On-line network reconfiguration for enhancement of voltage stability in distribution systems using artificial neural networks

Network reconfiguration for maximizing voltage stability is the determinati on of switching-options that maximize voltage stability the most for a part icular set of loads on the distribution systems, and is performed by alteri ng the topological structure of distribution feeders. Network reconfigurati on for time-varying loads is a complex and extremely nonlinear optimization problem which can be effectively solved by Artificial Neural Networks (ANN s), as ANNs are capable of learning a tremendous variety of pattern mapping relationships without having a priori knowledge of a mathematical function . In this paper a generalized ANN model is proposed for on-line enhancement of voltage stability under varying load conditio ns. The training sets for the ANN are carefully selected to cover the entire range of input space. F or the ANN model, the training data are generated from the Daily Load Curve s (DLCs). A 16-bus test system is considered to demonstrate the performance of the developed ANN model, The proposed ANN is trained using Conjugate Gr adient Descent Back-propagation Algorithm and tested by applying arbitrary input data generated from DLCs. The test results of the ANN model are found to be the same as that obtained by off-line simulation. The enhancement of voltage stability can be achieved by the proposed method without any addit ional cost involved for installation of capacitors, tap-changing transforme rs, and the related switching equipment in the distribution systems. The de veloped ANN model can be implemented in hardware using the neural chips cur rently available.