Power signal with high sample rate is more required for complex tasks including demand response and power disaggregation. For this issue, a signal compression method based on compressed sensing (CS) is designed, to support signal transformation under low energy consumption and narrow bandwidth condition in smart meters. On the basis of analysis on characteristics of different types of loads, an empirical model for power signal of smart meters is proposed. Experimental results demonstrate the validity of the empirical model, the best method of generating the representation matrix and projection matrix, and the superiority of compressed sensing based method compared to wavelet transform.
