Digital cloud online monitoring and diagnosis method and system

Abstract

The invention, which relates to the technical field of transformer monitoring, discloses a digital cloud-based online monitoring and diagnosis method comprising the following steps: acquiring real-time operation parameters of a transformer; uploading the real-time operation parameters to a cloud platform, and establishing a transformer digital twinborn model; calculating a deviation value between the real-time operation parameter and a preset standard parameter by using a transformer digital twinborn model, and generating a transformer health state index; based on the transformer health state index, adopting an adaptive multi-mode deep learning framework to predict the health state of the transformer to obtain a fault prediction result; and formulating a preventive maintenance strategy according to the fault prediction result and pushing the preventive maintenance strategy to the terminal equipment. According to the method, the transformer digital twinborn model is constructed, so that the operation state of the transformer is accurately simulated and monitored in real time, and the accuracy of state evaluation is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of transformer monitoring, and in particular to a digital cloud-based online monitoring and diagnosis method and system. Background Art

[0002] As the power system develops towards large capacity, high voltage and intelligence, the operating status of transformers, as key equipment in the power grid, directly affects the safe and stable operation of the power system. Traditional transformer status monitoring methods mainly rely on regular offline detection and manual inspections, which have problems such as long monitoring cycles, poor real-time performance, and insufficient data integrity. This method is difficult to meet the needs of modern power grids for real-time monitoring of equipment status, fault warning, and intelligent operation and maintenance.

[0003] In recent years, with the rapid development of emerging technologies such as the Internet of Things, big data, and cloud computing, transformer online monitoring technology has made significant progress. By installing various sensors on the transformer, key parameters such as oil temperature, winding temperature, load current, and partial discharge can be collected in real time, providing a rich data basis for transformer status assessment. However, the current online monitoring system still has problems such as limited data analysis capabilities, low accuracy of diagnostic results, and poor system scalability. Summary of the invention

[0004] In view of the problems existing in the existing transformer online monitoring system in terms of data processing, diagnosis accuracy and system scalability, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is how to use cloud computing and artificial intelligence technology to achieve real-time monitoring, accurate diagnosis and fault prediction of the transformer operating status, thereby improving the operating reliability and maintenance efficiency of the transformer.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, an embodiment of the present invention provides a digital cloud-based online monitoring and diagnosis method, which includes collecting real-time operating parameters of the transformer, which include electrical parameters, thermal parameters and insulation parameters; uploading the real-time operating parameters to a cloud platform to establish a digital twin model of the transformer; using the digital twin model of the transformer to calculate the deviation between the real-time operating parameters and preset standard parameters, and generate a transformer health status index; based on the transformer health status index, an adaptive multimodal deep learning framework is used to predict the health status of the transformer to obtain a fault prediction result; based on the fault prediction result, a preventive maintenance strategy is formulated and pushed to the terminal device.

[0008] As a preferred solution of the digital cloud-based online monitoring and diagnosis method of the present invention, wherein: establishing a transformer digital twin model includes extracting nameplate parameters, structural characteristics and material property data of the transformer from a cloud database; importing a pre-established transformer multi-physics field coupling model framework, the transformer multi-physics field coupling model framework includes an electrical sub-model, a thermal sub-model and an insulator model;

[0009] The basic parameters of the electrical sub-model, thermal sub-model and insulator model are initialized using nameplate parameters, structural characteristics and material property data; a coupling matrix between sub-models is constructed, and the coupling matrix is ​​adaptively optimized using a deep reinforcement learning algorithm; the electrical sub-model, thermal sub-model and insulator model are integrated based on the optimized coupling matrix to form an initial transformer digital twin model.

[0010] As a preferred solution of the digital cloud-based online monitoring and diagnosis method described in the present invention, the coupling matrix between sub-models is constructed, including: establishing a coupling matrix framework; filling the self-influence coefficient of each sub-model in the diagonal elements of the coupling matrix, and filling the initial cross-influence coefficient in the non-diagonal elements; introducing a time lag factor to expand the coupling matrix into a four-dimensional tensor; establishing a group of physical association equations between sub-models; combining the sub-model parameters and the group of physical association equations to calculate the initial values ​​of the coupling matrix elements; designing an adaptive weight adjustment algorithm to dynamically update the coupling matrix element values ​​according to real-time operating parameters.

