I rather use stable microRNAs, and preferably more than one, as endogenous controls (using the geometric mean as reference value). One good tool for identifying the most stable miRNAs using your high throughput profiling data is the software NormFinder (https://www.moma.dk/normfinder-software). In that webpage you can find documentation for Excel and R. If you are familiar with the programing environment R, you can include the NormFinder R code in your script or follow the instructions on the attached file (downloaded from the above referenced website). The authors of NormFinder also have a paper (Andersen et al. Cancer Research 2004(64): 5245-5250) describing the statistical framework.
As Yury O. Nunez Lopez stated, it is better to use two or three stable miRNAs or snRNAs. They depend on the type of your sample, but maybe someone already tested reference genes for your type of samples (google it). Another useful software for reference gene identification is BestKeeper (here is the article about it https://www.gene-quantification.de/pfaffl-bestkeeper-2004.pdf).