The development of a power transformer fault diagnosis and management system, based on common methods for transformer fault diagnosis, describes the implementation method of the developed 8-structure reasoning mechanism and the main functions of the diagnosis and management system. Field tests show that the structure is reasonable, the reasoning is correct, and it can effectively diagnose internal faults in operating transformers.
Ensuring a safe, high-quality, and economical power supply is crucial for the operation of modern power systems. Safety is paramount, especially for key equipment in the power supply system. Power transformers, due to the long-term effects of various external environmental factors, inevitably age and eventually lose their functional purpose. If these defects are not detected and repaired in time, they will continue to develop, leading to operational accidents. Therefore, it is essential to manage and monitor transformers in operation using appropriate management methods and means. This is a vital task for safe production in the power industry.
Currently, the maintenance strategy for power transformers is based on periodic maintenance. Generally, preventative maintenance is carried out according to a set schedule. However, the cycle of this maintenance is not inherently related to the actual condition of the equipment, as it is not consistent with national regulations and manufacturer recommendations. Therefore, the following situations can occur: maintenance is performed when it is not needed, thus accelerating the aging and damage of the equipment; in urgent situations, the fixed maintenance cycle delays maintenance, causing damage to the equipment and even serious accidents.
All of the above situations will lead to unnecessary repairs.
Therefore, it is necessary to conduct in-depth research on the condition monitoring and fault diagnosis technology of Tunli transformers in order to improve the diagnostic and management capabilities of operators as soon as possible.
For the reasons mentioned above, with the support of the Shanghai Science and Technology Rising Star Program, we developed a power transformer fault diagnosis and management system, and further developed application technologies based on enriching and improving the original expert system.
1. A Brief Description of the Fault Analysis Mechanism of Power Transformers
1.1 Gas Chromatography in Oil: It is believed that the most effective method for early diagnosis of electrical equipment malfunctions is gas chromatography in oil. Under normal circumstances, the insulating materials in oil-filled electrical equipment gradually age and decompose under thermoelectric effects, producing a large portion of various low-molecular-weight hydrocarbons such as 0, 2, and 0.
1. When there is a latent overheating or discharge fault inside the transformer, the rate of gas production will be accelerated. Generally speaking, the gases produced by the decomposition of insulation differ depending on the nature of the fault; and for faults of the same nature, the amount of gas produced also differs due to differences in degree.
The composition and content of fault gases are closely related to the type and severity of the fault. There are various methods for interpreting gas composition data. The most commonly used diagnostic methods include static characteristic gas methods, ratio methods, and comparison methods, with the Rogers method being the most frequently used. my country uses similar ratio methods such as the David angle method and dynamic gas generation rate methods, including absolute gas generation rate methods and relative gas generation rate methods. (This is a special manuscript, received on [date], funded by the Shanghai Science and Technology Rising Star Program, project number: 沪科技,4,2612). Insulation Prevention Test Diagnosis Method: After a period of normal aging, especially after a fault occurs, the DC resistance, insulation resistance, absorption ratio, dielectric loss tangent, and DC leakage current of the windings of a transformer will change, showing significant differences from the factory specifications. Therefore, methods such as comparing the current test data with the test procedure data, comparing with the previous test data, and comparing with data from different phases can be used to determine the nature, degree, and location of internal faults in the transformer, thus providing a basis for the comprehensive diagnosis of the transformer.
1.3 Winding Deformation Test Diagnosis During long-distance transportation, transformers may experience impacts or short-circuit current surges, potentially causing winding deformation or displacement. Winding deformation or displacement alters the insulation distance and damages the solid insulation, potentially leading to sudden insulation failures. Furthermore, a decline in the winding's mechanical properties may result in damage under the influence of electrodynamic forces.
Currently, the frequency response method is the main method for detecting whether transformer windings are deformed. The method involves comparing the measured frequency response curve of the windings with the pre-programmed frequency response curve of the transformer under normal conditions at the factory. Based on whether there are changes, it can be determined whether the windings have been subjected to impact and the degree of deformation. Initially, this is indicated by the following: a lifting cover inspection and the implementation of protective measures such as capping.
1.4 Aging Assessment Test Furfural is the final product of insulating paper after a series of complex chemical reactions including thermal degradation, chemical degradation, and water degradation. At operating temperatures, there is a relatively accurate correlation between the furfural content dissolved in the oil and the uniform viscosity of the insulating paper. Therefore, the furfural content in the oil can reflect the overall deterioration degree of paper insulation, including the deep insulating paper. Using high-performance liquid chromatography (HPLC) to detect the furfural content in the insulating oil to determine the degree of insulation paper deterioration is a useful method for assessing the aging state of transformers. If necessary, the degree of polymerization and tensile strength of the paper sample can also be measured. The main method is to estimate the aging degree of the insulating paper by analyzing the furfural content and to infer the remaining life of the insulation by calculating the aldehyde gas production rate.
