Oil-immersed transformer gas on-line monitoring system and method

Abstract

The invention relates to the field of intelligent detection, and particularly discloses an oil-immersed transformer gas on-line monitoring system and method, and the method comprises the steps: collecting a hydrogen sulfide concentration value, a methane concentration value and an ethane concentration value in real time through a gas sensor group disposed in an oil tank of a monitored oil-immersed transformer; the hydrogen sulfide concentration value, the methane concentration value and the ethane concentration value are subjected to time sequence correlation and interactive analysis by using a data processing and analysis algorithm based on a deep learning neural network, so that whether the gas state is abnormal or not is automatically judged based on the real-time state of each gas in the oil tank; and a fault early warning prompt of the monitored oil-immersed transformer is generated when the gas state is abnormal. Thus, the gas state can be monitored in real time, abnormal conditions can be found in time to take measures to prevent potential faults of the oil-immersed transformer, the requirement for human intervention is reduced, and therefore the reliability and safety of the oil-immersed transformer are improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent monitoring, and more specifically, to an online monitoring system and method for oil-immersed transformer gas. Background Art

[0002] The oil-immersed transformer gas monitoring system is a critical safety device used to monitor the composition and concentration of dissolved gases within the oil-immersed transformer. By analyzing the type and content of these dissolved gases, it can detect faults or abnormalities within the transformer at an early stage, providing important information for safe operation and maintenance of the transformer. By monitoring changes in gas concentration within the oil-immersed transformer tank, potential problems can be identified promptly, allowing preventive measures to be taken to avoid catastrophic failures and power outages.

[0003] However, traditional methods for monitoring gas in oil-immersed transformers typically rely on periodic manual sampling and analysis of dissolved gas concentrations using gas chromatography. This is time-consuming and can delay early fault detection. Rapid changes in transformer status may not be captured promptly, impacting normal equipment operation and maintenance costs. Furthermore, traditional methods only monitor a few gas components and fail to fully account for potential fault gases, resulting in low accuracy in detecting gas anomalies.

[0004] Therefore, an online monitoring system for oil-immersed transformer gas is desired. Summary of the Invention

[0005] In order to solve the above-mentioned technical problems, the present application is proposed. The embodiments of the present application provide an online gas monitoring system and method for oil-immersed transformers. The system collects hydrogen sulfide concentration values, methane concentration values, and ethane concentration values ​​in real time by a gas sensor group deployed inside the oil tank of the monitored oil-immersed transformer, and uses a data processing and analysis algorithm based on a deep learning neural network to perform time-series correlation and interactive analysis on the hydrogen sulfide concentration values, methane concentration values, and ethane concentration values. Based on the real-time status of each gas inside the oil tank, the system automatically determines whether the gas state is abnormal, and generates a fault warning prompt for the monitored oil-immersed transformer when the gas state is abnormal. In this way, real-time monitoring of the gas state can be achieved, and abnormal conditions can be detected in time to take measures to prevent potential faults of the oil-immersed transformer, reducing the need for human intervention, thereby improving the reliability and safety of the oil-immersed transformer.

[0006] According to one aspect of the present application, there is provided an oil-immersed transformer gas online monitoring system, comprising:

[0007] The monitored oil-immersed transformer parameter acquisition module is used to obtain the time series of hydrogen sulfide concentration values, methane concentration values, and ethane concentration values ​​collected by the gas sensor group deployed inside the oil tank of the monitored oil-immersed transformer;

[0008] a time sample dimension arrangement module, configured to arrange the time series of the hydrogen sulfide concentration values, the methane concentration values, and the ethane concentration values ​​into a gas concentration time series input matrix according to the time dimension and the gas sample dimension;

[0009] a nonlinear compensation module, configured to perform nonlinear compensation on the gas concentration time series input matrix to obtain a compensated gas concentration time series input matrix;

[0010] a gas concentration time series correlation feature extraction module, configured to extract gas concentration time series correlation pattern features from the compensated gas concentration time series input matrix to obtain a gas concentration time series correlation feature graph;

[0011] a time series feature interaction enhancement module, configured to perform gas concentration time series feature interaction enhancement on the gas concentration time series correlation feature graph to obtain a gas concentration time series enhancement correlation feature;

[0012] The gas state judgment and warning module is used to determine whether there is any abnormality in the gas state based on the gas concentration time series enhanced correlation feature, and determine whether to generate a fault warning prompt for the monitored oil-immersed transformer.

