In this work, we apply multifidelity Monte Carlo sampling, a technique for uncertainty quantification, to two test cases from plasma microturbulence analysis, which is an important field of research into fusion power. Multifidelity Monte Carlo sampling aims to increase the accuracy of standard Monte Carlo estimators for statistical moments of model output based on uncertain input parameters. Standard Monte Carlo sampling estimates statistics by evaluating the underlying model for a number of samples and computing estimators from the results. In order to obtain acceptable precision, a pro- hibitive amount of samples is often required and simulations become computationally infeasible, especially if the underlying model is expensive to evaluate. To overcome this shortcoming, multifidelity Monte Carlo sampling distributes the model evaluations be- tween the model under consideration and one or more computationally less expensive surrogate models. The obtained model evaluations are then combined into estimators using a control variate approach. In our test scenarios, we consider the Cyclone Base Case, a popular benchmark in the analysis of plasma turbulence. Multifidelity Monte Carlo sampling is applied to a three-dimensional and an eight-dimensional version of this scenario. For our simulations, we use the plasma microturbulence code Gene and construct a sparse grid approximation and a machine learning surrogate. In both test cases, we find that the application of multifidelity Monte Carlo sampling leads to a reduction of the estimators’ mean squared error by orders of magnitude. Moreover, we show that combining different low-fidelity models can further decrease the error.     «

In this work, we apply multifidelity Monte Carlo sampling, a technique for uncertainty quantification, to two test cases from plasma microturbulence analysis, which is an important field of research into fusion power. Multifidelity Monte Carlo sampling aims to increase the accuracy of standard Monte Carlo estimators for statistical moments of model output based on uncertain input parameters. Standard Monte Carlo sampling estimates statistics by evaluating the underlying model for a number o...     »