality, thereby improving the accuracy of early warning level classification.

[0020] Preferably, the second aspect of the present invention provides a comprehensive monitoring system for transformer oil quality, including a data acquisition module, a data processing module, an oil quality assessment module, and a blockchain trusted management module; The data acquisition module is used to collect the refractive index, temperature, and oil flow pressure of the transformer oil in real time; The data processing module is used to process and analyze the detection data; The oil quality assessment module is based on the LSTM-Attention model to evaluate and give early warnings for oil quality deterioration; The blockchain trusted management module is used to perform trusted management on the detection data.

[0021] Preferably, the data upload process of the blockchain trusted management module is that the edge device uploads the monitored data to the side chain and generates a unique identifier. The smart contract in the main chain records the unique identifier and the corresponding timestamp, and uses a dynamic consensus mechanism for data verification and archiving.

[0022] In the present invention, the edge device uploads the monitoring data to the side chain and generates a unique identifier. This identifier is like the "ID card" of the data and is tightly bound to the data once generated. During subsequent transmission and storage, if the data is maliciously tampered with, its corresponding unique identifier will not match, so that data anomalies can be quickly discovered, effectively preventing data tampering and ensuring the integrity of the data; By adopting the combination of the side chain and the main chain, the side chain is responsible for receiving the data uploaded by the edge device, reducing the burden on the main chain, and at the same time providing an additional layer of security protection for the data. The main chain records the unique identifier and the timestamp through the smart contract, further ensuring the immutability and security of the data. Even if the side chain is attacked, the records on the main chain can still be used as reliable evidence for the authenticity of the data.

[0023] Preferably, the third aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the comprehensive monitoring method for transformer oil quality.

[0024] Compared with the prior art, the beneficial effects of the present invention are: The present invention adopts a dual-parameter composite probe combined with a matrix decoupling algorithm to achieve high-precision synchronous measurement of refractive index, temperature, and oil flow pressure parameters, solving the limitation of traditional sensors that only monitor a single parameter; and according to the LSTM-Attention multi-dimensional analysis model, through 8-dimensional feature vectors such as the refractive index change rate, temperature gradient, oil flow pulsation frequency, etc. and sliding window time-frequency analysis, combined with a 128-node short-term feature layer + 64-node long-term feature layer of a double-layer LSTM and the Attention mechanism, dynamic prediction of oil quality deterioration is realized; in addition, based on the main-chain - side-chain double-chain architecture and dynamic consensus mechanism of the blockchain, it ensures that the monitoring data is tamper-proof throughout the process, improves the efficiency of evidence storage, supports the dynamic consensus mechanism, and enhances the system throughput. The present invention combines fiber optic refractive index sensing, deep learning time series prediction, and blockchain evidence storage, solving the core problems such as poor real-time performance, low data credibility, and lagging early warning in traditional oil quality monitoring, ensuring that the system can achieve high-precision and real-time monitoring of transformer oil quality, while ensuring the reliable management of data and early warning of oil quality deterioration, providing a strong guarantee for the safe operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0026] Figure 1 It is a schematic flow chart of the method of the present invention; Figure 2 It is a schematic flow chart of the method for training the oil quality evaluation model of the present invention; Figure 3 It is a schematic structural diagram of the system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] The following will clearly and completely describe the technical solutions of the present invention in combination with the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0028] Please refer to Figure 1 , the first aspect embodiment of the present invention provides a comprehensive monitoring method for transformer oil quality, including: Using a fiber optic refractive index sensor to collect the oil quality parameters of transformer oil in real time; wherein, the oil quality parameters include refractive index, temperature, and oil flow pressure; and performing temperature compensation on the collected refractive index to obtain the actual refractive index; Specifically, the fiber optic refractive index sensor uses a dual-parameter composite probe. The dual-parameter composite probe includes a fiber Bragg grating and a fiber grating tilt grating; For the fiber Bragg grating, the phase mask method is used for inscription. During the inscription process, parameters such as the power of the ultraviolet laser, the exposure time, and the period of the phase mask are precisely controlled. For example, the ultraviolet laser power is stabilized at 50 mW, the exposure time is set to 30 s, and the phase mask period is 1060 nm, so as to ensure the accuracy and stability of the reflection wavelength of the fiber Bragg grating, enabling it to accurately sense the changes in the refractive index and temperature of the transformer oil.