Of course, the condition monitoring and fault diagnosis methods for power transformers also include various conventional methods such as external characteristic inspection, temperature, oil level, sound, current operating status, insulating oil characteristic inspection, acid value, resistivity, water content, surface tension, dielectric loss and breakdown voltage, partial discharge test, and oil flow charging test.
2 Power Transformer Fault Diagnosis and Management System
2.1 Transformer Fault Diagnosis Expert System 2.1.1 Expert System Structure The transformer fault diagnosis expert system consists of 7 parts: knowledge base, database, inference engine, knowledge acquisition and maintenance, data interpretation mechanism, and human-computer interface. 1. Its structure 1.
The knowledge base is a crucial component of an expert system, storing domain knowledge obtained from transformer experts, including classic common theoretical knowledge and procedures, as well as field-specific knowledge and experience from field experts. A complete expert system must enrich the necessary domain knowledge as much as possible. Since the power transformer fault diagnosis and management system Ding03 adopts a modular structure, it facilitates the expansion and maintenance of the knowledge base.
The database can be divided into three parts: module data (transformer characteristic parameter data, transformer benchmark data), and module database (context tree), which stores the current known work of the expert system, as well as facts provided by the user and intermediate conclusions derived from reasoning. It is dynamically updated during the diagnostic process, thus enabling longitudinal comparisons between collected data. The module database mainly contains gas analysis database, insulating oil characteristic test data, aging test database, insulation preventive test data, and winding test database. Its main functions include storing transformer characteristic data, performing horizontal comparisons and trend analysis of transformer characteristic data to improve the operation, maintenance, and diagnostic capabilities of operators. The transformer characteristic parameter database enables computer management of transformer information. The transformer benchmark database enables insulation preventive test data.
The inference engine is the core component of an expert system. It intelligently deduces conclusions based on acquired knowledge. Therefore, it employs a reverse reasoning mechanism and a depth-first search strategy for reasoning. Knowledge acquisition and maintenance are primarily achieved through communication between transformer diagnostic experts and knowledge engineers, utilizing a modular structure to enrich and complete the knowledge base. In addition to enriching the diagnostic module, in the maintenance of the oil chromatography analysis knowledge base, we also established an artificial neural network based on the characteristic gas method and ratio method, which greatly improves the expert system's self-learning ability.
The main tasks of data management are to maintain the established database, including database creation, modification, addition, deletion, retrieval, classification, sorting, formatting, printing, and comparison of various experimental data in both horizontal and vertical directions, curve simulation analysis, and classification and archiving of historical data.
The function of the explanation mechanism is to explain the final result of the system's reasoning process, provide suggestions to answer users' questions, and facilitate the training of on-site personnel and the transfer of expert experience.
The human-machine interface (HMI) converts user input into a simplified form within the system. This simplified form is then passed to the appropriate processing module. Our newly designed HMI uses a pre-compiled architecture, resulting in a user-friendly interface. Users simply click the mouse and input relevant information and basic data, greatly simplifying operation.
2.1.2 Knowledge Base Composition The knowledge base is the core of the expert system. Based on the national standards for various practical applications and the actual field conditions, we divide the entire knowledge base into two parts: a latent fault diagnosis module and a preventive test diagnosis module. The latent fault diagnosis module is further divided into a gas chromatography analysis module, a gas generation rate analysis module, an external inspection module (including insulation oil special test module), a connection oil special test module, an aging test module, and a comprehensive diagnosis module. It is mainly used to determine whether the transformer has latent faults, their nature and extent, and whether it is necessary to shut down the system. The preventive test diagnosis module is further divided into a DC resistance test module, an insulation resistance test module, a leakage current test module, a winding and bushing dielectric loss test module, a partial discharge test module, a spectrum analysis module, and a comprehensive analysis module. It is mainly used to determine the nature, extent, and location of faults. Because the knowledge base adopts a modular structure, each module has good independence, which is beneficial for program maintenance and expansion. The following only uses the gas chromatography analysis module as an example to introduce its knowledge base composition.
Based on the knowledge of the gas chromatography field, the sample can be divided into 33 parts (4), namely gas 12, gas ratio method or David's angle method.
Considering that the ratio method and the large angle method are more accurate, this system prioritizes the ratio method and the large angle method for diagnosis, while using the characteristic gas method for parallel auxiliary analysis. Gas production rate diagnosis includes static and dynamic analysis. For relative gas production rate analysis, the system saves basic factory data and historical test data, thus enabling comparison of data collected from different times. When the relative gas production rate exceeds a certain threshold each month, attention should be paid. The system also provides a curve description of dynamic data. This significantly improves the intuitiveness, effectiveness, and accuracy of the diagnosis.