[0013] According to another aspect of the present application, a method for online monitoring of gas in an oil-immersed transformer is provided, comprising:

[0014] Obtaining the time series of hydrogen sulfide concentration values, methane concentration values, and ethane concentration values ​​collected by the gas sensor group deployed inside the oil tank of the monitored oil-immersed transformer;

[0015] Arranging the time series of the hydrogen sulfide concentration values, the methane concentration values, and the ethane concentration values ​​into a gas concentration time series input matrix according to the time dimension and the gas sample dimension;

[0016] performing nonlinear compensation on the gas concentration time series input matrix to obtain a compensated gas concentration time series input matrix;

[0017] Performing gas concentration time series correlation pattern feature extraction on the compensated gas concentration time series input matrix to obtain a gas concentration time series correlation feature graph;

[0018] Performing gas concentration time series feature interactive enhancement on the gas concentration time series correlation feature graph to obtain a gas concentration time series enhancement correlation feature;

[0019] Based on the gas concentration time series intensified correlation feature, it is determined whether there is any abnormality in the gas state, and whether a fault warning prompt for the monitored oil-immersed transformer is generated.

[0020] Compared to the prior art, the present application provides an online gas monitoring system and method for oil-immersed transformers. This system uses a gas sensor group deployed inside the oil tank of the monitored oil-immersed transformer to collect hydrogen sulfide, methane, and ethane concentrations in real time. It then uses a data processing and analysis algorithm based on a deep learning neural network to perform time-series correlation and interactive analysis on these hydrogen sulfide, methane, and ethane concentrations. This system automatically determines whether the gas state is abnormal based on the real-time status of each gas inside the oil tank, and generates a fault warning for the monitored oil-immersed transformer when the gas state is abnormal. This system enables real-time monitoring of the gas state, allowing for timely detection of abnormalities and the implementation of measures to prevent potential faults in the oil-immersed transformer, reducing the need for human intervention and thereby improving the reliability and safety of the oil-immersed transformer. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0022] Figure 1 4 is a block diagram of an oil-immersed transformer gas online monitoring system according to an embodiment of the present application.

[0023] Figure 2 Schematic diagram of the architecture of an online gas monitoring system for an oil-immersed transformer according to an embodiment of the present application.

[0024] Figure 3 This is a block diagram of a gas status judgment and warning module in an oil-immersed transformer gas online monitoring system according to an embodiment of the present application.

[0025] Figure 4 4 is a block diagram of a training module in an oil-immersed transformer gas online monitoring system according to an embodiment of the present application.

[0026] Figure 5 Flowchart of an online gas monitoring method for an oil-immersed transformer according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] The following describes embodiments of the present disclosure in more detail with reference to the accompanying drawings. While the drawings illustrate certain embodiments of the present disclosure, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of protection of the present disclosure.

[0028] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in a different order and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0029] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, i.e., "including but not limited to." The term "based on" should be understood as "based at least in part on." The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment." The terms "first," "second," etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0030] It should be noted that the modifications of "one" and "multiple" mentioned in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, they should be understood as "one or more".

[0031] The oil-immersed transformer gas monitoring system is a critical safety device used to monitor the composition and concentration of dissolved gases within oil-immersed transformers. Specifically, the system operates on the principle that when a fault or abnormality occurs within the transformer, specific dissolved gases are generated. By monitoring the type and content of these dissolved gases, the type and severity of the fault or abnormality can be determined. This allows for the timely identification of potential problems and the implementation of preventive measures, providing crucial information for the safe operation and maintenance of the transformer.

[0032] However, traditional methods for monitoring gas in oil-immersed transformers typically rely on periodic manual sampling and analysis of dissolved gas concentrations using gas chromatography. This is time-consuming and can delay early fault detection. Rapid changes in transformer status may not be captured promptly, impacting normal equipment operation and maintenance costs. Furthermore, traditional methods only monitor a few gas components and fail to fully account for potential fault gases, resulting in low accuracy in detecting gas anomalies.

[0033] Therefore, in response to the above technical problems, the technical concept of this application is to collect hydrogen sulfide concentration values, methane concentration values, and ethane concentration values ​​in real time by using a gas sensor group deployed inside the oil tank of the monitored oil-immersed transformer, and use a data processing and analysis algorithm based on a deep learning neural network to perform time series correlation and interactive analysis on the hydrogen sulfide concentration values, methane concentration values, and ethane concentration values. In this way, based on the real-time status of each gas inside the oil tank, it is automatically determined whether there is an abnormality in the gas state, and when there is an abnormality in the gas state, a fault warning prompt is generated for the monitored oil-immersed transformer. In this way, real-time monitoring of the gas state can be achieved, and abnormal conditions can be discovered in time to take measures to prevent potential faults of the oil-immersed transformer, reducing the need for human intervention, thereby improving the reliability and safety of the oil-immersed transformer.

