APPLICATION OF ARTIFICIAL NEURAL NETWORKS IN REACTIVE POWER OPTIMIZATION

The reactive power optimization problem is one of the most important problems facing dispatching engineers when they are operating large scale power systems. Reactive power optimization is mathematical approach of the power system optimization problem which is to determine the least control movements to keep power system at a most desired state. It is very flexible and powerful tool which cane address a wide range of planning and operation studies. However the complexity of reactive power optimization often discourages the user. This paper proposes a new artificial neural network-based approach (ANN) for reactive power optimization of interconnected power systems. Feed-forward ANN with Back Propagation training algorithm is used and the training data is obtained by solving several abnormal conditions using Linear Programming (LP). Considering generator voltages, reactive power sources and transformer taps as control variables, and load bus voltages and generator reactive powers as dependent variables. The relations are derived according to sensitivity relations based on Newton-Raphson load flow equations. The 220 kV network of the Unified Power System (UPS) of Egypt was used in this paper to test the suggested approach. Results proved the high accuracy and the suitability for the on line applications.