[0029] The fiber grating tilt grating precisely adjusts the tilt angle of the grating to 8° to detect the oil flow pressure of the transformer oil. This angle has been verified through a large number of experiments, which can ensure sensitivity to the oil flow shear force while not affecting the monitoring effect of the refractive index. The specific detection process is as follows: A clear oil flow direction arrow is marked on the surface of the fiber optic probe to guide the orientation calibration during installation, ensuring that the probe maintains a preset angle (such as parallel or perpendicular) with the oil flow direction. The arrow direction is aligned with the fiber deformation sensitive axis, enabling the lateral pressure change (generated by the oil flow impact) to be accurately captured by the fiber grating tilt grating. When the transformer oil flows along the arrow direction, the oil flow exerts a lateral shear force on the probe, causing the fiber to undergo micron-level deformation (such as bending or stretching). The grating period changes due to the deformation, and the characteristic wavelength of the reflection spectrum shifts. The shift amount is proportional to the oil flow pressure. By the relationship between the direction marked by the arrow and the deformation direction, the forward / backward oil flow can be distinguished.

[0030] Furthermore, the various oil quality parameters collected within the analysis period are respectively fitted to obtain the curves of the various oil quality parameters; the curves of the various oil quality parameters and the corresponding preset standard parameters are respectively placed in the same coordinate system; An abnormal area is obtained from the areas enclosed by the oil quality curve and the preset standard curve, and the area of the abnormal area is calculated. It is judged whether the area of the abnormal area is less than the preset area; if so, the various oil quality parameters in the abnormal area are removed; if not, the various oil quality parameters in the abnormal area are retained; Among them, the abnormal area is the area enclosed by the oil quality curve above the preset standard curve; the preset standard curve is obtained by fitting a number of historical normal oil quality parameters; among them, the normal oil quality parameters are the oil quality parameters within the preset range.

[0031] For example: within an analysis period (such as 24 hours), the oil quality parameters, including the actual refractive index, temperature, and oil flow pressure, are collected once every 1 hour through the sensors installed on the transformer, and a total of 24 groups of data are collected. The analysis period can be set according to the actual situation; Using curve fitting methods such as the least squares method, fit the 24 groups of data of the actual refractive index, temperature, and oil flow pressure respectively to obtain the curve equations of each oil quality parameter. For example, for the actual refractive index data, the obtained curve equation is n = f1(t) (t is time), the temperature curve equation is T = f2(t), and the oil flow pressure curve equation is P = f3(t).

[0032] Suppose the oil quality parameters of this transformer in the normal operating state in the past several years are collected. According to the preset normal value ranges of each oil quality parameter, filter out the data within the normal range to obtain several groups of historical normal oil quality parameters; Using the filtered historical normal oil quality parameters, perform curve fitting on the actual refractive index, temperature, and oil flow pressure respectively to obtain the preset standard curve equations. Suppose the standard curve equation of the actual refractive index is ns = g1(t), the standard curve equation of the temperature is Ts = g2(t), and the standard curve equation of the oil flow pressure is Ps = g3(t).

[0033] Plot the actual refractive index curve n = f1(t) and the standard curve ns = g1(t), the temperature curve T = f2(t) and the standard curve Ts = g2(t), and the oil flow pressure curve P = f3(t) and the standard curve Ps = g3(t) in the same coordinate system respectively. For each oil quality parameter, find the area enclosed by the oil quality curve above the preset standard curve. These areas are the abnormal areas.

