Converter transformer active early warning system and method based on oil chromatography online monitoring

Abstract

The invention discloses a converter transformer active early warning system and method based on oil chromatography on-line monitoring, and the system comprises a data collection module which is used for obtaining the time series historical data of insulating oil dissolved gas; the data preprocessing module is used for obtaining the standardized time sequence historical data of the insulating oil dissolved gas; the feature extraction module is used for obtaining a feature value of the time sequence historical data of the insulating oil dissolved gas; the model training module is used for obtaining a fault diagnosis model and a fault diagnosis feature weight of the time sequence historical data of the insulating oil dissolved gas; and the monitoring and early warning module is used for judging whether the data of the insulating oil dissolved gas collected at the current time point has a fault risk or not. According to the method, the Angle-Transform model is used for fault diagnosis of the converter transformer, features are extracted through the time convolution network, fluctuation of the accuracy and stability of abnormal recognition caused by a traditional monitoring method is solved, and safety and reliability of the converter transformer are ensured.

Description

Technical Field

[0001] The invention relates to the technical field of converter transformers, and in particular to an active early warning system and method for converter transformers based on online monitoring of oil chromatography. Background Art

[0002] In the field of power grid systems, as the scale of power grids gradually expands, the structure of power grids becomes increasingly complex. The huge power grid and the need to ensure continuous and safe electricity use have put forward higher requirements for the safe and reliable operation of power grids. As the core equipment of the power system, the converter transformer is a hub equipment that plays the role of power transmission, voltage rise and fall, electric energy distribution and transfer insulation isolation in the power system. Whether the converter transformer operates safely and reliably will directly affect the safety and stability of the entire power system. Therefore, it is of great significance to discover the latent faults of the converter transformer as early as possible or to detect the type and severity of the fault, which is of great significance for the safe and reliable operation of the entire power system, reducing the time for fault repair and reducing economic losses.

[0003] When a converter transformer has internal thermal faults, discharge faults, or oil or paper aging during operation, the insulating oil will crack to produce dissolved gases such as hydrogen (H2), methane (CH4), ethane (C2H6), ethylene (C2H4), acetylene (C2H2), carbon monoxide (CO), and carbon dioxide (CO2). The content of dissolved gases is closely related to the type and severity of the fault. By online monitoring of the weak dissolved gas concentration in the converter transformer insulating oil and analyzing the changes in the gas dissolved in the converter transformer oil, early latent faults inside the transformer can be discovered early, making it easier to grasp the development trend of the fault and take maintenance measures in a timely manner.

[0004] At present, the status monitoring data of converter transformers mainly include oil chromatography, single hydrogen, single acetylene, partial discharge, etc. Among them, oil chromatography online monitoring technology can detect rapid faults due to its continuous monitoring and real-time control of the status of converter transformer monitoring equipment, making it the main technical means to realize transformer status maintenance. Oil chromatography online monitoring usually analyzes the concentration of different gases in the oil chromatogram of the converter transformer based on the three-ratio method to judge the internal operation of the converter transformer. However, due to the large limitations of the three-ratio method, it can only include the faults corresponding to its coding combination, and may not find the corresponding ratio coding combination. The fault classification coverage is not comprehensive. For the fault types with strong dispersion, it is difficult to accurately extract fault features; converter transformer fault warning based on neural network deep learning is still in its infancy. According to known abnormal samples, the early warning system is established through supervised learning training. Its accuracy is significantly higher than that of the ratio method transformer fault diagnosis. However, in reality, the fault rate of converter transformers is extremely low, and the number of fault cases and abnormal samples that can be collected is limited. It is difficult to meet the requirements of large-scale training data required for training, resulting in large fluctuations in the recognition accuracy and stability of converter transformer abnormalities, affecting the results and credibility of converter transformer fault warning.

[0005] In order to eliminate the problems in the process of online monitoring and fault warning of converter transformers, such as the inability to reliably and accurately identify converter transformer faults and insufficient timely fault warning due to the inability of the monitoring method to accurately reflect the operating status of the converter transformer, technical personnel in this field have been seeking effective online monitoring and fault warning methods for converter transformers, so as to make full use of large-scale data for abnormal monitoring and early warning during the operation of the converter transformer, and ensure the stability and reliability of online monitoring of the converter transformer. Summary of the invention

[0006] The purpose of the present invention is to provide an active early warning system and method for converter transformers based on online monitoring of oil chromatography, use the Anomaly-Transformer model for fault diagnosis of converter transformers, extract features through a temporal convolutional network, improve the accuracy of transformer fault diagnosis, solve the accuracy and stability fluctuations of abnormal identification caused by traditional monitoring methods, and ensure the safety and reliability of converter transformers.

