Sample Size Selection for Monte Carlo Resource Adequacy Assessment of Power Systems

Choosing the number of Monte Carlo samples to use in a probabilistic resource adequacy assessment requires striking a balance between computational expense and the statistical precision of study results. This paper reviews the conventional Central Limit Theorem approach to quantifying result confidence in Monte Carlo analysis, and then introduces two new methods that address shortcomings in this approach when applied in the context of resource adequacy assessment.The first alternative provides a lower bound on adequacy confidence in the case of all-zero sample observations, a common occurrence in resource adequacy analysis that the Central Limit Theorem method is ill-equipped to handle. The second alternative uses Bayesian conjugate priors to learn from outcome samples more efficiently, reducing sample size requirements if an appropriate prior distribution can be provided. Both of these methods’ potential advantages over the traditional approach are demonstrated via empirical comparisons of sample size requirements and inference error in resource adequacy assessment of a test system.