The Research Of Fault Diagnosis Based On Quantum Genetic Algorithm And State Comprehensive Evaluation Of Power Transformer

Transformer condition-based maintenance, which made corresponding maintenance strategy based on equipment actual condition, had significant meaning on ensure the safe and economic operation of power system. This paper begins from the analysis of power transformer common fault configuration and its engender causes analysis, on one hand find the most appropriate characteristic values of the transformer faults, identify the latent faults from the changes of these characteristic, and deduced the power transformer fault diagnosis model which based on the quantum genetic algorithm; on the other hand, some related parameters of transformer which most capable performance the characterization are found out, evaluate the state of transformer by these parameters, and proposed state evaluation method which is based on multiple source information, this method can effectively improve veracity of transformer operation condition assessment and latent fault diagnosis, and have great meaning on formulate reasonable examine and repair strategy and improve condition-based maintenance efficiency. The main research contents and innovation is:(1)This article researched the analysis method of the transformer short-circuit tolerant capability, and it was proposed that the safety margin of the transformer short-circuit tolerant capability was as the characteristic parameters of the transformer short-circuit tolerant capability. These provided a new way for the transformer comprehensive evaluation. The calculation of the characteristic parameters of the transformer short-circuit tolerant capability was put out, and 50 transformers in operation which belongs to Hebei Electric Power Company was analyzed by this method. The actual results demonstrated that the method which applied for transformer condition assessment was feasible and correct. The characteristic parameters of the transformer not only reflected the state of evaluation of the effective parameters of mechanical properties, but also provided an important basis for mechanical fault diagnosis of transformer.(2) The FDS was found to measure the water in the oil-paper. The characteristic parameters of the moisture content in the insulation paper of transformer which was detected by the FDS:there was an actual curve while the insulation was tested at different frequency of dielectric loss trend. By the mathematical methods, a standard curve and the actual curve were measured to fit. The final fitting curve was analyzed to determine the moisture content of oil-paper. The application shows that the characteristic parameters of the moisture content in the diagnosis of transformer damping insulation were very effective.(3) It was researched that a new method of positioning in oil based on supersonic phased-array and broadband array signal processing. A set of partial discharge detection system based on supersonic phased-array sensor was builded, and partial discharge (PD) broadband ultrasonic signal could be directly detected, while the array model formationing. According to the shortcoming of the PD supersonic narrowband array signal goniometric algorithm based on MUSIC, this paper proposed a consistent focusing Wideband goniometric measurements combined with cross-goniometric principle to achieve the positioning. Finally, through several positioning experiments, the results show the effectiveness and accuracy of this method.(4) This paper proposed a electric transformer oil dissolved gas analytical method based on the quantum genetic neural network, and the method used in quantum genetic algorithm in transformer oil dissolved gas analysis of the initial neural network weights and thresholds obtained. The global optimal solution was accessed to a multilayer feedforward neural network near the variable parameters, so the problem that ordinary network training algorithm is easy converge to local minimum when the sample was large and the network was smaller was settled out. At the same time the network generalization (of the accuracy and reliability) was better.(5) Through extensive collection of information, a variety of power transformer test data integrated. A more comprehensive set of fault symptoms and fault location feature set were obtained. This paper presented power transformer maintenance strategies optimization model based on Bayesian network, and Bayesian network method was applied in transformer maintenance program. Through rough set theory Bayesian network was effectively reduced. Finally, it demonstrated the effectiveness of the method through an example.(6) Full account of the seven categories of status parameters of the basic information, running tour information, test information, maintenance information, fault (defect) information, poor working conditions and family defect of transformer, this paper putted out a comprehensive evaluation method of the transformer condition. By this method the transformer was divided into the transformer body, casing, tap, cooling systems and other components, and they were evaluated individually. Finally, the results of comprehensive evaluation of the various parts of the transformer were given the overall evaluation results, effectively improving the sensitivity of the evaluation. This paper constructed an accurate and reasonable state rating system for transformers, and the Marking Scheme reflected the status of the state of deterioration more intuitively. At the same time the amount of the state which divided into an important, general, reference three types was scored. Based on the degree of deterioration, the state would be divided into four state variables to improve the accuracy of the transformer condition assessment.