Power Equipment Fault Diagnosis Method and System

Abstract

An embodiment of the present invention provides a power equipment fault diagnosis method and system, belonging to the technical field of power equipment fault diagnosis. The method includes: obtaining decomposition gas parameters of each decomposed gas in the used oil and gas of the power equipment; filling the decomposition gas parameters into a preset matrix, and using the filled matrix as an input parameter to execute a fault recognition model to output a fault recognition result; outputting and reporting the fault recognition result. The solution of the present invention takes each decomposed gas as a neuron, trains by constructing a lightweight deep neural network model, and uses small sample data training technology to obtain an inference model, which is deployed to an intelligent fusion terminal. The edge computing function of the fusion terminal is used to dynamically identify the types of power equipment faults, so as to realize intelligent, Internet of Things-based and online monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment fault diagnosis, and particularly to a power equipment fault diagnosis method and a power equipment fault diagnosis system. Background Art

[0002] When evaluating the performance of power equipment, reliability is one of the important performance indicators. In the power grid system, there are various power equipment, and the reliability of various power equipment is related to the reliability of the entire power grid system. Among them, transformers and gas insulated switchgear (GIS) equipment are the most important, and their operating states have an important impact on system safety. With the continuous improvement of the requirements for the operation and maintenance of transformers and GIS equipment, the research work on on-line fault diagnosis technology for equipment has received more and more attention.

[0003] Currently, there are methods for judging faults based on the decomposed gases of the used oil and gas of power equipment. However, most of these methods for judging faults based on decomposed gases are judged manually or obtain data from a database through software and judge the cause of the fault according to the detection standard. However, the degree of intelligence and Internet of Things is still insufficient, resulting in a reduction in the efficiency of equipment monitoring. In view of the above problems, a new power equipment fault diagnosis method needs to be created. Summary of the Invention

[0004] The purpose of the embodiments of the present invention is to provide a power equipment fault diagnosis method and a power equipment fault diagnosis system to at least solve the problem of insufficient intelligence and Internet of Things degree of the current diagnosis method.

[0005] To achieve the above purpose, in the first aspect of the present invention, a power equipment fault diagnosis method is provided, the method includes: obtaining the decomposition gas parameters of each decomposed gas in the used oil and gas of the power equipment; filling the decomposition gas parameters into a preset matrix, and using the filled matrix as an input parameter to execute a fault recognition model to output a fault recognition result; outputting and reporting the fault recognition result.

[0006] Optionally, the power equipment is an oil-immersed transformer or a gas insulated switchgear.

[0007] Optionally, the decomposed gases of the oil-immersed transformer include at least one of hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, and carbon dioxide; the decomposed gases of the gas insulated switchgear include at least one of a mixture of sulfur dioxide and sulfuryl fluoride, hydrogen sulfide, carbon monoxide, and hydrogen fluoride.

[0008] Optionally, the decomposition gas parameter is the volume fraction of the decomposed gas.

[0009] Optionally, the method further includes: constructing a fault identification model, including: simulating multiple sets of decomposition gas parameters according to the decomposition rules of the oil and gas used by the power equipment; filling each set of decomposition gas parameters into a preset matrix, and using the obtained multiple filled matrices as training samples; performing model training on the ResNet18 deep neural network based on the training samples to obtain an initial model; extracting the historical operation parameters of the power equipment, and performing re-simulation and correction on the initial model based on the historical operation parameters to obtain a fault identification model.

[0010] Optionally, the preset matrix is an 8*8 blank matrix; the step of filling each set of decomposition gas parameters into the preset matrix includes: filling data in the 4*4 matrix at the center position of the 8*8 blank matrix, and filling 0 in the remaining unfilled areas of the 8*8 blank matrix.

[0011] Optionally, when the power equipment is an oil-immersed transformer, the fault identification results include: normal, high-energy discharge, low-energy discharge, partial discharge, high-temperature overheating, medium-low temperature overheating.

