基于PCA和粗糙集构建决策树的变电站故障诊断

Fault diagnosis of substation by the constructed decision tree based on principal component analysis(PCA) and rough set

提出一种基于主元分析(PCA)和粗糙集理论结合继而构建决策树的故障诊断方法。该方法利用PCA对原始故障决策表的条件属性集进行降维处理,得到由主元变量构成的故障决策表,采用等频分割方法对这一决策表的数据离散化,进而采用基于主元属性重要度的粗糙集属性约简算法得到离散后的决策表的最小约简,以约简数据集为样本基于核属性采用一种改进的决策树算法训练学习,构建故障决策树进行诊断决策。测试实例证明了该方法能简化故障诊断系统,提取容错性较强的诊断规则,提高了故障的识别率。

A method for substation fault diagnosis based on principal component analysis（PCA） and rough set theory,then to constructed decision tree,is proposed.By this method,PCA is used to decrease the dimension of all the condition attributes of the original fault decision table and get fault decision table that consists of principle component variables.Then,an equal-frequency devision method is used to discrete the value of the above decision table.The next is that a rough attribute reduction based on the significance of principle component variables is applied to obtain a minimum reduction of the discrete decision table.Finally,based on the core attributes,an improved decision tree algorithm is used to train the reducted data set and construct a decision tree for the diagnosis.Test proved that this method can simplify the fault diagnosis system,extract the diagnosis rules of better fault tolerance and increase the fault recognition rate.