Distribution system reliability improvement using predictive models

This paper describes the methodology used by the authors for performing a predictive reliability assessment of a real distribution system of the USA. In the first stage of this approach, a Predictive Reliability Model (PRM) was developed by using five-year historical outage data, information about reliability improvement projects recently implemented by the distribution utility, distribution feeder models, and customer and feeder load data. Then the PRM parameters were iteratively adjusted to match the estimated current reliability values of the real feeders. Next, a comprehensive set of reliability improvement projects was evaluated and prioritized according to a cost-effectiveness metric based on SAIFI. After modeling the implementation of the selected projects, the SAIFI, SAIDI, CAIDI and MAIFI values of the study area were estimated. Finally, the reductions achieved by implementing the projects under fuse-clearing and fuse-saving overcurrent protection schemes were compared. The results demonstrate the advantages, versatility and potential of predictive reliability modeling.