The invention relates to an adaptive serial estimation method for the water content of a two-phase flow. Estimation is performed by virtue of measurement data of different sensors at upstream and downstream, and the method comprises the following steps: performing time alignment on the measurement data of the upstream and downstream sensors; performing state estimation initialization; performing linear Kalman state estimation and updating on a water content measurement value of the upstream sensor to obtain prediction of a state vector and prediction of a covariance matrix; calculating a process state switching matrix of the two-phase flow by virtue of the measurement data of the downstream sensor: performing unscented Kalman filtering-based nonlinear state estimation on a downstream water content measurement value.