A new approach to ANN-based real time voltage stability monitoring and reactive power management

A new approach to ANN-based real time voltage stability monitoring and reactive power management in a power system has been proposed in this paper. In this approach a separate ANN has been trained for each of the vulnerable load buses in the system from the voltage stability point of view. The vulnerable load buses of the system are identified by modal analysis of the system reduced Jacobian (of Q-V coupling) matrix. The output obtained from each of these ANNpsilas is the ldquodistance to voltage collapserdquo in terms of the MVAR margin available at the specific vulnerable load bus corresponding to the current operating point (loading condition) of the system. The novel inputs proposed for these ANNpsilas consist of the complex power contributions of each of the generators and other controllable reactive power sources provided in the system in meeting the load at the particular load bus at the current loading condition and the electrical distances between these sources and the load bus, in addition to the conventionally used inputs of reactive power margins at the sources and the complex power drawn and voltage magnitude at the particular load bus. Special care has been taken in generating a comprehensive set of realistic loading conditions in the system, including generating unit outage conditions and relevant ldquoN-1rdquo network element outage contingency conditions.