The information generated from Smart Meters in advanced measurement infrastructure (AMI) in a Smart Grid, needs real-time data acquisition and processing, this could generate high redundancy of data in an uplink in the communication network compared to the number of smart meters in a city. In this sense, the application of Compressive Sensing (CS) can improve the efficiency in the transfer of data and reduce a possible overload in the system. However, it is necessary to establish the minimum conditions required for the use of CS for data related to the Electrical Energy Consumption Profile. Thus, this article presents the analysis of the use of CS as a source compression method to reduce the data to be transmitted on a network. So, in order to evaluate the quality of the load profile reconstruction using CS despite the transformations that the data undergoes in a real wireless communication system, this technique has been implemented on a Software Defined Radio System (SDR). For this, the measurement matrix implemented with Gauss and Bernoulli distributions is analyzed. The algorithms for the reconstruction are: OMP (Paired orthogonal search), BP (Base search) and GPSR (Gradient Projection Sparse Recovery) in different compression ratios (10 - 80%). It also shows the comparison of errors generated in the uses for the representation of signals in the Gaussian campaign, the DCT transform, the Wavelet Meyer transform and Wavelet Biorthogonal 3.9.