Sample size reduction in stochastic production simulation

Abstract

A combined control variable and stratified sampling method is proposed for Monte Carlo production simulation. Using the production cost obtained from the load duration curve type simulation as the control variable, both the theory and a sample study suggest that the required (for an acceptable coefficient of variation) number of samples can be reduced by approximately a factor of 1000. In the example study, 8000 samples were replaced with 10 samples; in addition, the proposed method using the 10 samples produces an estimator of the mean production cost that has a much smaller variance than an estimator based on a traditional Monte Carlo study which uses 8000 samples. It is believed that this significant reduction in running time will enable chronological simulation to be used in a number of long range planning studies. The advantage of this method is additional insight into actual operation through chronological simulation as well as less bias (or systematic error). With this proposed method, it is believed that the computational time requirement of stochastic production cost simulation has been reduced to the point that its advantages outweigh its additional (over probabilistic simulation) running time.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>