Dynamic security control using secure regions derived from a decision tree technique

This paper describes a real time dynamic security control technique in which rules derived from a decision tree learner are used to determine the necessary control actions to make an insecure power system state secure. A representative set of system operating states are classified as transiently stable or unstable by the extended equal area criterion (EEAC) for fast evaluation of transient stability and used as a training set for the decision tree, which outputs a set of classification rules. Security control is achieved by selecting optimal control actions to change the current, insecure state into a state within a secure region, without violating the acquired security rules. The technique is demonstrated on the 39 bus New England test system.