[0034] Figure 1 4 is a block diagram of an oil-immersed transformer gas online monitoring system according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of an oil-immersed transformer gas online monitoring system according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to an embodiment of the present application, the oil-immersed transformer gas online monitoring system 100 includes: a monitored oil-immersed transformer parameter acquisition module 110, which is used to obtain the time series of hydrogen sulfide concentration values, methane concentration values ​​and ethane concentration values ​​collected by the gas sensor group deployed inside the oil tank of the monitored oil-immersed transformer; a time sample dimension arrangement module 120, which is used to arrange the time series of the hydrogen sulfide concentration values, methane concentration values ​​and ethane concentration values ​​according to the time dimension and the gas sample dimension into a gas concentration time series input matrix; a nonlinear compensation module 130, which is used to perform nonlinear compensation on the gas concentration time series input matrix to obtain a compensation matrix. A post-compensation gas concentration time series input matrix; a gas concentration time series correlation feature extraction module 140, used to perform gas concentration time series correlation pattern feature extraction on the post-compensation gas concentration time series input matrix to obtain a gas concentration time series correlation feature diagram; a time series feature interaction enhancement module 150, used to perform gas concentration time series feature interaction enhancement on the gas concentration time series correlation feature diagram to obtain a gas concentration time series enhancement correlation feature; and a gas state judgment and early warning module 160, used to determine whether there is an abnormality in the gas state based on the gas concentration time series enhancement correlation feature, and determine whether to generate a fault early warning prompt for the monitored oil-immersed transformer.

[0035] In an embodiment of the present application, the monitored oil-immersed transformer parameter acquisition module 110 is used to obtain a time series of hydrogen sulfide concentration values, methane concentration values, and ethane concentration values ​​collected by a gas sensor group deployed inside the oil tank of the monitored oil-immersed transformer. It should be understood that, considering that hydrogen sulfide, methane, and ethane are common gases generated inside oil-immersed transformers, they have different chemical properties and degrees of harm. Specifically, hydrogen sulfide is highly corrosive and may cause corrosion of metal components inside the oil-immersed transformer, thereby affecting the performance and life of the equipment. In addition, high concentrations of hydrogen sulfide are toxic gases that pose a serious threat to human health and may also cause the risk of fire and explosion. Methane and ethane are flammable gases and pose a risk of explosion and fire. In particular, at high concentrations, they may affect the insulation performance inside the transformer and the operational stability of the transformer, thereby causing performance degradation or failure of the oil-immersed transformer. Based on this, in the technical solution of the present application, the time series of hydrogen sulfide concentration values, methane concentration values ​​and ethane concentration values ​​collected by the gas sensor group deployed inside the oil tank of the monitored oil-immersed transformer are obtained, and the hydrogen sulfide concentration values, the methane concentration values ​​and the ethane concentration values ​​are subjected to time series collaborative analysis and mutual interaction. This allows a comprehensive analysis of whether the gas status is normal, so as to promptly discover potential abnormal situations and take necessary measures to ensure the safe operation of the oil-immersed transformer.

[0036] In an embodiment of the present application, the time sample dimension arrangement module 120 is configured to arrange the time series of the hydrogen sulfide concentration values, methane concentration values, and ethane concentration values ​​according to the time dimension and the gas sample dimension into a gas concentration time series input matrix. Accordingly, considering that the hydrogen sulfide concentration, the methane concentration, and the ethane concentration all exhibit dynamic trends and fluctuations over time, that is, the hydrogen sulfide concentration, the methane concentration, and the ethane concentration have certain time series characteristic information along the time dimension. Furthermore, the hydrogen sulfide concentration, the methane concentration, and the ethane concentration are interrelated along the time dimension. For example, increased hydrogen sulfide production may affect the degradation of organic matter in transformer oil, resulting in reduced methane and ethane production. This interrelated relationship is crucial for detecting abnormal gas conditions. Therefore, in the technical solution of the present application, the time series of the hydrogen sulfide concentration values, the methane concentration values, and the ethane concentration values ​​are arranged according to the time dimension and the gas sample dimension into a gas concentration time series input matrix. Specifically, arranging the time series of hydrogen sulfide, methane, and ethane concentrations along the time dimension preserves their temporal variations and fluctuation trends. Arranging them along the gas sample dimension integrates the temporal correlations among the hydrogen sulfide, methane, and ethane concentrations. This allows for a better understanding of the gas status in oil-immersed transformers, facilitating more accurate gas anomaly detection.

[0037] In an embodiment of the present application, the nonlinear compensation module 130 is used to perform nonlinear compensation on the gas concentration time-series input matrix to obtain a compensated gas concentration time-series input matrix. It should be understood that the response characteristics of a gas sensor or monitoring device may be nonlinear, meaning that the relationship between changes in gas concentration and sensor output is not a simple linear relationship. For example, some sensors may respond weakly at low concentrations and more strongly at high concentrations. Furthermore, gas sensors may experience drift or errors over long periods of use and may be affected by environmental factors and other gas components, resulting in cross-interference and deviations between the output signal and the actual gas concentration. Therefore, in order to more accurately reconstruct the temporal variation trend of actual gas concentration and better monitor and analyze gas concentration changes, the technical solution of the present application performs nonlinear compensation on the gas concentration time-series input matrix to obtain a compensated gas concentration time-series input matrix. It is worth noting that nonlinear compensation is a correction method used to address the nonlinear response characteristics of a sensor or system. Specifically, when a nonlinear relationship exists between the sensor output signal and the input signal, nonlinear compensation can correct this nonlinear response through a series of mathematical operations or algorithms, so that the output more accurately reflects the actual changes in the input. It should be understood that performing nonlinear compensation on the gas concentration timing input matrix can help correct the nonlinear part in the gas sensor output so as to more accurately estimate the actual gas concentration timing characteristics, thereby improving the accuracy and reliability of gas concentration monitoring and analysis.

