Adaptive Compressive Sensing based Data Compression on Energy Consumption in Smart Grid

With increased interests on efficient energy management in the smart grid, a large amount of distributed data points need to be monitored and corresponding energy big data need to be captured. To manage the large volume of data efficiently, the notion of compressive sensing (CS) is used for the compression with low computational complexity. However, the CS based data compression incurs an inevitable reconstruction error. For mitigating of the reconstruction error for a power signal, we present an adaptive CS (ACS) based compression scheme in the smart grid. In the proposed scheme, a representation matrices is adaptively selected to improve the reconstruction performance in accordance with a power profile, while conventional CS based compression is performed with a deterministic representation matrix regardless of a structure of a power signal. To this end, machine learning (ML) based classifier is used to choose the optimal representation matrix with statistical features of a power profile. Through simulation results, it is shown that the proposed method has improved reconstruction performance compared to conventional approaches.