[0007] The active early warning system for converter transformers based on online monitoring of oil chromatography designed by the present invention to achieve one of the above purposes is special in that it includes:

[0008] A data collection module is used to collect historical data of gas dissolved in insulating oil from historical data of the converter transformer insulating oil chromatogram, and associate the collected historical data of gas dissolved in insulating oil with a timestamp mark to obtain time series historical data of gas dissolved in insulating oil;

[0009] A data preprocessing module is used to identify abnormal values ​​in the time series historical data of the insulating oil dissolved gas, replace the abnormal values ​​in the time series historical data of the insulating oil dissolved gas with normal values ​​in the time series historical data of the insulating oil dissolved gas, obtain the preprocessed time series historical data of the insulating oil dissolved gas, and standardize the preprocessed time series historical data of the insulating oil dissolved gas to obtain the standardized time series historical data of the insulating oil dissolved gas;

[0010] A feature extraction module is used to extract features from the standardized time series historical data of the gas dissolved in the insulating oil through a time convolution network to obtain feature values ​​of the time series historical data of the gas dissolved in the insulating oil;

[0011] A model training module is used to train a fault diagnosis model using characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil, extract fault features from the characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil, and obtain a fault diagnosis model and fault diagnosis feature weights of the time series historical data of the gas dissolved in the insulating oil;

[0012] The monitoring and early warning module is used to determine whether the data of dissolved gas in insulating oil collected at the current time point has a fault risk based on the fault diagnosis model and fault diagnosis feature weights of the time series historical data of dissolved gas in insulating oil.

[0013] Furthermore, the data collection module collects historical data of gas dissolved in insulating oil from historical data of converter transformer insulating oil chromatogram, including gas concentrations of hydrogen, methane, ethane, ethylene and acetylene dissolved in converter transformer insulating oil.

[0014] Furthermore, the data preprocessing module preprocesses the time series historical data of the dissolved gas in the insulating oil, including: identifying abnormal values ​​in the time series historical data of the dissolved gas in the insulating oil collected at fixed intervals, and using the normal value in the time series historical data of the dissolved gas in the insulating oil at the previous moment and the normal value in the time series historical data of the dissolved gas in the insulating oil at the next moment, taking the average value to replace the abnormal value in the time series historical data of the dissolved gas in the insulating oil, and obtaining the preprocessed time series historical data of the dissolved gas in the insulating oil, as described in the following formula:

[0015]

[0016] Among them, t represents the current time, x t-1 represents the normal value of the time series historical data of the gas dissolved in the insulating oil at the moment before the current moment t, x t+1 represents the normal value of the time series historical data of the gas dissolved in the insulating oil at the moment after the current moment t, x t* Represents the time series historical data of dissolved gas in insulating oil after pretreatment.

[0017] Furthermore, the data preprocessing module performs standardization processing on the time series historical data of the gas dissolved in the insulating oil, including: performing normalization processing on the preprocessed time series historical data of the gas dissolved in the insulating oil, to obtain standardized time series historical data of the gas dissolved in the insulating oil, as specifically described in the following formula:

[0018]

[0019] Among them, x i represents the time series historical data of dissolved gas in insulating oil after pretreatment, max(x i ) represents the maximum value in the time series historical data of the dissolved gas in the pre-treated insulating oil, min(x i ) represents the minimum value of the time series historical data of the dissolved gas in the pre-treated insulating oil, x i * Represents the time series historical data of the standardized dissolved gas in insulating oil.

[0020] Furthermore, the feature extraction module extracts features of the standardized time series historical data of the gas dissolved in the insulating oil, including: extracting local features of the time series historical data of the gas dissolved in the insulating oil by one-dimensional convolution, as described in the following formula:

[0021]

[0022] Wherein, k(m) represents the local features of the time series historical data of the dissolved gas in the insulating oil, x(t) represents the standardized time series historical data of the dissolved gas in the insulating oil, w represents the number of weights in the convolution kernel, and y(t) represents the one-dimensional convolution output at time t;

[0023] The feature extraction module extracts features of the standardized time series historical data of the dissolved gas in the insulating oil, and further includes: extracting dependency features of the time series historical data of the dissolved gas in the insulating oil within a longer time span by dilated causal convolution, as described in the following formula:

[0024]

[0025] Among them, f(i) represents the dependency characteristics within a longer time span of the time series historical data of dissolved gas in insulating oil, y(s) represents the output of one-dimensional convolution, w represents the number of weights in the convolution kernel, d represents the dilation factor, and f(s) represents the dilated causal convolution output at time t;

[0026] The feature extraction module extracts features of the standardized time series historical data of the dissolved gas in the insulating oil, and further includes: integrating the features of the time series historical data of the dissolved gas in the insulating oil through a fully connected layer using the convolution output after local feature extraction and dependency extraction within a longer time span to obtain feature values ​​of the time series historical data of the dissolved gas in the insulating oil, as described in the following formula:

[0027] y=σ(W x +b)

[0028] Among them, σ represents the activation function, W represents the weight matrix, x represents the convolution output after local feature extraction and dependency extraction over a longer time span, b represents the bias term, and y represents the eigenvalue of the time series historical data of the dissolved gas in the insulating oil.

