基于BP网络的变压器油中溶解气体在线监测

NEURAL NETWORK BASED ON-LINE DETECTION OF DISSOLVED GAS IN TRANSFORMER OIL

研究了对变压器油中溶解气体及温度进行微机在线监测的方法 ,介绍了传感器与油气分离装置 ,气体在线监测的硬、软件特性及系统的控制方式 ,重点阐述了如何利用人工神经网络拟合混合气体之间的非线性关系 ,有效地防止气体传感器之间的“交叉敏感”,然后利用训练好的神经网络对混合气体浓度进行在线监测 ,并给出了实验结果。

An on-line method for detecting dissolved gases in transformer oil is studied. It directly senses the thickness of the mixed gases appeared with transformer faults by using a trained neural network to eliminate cross-sensibility among them. The sensor characteristics and device of separating oil and gases, the system control approach and its hardware and software are analyzed. It is emphasized on how to use artificial neural network to approximate nonlinear relation between input and output in order to avoid cross-sensibility. Tests on experimental device show that the proposed method is effective and can be used for on-line detection of the hidden fault in transformer.