Consensus clustering algorithms for Asset Management in power systems

This paper aims to present an intelligent approach to the Asset Management (AM) in power systems. Three existing consensus functions (i.e., Non-negative matrix Factorisation based, Weighted Partition via Kernel and Information Theory-based) have been compared and applied to analyse a power system document depository (PSD). In addition, ontology is employed in this research to implement a document dataset modification, which shows a significant improvement for the single consensus clustering algorithm. Moreover, Genetic Algorithm (GA) is utilised to solve the weighted kernel-based consensus function. The influences of different settings of GA are analysed. Relevant improvements of GA are also presented.