[0011] As a preferred solution of the digital cloud-based online monitoring and diagnosis method described in the present invention, generating a transformer health status index includes the following steps: extracting transformer preset standard parameters from a cloud platform database; inputting real-time operating parameters into the transformer digital twin model to calculate the key state quantities of each sub-model; comparing the key state quantities with the preset standard parameters to calculate various deviation values; quantifying and scoring various deviation values ​​based on a health assessment rule library to obtain a health score for each sub-model; and calculating the transformer health status index based on the health score of each sub-model and combined with the preset weights of the sub-model.

[0012] As a preferred solution of the digital cloud-based online monitoring and diagnosis method described in the present invention, the calculation of the key state quantities of each sub-model includes the following steps: classifying the real-time operating parameters and inputting them into the electrical sub-model, thermal sub-model and insulator model of the transformer digital twin model; using the electrical sub-model to calculate the impedance and loss of the transformer; using the thermal sub-model to calculate the hot spot temperature of the transformer; using the insulator model to calculate the dielectric strength of the transformer; executing data exchange between sub-models, and adjusting the calculation results of each sub-model according to the influencing factors; time-stamping and data verification of impedance, loss, hot spot temperature and dielectric strength; integrating the calculation results of each sub-model to generate a transformer overall state data set containing key state quantities.

[0013] As a preferred solution of the digital cloud-based online monitoring and diagnosis method described in the present invention, the adaptive multimodal deep learning framework includes a convolutional neural network module, a long short-term memory network module and an adaptive fusion module.

[0014] As a preferred solution of the digital cloud-based online monitoring and diagnosis method described in the present invention, the following steps are included: using an adaptive multimodal deep learning framework to predict the health status of a transformer: extracting a transformer health status index, real-time operating parameters and historical fault data from a cloud platform to generate a standardized feature vector; constructing a multimodal input layer, inputting the standardized feature vector into a convolutional neural network module, and inputting the historical fault data into a long short-term memory network module; the convolutional neural network module extracts the time series features of the standardized feature vector, and the long short-term memory network module extracts the long-term dependency features of the historical fault data; the adaptive fusion module calculates the weight coefficients of the output of the convolutional neural network module and the output of the long short-term memory network module based on feature importance analysis; according to the weight coefficient, the output features of the convolutional neural network module and the long short-term memory network module are fused to obtain fused features; the fused features are input into a multi-task learning network, and a fault prediction result is generated through three parallel output branches.

[0015] In the second aspect, an embodiment of the present invention provides a digital cloud-based online monitoring and diagnosis system, which includes a data acquisition module for collecting real-time operating parameters of the transformer; a model building module for uploading the real-time operating parameters to the cloud platform to establish a digital twin model of the transformer; a health assessment module for using the digital twin model of the transformer to calculate the deviation value between the real-time operating parameters and the preset standard parameters, and generate a transformer health status index; a fault prediction module for predicting the health status of the transformer based on the transformer health status index using an adaptive multimodal deep learning framework to obtain a fault prediction result; a maintenance strategy module for formulating a preventive maintenance strategy based on the fault prediction result and pushing it to the terminal device.

[0016] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the digital cloud-based online monitoring and diagnosis method as described in the first aspect of the present invention are implemented.

[0017] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the digital cloud-based online monitoring and diagnosis method as described in the first aspect of the present invention are implemented.

[0018] The beneficial effects of the present invention are as follows: by constructing a digital twin model of the transformer, the present invention realizes accurate simulation and real-time monitoring of the transformer operating status, thereby improving the accuracy of status assessment. The adaptive multimodal deep learning framework is adopted to effectively integrate multi-source data information, significantly improving the accuracy and reliability of fault prediction. The data processing and model calculation based on the cloud platform greatly improve the scalability and computing efficiency of the system, making online monitoring of large-scale transformer groups possible. By pushing preventive maintenance strategies to terminal devices, the intelligent and proactive maintenance of transformers is realized, effectively reducing the equipment failure rate and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 This is a framework flow chart of the digital cloud-based online monitoring and diagnosis method.

