大型变压器的并行诊断技术研究

Research on large power transformer fault parallel diagnosis technology

为了提高采用多种方法组合对大型电力变压器故障诊断的速度,提出了一种变压器故障诊断的并行技术,采用性价比高的PC机群(PC-cluster)和数值计算能力强大的Matlab分布式计算引擎(MDCE)和分布式计算工具箱(DCT)搭建分布式计算平台,介绍了基于该平台的大型变压器并行诊断技术的实现过程,使模糊C均值聚类诊断法、BP神经网络诊断法、粗糙集理论故障诊断法和IEC60599推荐的三比值法等多种诊断方法并行计算,同时处理。仿真表明:并行计算与单项方法相比,其诊断误判率大大降低;与串行计算相比,节省了大量的计算时间,满足故障诊断正确性和快速性的工程要求。

In order to reduce the computing time for large power transformer diagnosis by synthesizing several conventional transformer fault diagnosis methods,a new concurrent diagnosis technology for large power transformer is proposed.A distributed computing platform based on high performance price PC-cluster coupled with a high performance MATLAB distributed computing engine（MDCE） and distributed computing toolbox（DCT） was established.The implemental process of the parallel diagnosis technology was also introduced.Several diagnosis methods for example,Fuzzy C-means clustering（FCM） algorithm,BP neural network,rough set theory and three-ration method recommended by IEC 60599 could be used concurrently.Simulation showed that this concurrent diagnosis method reduced the misdiagnosis rate than that of any of above single methoddid did.Compared with serial computing method,this new method can save a lot of time and can meet the engineering requirement of accuracy and fastness.