[0029] As a preferred solution, the model training module includes:

[0030] An Anomaly-Transformer model training unit, used for inputting characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil into the Anomaly-Transformer model for training;

[0031] The self-attention mechanism unit is used to implement the self-attention mechanism in the Anomaly-Transformer model to capture the local dependencies in the eigenvalues ​​of the time series historical data of the dissolved gas in the insulating oil, as described in the following formula:

[0032]

[0033] Among them, σ i represents a learnable scale parameter, i represents the first index point in the time series historical data of the gas dissolved in the insulating oil, j represents the second index point in the time series historical data of the gas dissolved in the insulating oil, and {1, 2, ..., N} represents the time series historical data of the gas dissolved in the insulating oil;

[0034] The self-attention mechanism unit also captures the long-term dependencies in the feature values ​​of the time series historical data of the gas dissolved in the insulating oil, as described in the following formula:

[0035]

[0036] Among them, Q represents the query matrix, K represents the key matrix, and V represents the value matrix;

[0037] The model optimization unit is used to optimize the parameters of the Anomaly-Transformer model through the back-propagation algorithm to minimize the reconstruction loss;

[0038] The feature weight extraction unit is used to extract the fault diagnosis feature weight of the time series historical data of the dissolved gas in the insulating oil from the trained Anomaly-Transformer model as the basis for monitoring and early warning.

[0039] As a preferred solution, the monitoring and early warning module calculates the abnormal score of the data of the dissolved gas in the insulating oil collected at the current time point based on the fault diagnosis model and fault diagnosis feature weight of the time series historical data of the dissolved gas in the insulating oil, and determines whether the data of the dissolved gas in the insulating oil collected at the current time point is abnormal data according to the abnormal score threshold. The abnormal score AnomaltScore (X) of the data of the dissolved gas in the insulating oil collected at the current time point is calculated using the following formula:

[0040]

[0041] Where X represents the time series historical data of the gas dissolved in insulating oil, represents the data of dissolved gas in insulating oil predicted by the fault diagnosis model based on the time series historical data of dissolved gas in insulating oil, P represents the local dependency in the characteristic values ​​of the time series historical data of dissolved gas in insulating oil, S represents the long-term dependency in the characteristic values ​​of the time series historical data of dissolved gas in insulating oil, AssDis(P, S; λ; X) represents the difference between the local dependency in the characteristic values ​​of the time series historical data of dissolved gas in insulating oil and the long-term dependency in the characteristic values ​​of the time series historical data of dissolved gas in insulating oil, and S0ftmax(-AssDis()) represents the conversion of the negative value of the dependency difference into a probability Distribution; Based on the anomaly score AnomaltScore(X) of the data of dissolved gas in insulating oil collected at the current time point, the data of dissolved gas in insulating oil collected at the current time point is compared with the anomaly score threshold. If the data of dissolved gas in insulating oil collected at the current time point is greater than the anomaly score threshold, the data of dissolved gas in insulating oil collected at the current time point is abnormal data. The data of dissolved gas in insulating oil collected at the current time point is compared with the characteristic values ​​of the time series historical data of dissolved gas in insulating oil for similarity, and the fault type of the converter transformer is determined. According to the fault type of the converter transformer determined by the comparison, the early warning information is sent to the monitoring and operation platform.

[0042] The active early warning method for converter transformer based on online monitoring of oil chromatography designed by the present invention to achieve the second objective above is special in that it comprises the following steps:

[0043] Collect historical data of the gas dissolved in the insulating oil from the historical data of the converter transformer insulating oil chromatogram, and associate the collected historical data of the gas dissolved in the insulating oil with a timestamp mark to obtain the time series historical data of the gas dissolved in the insulating oil;

[0044] Identify abnormal values ​​in the time series historical data of the gas dissolved in the insulating oil, use normal values ​​in the time series historical data of the gas dissolved in the insulating oil to replace the abnormal values ​​in the time series historical data of the gas dissolved in the insulating oil, obtain pre-processed time series historical data of the gas dissolved in the insulating oil, and standardize the pre-processed time series historical data of the gas dissolved in the insulating oil to obtain standardized time series historical data of the gas dissolved in the insulating oil;

[0045] The time convolution network is used to extract the features of the standardized time series historical data of the gas dissolved in the insulating oil, and the feature values ​​of the time series historical data of the gas dissolved in the insulating oil are obtained;

[0046] The fault diagnosis model is trained by using the characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil, and the fault characteristics in the characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil are extracted to obtain the fault diagnosis model and fault diagnosis characteristic weights of the time series historical data of the gas dissolved in the insulating oil;

[0047] Based on the fault diagnosis model and fault diagnosis feature weights of the time series historical data of the dissolved gas in the insulating oil, it is judged whether the data of the dissolved gas in the insulating oil collected at the current time point has a fault risk.