[0021] Figure 2 A flow chart for building a transformer digital twin model for the digital cloud-based online monitoring and diagnosis method.

[0022] Figure 3 A diagram of computer equipment for a digital cloud-based online monitoring and diagnosis method. DETAILED DESCRIPTION

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0024] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0026] Example 1

[0027] Reference Figure 1 to Figure 3, which is the first embodiment of the present invention, and provides a digital cloud-based online monitoring and diagnosis method, comprising:

[0028] S1: Collect real-time operating parameters of transformer.

[0029] It should be noted that the real-time operating parameters include electrical parameters, thermal parameters and insulation parameters. Among them, electrical parameters include high and low voltage side voltage, current, active power, reactive power, and power factor; thermal parameters include top oil temperature, bottom oil temperature, ambient temperature and cooling equipment operating status; insulation parameters include partial discharge signal, insulating oil dielectric strength and dissolved gas content. These parameters fully reflect the operating status of the transformer and provide the necessary data basis for the subsequent establishment of the digital twin model.

[0030] S2: Upload the real-time operating parameters to the cloud platform and establish a digital twin model of the transformer on the cloud platform.

[0031] S2.1: Import the real-time operation parameters into the cloud platform and pre-process the real-time operation parameters.

[0032] For example, preprocessing includes data cleaning and format unification.

[0033] S2.2: Extract the transformer nameplate parameters, structural characteristics, and material property data from the cloud database.

[0034] S2.3: Import the pre-established transformer multi-physics coupling model framework, which includes an electrical sub-model, a thermal sub-model and an insulator model.

[0035] S2.4: Using nameplate parameters, structural characteristics and material property data, initialize the basic parameters of the electrical sub-model, thermal sub-model and insulator model respectively.

[0036] S2.5: Construct the coupling matrix between sub-models and define the interaction relationships and influencing factors between electrical, thermal and insulation.

[0037] Specifically, a 3x3 coupling matrix framework is established, and the rows and columns of the matrix correspond to the electrical sub-model, thermal sub-model and insulator sub-model respectively; the self-influence coefficient of each sub-model is filled in the diagonal elements of the coupling matrix, and the initial cross-influence coefficient is filled in the non-diagonal elements; the time lag factor is introduced to expand the coupling matrix into a four-dimensional tensor, and the fourth dimension represents the change of influence intensity at different time scales; based on the transformer nameplate parameters, structural characteristics and material property data extracted in S2.2, a set of physical association equations between sub-models is established; using the basic parameters of the sub-models initialized in S2.4, combined with the set of physical association equations, the initial values ​​of the coupling matrix elements are calculated; an adaptive weight adjustment algorithm is designed to dynamically update the values ​​of the coupling matrix elements according to the real-time operating parameters; a coupling strength evaluation index system is constructed, including correlation coefficients, sensitivity and time response characteristics, which are used to quantify the interaction relationship between sub-models; based on the coupling strength evaluation index, the key influencing factors between electrical-thermal-insulation are defined, and these key influencing factors are marked in the coupling matrix.

[0038] Preferably, a time lag factor is introduced to expand the coupling matrix into a four-dimensional tensor, which more accurately describes the time-varying interactions between sub-models. An adaptive weight adjustment algorithm is designed to enable the coupling matrix to be dynamically updated according to real-time data, thereby improving the accuracy and adaptability of the model. A coupling strength evaluation index system is constructed to provide a scientific basis for quantifying the interaction relationship between sub-models.

[0039] S2.6: Use deep reinforcement learning algorithm to adaptively optimize the influencing factors in the coupling matrix.

[0040] S2.7: Integrate the electrical sub-model, thermal sub-model and insulator model based on the optimized coupling matrix to form an initial transformer digital twin model.

