Investigation into the development of a new ANN-based relay for detecting loss of mains of embedded generation

This paper proposes a new method for detecting loss of mains (LOM), which constitutes one of the principle protection requirements for an embedded generator (EG). It particularly consider the case when the EG is "islanded" with part of the utility load which has a capacity close to that of the generator. The proposed method is based on the application of artificial neural network (ANN). Various ANN architectures, using different training and test data and different ways of presenting the data were investigated. It was possible to achieve a very high correct classification from the developed ANN network which only requires sampled inputs from the instantaneous values of three phase voltages and currents signals. Using this proposed method it has also been found that the ANN-based relay not only correctly classified the LOM cases when an EG and the "islanded" load have matching capacities, but also able to differentiate between the LOM and loss of parallel line which previously reported as being a difficult problem.