Reverse transcription quantitative real-time PCR (RT-qPCR) has become one of the most widely used methods for gene expression analysis, but its successful application depends on the stability of suitable reference genes used for data normalization. In plant studies, the choice and optimal number of reference genes must be experimentally determined for the specific conditions, plant species, and cultivars. In this study, ten candidate reference genes of sweetpotato (Ipomoea batatas) were isolated and the stability of their expression was analyzed using two algorithms, geNorm and NormFinder. The samples consisted of tissues from four sweetpotato cultivars subjected to four different environmental stress treatments, i.e., cold, drought, salt and oxidative stress. The results showed that, for sweetpotato, individual reference genes or combinations thereof should be selected for use in data normalization depending on the experimental conditions and the particular cultivar. In general, the genes ARF, UBI, COX, GAP and RPL were validated as the most suitable reference gene set for every cultivar across total tested samples. Interestingly, the genes ACT and TUB, although widely used, were not the most suitable reference genes in different sweetpotato sample sets. Taken together, these results provide guidelines for reference gene(s) selection under different experimental conditions. In addition, they serve as a foundation for the more accurate and widespread use of RT-qPCR in various sweetpotato cultivars.