An Improved Association Rules Mining Algorithm in Fault Diagnosis Expert System of Power System
ES (Expert System) is the most widely used AI(Artificial Intelligence) technique currently. However, now there are only a few ES operated successfully in live field for it suffers from some bottlenecks. There are much considerable association information hidden among the fault information of power system, so ARM (Association Rule Mining) methods is discussed, and FP-Growth algorithm is improved in the paper. The improvements include: 1) To mine rules of time sequence, to avoid mining invalid rules with latter events leading earlier events; 2) Weighted support is adopted, weight of support of serious fault or unusual fault is bigger than that of general fault, so as to mine rules about serious fault or unusual fault; 3) Fault Elementary Events are divided into 2 types: FSEE (Fault Symptom Elementary Event) and FTEE (Fault Type Elementary Event). FSEE represents symptom of fault, and FTEE represents type of fault. As to simple fault, FSEE is always antecedent, while FTEE is always consequent. As to cascading fault, FTEE can be antecedent too. Mining is grouped by fault type, so as to avoid mining invalid rules with jumbled association information; 4) Prune technique of FP- tree is improved. If an item set includes a type event, then the item set can’t be pruned; 5) OR logical function of rules is added, so as to heighten generality. For example, current protection is used in some places, while voltage protection is used in some other places, the rule can be expressed as current OR voltage. Space complexity and time complexity of the improved algorithm can be decreased obviously. Use case shows velocity and accuracy of the improved algorithm is heightened prominently, rules obtained by DM (Data mining) are valid and practicable, the DM module can operate online.
Data Mining Association Rule FP-Growth Algorithm Fault Diagnosis Expert System
LI Zaihua BAI Xiaomin ZHOU Ziguan XU Jing LI Xiaojun Zhao Wei
China Electric Power Research Institute China North of China Power Grid Corporation China
国际会议
桂林
英文
1-7
2009-10-28（万方平台首次上网日期，不代表论文的发表时间）