In real-time operation, it is inevitable that large power source failures may occur randomly, then the frequency of the power system will drastically drop, possibly causing problems for the power system as well as electrical equipment. Therefore, estimating the lowest point of power system frequency and warning the dispatcher of the risk of load shedding (F81U under-frequency relay) in response to typical power failure scenarios in real-time operation is extremely essential, especially in the context of high penetration of renewable energy sources into Vietnam power system. To achieve that goal, the authors have researched and built an online tool into the existing SCADA/EMS system, using a multivariate linear regression model methodology improved from international experience [1]. This tool has given reliable results and is currently being applied regularly and effectively in monitoring and dispatching of the power system at the National Load Dispatch Center of Vietnam (EVNNLDC).