Research Of On-line Monitoring System For Moisture Content In Transformer Oil

The transformer oil is an important liquid dielectric material in the transformer; it plays the roles of insulation, cooling and extinction of arc. The presence of moisture can deteriorate transformers'insulation performance drastically. Severely, it can lead to some catastrophic accidents such as insulation breakdown and transformer burnout. Through the control of moisture content in oil, it can not only prevent insulation strength from decreasing to dangerous level, but also evaluate the whole insulation status of transformers, and estimate the leak tightness of the apparatuses. Hence, the on-line monitoring of moisture in transformer oil is of great significance to the operation safety of transformers.The generation reasons of moisture in oil, the mechanism of moisture dynamics in the transformer oil, and paper-oil moisture equilibrium is studied in the paper. Also the moisture states existing in oil and the hazard to transformer insulation property is analyzed. Considering the disadvantage and limitation of current measurements of moisture in oil, a detailed scheme of on-line monitoring for moisture in oil is proposed.Considering the study status of sensors and the requirement of on-line monitoring, the polyimide-based capacitive humidity sensor is selected as moisture sensor. The on-line monitoring system for moisture in transformer oil is developed. Through the conversion of the output of sensor, data sampling of moisture and temperature signals are realized with the computer serial ports. The performance of sensor in the transformer oil condition is studied.According to the analysis of the relevant theoretical foundation of moisture in transformer oil, a moisture-in-oil monitoring model based on neural network is put forward in the paper. This model can not only estimate moisture in transformer oil, but also can detect faults which can cause the abnormal variation of water content in the transformer oil through comparing the value of estimated moisture with the measured ones. And the validity of the model confirmed through simulation experiments.