[0012] Optionally, the power equipment is a gas-insulated switchgear, and the fault identification results include: normal, arc discharge, spark discharge, partial discharge.

[0013] Optionally, the step of outputting the fault identification result includes: outputting an alarm message while outputting the fault identification result.

[0014] A second aspect of the present invention provides a power equipment fault diagnosis system, including: a collection unit for acquiring the decomposition gas parameters of each decomposed gas in the oil and gas used by the power equipment; a processing unit for filling the decomposition gas parameters into a preset matrix, and using the filled matrix as an input parameter to execute a preset network fault identification model to obtain a fault identification result; an output unit for outputting and reporting the fault identification result.

[0015] Optionally, the processing unit is further configured to: construct a fault identification model, including: simulating multiple sets of decomposition gas parameters according to the decomposition rules of the oil and gas used by the power equipment; filling each set of decomposition gas parameters into a preset matrix, and using the obtained multiple filled matrices as training samples; performing model training on the ResNet18 deep neural network based on the training samples to obtain an initial model; extracting the historical operation parameters of the power equipment, and performing re-simulation and correction on the initial model based on the historical operation parameters to obtain a fault identification model.

[0016] Optionally, the output unit is further configured to output an alarm message while outputting the fault identification result; the system further includes: an alarm unit for executing the alarm message.

[0017] On the other hand, the present invention provides a computer-readable storage medium, on which instructions are stored, and when the instructions run on a computer, the computer is caused to execute the above-mentioned power equipment fault diagnosis method.

[0018] Through the above technical solution, each decomposed gas in the present invention solution is used as a neuron, trained by constructing a lightweight deep neural network model, and a reasoning model is obtained by using the small sample data training technology, and then deployed to the intelligent fusion terminal. The edge computing function of the fusion terminal is used to dynamically identify the types of power equipment faults, so as to realize intelligent, Internet of Things-based, and online monitoring.

[0019] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent specific embodiments section. Description of the Drawings

[0020] The drawings are used to provide a further understanding of the embodiments of the present invention, and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the embodiments of the present invention, but do not constitute a limitation to the embodiments of the present invention. In the drawings:

[0021] Figure 1 is a flowchart of the steps of the power equipment fault diagnosis method provided by an embodiment of the present invention;

[0022] Figure 2 is a schematic diagram of the filling of the transformer oil decomposition gas data matrix provided by an embodiment of the present invention;

[0023] Figure 3 is a schematic diagram of the filling of the SF6 decomposition gas data matrix provided by an embodiment of the present invention;

[0024] Figure 4 is a schematic diagram of the training process of the transformer oil decomposition gas data provided by an embodiment of the present invention;

[0025] Figure 5 is a schematic diagram of the training process of the SF6 decomposition gas data provided by an embodiment of the present invention;

[0026] Figure 6 is a system structure diagram of the power equipment fault diagnosis system provided by an embodiment of the present invention. Detailed Description of the Embodiments

[0027] The following will describe in detail the specific embodiments of the present invention with reference to the drawings. It should be understood that the specific embodiments described herein are only for explaining and illustrating the present invention, and are not used to limit the present invention.

[0028] When evaluating the performance of power equipment, reliability is one of the important performance indicators. In the power grid system, there are various power equipment, and the reliability of various power equipment is related to the reliability of the entire power grid system. Among them, transformers and Gas Insulated Switchgear (GIS) equipment are the most important, and their operating states have an important impact on system safety. With the continuous improvement of the requirements for the operation and maintenance of transformers and GIS equipment, the research work on on-line fault diagnosis technology for equipment has received more and more attention. In recent years, with the rapid development of on-line monitoring technology for power transformers and GIS equipment, coupled with computer technology and communication technology, the detection data of power transformers and GIS equipment can be processed and transmitted in a timely manner, and real-time dynamic operating state data can be obtained, enabling on-line monitoring technology to be successfully applied to actual projects. Due to certain limitations of the detection technology and the complexity of internal faults in power transformers and GIS equipment, the reliability and stability of the current on-line monitoring systems in application are still insufficient. The evaluation results of on-line monitoring data for substation equipment directly affect the real-time operating state of the power grid. Through effective and reliable monitoring methods, latent faults inside the transformer can be detected in a timely manner and condition-based maintenance can be achieved to ensure safe and reliable operation in the substation.

