This paper presents a new approach in the evaluation of power quality indices for microgrids operating in islanded mode. The indices are estimated considering uncertainties associated with: load forecast error, generator outages and the stochastic behavior of the power output of renewable energy sources. These uncertainties were modeled by combining the following techniques: distributed slack bus power flow, point estimation and Gaussian mixtures. These combined techniques were used to provide probabilistic indices related to the frequency and bus voltages in the microgrid, for example: frequency and load cumulative probability distributions, partial and total load curtailment probabilities due to frequency and voltage deviations. The identification of the most vulnerable buses in relation to voltage problems was also carried out. The proposed analytical method was able to provide a significant reduction in the computational cost of the microgrid power quality assessment compared to the traditional Monte Carlo simulation.