[0034] Use numerical integration methods (such as the trapezoidal method) to calculate the area of each abnormal area. For example, for the abnormal area of the actual refractive index, its area Sn can be calculated by the following formula: ; where i is the serial number of each oil quality parameter. Suppose i = 0 represents temperature, i = 1 represents the actual refractive index, and i = 2 represents the oil flow pressure; is the change curve within the analysis period of oil quality parameter i, is the preset standard curve of oil quality parameter i; t1 and t2 are the corresponding times at the intersection of the two curves respectively.

[0035] Preset area setting: According to actual needs and experience, set the preset areas of the abnormal areas of the actual refractive index, temperature, and oil flow pressure to Sn, ST, and SP respectively; For the area S0 of the temperature abnormal area, if S0 < ST, then remove the temperature data of each time point within this abnormal area; if S0 ≥ ST, then retain the temperature data of each time point within this abnormal area; For the actual refractive index abnormal area S1, if S1 < Sn, the actual refractive index data at each time point within this abnormal area shall be removed; if S1 ≥ Sn, the actual refractive index data at each time point within this abnormal area shall be retained; For the oil flow pressure abnormal area S2, if S2 < SP, the oil flow pressure data at each time point within this abnormal area shall be removed; if S2 ≥ SP, the oil flow pressure data at each time point within this abnormal area shall be retained.

[0036] The actual refractive index value of the transformer oil is obtained by precisely measuring the wavelength shift through a demodulator and combining with a temperature compensation algorithm.

[0037] Since temperature changes will affect the refractive index of the optical fiber, the reference grating method is adopted for temperature compensation in this system. By measuring the wavelength drift caused by temperature and differentiating it from the refractive index change, the interference of temperature is eliminated. The calculation formula for the actual refractive index n is as follows: ; where, the refractive index collected in real time, RIU / o C, is the difference between the current measured temperature and the standard temperature.

[0038] Furthermore, for the mixed signal collected by the dual-parameter composite probe, the matrix decoupling algorithm is adopted for decoupling. By establishing a mathematical relationship model among the refractive index, temperature, and oil flow pressure signals, the corresponding decoupling algorithm program is written. During the program implementation process, matrix operations are performed on the collected signals to separate the independent refractive index, temperature, and oil flow pressure signals, providing accurate data for subsequent analysis. To remove the noise in the data, a digital filtering algorithm is adopted. For example, the Kalman filtering algorithm is used to process the data. This algorithm can perform optimal estimation on the data containing noise according to the state equation and observation equation of the system, effectively improving the quality of the data.

[0039] It should be noted that when installing the fiber optic refractive index sensor, a position with stable oil flow and uniform temperature in the transformer oil circulation pipeline is selected. Through a specially designed installation fixture, the dual-parameter composite probe is firmly fixed inside the pipeline to ensure that the probe is in full contact with the oil flow and will not be displaced or damaged due to the impact of the oil flow. During the installation process, a high-precision optical alignment device is used to ensure the accurate connection between the optical fiber and the probe, reducing the loss of optical signals. After the connection is completed, the sensor is calibrated using a standard refractive index oil sample, an environment at different temperatures, and a simulated oil flow device. By using an oil sample with a known refractive index, the output signal of the sensor is measured, and based on the theoretical relationship between the wavelength shift and the refractive index change, the measurement results of the sensor are calibrated and corrected. For temperature calibration, the output of the sensor is recorded in different temperature environments, and the wavelength drift caused by temperature is compensated through a temperature compensation algorithm (such as the reference grating method) to ensure the accuracy and reliability of the measured refractive index value. For the oil flow parameters, a simulated oil flow device is used to generate oil flows with different flow rates and pressures, and the transverse pressure change caused by the oil flow monitored by the fiber grating tilted grating and the function of monitoring the oil flow shear force are calibrated to determine the sensitivity and response characteristics of the sensor.