[0041] Specifically, a unified data interface is constructed to connect the electrical sub-model, thermal sub-model and insulator model; the optimized coupling matrix is ​​applied to formulate data exchange rules and calculation sequence between sub-models; a sub-model iterative calculation framework is established, which uses the time stepping method to perform the calculation of the electrical-thermal-insulator sub-model in a predetermined order within each time step, and transfers intermediate results between sub-models to reflect the dynamic coupling process of each physical field; a model state synchronization mechanism is developed, which performs unified time stamping and data verification on the output of each sub-model at the end of each calculation cycle to maintain time consistency and data coherence between sub-models; a sub-model output summary algorithm is developed to generate a transformer overall state data set; a model self-calibration mechanism is constructed to dynamically adjust the coupling strength between sub-models according to real-time operating parameters; the above functions are integrated to form a complete initial transformer digital twin model, and the function is verified.

[0042] Preferably, firstly, the unified data interface and optimized coupling matrix ensure seamless integration and efficient data exchange between sub-models, greatly improving the overall performance and computational efficiency of the model. Secondly, the iterative calculation framework using the time-stepping method can accurately capture the dynamic coupling process between the physical fields, significantly improving the simulation accuracy of the model for the actual operating state of the transformer. The model state synchronization mechanism and output summary algorithm ensure the temporal consistency and integrity of the data, providing a reliable basis for subsequent analysis. In particular, the introduction of the model self-calibration mechanism enables the digital twin model to dynamically adjust the coupling strength according to the real-time operating parameters, greatly enhancing the model's adaptability and long-term accuracy.

[0043] S2.8: Use the electrical parameters, thermal parameters and insulation parameters in the real-time operating parameters to update the corresponding sub-model parameters of the initial transformer digital twin model.

[0044] S2.9: Store the updated transformer digital twin model on the cloud platform.

[0045] S3: Use the transformer digital twin model to evaluate the health status of the transformer and generate a health status index.

[0046] S3.1: Extract transformer preset standard parameters from the cloud platform database.

[0047] It should be noted that the preset standard parameters of the transformer include rated voltage, rated current, normal oil temperature range and partial discharge threshold.

[0048] S3.2: Input the real-time operating parameters into the transformer digital twin model and calculate the key state quantities of each sub-model.

[0049] It should be noted that the key state quantities include the impedance and loss of the electrical sub-model, the hot spot temperature of the thermal sub-model, the dielectric strength of the insulator model, etc.

[0050] Specifically, the real-time operating parameters are classified according to electrical parameters, thermal parameters and insulation parameters, and are respectively input into the electrical sub-model, thermal sub-model and insulator sub-model of the transformer digital twin model; the electrical sub-model is used to calculate the impedance and loss of the transformer; the thermal sub-model is applied to calculate the hot spot temperature of the transformer; the insulator model is used to calculate the dielectric strength of the transformer; the electrical-thermal-insulation interaction relationship defined in the coupling matrix is ​​triggered to execute data exchange between sub-models; the calculation results of each sub-model are adjusted according to the influencing factors; the impedance, loss, hot spot temperature and dielectric strength output by each sub-model are time-stamped and data-verified through the model state synchronization mechanism; the sub-model output summary algorithm is used to integrate the calculation results of each sub-model to generate a transformer overall state data set containing key state quantities.

[0051] S3.3: Compare key state quantities with preset standard parameters and calculate various deviation values.

[0052] Specifically, the deviation values ​​include voltage deviation, current deviation, oil temperature deviation and partial discharge intensity deviation.

[0053] S3.4: Based on the pre-established health assessment rule base, each deviation value is quantitatively scored to obtain the health score of each sub-model.

[0054] S3.5: Based on the health score of each sub-model and the preset weight of the sub-model, calculate the transformer health status index.

[0055] S3.6: Store the transformer health status index in the cloud platform and update the transformer health status database.

[0056] S4: Based on the transformer health status index, an adaptive multimodal deep learning framework is used to predict the transformer health status and obtain the fault prediction results.

[0057] Specifically, the steps include:

[0058] S4.1: Extract transformer health status index, real-time operating parameters, and historical fault data from the cloud platform.

[0059] S4.2: Perform normalization processing on the transformer health status index and real-time operating parameters to generate a standardized feature vector.

