In the last years, the fuzzy techniques started to be used to model the uncertainties from distribution systems. The qualitative assessments for power quality indices can be mathematically translated into membership functions and thus integrated in several methodologies used for planning or operation purposes. This paper focuses on the voltage quality assessment of distribution system under uncertain conditions and proposes an approach based on fuzzy techniques. Thus, the indices for evaluation of slow voltage variations (mean value of voltage and mean square deviation voltage) and mean daily load are expressed in linguistic terms and fuzzy numbers. The linguistic terms characterize categories of electric substations, obtained with Ward clustering method. The algorithm was tested using an electric distribution system with 117 substations (20/0.4 kV) that supplies the rural consumers. The results are comparable to the real values.