[0038] Specifically, in an embodiment of the present application, the nonlinear compensation module is configured to perform nonlinear compensation on the gas concentration time series input matrix using the following compensation formula to obtain the compensated gas concentration time series input matrix; wherein the correction formula is:

[0039]

[0040] Among them, f i,j is the eigenvalue of the (i, j)th position of the gas concentration time series input matrix, F i,j is the eigenvalue of the (i, j)th position of the compensated gas concentration time series input matrix, and A, B, C and D are corresponding adjustment parameters.

[0041] In an embodiment of the present application, the gas concentration time series correlation feature extraction module 140 is used to extract the gas concentration time series correlation pattern features of the compensated gas concentration time series input matrix to obtain a gas concentration time series correlation feature diagram. Specifically, in an embodiment of the present application, the gas concentration time series correlation feature extraction module is used to: pass the compensated gas concentration time series input matrix through a gas concentration time series correlation pattern feature extractor based on a convolutional neural network model to obtain the gas concentration time series correlation feature diagram. Accordingly, considering that the compensated gas concentration time series input matrix has certain time series correlation patterns and feature information, and the convolutional neural network has excellent feature extraction capabilities when processing time series data and extracting implicit correlation relationships between time series feature data. Therefore, in the technical solution of the present application, the compensated gas concentration time series input matrix is ​​passed through a gas concentration time series correlation pattern feature extractor based on a convolutional neural network model to extract the implicit feature information about the gas concentration time series of the compensated gas concentration time series input matrix, thereby obtaining a gas concentration time series correlation feature diagram with more feature expression significance.

[0042] In an embodiment of the present application, the time series feature interaction enhancement module 150 is used to perform gas concentration time series feature interaction enhancement on the gas concentration time series correlation feature graph to obtain a gas concentration time series enhanced correlation feature. Specifically, in an embodiment of the present application, the time series feature interaction enhancement module is used to: pass the gas concentration time series correlation feature graph through a gas concentration time series feature interaction enhancer based on a triple interactive attention module to obtain a gas concentration time series enhanced correlation feature graph as the gas concentration time series enhanced correlation feature. It should be understood that, considering that the gas concentration time series correlation feature graph has correlation feature information about gas concentrations and different time series in both the channel dimension and the spatial dimension, and there is mutual interaction and influence between these correlation feature information. Therefore, in order to strengthen the feature interaction between these correlation features, thereby improving the model's understanding of gas concentration changes and its ability to predict abnormal states, in the technical solution of the present application, the gas concentration time series correlation feature graph is passed through a gas concentration time series feature interaction enhancer based on a triple interactive attention module to obtain a gas concentration time series enhanced correlation feature graph. It is worth mentioning that the triple interactive attention module captures the feature interaction across dimensions through three branches, and these three branches focus on different dimensions of the input features respectively. Specifically, the first branch establishes an interaction between the channel and the height, the second branch establishes an interaction between the channel and the width, and the third branch establishes an interaction between the height and the width. Finally, the features output by these three branches are added and averaged to obtain the features after interaction. That is, through the triple interactive attention module, the gas concentration time series correlation feature map can realize the interaction of gas concentration time series correlation features between different levels, different positions and different channels, so as to better capture the key gas concentration time series features in the gas concentration time series correlation feature map, and realize the fine control and interaction enhancement of the features, so as to better model the complex time series correlation in the gas concentration time series data, so as to improve the gas anomaly detection capability.

[0043] Specifically, in an embodiment of the present application, the temporal feature interaction enhancement module is used to: use the gas concentration temporal feature interaction enhancer based on the triple interactive attention module to process the gas concentration temporal correlation feature graph using the following interaction formula to obtain the gas concentration temporal enhancement correlation feature graph; wherein the interaction formula is:

[0044] α1=δ(W 2 *σ(W 1 *X))

[0045]

[0046] α2=δ(W 2 *σ(W 1 *X1))

[0047]

[0048] α3=δ(W 2 *σ(W 1 *X2))

[0049]

[0050]

[0051] Wherein, X represents the gas concentration time series correlation characteristic diagram, W 1 represents a 1×1 convolution, W 2 represents a 7×7 convolution, * represents a convolution operation, σ(·) represents a ReLU function, δ(·) represents a Sigmoid function, α1 represents the first weight matrix, represents the Hadamard product, is the first feature map of interest, X1 represents the first rotated feature map obtained by rotating the gas concentration time series correlation feature map along the width dimension, α2 represents the second weight matrix, Trans(·) represents the transposition transformation, is the second feature map of interest, X2 represents the second rotation feature map obtained by rotating the gas concentration time series correlation feature map along the height dimension, α3 represents the third weight matrix, is the second attention feature map, f(·) represents the summing and averaging operation, A characteristic diagram showing the temporal enhancement correlation of the gas concentration.