[0048] A computing device designed by the present invention to achieve the third objective above includes a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0049] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned converter transformer active early warning method based on oil chromatography online monitoring.

[0050] In order to achieve the fourth purpose of the present invention, a computer storage medium is designed, in which at least one executable instruction is stored, and the executable instruction enables the processor to execute operations corresponding to the active early warning method for converter transformers based on online monitoring of oil chromatography as described above.

[0051] The present invention has the following beneficial effects:

[0052] (1) The active early warning system and method for converter transformers based on online monitoring of oil chromatography collects and preprocesses the time series historical data of dissolved gas in insulating oil, uses a temporal convolutional network for feature extraction, and uses the Anomaly-Transformer model to train to obtain a fault diagnosis model and feature weights. This can more accurately capture potential fault characteristics, thereby detecting abnormal conditions and issuing early warnings in a timely manner.

[0053] (2) The active early warning system and method for converter transformers based on online monitoring of oil chromatography adopts neural network deep learning, such as time convolution network and Anomaly-Transformer model, which can automatically learn the time series characteristics and dependencies of complex gas dissolved in insulating oil without manual setting of rules, thereby improving the intelligence level and work efficiency of online detection of converter transformers. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 The module diagram of an active early warning system for converter transformers based on online monitoring of oil chromatography is shown in FIG.

[0055] Figure 2 The present invention is a flow chart of an active early warning method for converter transformers based on online monitoring of oil chromatography.

[0056] Figure 3 The figure is a schematic diagram of the structure of a computing device of the present invention. DETAILED DESCRIPTION

[0057] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0058] Figure 1 FIG. 4 shows a schematic diagram of a module of an active early warning system for a converter transformer based on online monitoring of oil chromatography according to an embodiment of the present invention. Figure 1 As shown, the system includes:

[0059] A data collection module is used to collect historical data of gas dissolved in insulating oil from historical data of the converter transformer insulating oil chromatogram, and associate the collected historical data of gas dissolved in insulating oil with a timestamp mark to obtain time series historical data of gas dissolved in insulating oil;

[0060] A data preprocessing module is used to identify abnormal values ​​in the time series historical data of the insulating oil dissolved gas, replace the abnormal values ​​in the time series historical data of the insulating oil dissolved gas with normal values ​​in the time series historical data of the insulating oil dissolved gas, obtain the preprocessed time series historical data of the insulating oil dissolved gas, and standardize the preprocessed time series historical data of the insulating oil dissolved gas to obtain the standardized time series historical data of the insulating oil dissolved gas;

[0061] A feature extraction module is used to extract features from the standardized time series historical data of the gas dissolved in the insulating oil through a time convolution network to obtain feature values ​​of the time series historical data of the gas dissolved in the insulating oil;

[0062] A model training module is used to train a fault diagnosis model using characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil, extract fault features from the characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil, and obtain a fault diagnosis model and fault diagnosis feature weights of the time series historical data of the gas dissolved in the insulating oil;

[0063] The monitoring and early warning module is used to determine whether the data of dissolved gas in insulating oil collected at the current time point has a fault risk based on the fault diagnosis model and fault diagnosis feature weights of the time series historical data of dissolved gas in insulating oil, and provide early warning information to the monitoring and operation and maintenance platform.

[0064] The data collection module collects historical data of gas dissolved in the insulating oil from the historical data of the converter transformer insulating oil chromatogram, including gas concentrations of hydrogen, methane, ethane, ethylene and acetylene dissolved in the converter transformer insulating oil.

[0065] Specifically, the data preprocessing module preprocesses the time series historical data of the dissolved gas in the insulating oil, including: identifying abnormal values ​​in the time series historical data of the dissolved gas in the insulating oil collected at fixed intervals, including: complete failure abnormality, volatility abnormality, periodic abnormality, outlier abnormality, regularity abnormality, and using the normal value in the time series historical data of the dissolved gas in the insulating oil at the previous moment and the normal value in the time series historical data of the dissolved gas in the insulating oil at the next moment, taking the average value to replace the abnormal value in the time series historical data of the dissolved gas in the insulating oil, and obtaining the preprocessed time series historical data of the dissolved gas in the insulating oil, as described in the following formula:

[0066]

[0067] Among them, t represents the current time, x t-1 represents the normal value of the time series historical data of the gas dissolved in the insulating oil at the moment before the current moment t, x t+1 represents the normal value of the time series historical data of the gas dissolved in the insulating oil at the moment after the current moment t, x t * Represents the time series historical data of dissolved gas in insulating oil after pretreatment.

