基于粗糙集理论多区域并行神经分类器在变电站故障诊断中的应用

针对变电站故障诊断中不确定信息多和实时性要求高的特点,以变电站的开关保护信息为基础,提出了一种基于粗糙集理论和神经网络理论的多区域并行神经分类器的变电站故障诊断方法.该方法首先将变电站故障划分为多个独立的故障单元,针对每个区域故障单元建立故障模式库,利用粗糙集的知识约简和不确定信息的处理能力,对故障模式库并行挖掘,实行属性优选,再运用神经网络对故障诊断知识进行模式识别.将其应用于变电站故障诊断专家系统中,应用结果显示该方法不仅能缩小问题求解规模,实时性高,而且具有较强的抗干扰能力,是一种有效的变电站故障诊断方法.

Multi-area parallel neural classifiers for substation fault diagnosis based on rough set theory

In view of more indeterminate information and higher speed request in substation fault diagnosis,on the basis of switch protection information of substation,an approach called multi-area parallel neural classifiers for substation fault diagnosis is proposed.The method firstly partitions substation fault into many small independent area fault cells,fault mode is established for each area,rough set then is applied to mine knowledge and select perfect attributes from fault mode base based on its abilities of knowledge reduction and disposing indeterminate information,then fault is identified through parallel neural networks.Being used in fault diagnosis expert systems of substation,the results show that the approach not only reduces the scale of solution and improves diagnosis speed,but also owns high anti-inference ability.Therefore it is an effective method for substation fault diagnosis.