[0029] The application requirements for gas monitoring of substation equipment mainly come from two aspects: dissolved gas monitoring in transformer oil and detection of gas decomposition products in GIS cabinets. It is judged according to the content of the detected components of the insulating oil decomposition gas of oil-immersed transformers and reactors, and the volume fraction of the gas components decomposed from sulfur hexafluoride (SF6) gas in GIS equipment.

[0030] For transformer equipment, there are currently some basic theories for judging transformer faults based on gas concentration, namely the gas production rate method, the three-ratio method, and the David triangle method. The main idea is that the analysis results can be obtained by injecting the seven gases of carbon monoxide, carbon dioxide, methane, ethane, acetylene, ethylene, and hydrogen in the insulating oil at one time; according to the mutual dependence relationship between the relative concentration of the gas components generated by the cracking of the insulating oil and insulating gas in the oil-filled electrical equipment under faults and temperature, two gases with similar solubility diffusion coefficients are selected from the five characteristic gases to form three pairs of ratios, which are represented by different codes.

[0031] For GIS equipment filled with SF6 (sulfur hexafluoride) gas, discharge faults are the main reasons for the decomposition of SF6 gas. The decomposition characteristics of SF6 gas under different discharge conditions, and there are mainly three forms of discharge: arc discharge, spark discharge, corona or partial discharge.

[0032] Existing methods for fault judgment based on decomposed gases mostly rely on manual judgment or obtain data from a database through software, and judge the cause of the fault according to the detection standard. However, the degree of intelligence and Internet of Things is still insufficient, resulting in a reduction in the efficiency of equipment monitoring.

[0033] In view of the above problems, with the increasing maturity of artificial intelligence technology, deep learning has shown excellent performance in various fields such as image, speech, and natural language processing. According to the specific structure of the decomposed gas content and volume fraction data of the insulating oil or insulating gas of power equipment, this invention patent uses the deep neural network model of deep learning and combines the power Internet of Things technology to propose a power equipment fault intelligent diagnosis method and system based on an intelligent fusion terminal. This method takes each decomposed gas as a neuron, trains by constructing a lightweight deep neural network model, and uses the small sample data training technology to obtain an inference model, which is deployed to the intelligent fusion terminal. The edge computing function of the fusion terminal is used to dynamically identify the types of power equipment faults, so as to realize intelligent, Internet of Things-based, and online monitoring.

[0034] Figure 6 It is the system structure diagram of the power equipment fault diagnosis system provided by an embodiment of the present invention. As Figure 6 shown, an embodiment of the present invention provides a power equipment fault diagnosis system, which includes: a collection unit for obtaining the decomposed gas parameters of each decomposed gas in the used oil and gas of the power equipment; a processing unit for using the decomposed gas parameters as input parameters to execute a preset network fault identification model to obtain a fault identification result; an output unit for outputting and reporting the fault identification result.

[0035] Preferably, the processing unit is further used for: constructing a fault identification model, including: simulating multiple groups of decomposed gas parameters according to the decomposition rules of the used oil and gas of the power equipment; filling each group of decomposed gas parameters into a preset matrix, and using the obtained multiple filled matrices as training samples; performing model training on the training samples in the ResNet18 deep neural network to obtain an initial model; extracting the historical operation parameters of the power equipment, and performing re-simulation and correction on the initial model based on the historical operation parameters to obtain a fault identification model.

[0036] Preferably, the output unit is further used for outputting an alarm message while outputting the fault identification result; the system further includes: an alarm unit for executing the alarm message.

[0037] Figure 1 It is the method flow chart of the power equipment fault diagnosis method provided by an embodiment of the present invention. As Figure 1 shown, an embodiment of the present invention provides a power equipment fault diagnosis method, which includes:

[0038] Step S10: Obtain the decomposition gas parameters of each decomposed gas in the used oil and gas of the power equipment.

