Evaluating the Impact of Low Discrepancy Sequences on the Probabilistic Evaluation of Composite Power System Reliability

Evaluating the reliability of composite power systems using probabilistic means can quickly become a computationally intensive task as the size of the system grows. Different efforts like state space decomposition and population based intelligent search have focused on reducing computational cost by improving the way in which the state space is sampled. This work investigates the use of low discrepancy sequences (LDS) in order to improve the sampling process used when applying Monte Carlo simulation (MCS) to this problem. Low discrepancy (or quasi-random) sequences are deterministic, dependent sequences that are used for sampling a state space in a uniform fashion. Three examples of these sequences that are examined in this study include the Halton, Hammersley, and Faure sequences. Each of these sequences may be used in place of the typical random sampling when applying MCS to such a problem. This study examines conceptual and empirical differences between these LDS techniques and MCS, and discussions are made in terms of the convergence characteristics of the methods. Results demonstrate that the while LDS methods do perform well, their overall performance is comparable to MCS when fully converged results are needed.