Islanding Detection for Distributed Generators Based on Artificial Neural Network and S-transform

The increasing connection of Distributed Generator (DG) in electrical power systems brings benefits such as redundancy; however, problems like islanding are also present. Passive techniques are the most used for islanding detection because they are inexpensive and easy to implement. Nevertheless, these techniques are rather dependent on the power unbalance between the loads and generation at the time of islanding to detect it properly. With the aim of reducing this problem, techniques based on machine learning have received much attention recently. This work proposes a method for islanding detection based on the frequency spectrum at the terminals of the DG, which is obtained using the S-transform. After that, an Artificial Neural Network (ANN) is used to identify whether or not the event is an islanding. Using a sampling rate of 64 samples per cycle, the method can be easily implemented in digital relays already present in the system. In conclusion, the proposed method obtained a greater precision than conventional methods, and with a smaller detection time, being around 26 ms.