[0039] Specifically, as described above, the method of the present invention is based on the decomposition gas parameters of the used oil and gas of the power equipment for fault reasoning. Different fault conditions will cause the oil and gas to decompose in different directions and at different rates. For example, in a transformer, when there is a high-temperature overheating fault in the transformer, the temperature of its used oil and gas will inevitably increase, and the corresponding gas decomposition rate will also increase, and the volume fraction of the decomposed gas may also change. It is precisely based on this corresponding linkage relationship between the fault state and the specific fault cause that the specific fault cause can be inferred based on the manifested fault state. Therefore, as long as the volume parameters of the decomposed gas mixed in the current oil and gas are accurately known, the fault cause can be inferred.

[0040] Based on the above, first, the acquisition unit set at the power equipment end collects the decomposition gas parameters, which are the volume parameters corresponding to each decomposed gas. For the used oil and gas of different power equipment, different sensor modules are selected correspondingly to ensure that the target gas can be accurately collected. For example, for an oil-immersed transformer, its decomposed gases include: H2 (hydrogen), CH4 (methane), C2H6 (ethane), C2H4 (ethylene), C2H2 (acetylene), CO (carbon monoxide), CO2 (carbon dioxide). And the decomposed gases of GIS equipment include: H2S (hydrogen sulfide), SO2 + SOF2 (a mixture of sulfur dioxide and sulfuryl fluoride), CO (carbon monoxide), HF (hydrogen fluoride). Only two relatively typical power equipment decomposed gases are exemplified in the specific implementation manner, but the solution of the present invention is not limited to these two power equipment. As long as there is used oil and gas and there is an associated relationship between the oil and gas decomposed gas and the fault cause, the solution of the present invention can be implemented.

[0041] Based on the target gases preset for the corresponding power equipment, these gas parameters are collected and sent to the processing unit after the collection is completed. The processing unit synchronizes and integrates the gas parameter information uploaded by these sensors to ensure that the data used for subsequent training is the data collected at the same moment, avoiding the situation where the gas components are different from the actual situation due to data collected at different times.

[0042] Step S20: Use the decomposition gas parameters as input parameters, execute the preset network fault recognition model, and output the fault recognition result.

[0043] Specifically, the solution of the present invention is to pre-train a network fault identification model, and then in the subsequent process, based on the real-time collected data, use the pre-trained model for fault reasoning to obtain the fault reasoning result in real time. Therefore, before processing the current gas parameters, it is first necessary to construct a fault identification model.

[0044] First, through basic theoretical knowledge, dynamic simulation of oil and gas decomposition gases in different fault states is carried out to obtain multiple simulation results, such as training 10,000 simulation results. These simulation results are used as training samples for model training. To improve the efficient extraction of data from the training samples, preferably, before the model training of the present invention, the format of the sample data is first prepared. From the perspective of probability theory, all gas contents obtained by one-time decomposition are considered as a probability whole. Therefore, different from the traditional method, in data preprocessing, such as Figure 2 and Figure 3 , the gas component contents are arranged in the format of a 4*4 matrix in sequence, and the unfilled places are filled with 0. In addition, to enable the constructed neural network model to process more precisely, the dimension of the matrix is expanded, that is, 0 elements are filled in 2 rows and 2 columns around the matrix. Finally, the data format becomes an 8*8 matrix form.

[0045] Then, a ResNet18 deep neural network is constructed, which is a lightweight deep neural network model for data. Among them, 18 specifies 18 layers with weights, including convolutional layers and fully connected layers, excluding pooling layers and BN layers. ResNet introduces a residual network structure, and using this structure can avoid the problem of model performance degradation. The deeper the network, the more information we can obtain and the richer the features are. ResNet solves the problems brought by deepening the network, so that the network can be deepened to a deeper level, that is to say, we can utilize more information. By training the fault identification model through the ResNet18 deep neural network, both the training efficiency and the training accuracy can be greatly guaranteed.

