Abstract: The on-line monitoring and forecasting techniques for the nuclear power plant reactor coolant system are studied. Based on the analysis of the characteristics of the reactor coolant system components and fault characteristics, multi-state variable estimation techniques（MSET） and sequential probability ratio technology（SPRT） was used to estimate and predict the system and to determine whether the system is abnormal. The anomaly can trigger a fault diagnosis unit or give warning. Test results show that MSET and SPRT combined technology make a good performance in the abnormal condition monitoring tasks.