[0060] S4.3: Construct a multimodal input layer, input the standardized feature vector into the convolutional neural network module, and input the historical fault data into the long short-term memory network module.

[0061] It should be noted that the adaptive multimodal deep learning framework consists of a convolutional neural network module, a long short-term memory network module, and an adaptive fusion module. The multimodal input layer is divided into a feature input branch and a time series input branch. The feature input branch receives the standardized transformer health status index and real-time operating parameters, and the time series input branch receives the time series of historical fault data.

[0062] Furthermore, the convolutional neural network module consists of multiple convolutional layers, each followed by a ReLU activation function, a maximum pooling operation, and a batch normalization process. The last layer of this module is a fully connected layer that outputs a temporal feature vector that captures spatial features. The long short-term memory network module adopts a two-layer LSTM structure. The first layer contains 128 hidden units and outputs the complete sequence, and the second layer contains 64 hidden units and only outputs the results of the final time step. To prevent overfitting, a dropout mechanism is introduced between LSTM layers. The module finally outputs a temporal feature vector that represents long-term dependencies.

[0063] Furthermore, the adaptive fusion module uses the attention mechanism to dynamically calculate the importance weights of CNN and LSTM output features, and realizes adaptive fusion of features based on these weights. The output of the network adopts a multi-task learning paradigm, and predicts potential fault types, fault occurrence probability, and fault development trend through three parallel output branches (fault type branch, fault probability branch, and fault trend branch).

[0064] Preferably, this structural design makes full use of the advantages of CNN in spatial feature extraction and LSTM in temporal dependency modeling, realizes the effective integration of multimodal information through adaptive fusion, and provides comprehensive and accurate prediction capabilities for transformer health status assessment and fault prediction.

[0065] S4.4: The convolutional neural network module extracts the temporal features of the standardized feature vector, and the long short-term memory network module extracts the long-term dependency features of the historical fault data.

[0066] S4.5: The adaptive fusion module calculates the weight coefficients of the output of the convolutional neural network module and the output of the long short-term memory network module based on feature importance analysis.

[0067] S4.6: According to the weight coefficient, the output features of the convolutional neural network module and the long short-term memory network module are fused to obtain the fused features.

[0068] S4.7: Input the fused features into the multi-task learning network and generate fault prediction results through three parallel output branches.

[0069] It should be noted that the fault prediction results include a list of potential fault types of transformers, the probability of occurrence of fault types, and the expected deterioration rate, which are output by the fault type branch, fault probability branch, and fault trend branch, respectively. Each branch contains a task-specific attention mechanism to dynamically adjust the focus of fusion features. Information interaction is achieved between branches through soft parameter sharing, which utilizes the potential correlation between tasks while maintaining task independence. Finally, the outputs of the three branches are combined to form a comprehensive fault prediction result.

[0070] S4.8: Feedback the fault prediction results to the cloud platform and update the transformer health status database.

[0071] S5: Based on the fault prediction results, formulate a preventive maintenance strategy and push it to the terminal equipment.

[0072] Specifically, based on the fault prediction results, the preset maintenance strategy library is queried; in combination with the transformer health status index, corresponding maintenance measures are matched for each potential fault type; maintenance measures are prioritized according to the probability value of the fault type in the prediction results and the expected deterioration rate; based on the priority sorting results, a preliminary maintenance plan is generated including short-term and long-term maintenance recommendations, and the preliminary maintenance plan is submitted to the expert system for review, and the expert system adjusts and optimizes the preliminary maintenance plan according to preset rules and historical experience; the preliminary maintenance plan is revised according to the review results of the expert system to form a preventive maintenance strategy; a preventive maintenance strategy report is generated, including specific maintenance measures, execution priorities and recommended schedules, and the preventive maintenance strategy report is encrypted and pushed to authorized terminal devices.

[0073] Furthermore, the present embodiment also provides a digital cloud-based online monitoring and diagnosis system, including a data acquisition module for collecting real-time operating parameters of the transformer; a model building module for uploading the real-time operating parameters to the cloud platform to establish a digital twin model of the transformer; a health assessment module for using the digital twin model of the transformer to calculate the deviation value between the real-time operating parameters and the preset standard parameters, and generate a transformer health status index; a fault prediction module for predicting the health status of the transformer based on the transformer health status index using an adaptive multimodal deep learning framework to obtain a fault prediction result; a maintenance strategy module for formulating a preventive maintenance strategy based on the fault prediction result and pushing it to the terminal device.