2.1.3 Knowledge and Fuzzy Processing Appropriate knowledge is the guarantee for the accurate and efficient operation of an expert system. The widely used production rule method is adopted for small transformer fault diagnosis. It has strong modularity, good clarity, easy understanding, and easy modification and expansion. However, transformer fault diagnosis involves many uncertainties. Factors such as capacity, voltage, insulation level, cooling method, load characteristics, operating time, and insulation structure can all affect the diagnostic conclusion, and the mathematical correlation of these factors is currently unclear. Therefore, using precise boundary division rules to diagnose various critical faults in modules such as gas chromatography analysis is obviously unscientific. To address this, we introduce fuzzy processing technology. For example, in the ratio method, the original precise division is fuzzified. When 1/4 = 1, the code is 0. After fuzzy processing, the code remains the same, but a membership degree is added. Thus, each code can be encoded using two elements to determine the membership degree. Each code is not absolutely equal to 2, but rather equal to 0 and 2 to some extent. During system reasoning, several rules may simultaneously satisfy the preconditions. In this case, the rule with the largest membership degree is triggered.
2.2 The power transformer information management system is an important component of the expert system. It enables the expert system to perform fault diagnosis based on dynamic data, and also realizes computer management of transformer information, characteristic parameters, test data, diagnostic conclusions, etc., including a transformer characteristic data management system, a factory test data management system, a historical test database management system, and a transformer historical diagnostic conclusion management system. Each management system has functions such as adding, deleting, modifying, querying, sorting, browsing, printing, and analyzing, and a user-friendly human-computer interface. 3.
2.3 System Implementation The system is developed based on the 100898 Chinese operating platform, utilizing the latest intelligent expert reasoning language, 15.0, and database management language. Because 15.0 has strong logical reasoning capabilities and unique deductive reasoning abilities, expert knowledge is easy to access and maintain, and it has a good interpretation mechanism and user interface, making it the most successful computer language in expert systems. The core development of this expert system adopts the latest intelligent expert reasoning language, 15.0. Considering the need for on-site operation departments to manage various historical test data of power transformers by computer, we selected 15.0 in the dynamic database to complete the development of functions such as statistical management, curve description, and printing of various historical and dynamic data. We also adopted 15.0 language with artificial neural networks for knowledge acquisition and developed it, and integrated it with the expert system. 3. Diagnostic Example A certain factory's main transformer has a rated capacity of 63,000 cubic meters and a rated voltage of 0.3510.5 kWh. Its test data is as follows: 1. Gas chromatography analysis test: H0=217.
Resistance 1.5, water volume 33.5 kJ, clothing bending force 20, breakdown voltage 39.5 sand, dielectric loss 1.75; 3. Check the sound abnormality, there is a local rustling sound, the thermometer and oil level gauge are normal; 4. Preventive electrical test, high voltage side, medium voltage side, low voltage insulation resistance 15560, 20040, old, 2003.03020, piece, leakage current is 8, 12, 6, dielectric loss is normal value, transformer internal fault; ratio code is 102, transformer fault nature is high energy discharge.
2. External inspection analysis suggests that Ju Shengxiong often judges the problem to be an internal abnormality of the transformer.
3. Insulation performance test results: The insulation performance is good, and the product can continue to be used.
4. Based on the comprehensive analysis of latent fault diagnosis, it is recommended to immediately shut down the machine and conduct an internal inspection.
5. The preventive test analysis indicates abnormal insulation preventive test results, specifically abnormal DC resistance, and poor internal welding of the transformer. The abnormal DC resistance is also noted. Operators are advised to pay attention to the following five aspects: 1. Defective tap changer, mainly due to unclean tap changer, plating peeling, insufficient spring pressure, etc.; 2. Poor contact at the lead wire and coil welding points; 3. Poor connection and broken wires; 4. Inadequate contact between the lead rod and lead wire connection; 5. Inconsistent conductor type.
The on-site inspection after shutdown concluded that there was a transformer winding fault and high-energy contact.
4. Conclusion
The design and development of a power transformer fault diagnosis and management system is a highly complex task. It requires not only a thorough understanding of transformer operation, maintenance, electrical testing, and repair, but also mastery of key aspects of expert systems, the construction of knowledge bases, and the integration of technology with expert systems in transformer fault diagnosis. Furthermore, the analysis and handling of uncertainties in the fault diagnosis process necessitates knowledge of fuzzy mathematics. It is certain that by integrating key technologies into the power transformer fault diagnosis process, the diagnostic conclusions of the expert system will be reliable and accurate, which will significantly improve the level of transformer operation and maintenance in the power sector.
