Applying neural network and genetic algorithm for optimal placement of ultra-capacitors in metro systems

The aim of this paper is to find the optimum location for a set of ultra-capacitors connected to DC electric supply of a metro line (Stationary Energy Storage Systems S.E.S.S). In order to achieve minimum energy demand from electric mains, the genetic algorithm is used as the optimization method. The system behavior is modeled using neural networks as a surrogate model to formulate the mapping between the variables and the objective. Matlab Simulink model is used to generate the training and test sets for the neural network algorithm. Furthermore an Energy-management strategy based on a frequency approach is used for the on-board energy storage of trains. Based on this strategy the share of every on-board storage source (flywheel/battery) is specified. It is shown that application of the proposed method and optimum arrangement of ultra-capacitors leads to a considerable reduction of consumed energy, improvement of DC voltage profile, reduction of substation's peak current and also reduction of network losses. The method is highly effective when the electric system becomes more and more complex.