Islanding detection for grid integrated distributed generation using adaptive neuro-fuzzy inference system

Abstract

The use of distributed Generation (DG) for power supply and quality improvement is gaining more acceptance. Power Utilities connection standard for DG requires that islanding be quickly detected and DG sources isolated to prevent damage to power utility equipment, harm to power system operators, maintain power quality and ensure that protective devices do not mal-operate such that the power system security is maintained. Adaptive Neuro-fuzzy Inference System (ANFIS) is proposed for detection of islanding in power distribution networks with DGs connected. The ANFIS technique relies on the passive parameter during islanding for islanding detection. These passive parameters, Voltage, rate of change of frequency (ROCOF) and rate of change of power (ROCOP) forms the inputs for training and validation of the proposed islanding detection method. A segment of the South-East Nigeria 33/11KV power distribution network is used for generating the training data for training the ANFIS network and evaluating the performance of the islanding detection techniques. Performance evaluation was based on Time performance region and Non-detection Zone (NDZ). Results of simulations showed that the ANFIS islanding detection method detected islanding faster and in all case give a smaller NDZ compared with the frequency relay method.