[0074] This embodiment also provides a computer device, which is suitable for the case of a digital cloud-based online monitoring and diagnosis method, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the digital cloud-based online monitoring and diagnosis method proposed in the above embodiment.

[0075] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0076] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the digital cloud-based online monitoring and diagnosis method proposed in the above embodiment is implemented.

[0077] In summary, the present invention realizes accurate simulation and real-time monitoring of the transformer operating status by constructing a digital twin model of the transformer, thereby improving the accuracy of status assessment. The adaptive multimodal deep learning framework is adopted to effectively integrate multi-source data information, significantly improving the accuracy and reliability of fault prediction. Data processing and model calculation based on the cloud platform greatly improve the scalability and computing efficiency of the system, making online monitoring of large-scale transformer groups possible. By pushing preventive maintenance strategies to terminal devices, the intelligent and proactive maintenance of transformers is realized, effectively reducing equipment failure rates and maintenance costs.

[0078] Example 2

[0079] Reference Figure 1 to Figure 3 , which is the second embodiment of the present invention, and this embodiment provides a digital cloud-based online monitoring and diagnosis method. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0080] The above technical solution was applied to a 220kV / 110kV transformer operating in a provincial power company substation for simulation experiments. The transformer has a rated capacity of 180MVA and has been in operation for 15 years. The experimental period is 6 months, and the sampling frequency is once every 10 minutes.

[0081] First, the acquisition system obtained the real-time operating parameters of the transformer. In terms of electrical parameters, the average voltage on the high-voltage side is 221.8kV, the average voltage on the low-voltage side is 110.9kV, the average current on the high-voltage side is 458A, the average current on the low-voltage side is 917A, the average active power is 172.3MW, the average reactive power is 33.6Mvar, and the average power factor is 0.982. In terms of thermal parameters, the average top oil temperature is 57.6℃, the average bottom oil temperature is 52.1℃, and the average ambient temperature is 24.8℃. In terms of insulation parameters, the peak value of the partial discharge signal is 32pC, the dielectric strength of the insulating oil is 55kV / 2.5mm, and the dissolved gas analysis shows that the hydrogen content is 13ppm, the methane content is 11ppm, the ethane content is 7ppm, the ethylene content is 4ppm, and the acetylene content is 0.3ppm.

[0082] Then, the collected real-time operating parameters were uploaded to the cloud platform, and a digital twin model of the transformer was established. During the data preprocessing process, about 0.3% of abnormal data points were eliminated. The transformer nameplate parameters extracted from the cloud database include rated capacity 180MVA, rated voltage 220 / 110kV, rated current 472 / 945A, etc. In the multi-physics field coupling model framework, the electrical sub-model uses the equivalent circuit method, the thermal sub-model uses the thermal network method, and the insulator model is based on the dielectric response theory. After the coupling matrix is ​​initialized, the electrical-thermal coupling coefficient is 0.72, the electrical-insulation coupling coefficient is 0.63, and the thermal-insulation coupling coefficient is 0.78. After 100 iterations, the deep reinforcement learning algorithm optimized the coupling coefficients to 0.79, 0.68, and 0.83.

[0083] Subsequently, the system evaluated the health status of the transformer. The preset standard parameters extracted from the cloud database include: rated voltage ±5%, rated current ±10%, normal oil temperature range -25℃~105℃, and partial discharge threshold 50pC. The key state quantities calculated by the digital twin model include: impedance 0.054+j0.421Ω, loss 121kW, hot spot temperature 76.8℃, and dielectric strength 61kV / 2.5mm. Compared with the preset standard parameters, the voltage deviation is 0.82%, the current deviation is -2.97%, the oil temperature deviation is 2.6℃, and the partial discharge intensity deviation is -18pC. Based on the health assessment rule base, the health scores of each sub-model are: 93 points for the electrical sub-model, 89 points for the thermal sub-model, and 95 points for the insulator model. The transformer health status index obtained by comprehensive calculation is 92.3.

