In this talk, We propose two self-normalized statistics to jointly test the mean and variance change-points in long-memory time series. Under the null hypothesis that the series is a stationary long-memory process, we demonstrate that our test statistics converge to nondegenerate distributions. We derive the limit distributions of the two test statistics under three alternative hypotheses that the series has mean and variance change-points at different locations. The proposed test statistics perform well with finite samples in simulations. Moreover, we can distinguish whether a detected change-point is a mean or variance change-point, or if the mean and variance have changed together. We illustrate our tests through the analysis of two real datasets: the annual discharge volume of the Nile River, and the closing-price data of International Business Machines (IBM) common stock.