Research On Intelligent Identification And Themodynamics-Based Reaction Enthalpy Diagnosis Methodology For Transformer Compound Faults

As the key equipment for voltage regulation, power transmission and distribution, the power transformers play an indispensible role to guarantee safe, reliable and stable operation of the power systems. Once an accident happens, it may cause huge economic losses, or even results in severe social impacts. Therefore, incisive understanding of the transformers’ fault mechanism as well as effective diagnosis and evaluation of the early internal potential failures presents both theoretical and practical significance in reducing failure risk and ensuring reliable operation of the power grids. Presently there are a variety of fault diagnosis methodologies based on Dissolved Gas Analysis (DGA), however, a power transformer itself is basically a very complex system and the occurrence of an internal insulation fault is often the composition of a variety of fault types with uncertain and fuzzy features among the fault information.In addition, the physical and chemical process of the insulation deterioration and the mechanism of failure gases generation are still unclear so far, which renders as one of the principal reasons to account for why the available diagnosis methods are so restricted as only to identify a few kinds of fault types without giving any detailed assessment of the faults severity.At practice inside the transformer latent fault has a development process, namely between different types of failure is not a mutation, the severity of the same fault type is different also, and at present based on the fault diagnosis of transformer oil chromatogram in the level of fault is divided into several categories, the physical and chemical process and fault gas might not clear and incomplete insulation degradation mechanism. Chemical kinetics theory, the chemical reaction is essentially the old chemical bonds breaking and new chemical bond formation, this process requires certain energy destroy chemical bonds in the reactants, the energy for the transformer insulation cracking gas mainly comes from the discharge or overheating fault. Different bond rupture need different energy, and the cracking energy density, the greater the resulting hydrocarbon gas saturation should be higher. However, the theoretical foundation of the existing diagnosis methods are mainly based on the theoretical calculation of thermodynamic equilibrium points pressure relationship, different gas composition and temperature and gas characteristic is put forward according to experience, which ignores the dynamic changes of the chemical reaction process, also did not consider the chemical reaction of energy change. Therefore, to further estimate the severity of the fault in fault diagnosis, still need to in-depth study of the dynamic process of insulation cracking, explore the new characteristics of fault severity characterization, in order to more accurately assess the transformer running state.On the basis of comprehensive understanding of the related technical background on intelligent faults diagnosis and condition assessment for power transformers, diagnosis methodologies for power transformers’ compound faults are intensively studied in this thesis, including improvement of the intelligent identification method for compound faults, clarification of the relationship between microcosmic changes and macroscopic properties of the insulation based on thermal dynamic theories, as well as exploration of new indicators to characterize the failure severity. The innovative achievements obtained in this thesis are concluded as follows:As different diagnostic methods have different oil chromatographic characteristics and present sensible difference in reliability when it comes to diagnose different fault types, a double indexes-based fault diagnosis reliability evaluation method is hereby proposed. With the traditional accuracy index being included, a sensitivity index is introduced to represent the degree of missed diagnosis, which method can achieve a more comprehensive and quantitative assessment for the oil chromatogram diagnosis methods. Regarding the modified three-ratio method, which is so far the most reliable diagnosis method, still remains lower reliability in diagnosing the fault types such as low temperature overheating, spark discharges, discharges with overheating, etc. To address the above mentioned issues, this thesis proposed 01* type code combination, hydrocarbon ratio K and hydrogen A as the new criterion respectively. This lays a solid foundation for further improvement of the available diagnostic methods. As diagnosis information of single fault type is normally adopted in previous methods, the scheme of information fusion is introduced. With comprehensive utilization of feature information such as gas ratios and gas concentrations, as well as reasonable hierarchy, a hierarchical decision-making fault diagnosis method based on multiple characteristic information fusion is proposed in this thesis as to effectively fuse multiclass feature information for diagnosis.To further solve the complexity and fuzziness among the corresponding relationship during compound faults diagnosis, the membership degree incidence matrix is established by adopting membership function to improve the blurred boundary. in the meanwhile, a logical reasoning model with fuzzy decision and multiple-valued output is constructed through adopting intersection and union set in fuzzy logic inference instead of AND-OR logic in the two-valued logic, and ultimately a comprehensive diagnostic method based on information fusion and fuzzy logic is established to give multiple-valued output for compound faults. Based on the Microsoft Access platform, a diagnosis software system for transformer compound faults is developed to realize visualization and intelligent identification of fault diagnosis. Applications to the actual fault cases demonstrated that, the method has greatly raised diagnosis reliability significantly for the fault types of partial discharges, spark discharges, arc discharges as well as discharges with overheating. In addition, the fuzzy logical reasoning process in this method is simple, which ensures the fault diagnosis fast enough to meet the demand of online condition monitoring of transformer.As the physical and chemical mechanism of transformer faults is not clear enough, it is still quite difficult to realize accurate diagnosis by extracting the characteristic parameters from the macroscopic experimental results, conventional experience statistics and information deduction. Considering analysis in the microcosmic level is helpful to understand more deeply the mechanism behind the complex phenomena, reactive molecular dynamics simulation is introduced to clarify the pyrolysis process with the molecular model of transformer oil and insulating paper respectively. The ReaxFF reactive force field based on bond order is adopted to investigate the micro dynamic mechanism regarding the occurrence of dissolved gases in the oil. Referring to the trajectory files of the pyrolysis at the molecular level for the transformer oil and insulating paper with the element tracking method, the thesis tried to sort out the initial decomposition mechanism as well as the difference in the main products of different transformer oil and insulating paper, including C2H4, CH4, H2, CO2, etc. Through calculation of the chemical bond dissociation energy, i.e. the reaction enthalpy in the process of cracking, the correlativity between the pyrolysis products distribution over time in terms of amount and type and the microcosmic chemical bond dissociation is established, and the energy changing law during the pyrolysis is revealed. Through analysis of the dynamic process of oil/paper cracking, the microcosmic dynamic mechanism for gases generation of the transformer oil/paper pyrolysis is revealed at the atomic level, which provides theoretical boost for the fault severity assessment.In order to solve the problem of severity evaluation for the composite faults, based on the enthalpy theory of thermodynamics, an energy calculation model for fault gas generating in the transformer oil cracking is established. With contrastive analysis of the development trend for the fault characteristic parameters, a new index of energy difference rate is proposed as the indicator for fault severity from the view of reaction energy, which in energy characterization forms a quantitative assessment model to evaluate single fault severity. Consequently, based on the aforementioned information fusion and fuzzy logic scheme, together with weighted parameters as to characterize the information of occurrence probability of the fault types and the reliability of the diagnosis methods, a multi-index comprehensive weighted assessment methodology is proposed to evaluate the severity of the compound faults in the power transformers. The application to the actual transformer fault cases shows that the evaluation method gives higher accuracy compared with the previous diagnosis schemes.