[0084] Then, the system uses an adaptive multimodal deep learning framework for fault prediction. The convolutional neural network module contains three convolutional layers, each followed by a ReLU activation function, a maximum pooling operation, and a batch normalization process. The long short-term memory network module uses a two-layer LSTM structure, with the first layer containing 128 hidden units and the second layer containing 64 hidden units, and the dropout rate is set to 0.4. In the adaptive fusion module, the weight of the CNN output feature is 0.62, and the weight of the LSTM output feature is 0.38. The three output branches of the multi-task learning network predict the potential fault type, fault probability, and fault development trend respectively.

[0085] The prediction results show that the potential fault types of transformers are ranked from high to low in terms of probability: winding overheating (31%), insulation aging (26%), core overheating (17%), bushing failure (14%), and tap changer failure (12%). The expected deterioration rate of winding overheating is 0.4% / month, and the expected deterioration rate of insulation aging is 0.3% / month.

[0086] Finally, based on the above prediction results, the system formulated a preventive maintenance strategy. Short-term maintenance recommendations include: infrared thermal imaging inspection within 45 days, focusing on winding and core temperature; insulating oil chromatography analysis within 60 days to evaluate insulation condition. Long-term maintenance recommendations include: partial discharge test within 6 months; complete transformer overhaul within 18 months. After review by the expert system, the insulating oil chromatography analysis was brought forward to 30 days, and other recommendations remained unchanged. The final preventive maintenance strategy report is pushed to the authorized terminal device through an encrypted channel.

[0087] In order to evaluate the performance of the present invention, we compared it with the prior art, and the results are shown in Table 1.

[0088] Table 1 Comparison indexes of the present invention and the traditional method

[0089] Performance Indicators The present invention Traditional methods Fault prediction accuracy 91.2％ 84.7％ Forecast lead time 2-5 months 1-2 months False Positive Rate 3.5％ 7.2％ False negative rate 3.1％ 5.8％ Computational efficiency 0.6s / time 1.3s / time

[0090] As can be seen from Table 1, the present invention is superior to the prior art in key indicators such as fault prediction accuracy, prediction lead time, false alarm rate and missed alarm rate. In particular, in terms of prediction lead time, the present invention can detect potential faults 1-3 months earlier than the prior art, providing more sufficient preparation time for preventive maintenance. At the same time, the computational efficiency of the present invention is also better than the prior art, and it can process large amounts of real-time data more quickly to meet the needs of real-time monitoring of substations. These advantages are mainly due to the multi-physics field coupled digital twin model and adaptive multimodal deep learning framework adopted by the present invention, which can capture the operating status and fault characteristics of the transformer more comprehensively and accurately.

[0091] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A digital cloud-based online monitoring and diagnosis method, characterized in that: include, Collecting real-time operating parameters of the transformer, the real-time operating parameters including electrical parameters, thermal parameters and insulation parameters; Uploading the real-time operating parameters to the cloud platform to establish a transformer digital twin model; Utilizing the transformer digital twin model to calculate the deviation between the real-time operating parameters and the preset standard parameters, and generating a transformer health status index; Based on the transformer health status index, an adaptive multimodal deep learning framework is used to predict the transformer health status and obtain a fault prediction result; Based on the fault prediction results, a preventive maintenance strategy is formulated and pushed to the terminal device.

2. The digital cloud-based online monitoring and diagnosis method according to claim 1, characterized in that: The establishment of the transformer digital twin model includes: Extract transformer nameplate parameters, structural characteristics, and material property data from the cloud database; Importing a pre-established transformer multi-physics coupling model framework, wherein the transformer multi-physics coupling model framework includes an electrical sub-model, a thermal sub-model, and an insulator model; Initializing basic parameters of the electrical sub-model, the thermal sub-model, and the insulator model using the nameplate parameters, structural characteristics, and material property data; Constructing a coupling matrix between sub-models, and adaptively optimizing the coupling matrix using a deep reinforcement learning algorithm; The electrical sub-model, thermal sub-model and insulator model are integrated based on the optimized coupling matrix to form an initial transformer digital twin model.