[0052] In the embodiment of the present application, the gas state judgment and warning module 160 is used to determine whether there is an abnormality in the gas state based on the gas concentration time series enhanced correlation feature, and determine whether to generate a fault warning prompt for the monitored oil-immersed transformer. Figure 3 FIG. 1 is a block diagram of a gas state judgment and warning module in an oil-immersed transformer gas online monitoring system according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 3As shown, the gas state judgment and warning module 160 includes: a gas state judgment unit 161, configured to apply the gas concentration time-series enhanced correlation feature map to a classifier-based gas state detector to obtain a detection result, wherein the detection result indicates whether the gas state is abnormal; and a fault warning prompt unit 162, configured to generate a fault warning prompt for the monitored oil-immersed transformer in response to the detection result indicating an abnormal gas state. Specifically, the gas concentration time-series enhanced correlation feature obtained through the interaction of triple interactive attention features is used for classification processing. Based on the real-time status of each gas within the oil tank, the module automatically determines whether the gas state is abnormal. If an abnormal gas state is present, a fault warning prompt for the monitored oil-immersed transformer is generated. This enables real-time monitoring of the gas state, allowing for timely detection of abnormalities and the implementation of measures to prevent potential faults in the oil-immersed transformer, reducing the need for human intervention and improving the reliability and safety of the oil-immersed transformer.

[0053] It is worth mentioning that those skilled in the art should be aware that before applying a deep neural network model for inference, the deep neural network model must first be trained so that the deep neural network can implement specific functional capabilities.

[0054] Specifically, in the technical solution of the present application, the oil-immersed transformer gas online monitoring system also includes a training module for training the gas concentration time series correlation pattern feature extractor based on the convolutional neural network model, the gas concentration time series feature interaction enhancer based on the triple interactive attention module, and the classifier-based gas state detector.

[0055] Figure 4 FIG. 1 is a block diagram of a training module in an oil-immersed transformer gas online monitoring system according to an embodiment of the present application. Specifically, in the embodiment of the present application, Figure 4As shown, the training module 200 includes: a training data acquisition unit 210 for acquiring training data, wherein the training data includes a time series of training hydrogen sulfide concentration values, training methane concentration values ​​and training ethane concentration values ​​collected by a gas sensor group deployed inside the oil tank of the monitored oil-immersed transformer, and a real value of whether the gas state is abnormal; a training data dimension arrangement unit 220 for arranging the time series of training hydrogen sulfide concentration values, training methane concentration values ​​and training ethane concentration values ​​into a training gas concentration time series input matrix according to the time dimension and the gas sample dimension; a training data compensation unit 230 for performing nonlinear compensation on the training gas concentration time series input matrix to obtain a compensated training gas concentration time series input matrix; a training gas concentration time series association unit 240 for passing the compensated training gas concentration time series input matrix through the gas concentration time series association pattern feature extractor based on the convolutional neural network model to obtain a training gas concentration time series association feature map; A feature interaction enhancement unit 250 is used to pass the training gas concentration time series correlation feature graph through the gas concentration time series feature interaction enhancer based on the triple interactive attention module to obtain a training gas concentration time series enhancement correlation feature graph; a training feature optimization unit 260 is used to optimize the gas concentration time series enhancement correlation feature graph based on the feature matrix distribution of the training gas concentration time series enhancement correlation feature graph along the channel dimension to obtain an optimized training gas concentration time series enhancement correlation feature graph; a classification loss calculation unit 270 is used to pass the optimized training gas concentration time series enhancement correlation feature graph through the classifier-based gas state detector to obtain a classification loss function value; and a training unit 280 is used to train the gas concentration time series correlation pattern feature extractor based on the convolutional neural network model, the gas concentration time series feature interaction enhancer based on the triple interactive attention module, and the classifier-based gas state detector based on the classification loss function value and through back propagation of gradient descent.

[0056] Specifically, the training feature optimization unit 260 is configured to optimize the gas concentration time series enhanced correlation feature map based on the feature matrix distribution of the training gas concentration time series enhanced correlation feature map along the channel dimension to obtain an optimized training gas concentration time series enhanced correlation feature map. It should be understood that in the above technical solution, each feature matrix of the training gas concentration time series correlation feature map expresses the local correlation features of the time series-sample cross-dimension of the training hydrogen sulfide concentration value, the training methane concentration value, and the training ethane concentration value, and each feature matrix follows the channel distribution of the convolutional neural network model. Therefore, each feature matrix of the training gas concentration time series correlation feature map has a correlation feature distribution representation difference based on the difference in channel dimension distribution.

