SUMMARYWith continued increase in the electrical energy demand and tendency towards maximizing economic benefits in power transmission system, especially in the liberalized electricity markets, real-time voltage security analysis has become a growing concern in electric power utilities. However, static analysis methods, such as power flow based methods, have difficulty in evaluating voltage stability and some voltage stability feasible region boundaries may not be correctly analyzed by these methods due to the use of simple models for the system components. On the other hand, dynamic modeling and evaluation of voltage stability are complex, expensive, and time consuming. In this paper, a new feature selection technique combined with a probabilistic neural network (PNN) is proposed for this purpose. A major difference of our paper with the previous research works in the area is that this paper proposes voltage stability prediction, while the previous works usually focus on the voltage stability evaluation. The proposed dynamic voltage stability prediction method is examined on the IEEE 14-bus and New England 39-bus test systems and the effectiveness of the proposed method is demonstrated. Also, the effect of different load models, branch contingencies, and generator contingencies is evaluated. Another advantage of the proposed prediction method is that it can be used for varied topologies and configurations of the power system.