[0046] In addition to obtaining training samples, preferably, test samples will also be obtained. To ensure that the trained model conforms to the actual situation, the historical operation parameters of power equipment are selected as verification samples. The historical operation parameters include the decomposition gas parameters corresponding to the power equipment when a fault occurs during the historical operation process. By verifying with this part of the data, the trained model can be made to fit the actual situation of the power equipment in actual use. For example, the number of test data obtained from the power equipment factory is 10,000, and 1,000 pieces of on-site real data are taken.

[0047] When performing model training, based on the training samples in the ResNet18 deep neural network, under the Pytorch software, first train with the dynamically generated data to obtain an intermediate training model. Then, using the method of transfer learning, based on the intermediate training model, train with the test data of the power equipment factory. Finally, train with multiple (e.g., 800) pieces of real data available on-site to ultimately form an inference model, and use the remaining real data for testing.

[0048] In the embodiment of the present invention, the corresponding fault location model needs to finally convert the gas parameters into multiple categories, and these categories are the fault cause categories corresponding to the power equipment. Based on the common fault causes of power equipment in reality, determine the corresponding number of categories. That is, regardless of what gas parameters are, they can only be mapped to these preset categories in the end. Preferably, when the power equipment is an oil-immersed transformer, the fault recognition results include: normal, high-energy discharge, low-energy discharge, partial discharge, high-temperature overheating, medium-low temperature overheating. When the power equipment is a GIS device, the fault recognition results include: normal, arc discharge, spark discharge, partial discharge. That is, when the fault recognition model acts on the transformer, it finally needs to be processed into 6 categories, and when the fault recognition model acts on the GIS device, it finally needs to be processed into 4 categories. Based on this, as Figure 4 and Figure 5 , after completing the model training, a Softmax regression model is also needed to handle the multi-classification problem. The softmax activation function is used in the last fully connected layer to adjust the input to the number of categories 6 and 4.

[0049] After completing the model training, model optimization is also required. That is, if the recognition accuracy reaches the expectation, terminal deployment can be performed. If it does not reach the expectation, optimization is carried out and retraining is performed until the expectation is reached.

[0050] When the model meets the expectation and can be put into use, it is necessary to process the currently collected gas parameter information. Similar to model training, first fill the collected data with a preset matrix to make it suitable as the input parameter for the preset fault recognition model. Then, based on the training rules of the model, determine the fault cause corresponding to the current gas parameter information.

[0051] Step S30: Output the fault recognition result.

[0052] Specifically, after identifying the cause of the fault, the corresponding fault cause identification result is directly output. If the identification result is normal, it indicates that the current operating state of the power equipment is stable, and the original operating state is maintained and continued. If the identification result is not normal, no matter which other type of fault cause it is, it means that the corresponding power equipment is in an abnormal working state. To avoid evolving into a more serious fault, while pushing the fault identification result, an alarm message is also pushed to remind relevant personnel to intervene in a timely manner.

[0053] In another possible implementation, if a certain fault cause has a particularly huge impact on the power equipment, when it is identified that there is such a fault cause, in addition to outputting an alarm message, the processing unit will directly shut down the power equipment to avoid continuous operation under extremely high-risk conditions and ensure the operation stability of the power system.

[0054] In the embodiment of the present invention, aiming at the problems of low Internet of Things and intelligence levels and poor scalability in the traditional method of post-fault diagnosis, a deep learning method is adopted. By constructing and training a lightweight deep neural network model based on the ResNet network, the function of automatically identifying faults in the component content of transformer oil and sulfur hexafluoride decomposition gas is realized. This method can perform online identification, and the accuracy can reach the level of general experts, and the identification types can be dynamically adjusted.

[0055] The embodiment of the present invention also provides a computer-readable storage medium, on which instructions are stored, and when running on a computer, the computer is made to execute the above-mentioned power equipment fault diagnosis method.

