Research On Fault Synthetic Diagnosis Method Of Power Transformer Based-On Combining Quantities

Oil-gas and part-discharging signals are most frequently studied in transformer fault diagnosis now. However, Oil-gas only can reflect fault type but cannot point out the fault location and the degree. Part-discharging signals, which can realize fault location in theory, also cannot be carried out easily because of influence of circumstances. Therefore, the study of transformer fault diagnosis has not been consummate so far. In this background, the motive of this paper is to give study support for transformer synthetic diagnosis with combining availability information.Based on a math model of transformer winding founded, this paper carries out theory study of winding fault diagnosis with electric quantities. It mainly introduces the theory of least-square algorithm and recursive least-square algorithm and the application of them in winding fault diagnosis. The results of theory deducted indicate that it can gain the parameters (resistance, leakage inductance) of transformer winding by least-square technique and these parameters will be constant unless inner fault occur in transformer. To test it, this paper does series simulation by MATLAB. At the same time, in the aspect of fault diagnosis based-on oil gas, some study of application of neural network and data reliability analysis as well as fuzzy theory are also carried out in this paper.After theory study of above, this paper presents a thought of structuring a system of fault integrated diagnosis, and expounds its process to realize it. The system includes three models. They are electric model, fuzzy neural network model and data reliability analysis-neural network model. The three models can diagnose fault independently, the results of them can be confirmed and complemented by each other. And it will be more full-scale and more material.