[0057] Therefore, after the training gas concentration time series correlation feature map passes through the triple interactive attention module, since the spatial distribution difference of the feature matrix is ​​based on the channel distribution representation of the convolutional neural network model, under the spatial-channel distribution interactive attention mechanism of the triple interactive attention, the correlation feature distribution difference of each feature matrix of the training gas concentration time series reinforcement correlation feature map will be further enhanced based on the channel distribution difference, thereby affecting the classification regression effect of each feature matrix of the training gas concentration time series reinforcement correlation feature map based on the overall numerical distribution of the eigenvalues, and affecting the accuracy of the classification result obtained by the gas state detector based on the classifier of the gas concentration time series reinforcement correlation feature map. Based on this, in the technical solution of the present application, the gas concentration time series reinforcement correlation feature map is optimized based on the feature matrix distribution of the training gas concentration time series reinforcement correlation feature map along the channel dimension to obtain an optimized training gas concentration time series reinforcement correlation feature map.

[0058] Specifically, in an embodiment of the present application, the training feature optimization unit includes: a probability coefficient calculation subunit, which is used to perform scenario-based quasi-probability logical reasoning on each feature matrix of the training gas concentration timing reinforcement association feature map along the channel dimension to obtain a probability vector; and a weighted optimization subunit, which is used to perform weighted optimization on each feature matrix of the training gas concentration timing reinforcement association feature map along the channel dimension using the probability coefficient value of each position in the probability vector as a weighting coefficient to obtain the optimized training gas concentration timing reinforcement association feature map.

[0059] More specifically, in an embodiment of the present application, the probability coefficient calculation subunit is configured to perform scenario-based probabilistic logical reasoning on each feature matrix of the training gas concentration time series reinforcement correlation feature graph along the channel dimension using the following probability calculation formula to obtain the probability vector; wherein the probability calculation formula is:

[0060]

[0061] Among them, m i,j is the eigenvalue of the (i, j)th position of each feature matrix of the training gas concentration time series reinforcement correlation feature map, □(m i,j ) represents the probabilistic function of the eigenvalue, that is, the eigenvalue m i,j A probabilistic function mapped to the interval [0,1], S is the scale of each feature matrix of the training gas concentration time series reinforcement association feature map, that is, the width multiplied by the height, p is the class probability value obtained by the classifier of the training gas concentration time series reinforcement association feature map, α is the weight hyperparameter, and ω is the probability coefficient value of each position in the probability vector.

[0062] Specifically, for each feature matrix of the training gas concentration time-series reinforcement correlation feature map, the scenario-saturated class probabilistic reasoning logical association is adopted through probability distribution foreground constraints and relative probability mapping response assumptions, thereby endowing the feature set of each feature matrix of the training gas concentration time-series reinforcement correlation feature map with scenario-concept ontological cognition. In other words, the overall distribution is internally aligned with the scenario-based class probabilistic logical reasoning during the classification process, thereby improving the class-cognition understanding of the feature matrix scenario distribution of the training gas concentration time-series reinforcement correlation feature map. Thus, by weighted optimization of the corresponding feature matrix of the gas concentration time-series reinforcement correlation feature map using the probability coefficient values ​​of each position in the probability vector as weighting coefficients, the accuracy of the classification results obtained by the classifier-based gas state detector for the optimized gas concentration time-series reinforcement correlation feature map can be improved. This enables real-time monitoring of gas state, timely detection of abnormal conditions, and the implementation of measures to prevent potential failures of oil-immersed transformers, reducing the need for human intervention and thus improving the reliability and safety of oil-immersed transformers.

[0063] In summary, the oil-immersed transformer gas online monitoring system 100 based on the embodiment of the present application is explained. It uses a gas sensor group deployed inside the oil tank of the monitored oil-immersed transformer to collect hydrogen sulfide concentration values, methane concentration values, and ethane concentration values ​​in real time, and uses a data processing and analysis algorithm based on a deep learning neural network to perform time-series correlation and interactive analysis on the hydrogen sulfide concentration values, methane concentration values, and ethane concentration values. Based on the real-time status of each gas inside the oil tank, it automatically determines whether the gas state is abnormal, and generates a fault warning prompt for the monitored oil-immersed transformer when the gas state is abnormal. In this way, it is possible to achieve real-time monitoring of the gas state, and timely detect abnormal conditions to take measures to prevent potential faults of the oil-immersed transformer, reduce the need for human intervention, and thus improve the reliability and safety of the oil-immersed transformer.

[0064] As described above, the oil-immersed transformer gas online monitoring system 100 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server equipped with an oil-immersed transformer gas online monitoring algorithm. In one possible implementation, the oil-immersed transformer gas online monitoring system 100 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or hardware module. For example, the oil-immersed transformer gas online monitoring system 100 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the oil-immersed transformer gas online monitoring system 100 can also be one of the many hardware modules of the wireless terminal.