[0056] Those skilled in the art can understand that all or part of the steps in the method for implementing the above embodiments can be completed by instructing relevant hardware through a program. The program is stored in a storage medium, including several instructions to make a single-chip microcomputer, a chip or a processor execute all or part of the steps of the method described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks or optical discs that can store program codes.

[0057] The optional embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention. Additionally, it should be noted that in the above specific embodiments, the various specific technical features described can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not separately describe various possible combination methods.

[0058] In addition, any combination can be made among various different embodiments of the present invention as long as it does not violate the idea of the embodiments of the present invention, and it should also be regarded as the content disclosed by the embodiments of the present invention.

Claims

1. A method for diagnosing faults in power equipment, characterized in that The method includes: Obtaining the decomposition gas parameters of each decomposed gas in the used oil and gas of the power equipment; Filling the decomposition gas parameters into a preset matrix, and using the filled matrix as an input parameter to execute a fault identification model to output a fault identification result; where The construction rule of the fault identification model includes: According to the decomposition rule of the used oil and gas of the power equipment, simulating multiple groups of decomposition gas parameters; filling each group of decomposition gas parameters into the preset matrix, and using the obtained multiple filled matrices as training samples; performing model training on the training samples in the ResNet18 deep neural network to obtain an initial model; extracting the historical operation parameters of the power equipment, and performing re-simulation and correction of the initial model based on the historical operation parameters to obtain the fault identification model; Outputting and reporting the fault identification result.

2. The method according to claim 1, characterized in that The power equipment is an oil-immersed transformer or a gas-insulated switchgear.

3. The method according to claim 2, characterized in that The decomposed gases of the oil-immersed transformer include at least one of hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, and carbon dioxide; The decomposed gases of the gas-insulated switchgear include at least one of a mixture of sulfur dioxide and sulfuryl fluoride, hydrogen sulfide, carbon monoxide, and hydrogen fluoride.

4. The method according to claim 1, characterized in that The decomposition gas parameter is the volume fraction of the decomposed gas.

5. The method according to claim 1, characterized in that The preset matrix is an 8*8 blank matrix; The filling of each group of decomposition gas parameters into the preset matrix includes: Filling data in the 4*4 matrix at the center position of the 8*8 blank matrix, and filling 0 in the remaining unfilled areas of the 8*8 blank matrix.

6. The method according to claim 2, characterized in that When the power equipment is an oil-immersed transformer, the fault identification result includes: Normal, high-energy discharge, low-energy discharge, partial discharge, high-temperature overheating, medium-low temperature overheating.

7. The method according to claim 2, characterized in that When the power equipment is a gas-insulated switchgear, the fault identification result includes: normal, arc discharge, spark discharge, partial discharge.

8. The method according to claim 1, characterized in that Outputting the fault identification result includes: outputting an alarm message while outputting the fault identification result.

9. A power equipment fault diagnosis system, characterized in that The system includes: A collection unit for obtaining the decomposition gas parameters of each decomposed gas in the used oil and gas of the power equipment; A processing unit for filling the decomposition gas parameters into a preset matrix, and using the filled matrix as an input parameter to execute a fault identification model to obtain a fault identification result; The processing unit is further used for: constructing the fault identification model, including: According to the decomposition rule of the used oil and gas of the power equipment, simulating multiple groups of decomposition gas parameters; filling each group of decomposition gas parameters into the preset matrix, and using the obtained multiple filled matrices as training samples; performing model training on the training samples in the ResNet18 deep neural network to obtain an initial model; extracting the historical operation parameters of the power equipment, and performing re-simulation and correction of the initial model based on the historical operation parameters to obtain the fault identification model An output unit for outputting and reporting the fault identification result.

10. The system according to claim 9, characterized in that The output unit is further used for outputting an alarm message while outputting the fault identification result; The system further includes: an alarm unit for executing the alarm message.

11. A computer-readable storage medium, on which instructions are stored, and when the instructions are run on a computer, the computer is caused to execute the power equipment fault diagnosis method according to any one of claims 1-8.

Images