Sample size reduction in Monte Carlo based use-of-system costing of power systems

Accuracy of the results obtained through Monte Carlo based models greatly depends on the number of samples used in the simulation and the variance of the means of the associated random variables. Variance reduction techniques can be employed to reduce the sample size needed to achieve a given precision in the estimated values. Three such techniques, antithetic sampling, stratified sampling, and use of a control variable are investigated by the authors in the context of marginal costing of real powers. These methods have been implemented using a modified version of the IEEE 118 bus network and it is shown that the reduction in the sample size can exceed 75% in certain cases.