Broadband power line communications (PLC) is considered to be one of the most promising technologies to fulfill data transmission in smart grids, traditional linear models andstatistical properties can not understand welldescribenon-stationaryand nonlinear characteristics of PLC signal. In order to better studycharacteristicsofPLC signal, the mono-fractal and multi-fractal theories are introduced to understand the self-similarity characteristics of the PLC signals. Four common methods, namely, rescaled range analysis, variancetime plot method, periodic diagram analysis and wavelet-based method are used to study the nonlinear properties. Fractal analysis at different frequencies and times are also performed to verify further. The results reveal self-similarity of the PLC signals. Besides the mono-fractal properties, the paper tests the multi-fractal properties of PLC signals by the means of multi-fractal detrended fluctuation analysis (MFDFA). We estimated the multifractal spectrum of power low exponents from the measured PLC signals. We also proposed a new algorithm to improve the traditional MFDFA, where wavelet theory is integrated.