[0065] Alternatively, in another example, the oil-immersed transformer gas online monitoring system 100 and the wireless terminal may also be separate devices, and the oil-immersed transformer gas online monitoring system 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0066] Figure 5 FIG. 1 is a flow chart of an oil-immersed transformer gas online monitoring method according to an embodiment of the present application. Figure 5 As shown, according to the embodiment of the present application, the online monitoring method for oil-immersed transformer gas includes: S110, obtaining the time series of hydrogen sulfide concentration values, methane concentration values ​​and ethane concentration values ​​collected by the gas sensor group deployed inside the oil tank of the monitored oil-immersed transformer; S120, arranging the time series of the hydrogen sulfide concentration values, methane concentration values ​​and ethane concentration values ​​into a gas concentration time series input matrix according to the time dimension and the gas sample dimension; S130, performing nonlinear compensation on the gas concentration time series input matrix to obtain a compensated gas concentration time series input matrix; S140, performing gas concentration time series correlation pattern feature extraction on the compensated gas concentration time series input matrix to obtain a gas concentration time series correlation feature diagram; S150, performing gas concentration time series feature interactive enhancement on the gas concentration time series correlation feature diagram to obtain a gas concentration time series enhancement correlation feature; and, S160, determining whether there is an abnormality in the gas state based on the gas concentration time series enhancement correlation feature, and determining whether to generate a fault warning prompt for the monitored oil-immersed transformer.

[0067] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned oil-immersed transformer gas online monitoring method have been described in detail above. Figures 1 to 4 The description of the oil-immersed transformer gas online monitoring system has been introduced in detail, and therefore, its repeated description will be omitted.

[0068] While various implementations of the present disclosure have been described above, the foregoing description is intended to be illustrative and not exhaustive. The disclosure is not limited to the disclosed implementations, and numerous modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described implementations. The terminology used herein is selected to best explain the principles of the implementations, their practical applications, or improvements to existing technologies, or to enable others skilled in the art to understand the various implementations disclosed herein.

Claims

1. An oil-immersed transformer gas online monitoring system, characterized in that: include: The monitored oil-immersed transformer parameter acquisition module is used to obtain the time series of hydrogen sulfide concentration values, methane concentration values, and ethane concentration values ​​collected by the gas sensor group deployed inside the oil tank of the monitored oil-immersed transformer; a time sample dimension arrangement module, configured to arrange the time series of the hydrogen sulfide concentration values, the methane concentration values, and the ethane concentration values ​​into a gas concentration time series input matrix according to the time dimension and the gas sample dimension; a nonlinear compensation module, configured to perform nonlinear compensation on the gas concentration time series input matrix to obtain a compensated gas concentration time series input matrix; a gas concentration time series correlation feature extraction module, configured to extract gas concentration time series correlation pattern features from the compensated gas concentration time series input matrix to obtain a gas concentration time series correlation feature graph; a time series feature interaction enhancement module, configured to perform gas concentration time series feature interaction enhancement on the gas concentration time series correlation feature graph to obtain a gas concentration time series enhancement correlation feature; The gas state judgment and warning module is used to determine whether there is any abnormality in the gas state based on the gas concentration time series enhanced correlation feature, and determine whether to generate a fault warning prompt for the monitored oil-immersed transformer.

2. The oil-immersed transformer gas online monitoring system according to claim 1 is characterized in that: The nonlinear compensation module is used to: perform nonlinear compensation on the gas concentration time series input matrix using the following compensation formula to obtain the compensated gas concentration time series input matrix; Wherein, the correction formula is: Among them, f i,j is the eigenvalue of the (i, j)th position of the gas concentration time series input matrix, F i,j is the eigenvalue of the (i, j)th position of the compensated gas concentration time series input matrix, and A, B, C and D are corresponding adjustment parameters.

3. The oil-immersed transformer gas online monitoring system according to claim 2 is characterized in that: The gas concentration time series correlation feature extraction module is used to: pass the compensated gas concentration time series input matrix through a gas concentration time series correlation pattern feature extractor based on a convolutional neural network model to obtain the gas concentration time series correlation feature graph.

4. The oil-immersed transformer gas online monitoring system according to claim 3 is characterized in that: The temporal feature interaction enhancement module is used to: pass the gas concentration temporal correlation feature graph through a gas concentration temporal feature interaction enhancer based on a triple interactive attention module to obtain a gas concentration temporal enhancement correlation feature graph as the gas concentration temporal enhancement correlation feature.

