Load flow solution in electrical power systems with variable configurations by progressive learning networks

In recent years, interest in the application of soft computing techniques to electrical power systems has rapidly grown; in particular the application of artificial neural networks (ANN) and genetic algorithms (GA) in the solution of load-flow problem in wide electrical power systems, as valid alternative to the classical numerical algorithms, is an interesting research topic. In the present paper, a refined solution strategy based on statistical methods, on a particular the grouping genetic algorithm (GGA) and on progressive learning networks (PLN) is presented to solve load-flow problems in electrical power systems taking also into account configuration changes; in particular, a procedure to solve the system when a link is removed, or added, is described and implemented. Test results on the standard IEEE 118 bus network have demonstrated the good potential and efficiency of the procedure.