Temperature Rise And Moisture In Traction Transformer And Insulation Condition Prediction

The insulation of the traction power supply equipment is demand more reliable with the development of the high-speed passenger transportation line, and the heavy-loaded train transportation. At the same time, since the condition maintenance and life management are the developing trend, it is urgent to study on the insulation ageing, insulation failure mechanics, insulation condition monitoring and so on for the traction power supply equipments. But there are few reports about insulation ageing factors of traction transformer. It is found there are three characters for loads in traction transformer, i.e., impact properties, asymmetry, and non-linearity, affecting the insulation ageing factors, e.g., temperature rise, and disequilibrium moisture partition of oil-paper insulation.Two thermal circuits are developed for calculating and analyzing the inner temperature rise in ungrounded-wye delta traction transformer and Scott connected traction transformer, in order to minish the temperaure rise calculation error under anisomerous loads. And by employing heat transfer principles and hydrodynamics, another model for hot-spot temperature rise in non-directed flow windings has been developed to explain the thermal overshoot phenomenon for step changes in load, as occurs in traction transformers, in which the duct oil flow velocity is taken into account, and it is validated using experimental results obtained for step changing load current on a 40-MVA-ONAN-cooled unit. According to the developed model as applied to two transformer units (20MVA and 40MVA), the hot spot temperature rise overshoots because of the change in duct flow velocity lags the change in power dissipation after the change in load, and the overshoot factor is larger with larger difference between the initialized and steady velocity.A surface function for the temperature, the moisture in oil, and the moisture in paper is proposed, using the existed curves at the different temperatures, according to which, the equilibrium moisture partition curves in oil-paper insulation at arbitrary temperatures can be obtained. The method is validated by comparing the calculated curve and the published experiment data. And then a transient moisture equilibrium model of oil-paper insulation is designed using Fick's second law, which is validated by experiment. By the model, the transient moisture concentration at the coordinate in the direction of the thickness is obtained, and then average moisture concentration of paper is calculated, on basis of analyzing the affects of the temperature, the equilibrium concentration, the thickness of the paper and the penetrating mode of moisture on the equilibrium time. And by the model, a model based monitoring technique for moisture in oil-paper insulation of the traction transformer is proposed.Three hypotheses about steady-state responding, transient responding and resuming behavior of semiconductor gas sensor used in on-line monitoring system for dissolved gases in transformer oil insulting oil, according to which, the logarithm of the steady-state resistance value and the logarithm of the gases concentration obey to tangent function, the transient responding is decided by the current resistance of the sensor and the gases concentration, and the resuming behavior is decided by the current resistance of the sensor and gases concentration at the next time. The hypotheses have been validated by experiment.Two methods have been used for predicting insulation faults in transformer. In the first method, the gases concentrations are predicted firstly by grey model, and then the results will be calibrated by self-learning BP-neural networks, on basis of analyzing the affects of different types of gases each other, as well as the relationship of time sequences made up of fault gases. In the second method, the faults relative membership degrees are predicted using time series analysis technique based on gray theory, and the the faults are diagnosed by an improved threshold value principle. The two methods have been validated by analyzing history datum.