Neural Networks Applications for Balancing of Reactive Power between Generating and Demand Points

Abstract

The trained ANN has been tested with input data not previously seen by the ANN. Radial basis (exact fit) algorithm used for providing an estimate of the reactive power at collapse in the presence of synchronous compensator and co generator. The trained ANN provides reasonable results in an extremely short time (almost instantaneously) when compared with other existing methods utilizing successive power flow or optimization. The inputs/outputs for training is easily obtained via successive power flows and do not require extensive computation when compare with certain other energy based methods. Index Terms: Neural networks, voltage stability, reactive power, and compensation