[0068] Specifically, the data preprocessing module performs standardization processing on the time series historical data of the dissolved gas in the insulating oil, including: performing normalization processing on the preprocessed time series historical data of the dissolved gas in the insulating oil, so that the value of the time series historical data of the dissolved gas in the insulating oil is between 0 and 1, and obtaining the standardized time series historical data of the dissolved gas in the insulating oil, which is specifically described in the following formula:

[0069]

[0070] Among them, x i represents the time series historical data of dissolved gas in insulating oil after pretreatment, max(x i ) represents the maximum value in the time series historical data of the dissolved gas in the pre-treated insulating oil, min(x i ) represents the minimum value of the time series historical data of the dissolved gas in the pre-treated insulating oil, x i * Represents the time series historical data of the standardized dissolved gas in insulating oil.

[0071] Specifically, the feature extraction module extracts features of the standardized time series historical data of the dissolved gas in the insulating oil, including: extracting local features of the time series historical data of the dissolved gas in the insulating oil by one-dimensional convolution, as described in the following formula:

[0072]

[0073] Wherein, k(m) represents the local features of the time series historical data of the dissolved gas in the insulating oil, x(t) represents the standardized time series historical data of the dissolved gas in the insulating oil, w represents the number of weights in the convolution kernel, and y(t) represents the one-dimensional convolution output at time t;

[0074] The feature extraction module extracts features of the standardized time series historical data of the dissolved gas in the insulating oil, and further includes: extracting dependency features of the time series historical data of the dissolved gas in the insulating oil within a longer time span by dilated causal convolution, as described in the following formula:

[0075]

[0076] Among them, f(i) represents the dependency characteristics within a longer time span of the time series historical data of dissolved gas in insulating oil, y(s) represents the output of one-dimensional convolution, w represents the number of weights in the convolution kernel, d represents the dilation factor, and f(s) represents the dilated causal convolution output at time t;

[0077] The feature extraction module extracts features of the standardized time series historical data of the dissolved gas in the insulating oil, and further includes: integrating the features of the time series historical data of the dissolved gas in the insulating oil through the fully connected layer using the convolution output after local feature extraction and dependency extraction within a longer time span to obtain the feature value of the time series historical data of the dissolved gas in the insulating oil, as described in the following formula:

[0078] y=σ(W x +b)

[0079] Among them, σ represents the activation function, W represents the weight matrix, x represents the convolution output after local feature extraction and dependency extraction over a longer time span, b represents the bias term, and y represents the eigenvalue of the time series historical data of the dissolved gas in the insulating oil.

[0080] Specifically, the model training module includes:

[0081] An Anomaly-Transformer model training unit, used for inputting characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil into the Anomaly-Transformer model for training;

[0082] The self-attention mechanism unit is used to implement the self-attention mechanism in the Anomaly-Transformer model to capture the local dependencies in the eigenvalues ​​of the time series historical data of the dissolved gas in the insulating oil, as described in the following formula:

[0083]

[0084] Among them, σ i represents a learnable scale parameter, i represents the first index point in the time series historical data of the gas dissolved in the insulating oil, j represents the second index point in the time series historical data of the gas dissolved in the insulating oil, and {1, 2, ..., N} represents the time series historical data of the gas dissolved in the insulating oil;

[0085] The self-attention mechanism unit also captures the long-term dependencies in the feature values ​​of the time series historical data of the dissolved gas in the insulating oil, as described in the following formula:

[0086]

[0087] Among them, Q represents the query matrix, K represents the key matrix, and V represents the value matrix;

[0088] The model optimization unit is used to optimize the parameters of the Anomaly-Transformer model through the back-propagation algorithm to minimize the reconstruction loss;

[0089] The feature weight extraction unit is used to extract the fault diagnosis feature weight of the time series historical data of the dissolved gas in the insulating oil from the trained Anomaly-Transformer model as the basis for monitoring and early warning.

[0090] Specifically, the monitoring and early warning module calculates the anomaly score of the data of the dissolved gas in the insulating oil collected at the current time point based on the fault diagnosis model and fault diagnosis feature weight of the time series historical data of the dissolved gas in the insulating oil, and determines whether the data of the dissolved gas in the insulating oil collected at the current time point is abnormal data according to the anomaly score threshold. The anomaly score AnomaltScore(X) of the data of the dissolved gas in the insulating oil collected at the current time point is calculated using the following formula:

[0091]