[0040] Extract several oil quality feature vectors from historical data to train the LSTM-Attention model, and obtain an oil quality evaluation model through training. Specifically, before extracting the oil quality feature vectors, the collected oil quality parameters are preprocessed. First, the oil quality parameters are denoised to remove outliers caused by factors such as sensor noise and electromagnetic interference. Methods such as median filtering are used to smooth the data and improve the data quality. Then, the oil quality parameters are normalized to map parameter values in different ranges to the [0,1] interval, so that different features have the same weight, facilitating subsequent model training and analysis.

[0041] According to the requirements of oil quality deterioration evaluation, multi-dimensional oil quality feature vectors are extracted. Based on the collected oil quality parameters, the refractive index change rate, refractive index-temperature coupling parameter, cumulative refractive index change, oil flow-refractive index dynamic coupling, and refractive index sliding variance are calculated. Among them, the calculation formula for the refractive index change rate is ; the calculation formula for the refractive index-temperature coupling parameter is ; the calculation formula for the cumulative refractive index change is ; the calculation formula for the oil flow-refractive index dynamic coupling is ; the calculation formula for the refractive index sliding variance is .

[0042] Furthermore, the oil quality feature vectors are constructed as follows: ; Among them, is the analysis period, is the difference between the highest actual refractive index and the lowest actual refractive index within the analysis period, is the difference between the highest temperature and the lowest temperature within the analysis period, f is the oil flow pulsation frequency, j is the number of historical alarms, and the number of historical alarms is the cumulative number of times that the refractive index change rate of transformer oil exceeds the preset safety threshold during continuous monitoring.

[0043] Please refer to Figure 2 , and obtain a number of historical oil quality feature vectors and corresponding warning labels from historical data; among them, the warning labels correspond one-to-one with the warning levels; the warning labels can be set as natural numbers; Integrate a number of oil quality feature vectors and warning labels into a number of groups of training data and test data; use the training data to train the LSTM-Attention model; use the test data to test the trained LSTM-Attention model, and adjust the LSTM-Attention model according to the test results; finally obtain an oil quality evaluation model with the oil quality feature vector as the input and the warning label as the output.

[0044] For example: Divide the extracted historical data into training data and test data according to a ratio of 7:3. Among them, the LSTM-Attention model includes an input layer, an LSTM layer, an Attention layer, and an output layer; During the training process, the input layer is used to receive the oil quality feature vector; The double-layer network structure of LSTM is set as follows: The first layer extracts short-term features, configured as a short time series window processing unit with high forget gate sensitivity. By dynamically adjusting the weight ratio of the input gate to the forget gate (recommended range 1:1.5 - 2.0), it preferentially captures short-term dynamic features within the input sequence with a span of less than N time steps. The short-term features include, but are not limited to: signal instantaneous fluctuation patterns, local gradient change features, high-frequency component correlation characteristics, and there are 128 nodes; The second layer extracts long-term features: forms a cascade structure with the first layer, its memory unit capacity is expanded to 1.5 - 2 times that of the first layer, adopts a decreasing learning rate strategy (the initial value is set to 0.6 - 0.8 times that of the first layer), and extracts long-term dependence features within the input sequence with a span of more than 3N time steps by strengthening the propagation path of the historical state vector. The long-term features include: periodic law features, trend evolution features, cross-period correlation features, and there are 64 nodes; where N is a positive integer; The Attention layer determines the weight coefficients of the time steps, through the formula Calculate the weight coefficients of the time steps ; where \(t\) is the time step, softmax is the activation function, tanh is the hyperbolic tangent activation function, and \(h\) t is the hidden state at time step \(t\), \(W\) is the weight matrix, \(b\) is the bias vector, \(u\) is the attention vector, and \(T\) is the transpose of the attention vector.

