Rough Set Theory for Data Mining in an On-line Cable Condition Monitoring System

This paper presents the application of a new methodology, the Rough Set (RS) theory, for data mining or knowledge acquisition in an on-line cable condition monitoring system. The attractiveness of the RS theory is that it allows automated generation of knowledge models of clear semantics which offer explanations of the inferences performed when used in diagnostics. Following an introduction to a cable on-line condition monitoring system, where High Frequency Current Transformers (HFCT) are used for continuous monitoring of Partial Discharge (PD) activity, and the desired data mining or knowledge acquisition technology for such a system, the paper briefly described the concept of RS and the procedure of its implementation in practical applications. It then presents an example of the application of the method to knowledge deduction which can be easily adopted in the online 11-kV cable condition monitoring system. In such a system effective data mining proved to be the bottleneck due to the large volume of data generated by the system (sampling rate at 100 Mega Samples/second). The paper demonstrates that the RS theory is effective in mining knowledge rules from large volume of data. It requires little computing time and storage space, when compared with other data mining algorithms, as it simply removes redundant data and those data containing no information. It has the potential to be applied to automatically analyse utilities’ condition monitoring database in data mining and knowledge acquisition.