Global Sensitivity Analysis for Power Systems via Quasi-Monte Carlo Methods

There are many inputs with uncertainty in a power system, due to factors such as uncertainties in the distributed renewable generation, or natural disasters like hurricanes. The global sensitivity analysis of a model quantifies the importance of each input parameter to the model output when input parameters have uncertainty. In global sensitivity analysis, unlike local sensitivity analysis, all input factors are varied simultaneously, and as a consequence, one can assess the impact of the higher order interactions among the parameters. In this paper we will use global sensitivity analysis, in particular the Sobol' sensitivity indices, to assess the importance of input parameters in the IEEE 14-bus modified test system. By identifying unimportant input parameters, we will reduce the complexity of the model. We will use randomized quasi-Monte Carlo methods to estimate the sensitivity indices and perform uncertainty quantification for the output of the reduced model.