[0045] Both \(W\), \(b\), and \(u\) are trainable parameters, that is, variables automatically learned and adjusted by the model through training data, and their values are updated by optimization algorithms such as gradient descent during the backpropagation process; \(W\) and \(u\) jointly determine which hidden state features are more important for the current task, and \(b\) allows the model to apply a baseline shift to the features before the non-linear transformation, avoiding sensitive dependence on zero input.

[0046] Set the training parameters of the model, such as the learning rate of 0.001, the training batch size of 32, and the number of training epochs of 100; during the training process, use the validation data to validate the model, and adjust the parameters of the model according to the loss function value of the validation data to prevent overfitting of the model; after training is completed, input the oil quality feature vector obtained from the real-time collected oil quality parameters into the model, the model outputs the prediction result, and according to the prediction result and the preset warning level rules, generate the corresponding warning level.

[0047] Input the oil quality characteristic parameters of the transformer oil to be detected into the oil quality evaluation model, and output the warning level of the transformer oil quality.

[0048] The LSTM of the present invention is a deep learning model capable of processing time series data and is suitable for processing non-linear and time-varying data. Combining LSTM with the Attention mechanism can dynamically predict oil quality data, identify the trend of oil quality deterioration in advance, and provide early warnings for oil quality anomalies. As a distributed ledger technology, blockchain can provide reliable data storage and auditing functions for oil quality monitoring data. Through the decentralized characteristics and encryption algorithms of blockchain, ensure the security and immutability of oil quality data during the processes of collection, storage, transmission, and auditing, meeting the requirements of the power system for data transparency, reliability, and security.

[0049] Please refer to Figure 3 , the second aspect of the present invention provides a comprehensive transformer oil quality monitoring system, including a data acquisition module, a data processing module, an oil quality evaluation module, and a blockchain trusted management module; The data acquisition module collects the refractive index, temperature, and oil flow pressure of the transformer oil in real time; The data processing module processes and analyzes the collected data; It should be noted that a high-speed and high-precision data acquisition card is selected for the data acquisition module. The sampling frequency of this data acquisition card is set above 10 kHz, which can quickly collect the weak optical signals output by the sensor and convert them into digital signals. Its resolution reaches 16 bits, ensuring the accuracy of data acquisition and being able to accurately capture the minute changes in refractive index, temperature, and oil flow parameters.

[0050] The data acquisition card is connected to the sensor through high-speed data lines. To reduce electromagnetic interference, the data lines use cables with good shielding performance and are well grounded. In terms of data transmission, a combination of wired and wireless methods is adopted. Inside the transformer, due to the complex electromagnetic environment, shielded twisted pair cables are used for data transmission to ensure that the data is not affected by electromagnetic interference during transmission. While outside the transformer, to achieve remote data transmission, a 5G communication module is used to transmit the collected data to the data processing center in real time.

[0051] During the data transmission process, data verification and encryption technologies are adopted to ensure the integrity and accuracy of the data. For example, the cyclic redundancy check algorithm is used to verify the data to ensure that no errors occur during data transmission; the Advanced Encryption Standard encryption algorithm is used to encrypt the data to prevent the data from being stolen or tampered with.

[0052] The oil quality assessment module issues early warnings about the oil quality based on the LSTM-Attention model; The blockchain trusted management module conducts trusted management of the detection data.

[0053] In one embodiment, the blockchain architecture is built as follows: The blockchain trusted management module adopts a main chain-side chain double-chain architecture. The main chain is built based on Hyperledger Fabric, and multiple servers with stable performance and strong computing power are selected as the main chain nodes. Smart contracts are deployed on the main chain to store the hash values of key events, such as early warning triggers, operation and maintenance operations, etc. The smart contracts are written in Go language and follow the smart contract development specifications of Hyperledger Fabric to ensure their security and reliability. The side chain is built using IPFS (InterPlanetary File System), and the distributed storage characteristics of IPFS are utilized to store the original sensor data stream. In the IPFS network, multiple nodes are set up and distributed in different geographical locations to improve the storage security and access efficiency of the data. Through the Merkle Tree structure, the integrity verification of the data stored on the IPFS side chain is carried out. During the data storage process, the data is divided into multiple small pieces, the hash value of each small piece is calculated, and then the root hash value of the Merkle tree is calculated layer by layer and stored on the main chain to achieve the integrity verification of the data.