[0092] Where X represents the time series historical data of the gas dissolved in insulating oil, represents the data of dissolved gas in insulating oil predicted by the fault diagnosis model based on the time series historical data of dissolved gas in insulating oil, P represents the local dependency in the eigenvalues ​​of the time series historical data of dissolved gas in insulating oil, S represents the long-term dependency in the eigenvalues ​​of the time series historical data of dissolved gas in insulating oil, AssDis(P, S; λ; X) represents the difference between the local dependency in the eigenvalues ​​of the time series historical data of dissolved gas in insulating oil and the long-term dependency in the eigenvalues ​​of the time series historical data of dissolved gas in insulating oil, and Softmax(-AssDis()) represents the conversion of the negative value of the dependency difference into a probability Distribution; Based on the anomaly score AnomaltScore(X) of the data of dissolved gas in insulating oil collected at the current time point, the data of dissolved gas in insulating oil collected at the current time point is compared with the anomaly score threshold. If the data of dissolved gas in insulating oil collected at the current time point is greater than the anomaly score threshold, the data of dissolved gas in insulating oil collected at the current time point is abnormal data. The data of dissolved gas in insulating oil collected at the current time point is compared with the characteristic values ​​of the time series historical data of dissolved gas in insulating oil for similarity, and the fault type of the converter transformer is determined. According to the fault type of the converter transformer determined by the comparison, the early warning information is sent to the monitoring and operation platform.

[0093] Figure 2 FIG. 4 is a flow chart showing a converter transformer active early warning method based on oil chromatography online monitoring according to an embodiment of the present invention. Figure 2 As shown, the method comprises the following steps:

[0094] Step 1: Collect historical data of insulating oil dissolved gas from historical data of converter transformer insulating oil chromatogram, and associate the collected historical data of insulating oil dissolved gas with a timestamp mark to obtain time series historical data of insulating oil dissolved gas;

[0095] Step 2, identifying abnormal values ​​in the time series historical data of the insulating oil dissolved gas, replacing the abnormal values ​​in the time series historical data of the insulating oil dissolved gas with normal values ​​in the time series historical data of the insulating oil dissolved gas, obtaining the pre-processed time series historical data of the insulating oil dissolved gas, and standardizing the pre-processed time series historical data of the insulating oil dissolved gas to obtain standardized time series historical data of the insulating oil dissolved gas;

[0096] Step 3: extracting features from the standardized time series historical data of the gas dissolved in the insulating oil by using a time convolution network to obtain feature values ​​of the time series historical data of the gas dissolved in the insulating oil;

[0097] Step 4: training a fault diagnosis model using the characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil, extracting fault features from the characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil, and obtaining a fault diagnosis model and fault diagnosis feature weights of the time series historical data of the gas dissolved in the insulating oil;

[0098] Step 5: Based on the fault diagnosis model and fault diagnosis feature weights of the time series historical data of the dissolved gas in the insulating oil, determine whether the data of the dissolved gas in the insulating oil collected at the current time point has a fault risk, and provide early warning information to the monitoring and operation and maintenance platform.

[0099] Figure 3 A schematic diagram of the structure of a computing device according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computing device.

[0100] like Figure 3 As shown, the computing device may include: a processor, a communication interface, a memory, and a communication bus.

[0101] Wherein: the processor, the communication interface, and the memory communicate with each other via a communication bus.

[0102] Communication interface, used to communicate with other devices such as client or other server network elements.

[0103] The processor is used to execute the program, and specifically can execute the relevant steps in the embodiment of the active early warning method for converter transformer based on online monitoring of oil chromatography.

[0104] Specifically, the program may include program codes including computer operation instructions.

[0105] The processor may be a central processing unit (CPU), or a specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the computing device may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0106] The memory is used to store programs. The memory may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0107] The program can be specifically used to enable the processor to execute the active early warning method for converter transformers based on online monitoring of oil chromatography in any of the above-mentioned method embodiments. The specific implementation of each step in the program can refer to the corresponding descriptions in the corresponding steps and units in the above-mentioned embodiment of the active early warning method for converter transformers based on online monitoring of oil chromatography, which will not be repeated here. Technical personnel in the relevant field can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described equipment and modules can refer to the corresponding process description in the aforementioned method embodiment, which will not be repeated here.

[0108] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the present invention is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the description of the above specific languages ​​is for disclosing the best mode of the present invention.

[0109] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0110] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the intention that the claimed invention requires more features than those expressly recited in each claim. Rather, as reflected in the claims, inventive aspects lie in less than all of the features of the individual embodiments previously disclosed. Therefore, the claims that follow the detailed description are hereby expressly incorporated into the detailed description, with each claim itself serving as a separate embodiment of the present invention.

Claims

1. An active early warning system for converter transformers based on online monitoring of oil chromatography, characterized in that: include: A data collection module is used to collect historical data of gas dissolved in insulating oil from historical data of the converter transformer insulating oil chromatogram, and associate the collected historical data of gas dissolved in insulating oil with a timestamp mark to obtain time series historical data of gas dissolved in insulating oil; A data preprocessing module is used to identify abnormal values ​​in the time series historical data of the insulating oil dissolved gas, replace the abnormal values ​​in the time series historical data of the insulating oil dissolved gas with normal values ​​in the time series historical data of the insulating oil dissolved gas, obtain the preprocessed time series historical data of the insulating oil dissolved gas, and standardize the preprocessed time series historical data of the insulating oil dissolved gas to obtain the standardized time series historical data of the insulating oil dissolved gas; A feature extraction module is used to extract features from the standardized time series historical data of the gas dissolved in the insulating oil through a time convolution network to obtain feature values ​​of the time series historical data of the gas dissolved in the insulating oil; A model training module is used to train a fault diagnosis model using characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil, extract fault features from the characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil, and obtain a fault diagnosis model and fault diagnosis feature weights of the time series historical data of the gas dissolved in the insulating oil; The monitoring and early warning module is used to determine whether the data of dissolved gas in insulating oil collected at the current time point has a fault risk based on the fault diagnosis model and fault diagnosis feature weights of the time series historical data of dissolved gas in insulating oil.