3. The digital cloud-based online monitoring and diagnosis method according to claim 2, characterized in that: The construction of the coupling matrix between sub-models includes: Establish a coupling matrix framework; Fill in the self-influence coefficient of each sub-model in the diagonal elements of the coupling matrix, and fill in the initial cross-influence coefficient in the non-diagonal elements; The time lag factor is introduced to expand the coupling matrix into a four-dimensional tensor; Establish the physical correlation equations between sub-models; Calculating initial values ​​of coupling matrix elements in combination with sub-model parameters and the physical association equations; An adaptive weight adjustment algorithm is designed to dynamically update the coupling matrix element values ​​according to real-time operating parameters.

4. The digital cloud-based online monitoring and diagnosis method according to claim 1, characterized in that: Generating the transformer health status index comprises the following steps: Extracting transformer preset standard parameters from the cloud platform database; Input the real-time operating parameters into the transformer digital twin model and calculate the key state quantities of each sub-model; Compare the key state quantity with the preset standard parameter and calculate various deviation values; Quantify and score each deviation value based on the health assessment rule base to obtain the health score of each sub-model; Based on the health score of each sub-model and the preset weight of the sub-model, the transformer health status index is calculated.

5. The digital cloud-based online monitoring and diagnosis method according to claim 4, characterized in that: The calculation of the key state quantity of each sub-model comprises the following steps: Classify and input the real-time operating parameters into the electrical sub-model, thermal sub-model and insulator model of the transformer digital twin model; Utilizing the electrical sub-model, calculating the impedance and loss of the transformer; Utilizing the thermal sub-model, calculating the hot spot temperature of the transformer; Calculating the dielectric strength of the transformer using the insulator model; Execute data exchange between sub-models and adjust the calculation results of each sub-model according to the influencing factors; Time stamping and data verification of impedance, loss, hot spot temperature, and dielectric strength; The calculation results of each sub-model are integrated to generate a transformer overall state data set containing key state quantities.

6. The digital cloud-based online monitoring and diagnosis method according to claim 1, characterized in that: The adaptive multimodal deep learning framework includes a convolutional neural network module, a long short-term memory network module and an adaptive fusion module.

7. The digital cloud-based online monitoring and diagnosis method according to claim 1, characterized in that: The method of using an adaptive multimodal deep learning framework to predict the health status of a transformer includes the following steps: Extract transformer health status index, real-time operating parameters and historical fault data from the cloud platform to generate a standardized feature vector; Constructing a multimodal input layer, inputting the standardized feature vector into a convolutional neural network module, and inputting the historical fault data into a long short-term memory network module; The convolutional neural network module extracts the time series features of the standardized feature vector, and the long short-term memory network module extracts the long-term dependency features of the historical fault data; The adaptive fusion module calculates the weight coefficients of the output of the convolutional neural network module and the output of the long short-term memory network module based on feature importance analysis; According to the weight coefficient, the output features of the convolutional neural network module and the long short-term memory network module are fused to obtain a fused feature; The fused features are input into the multi-task learning network, and the fault prediction results are generated through three parallel output branches.

8. A digital cloud-based online monitoring and diagnosis system, based on the digital cloud-based online monitoring and diagnosis method according to any one of claims 1 to 7, characterized in that: Also includes, Data acquisition module, used to collect real-time operating parameters of transformers; A model building module, used to upload the real-time operating parameters to a cloud platform to establish a transformer digital twin model; A health assessment module, used to calculate the deviation value between the real-time operating parameters and the preset standard parameters using the transformer digital twin model, and generate a transformer health status index; A fault prediction module is used to predict the health status of the transformer based on the transformer health status index using an adaptive multimodal deep learning framework to obtain a fault prediction result; The maintenance strategy module is used to formulate a preventive maintenance strategy based on the fault prediction result and push it to the terminal device.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the digital cloud-based online monitoring and diagnosis method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the digital cloud-based online monitoring and diagnosis method according to any one of claims 1 to 7 are implemented.

Images