An unsupervised learning based MCDM approach for optimal placement of fault indicators in distribution networks

This paper proposes a novel integrated model based on multi-criteria decision-making (MCDM) method to assess and rank the feeder sections to optimally locate fault indicators in a large-scale distribution network. First, decision makers' weights are computed based on the subjective weighting information. Then, dataset including data calculated for every alternative under active power flow, failure rate, repair time, number of customers, and customer type criteria are grouped using mixed-type clustering algorithm. This leads to the number of clusters extracted from the dataset, then obtained number of clusters are utilized to set the required linguistic terms. Finally, the ranking order of alternative is assessed by using the fuzzy similarity measure-based technique for order of preference by similarity to ideal solution (TOPSIS). The newly proposed model combines the merits of integrated enhanced particle swarm optimization (EPSO) fuzzy analytic hierarchy process (AHP) model in dealing with fuzziness and vagueness of linguistic assessments, the merits of clustering algorithm focused on representing performance values of each alternative with respect to each decision criterion and the merits of similarity measure-based TOPSIS in solving complex decision-making problems. The proposed model is applied to the RBTS4 and then sensitivity analysis is performed in the optimal solution.