[0054] Data Upload and Consensus Mechanism: Write a data upload program on the edge device. After collecting new oil quality monitoring data, upload the data to the IPFS side chain. The IPFS side chain returns a unique identifier, and the edge device sends the unique identifier and the corresponding timestamp to the main chain smart contract for recording. During the data upload process, compress the data to reduce the data transmission volume and storage space occupancy.

[0055] To improve the system throughput and ensure the credibility of data, a dynamic consensus mechanism is adopted. According to seasonal changes, the network load will vary. For example, set the number of nodes in summer to 15 and the number of nodes in winter to 10. During the consensus verification process, use the zk-SNARK (Zero-Knowledge Succinct Non-Interactive Argument) algorithm. When a node verifies data, it generates a proof through the zk-SNARK algorithm, and other nodes can verify the integrity and authenticity of the data without knowing the specific data content. This method can not only improve the verification efficiency but also protect data privacy and enhance the credibility of the entire system.

[0056] The embodiment of this application also provides a computer-readable storage medium, including instructions, which when running on a computer, cause the computer to execute any of the above methods.

[0057] Some of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is obtained by software simulation of a large amount of collected data to get a formula closest to the actual situation; the preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0058] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A comprehensive monitoring method for transformer oil quality, characterized in that, Including: Obtain the oil quality parameters of transformer oil in real time and perform preprocessing; among them, the oil quality parameters include refractive index, temperature, and oil flow pressure; Perform temperature compensation on the refractive index after preprocessing based on temperature to obtain the actual refractive index; Construct an oil quality feature vector based on the actual refractive index, the preprocessed temperature, and the oil flow pressure; Extract several oil quality feature vector pairs from historical data to train the LSTM-Attention model, and obtain an oil quality evaluation model through training; the LSTM-Attention model is used to predict the warning level of transformer oil quality based on the oil quality feature vector of transformer oil quality; Input the oil quality feature vector of the transformer oil to be detected into the oil quality evaluation model, and output the warning level of the transformer oil quality.

2. The comprehensive monitoring method for transformer oil quality according to claim 1, characterized in that The oil quality parameters are obtained in real time through an optical fiber refractive index sensor; The optical fiber refractive index sensor is a dual-parameter composite probe including an optical fiber Bragg grating and an optical fiber grating tilt grating; The optical fiber Bragg grating is used to detect the refractive index and temperature, and the optical fiber grating tilt grating is used to monitor the oil flow pressure.

3. The comprehensive monitoring method for transformer oil quality according to claim 2, wherein The preprocessing process of the oil quality parameters includes: Respectively fit each oil quality parameter collected within the analysis period to obtain the curve of each oil quality parameter; place the curves of each oil quality parameter and the corresponding preset standard parameters in the same coordinate system; Obtain the abnormal area from the areas enclosed by the oil quality curve and the preset standard curve, calculate the area of the abnormal area, and determine whether the area of the abnormal area is less than the preset area; if yes, remove each oil quality parameter in the abnormal area; if not, retain each oil quality parameter in the abnormal area; The calculation formula for the area of the abnormal area is as follows: ； Among them, i is the serial number of each oil quality parameter, and i = 0, 1, 2, representing temperature, actual refractive index, and oil flow pressure respectively; is the change curve within the analysis period of the oil quality parameter i, is the preset standard curve of the oil quality parameter i; t1 and t2 are the corresponding times at the intersection of the two curves respectively.

4. A comprehensive monitoring method for transformer oil quality according to claim 3, characterized in that The abnormal area is the area enclosed by the oil quality curve above the preset standard curve, and the preset standard curve is obtained by fitting several historical normal oil quality parameters; among them, the normal oil quality parameters are oil quality parameters within the preset range.