2. The converter transformer active early warning system based on oil chromatography online monitoring according to claim 1 is characterized in that: The data collection module collects historical data of gas dissolved in the insulating oil from the historical data of the converter transformer insulating oil chromatogram, including gas concentrations of hydrogen, methane, ethane, ethylene and acetylene dissolved in the converter transformer insulating oil.

3. The converter transformer active early warning system based on oil chromatography online monitoring according to claim 1 is characterized in that: The data preprocessing module preprocesses the time series historical data of the insulating oil dissolved gas, including: identifying abnormal values ​​in the time series historical data of the insulating oil dissolved gas collected at fixed intervals, and using the normal value in the time series historical data of the insulating oil dissolved gas at the previous moment and the normal value in the time series historical data of the insulating oil dissolved gas at the next moment, taking the average value to replace the abnormal value in the time series historical data of the insulating oil dissolved gas, and obtaining the preprocessed time series historical data of the insulating oil dissolved gas, which is specifically described in the following formula: Among them, t represents the current time, x t-1 represents the normal value of the time series historical data of the gas dissolved in the insulating oil at the moment before the current moment t, x t+1 represents the normal value of the time series historical data of the gas dissolved in the insulating oil at the moment after the current moment t, x t * Represents the time series historical data of dissolved gas in insulating oil after pretreatment.

4. The converter transformer active early warning system based on oil chromatography online monitoring according to claim 1 is characterized in that: The data preprocessing module performs normalization processing on the time series historical data of the gas dissolved in the insulating oil, including: performing normalization processing on the preprocessed time series historical data of the gas dissolved in the insulating oil, to obtain the normalized time series historical data of the gas dissolved in the insulating oil, as specifically described in the following formula: Among them, x i represents the time series historical data of dissolved gas in insulating oil after pretreatment, max(x i ) represents the maximum value in the time series historical data of the dissolved gas in the pre-treated insulating oil, min(x i ) represents the minimum value of the time series historical data of the dissolved gas in the pre-treated insulating oil, x i * Represents the time series historical data of the standardized dissolved gas in insulating oil.

5. The converter transformer active early warning system based on oil chromatography online monitoring according to claim 1 is characterized in that: The feature extraction module extracts features of the standardized time series historical data of the gas dissolved in the insulating oil, including: extracting local features of the time series historical data of the gas dissolved in the insulating oil by one-dimensional convolution, as described in the following formula: Wherein, k(m) represents the local features of the time series historical data of the dissolved gas in the insulating oil, x(t) represents the standardized time series historical data of the dissolved gas in the insulating oil, w represents the number of weights in the convolution kernel, and y(t) represents the one-dimensional convolution output at time t; The feature extraction module extracts features of the standardized time series historical data of the dissolved gas in the insulating oil, and further includes: extracting dependency features of the time series historical data of the dissolved gas in the insulating oil within a longer time span by dilated causal convolution, as described in the following formula: Among them, f(i) represents the dependency characteristics within a longer time span of the time series historical data of dissolved gas in insulating oil, y(s) represents the output of one-dimensional convolution, w represents the number of weights in the convolution kernel, d represents the dilation factor, and f(s) represents the dilated causal convolution output at time t; The feature extraction module extracts features of the standardized time series historical data of the dissolved gas in the insulating oil, and further includes: integrating the features of the time series historical data of the dissolved gas in the insulating oil through a fully connected layer using the convolution output after local feature extraction and dependency extraction within a longer time span to obtain feature values ​​of the time series historical data of the dissolved gas in the insulating oil, as described in the following formula: y=σ(W x +b) Among them, σ represents the activation function, W represents the weight matrix, x represents the convolution output after local feature extraction and dependency extraction over a longer time span, b represents the bias term, and y represents the eigenvalue of the time series historical data of the dissolved gas in the insulating oil.

