Over the past few years, power quality (PQ) monitoring has become an important topic because of the negative impact of different machines to the electrical network and to the susceptibility of critical equipment. There are different disturbances that affect the PQ; therefore, in order to apply a proper solution, these have to be correctly detected and classified. In general, signal processing techniques are applied for their detection. Recently, several approaches based on empirical mode decomposition (EMD) method have been reported; however, the selection of the best-suited method in terms of processing and performance for a particular case can be a complicated decision-making process. In this paper, a quantitative and qualitative comparative study using EMD methods such as conventional EMD, ensemble EMD, and complete ensemble EMD is presented. The study is applied to synthetic and real PQ signals, in which aspects of the computational cost and decomposition accuracy are discussed.