Artificial neural network-based dynamic equivalents for distribution systems containing active sources

An approach to identify generic dynamic equivalents to distribution systems using recurrent artificial neural networks (ANN) is presented. It is expected that in the near future a large number of active sources will be utilised within distribution systems and thus, neither detailed modelling nor lumped-load representation for distribution areas will be acceptable. Therefore, the paper suggested training a recurrent ANN to represent the dynamic behaviour of the distribution network. To involve the dynamic characteristics in the ANN, values of the features that are involved are also introduced at the input layer, thereby defining the order of the dynamic equivalent. The approach depends on variables at the boundary buses, hence no knowledge of the parameters and the topology of the distribution system is needed. At the same time, the computational requirements and the accuracy of the proposed technique are independent of the size and complexity of the network. A 16-machine test network with 112 active distributed sources in the low-voltage area is used to verify the suggested method. Comparisons between the response of the original system and the ANN-based dynamic equivalent show the accuracy of the equivalent model and the validity of the proposed method.