A Technique for Reducing Computational Effort in Monte-Carlo Based Composite Reliability Evaluation

There is a growing need within the electrical utility industry to refine the methods used in the allocation of capital resources based on reliability considerations, and to determine what are the tradeoffs between cost and reliability. The evaluation of the steady state adequacy of bulk power (composite) systems is widely recognized as an important part of this process. The computational tools currently developed for bulk system reliability evaluation differ basically with respect to the method of selecting the system scenario (step (1)) and the network model used in the adequacy assessment (step (2)). For example, models such as SYREL, GATOR, RECS and COMREL, are based on the successive enumeration of severe/likely scenarios and use an ac power flow for adequacy assessment. In turn, models such as SICRET, MEXICO and CONFTRA are based on Monte-Carlo sampling of scenarios, and use a linearized power flow model ("dc" power flow) for adequacy assessment. One possible limitation of Monte-Carlo methods is the strong dependence of computational effort (proportional to the number of samplings) with respect to the desired accuracy of the estimates. For example, a sample size of 104 would be enough to estimate a LOLP of 10-3 with a relative uncertainty of 30%. However, the same estimate would require 106 samplings if the desired accuracy was 3%. Some methods have been proposed to reduce this computational effort, such as stratification, Importance Sampling and bounding methods. This paper describes a new technique for reducing the computational effort in Monte-Carlo based composite reliability evaluation.