Artificial Neural Network Approach for Islanding Detection in Inverter Based Distributed Generator with a Forced Transient in System Frequency

In this paper، an intelligent islanding detection method is presented for inverter based distributed generation (DG) using probabilistic neural network (PNN) and wavelet transform. The presented method is based on the change of DG reactive power reference (Qref) in inverter control interface to create a small forced transient in frequency and its derivative. Changing the Qref causes a forced transient in system frequency and also in its derivative in islanding conditions. The main idea is to use the created transient in frequency derivative in islanding conditions. The PNN is trained by features extracted from the frequency derivative data through the discrete wavelet transformation (DWT) in islanding and non-islanding conditions. The proposed method is evaluated in islanding and non-islanding conditions، using PSCAD/EMTDC and MATLAB software. Simulation results show that the proposed method has a proper operation in the islanding and non-islanding conditions.