Realization of a power transformer on-line monitoring and diagnosis system based on DGA and PNN

A novel on-line monitoring and fault diagnosis system of power transformer using basic principle of three ratio method based on DGA (Dissolved Gases Analysis) and PNN(Probabilistic Neural Networks) for engineering application is proposed. According to tested dissolved gases contents changing characteristic of transformer oils under different transformer working conditions, the five characteristics of gases (C₂H₂,C₂H₄,CH₄,H₂,C₂H₆) in transformer insulating oil will be as the objects of monitoring, through the continuous detection of these gases, the three pairs of gases content ratio in these key/characteristic gases are got as one of extracted features. Then, the extracted feature parameters are used as inputs to classifiers based on a Probabilistic Neural Networks model for six-class fault recognition. The system realization platform is based on PXI and cRIO in Virtual Instruments technology. The software adopts Labview and the hardware platform is PXI for development and management and cRIO for working. The simulation results show that the system can on-line provide and display the changing process of dissolved gases, the corresponding diagnosis and discover the hidden faults timely during operation. It proposes a new way for power transformer on-line monitoring and fault diagnosis.