Aiming at the problem of efficient measurement of dynamic error of intelligent electricity meters, firstly, in amplitude domain, the characteristics of fast random fluctuation of complex dynamic load signals are analyzed over a long period of time; Secondly, the model of compressed sensing pseudo-random dynamic load test signal is established, and then the measurement matrix construction method combining Hadamard matrix and the down-sampling matrix is adopted to determine the model parameters of the test signal, and a scheme for generating compressed sensing measurement signal is proposed. Finally, by means of the experimental test method, the dynamic error data of the electricity meter under the conditions of compressed sensing and m-series dynamic test signals are compared. The results show that the dynamic error test efficiency of the compressed sensing dynamic test signal is 4 times higher than that of the m-sequence pseudo-random test signal.