5. A comprehensive monitoring method for transformer oil quality according to claim 1, characterized in that, The process of performing temperature compensation on the refractive index after preprocessing based on temperature to obtain the actual refractive index includes: Obtain the actual refractive index n of transformer oil through a temperature compensation algorithm, and the calculation formula is as follows: ； Among them, The refractive index collected in real time, RIU / o C, is the difference between the current measured temperature and the standard temperature.

6. The comprehensive monitoring method for transformer oil quality according to claim 1, characterized in that The process of constructing an oil quality feature vector based on the actual refractive index, the preprocessed temperature, and the oil flow pressure includes: ； Among them, is the analysis period, is the difference between the highest actual refractive index and the lowest actual refractive index within the analysis period, is the actual refractive index change rate, is the difference between the highest temperature and the lowest temperature within the analysis period, f is the oil flow pulsation frequency, is the refractive index-temperature coupling parameter, is the cumulative amount of refractive index change, is the oil flow-refractive index dynamic coupling, is the refractive index sliding variance, j is the historical alarm times, and the historical alarm times is the cumulative number of times that the refractive index change rate of the transformer oil exceeds the preset safety threshold during continuous monitoring.

7. A comprehensive monitoring method for transformer oil quality according to claim 1, characterized in that, The training process of the oil quality evaluation model is as follows: Obtain several groups of historical oil quality feature vectors and corresponding warning level labels from historical data, divide the historical oil quality feature vectors into a training data set and a validation data set, and train the LSTM-Attention model through the training data set and the validation data set to obtain an oil quality evaluation model.

8. The comprehensive monitoring method for transformer oil quality according to claim 7, wherein The LSTM-Attention model includes an input layer, an LSTM layer, an Attention layer, and an output layer; The input layer is used to receive the oil quality feature vector; The LSTM layer includes a first layer and a second layer; the first layer is used to extract short-term dynamic features with a span less than N time steps in the oil quality feature vector; the second layer is used to extract long-term dynamic features with a span greater than 3N time steps in the oil quality feature vector; where N is a positive integer; The Attention layer is used to determine the weight coefficient of the time step; The output layer performs weighted fusion on the short-term dynamic features and long-term dynamic features extracted by the LSTM layer based on the time-step weight coefficients determined by the Attention layer, generates a comprehensive feature vector, maps it to a preset classification space for transformer oil quality warning levels according to this comprehensive feature vector, and finally outputs the warning level of the transformer oil quality.

9. The comprehensive monitoring method for transformer oil quality according to claim 8, wherein The weight coefficients of the time steps are obtained through the following methods, including: Through the formula the weight coefficient of the time step is calculated ; where t is the time step, softmax is the activation function, tanh is the hyperbolic tangent activation function, h t is the hidden state at time step t, W is the weight matrix, b is the bias vector, u is the attention vector, and T is the transpose of the attention vector.

10. A comprehensive monitoring system for transformer oil quality operates based on the comprehensive monitoring method for transformer oil quality described in any one of claims 1-9, and is characterized in that, It includes a data acquisition module, a data processing module, an oil quality assessment module, and a blockchain trusted management module; The data acquisition module is used to collect the refractive index, temperature, and oil flow pressure of the transformer oil in real time; The data processing module is used to process and analyze the detection data; The oil quality assessment module is used to give early warnings about the oil quality; The blockchain trusted management module is used to manage the trust of the detection data.

11. The integrated monitoring system for transformer oil quality according to claim 10, characterized in that, The data uploading process of the blockchain trusted management module is that the edge device uploads the monitored data to the side chain and generates a unique identifier. The smart contract in the main chain records the unique identifier and the corresponding timestamp, and uses a dynamic consensus mechanism for data verification and archiving.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the transformer oil quality comprehensive monitoring method described in any one of claims 1-9.

Images