6. The converter transformer active early warning system based on oil chromatography online monitoring according to claim 1 is characterized in that: The model training module includes: An Anomaly-Transformer model training unit, used for inputting characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil into the Anomaly-Transformer model for training; The self-attention mechanism unit is used to implement the self-attention mechanism in the Anomaly-Transformer model to capture the local dependencies in the eigenvalues ​​of the time series historical data of the dissolved gas in the insulating oil, as described in the following formula: Among them, σ i represents a learnable scale parameter, i represents the first index point in the time series historical data of the gas dissolved in the insulating oil, j represents the second index point in the time series historical data of the gas dissolved in the insulating oil, and {1,2,…,N} represents the time series historical data of the gas dissolved in the insulating oil; The self-attention mechanism unit also captures the long-term dependencies in the feature values ​​of the time series historical data of the gas dissolved in the insulating oil, as described in the following formula: Among them, Q represents the query matrix, K represents the key matrix, and V represents the value matrix; The model optimization unit is used to optimize the parameters of the Anomaly-Transformer model through the back-propagation algorithm to minimize the reconstruction loss; The feature weight extraction unit is used to extract the fault diagnosis feature weight of the time series historical data of the dissolved gas in the insulating oil from the trained Anomaly-Transformer model as the basis for monitoring and early warning.

7. The converter transformer active early warning system based on oil chromatography online monitoring according to claim 1 is characterized in that: The monitoring and early warning module calculates the anomaly score of the data of the dissolved gas in the insulating oil collected at the current time point based on the fault diagnosis model and fault diagnosis feature weight of the time series historical data of the dissolved gas in the insulating oil, and determines whether the data of the dissolved gas in the insulating oil collected at the current time point is abnormal data according to the anomaly score threshold. The anomaly score AnomaltScore(X) of the data of the dissolved gas in the insulating oil collected at the current time point is calculated using the following formula: Where X represents the time series historical data of the gas dissolved in insulating oil, represents the data of dissolved gas in insulating oil predicted by the fault diagnosis model based on the time series historical data of dissolved gas in insulating oil, P represents the local dependency in the eigenvalues ​​of the time series historical data of dissolved gas in insulating oil, S represents the long-term dependency in the eigenvalues ​​of the time series historical data of dissolved gas in insulating oil, AssDis(P,S;λ;X) represents the difference between the local dependency in the eigenvalues ​​of the time series historical data of dissolved gas in insulating oil and the long-term dependency in the eigenvalues ​​of the time series historical data of dissolved gas in insulating oil, Softmax(-AssDis()) represents the conversion of the negative value of the dependency difference into a probability Distribution; Based on the anomaly score AnomaltScore(X) of the data of dissolved gas in insulating oil collected at the current time point, the data of dissolved gas in insulating oil collected at the current time point is compared with the anomaly score threshold. If the data of dissolved gas in insulating oil collected at the current time point is greater than the anomaly score threshold, the data of dissolved gas in insulating oil collected at the current time point is abnormal data. The data of dissolved gas in insulating oil collected at the current time point is compared with the characteristic values ​​of the time series historical data of dissolved gas in insulating oil for similarity, and the fault type of the converter transformer is determined. According to the fault type of the converter transformer determined by the comparison, the early warning information is sent to the monitoring and operation platform.

8. An active early warning method for converter transformer based on online monitoring of oil chromatography, characterized in that: The steps include: Collect historical data of the gas dissolved in the insulating oil from the historical data of the converter transformer insulating oil chromatogram, and associate the collected historical data of the gas dissolved in the insulating oil with a timestamp mark to obtain the time series historical data of the gas dissolved in the insulating oil; Identify abnormal values ​​in the time series historical data of the gas dissolved in the insulating oil, use normal values ​​in the time series historical data of the gas dissolved in the insulating oil to replace the abnormal values ​​in the time series historical data of the gas dissolved in the insulating oil, obtain pre-processed time series historical data of the gas dissolved in the insulating oil, and standardize the pre-processed time series historical data of the gas dissolved in the insulating oil to obtain standardized time series historical data of the gas dissolved in the insulating oil; The time convolution network is used to extract the features of the standardized time series historical data of the gas dissolved in the insulating oil, and the feature values ​​of the time series historical data of the gas dissolved in the insulating oil are obtained; The fault diagnosis model is trained by using the characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil, and the fault characteristics in the characteristic values ​​of the time series historical data of the gas dissolved in the insulating oil are extracted to obtain the fault diagnosis model and fault diagnosis characteristic weights of the time series historical data of the gas dissolved in the insulating oil; Based on the fault diagnosis model and fault diagnosis feature weights of the time series historical data of the dissolved gas in the insulating oil, it is judged whether the data of the dissolved gas in the insulating oil collected at the current time point has a fault risk.

9. A computing device comprising a processor, a memory, a communication interface and a communication bus, characterized in that: The processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the active early warning method for converter transformers based on online monitoring of oil chromatography as described in claim 8.

10. A computer storage medium, wherein at least one executable instruction is stored in the computer storage medium, wherein the executable instruction enables a processor to execute operations corresponding to the converter transformer active early warning method based on oil chromatography online monitoring as claimed in claim 8.

Images