基于粗糙集的改进约简算法和决策树的电网故障诊断模型

采用基于粗糙集的决策树方法建立电网故障诊断模型,可以方便地处理含有非数值特征的、不精确的故障样本,且当丢失或出错的故障信息不是关键信号时,不会影响诊断结果,具有较强的容错能力和适应性。该文基于粗糙集理论,首先利用可辨识矩阵的改进算法对由断路器和保护为条件属性、考虑各种故障情况所组成的诊断决策表进行简化;然后采用加权平均粗糙度的概念,作为选择分离属性的标准,构造电网故障决策树,从而实现对电网的故障诊断。通过实例表明,该方法能有效地约简知识,具有很强的容错能力,能准确地判断电网故障以及对其进行定位。与规则表示相比,决策树直观、易于理解、维护和修改。

Application of rough set to build electric power grid fault diagnosis model based on decision tree

By employing decision tree to build the model of electric power grid fault diagnosis, fault sample with non numerical and inaccuracy values can be processed, diagnosis result can be learned when information is corrupted, erroneous and even missing because of its strong fault tolerant ability and better adaptive capability. Based on rough set theory., the paper proposes an improved discernibility matrix to reduce the decision table which includes all kinds of fault cases with the signals of protection relays and takes circuit breakers as condition attributes and fault sections as value attributes. Then the decision tree of electric power grid diagnosis is built by using weighted mean roughness as separating attributes standard to realize electric power grid fault diagnosis. An example shows that the presented method can reduce attributes effectively with strong fault tolerant ability and the fault of electric power grid can be identified accurately. Compared with rnle sets, decision tree is easier to maintain and modify. Furthermore, the decision tree is simple and easy to comprehend.