5. The oil-immersed transformer gas online monitoring system according to claim 4 is characterized in that: The temporal feature interaction enhancement module is used to: use the gas concentration temporal feature interaction enhancer based on the triple interactive attention module to process the gas concentration temporal correlation feature graph using the following interaction formula to obtain the gas concentration temporal enhancement correlation feature graph; The interaction formula is: α1=δ(W 2 *σ(W 1 *X)) α2=δ(W 2 *σ(W 1 *X1)) α3=δ(W 2 *σ(W 1 *X2)) Wherein, X represents the gas concentration time series correlation characteristic diagram, W 1 represents a 1×1 convolution, W 2 represents a 7×7 convolution, * represents a convolution operation, σ(·) represents a ReLU function, δ(·) represents a Sigmoid function, α1 represents the first weight matrix, represents the Hadamard product, is the first feature map of interest, X1 represents the first rotated feature map obtained by rotating the gas concentration time series correlation feature map along the width dimension, α2 represents the second weight matrix, Trans(·) represents the transposition transformation, is the second feature map of interest, X2 represents the second rotation feature map obtained by rotating the gas concentration time series correlation feature map along the height dimension, α3 represents the third weight matrix, is the second attention feature map, f(·) represents the summing and averaging operation, A characteristic diagram showing the temporal enhancement correlation of the gas concentration.

6. The oil-immersed transformer gas online monitoring system according to claim 5, characterized in that: The gas state judgment and warning module includes: a gas state judgment unit, configured to pass the gas concentration time series enhanced correlation feature diagram through a classifier-based gas state detector to obtain a detection result, wherein the detection result is used to indicate whether the gas state is abnormal; The fault warning prompt unit is used to generate a fault warning prompt of the monitored oil-immersed transformer in response to the detection result that the gas state is abnormal.

7. The oil-immersed transformer gas online monitoring system according to claim 6, characterized in that: It also includes a training module for training the gas concentration time series association pattern feature extractor based on the convolutional neural network model, the gas concentration time series feature interaction enhancer based on the triple interactive attention module, and the gas state detector based on the classifier.

8. The oil-immersed transformer gas online monitoring system according to claim 7, characterized in that: The training module includes: a training data acquisition unit, configured to acquire training data, the training data including a time series of training hydrogen sulfide concentration values, training methane concentration values, and training ethane concentration values ​​collected by a gas sensor group deployed inside an oil tank of a monitored oil-immersed transformer, and a true value indicating whether the gas state is abnormal; A training data dimension arrangement unit is used to arrange the time series of the training hydrogen sulfide concentration values, the training methane concentration values, and the training ethane concentration values ​​into a training gas concentration time series input matrix according to the time dimension and the gas sample dimension; a training data compensation unit, configured to perform nonlinear compensation on the training gas concentration time series input matrix to obtain a compensated training gas concentration time series input matrix; A training gas concentration time series association unit, configured to pass the compensated training gas concentration time series input matrix through the gas concentration time series association pattern feature extractor based on the convolutional neural network model to obtain a training gas concentration time series association feature graph; a training feature interaction enhancement unit, configured to pass the training gas concentration time series correlation feature map through the gas concentration time series feature interaction enhancer based on the triple interactive attention module to obtain a training gas concentration time series reinforcement correlation feature map; a training feature optimization unit, configured to optimize the gas concentration time series enhancement correlation feature map based on the feature matrix distribution of the training gas concentration time series enhancement correlation feature map along the channel dimension to obtain an optimized training gas concentration time series enhancement correlation feature map; a classification loss calculation unit, configured to pass the optimized training gas concentration time series reinforcement correlation feature map through the classifier-based gas state detector to obtain a classification loss function value; A training unit is used to train the gas concentration time series association pattern feature extractor based on the convolutional neural network model, the gas concentration time series feature interaction enhancer based on the triple interactive attention module, and the classifier-based gas state detector based on the classification loss function value and through back propagation of gradient descent.

9. The oil-immersed transformer gas online monitoring system according to claim 8, characterized in that: The training feature optimization unit includes: a probability coefficient calculation subunit, configured to perform scenario-based probabilistic logic reasoning on each feature matrix of the training gas concentration time series reinforcement correlation feature graph along the channel dimension to obtain a probability vector; The weighted optimization subunit is used to perform weighted optimization on each feature matrix of the training gas concentration time series reinforcement association feature map along the channel dimension using the probability coefficient value of each position in the probability vector as a weighting coefficient to obtain the optimized training gas concentration time series reinforcement association feature map.

10. A method for online monitoring of gas in an oil-immersed transformer, characterized in that: include: Obtaining the time series of hydrogen sulfide concentration values, methane concentration values, and ethane concentration values ​​collected by the gas sensor group deployed inside the oil tank of the monitored oil-immersed transformer; Arranging the time series of the hydrogen sulfide concentration values, the methane concentration values, and the ethane concentration values ​​into a gas concentration time series input matrix according to the time dimension and the gas sample dimension; performing nonlinear compensation on the gas concentration time series input matrix to obtain a compensated gas concentration time series input matrix; Performing gas concentration time series correlation pattern feature extraction on the compensated gas concentration time series input matrix to obtain a gas concentration time series correlation feature graph; Performing gas concentration time series feature interactive enhancement on the gas concentration time series correlation feature graph to obtain a gas concentration time series enhancement correlation feature; Based on the gas concentration time series intensified correlation feature, it is determined whether there is any abnormality in the gas state, and whether a fault warning prompt for the monitored oil-immersed transformer is generated.

Images