TY - JOUR
T1 - Islanding detection techniques for grid-connected photovoltaic systems-A review
AU - Khan, Mohammed Ali
AU - Haque, Ahteshamul
AU - Kurukuru, VS Bharath
AU - Saad, Mekhilef
PY - 2022
Y1 - 2022
N2 - Photovoltaic (PV) systems are increasingly assuming a significant share in the power generation capacity in many countries, and their massive integration with existing power grids has resulted in critical concerns for the distribution system operators. Most of these critical concerns arise due to abnormalities at the grid side (undervoltage and short-circuit events), and as a consequence of failing islanding detection (ID). Generally, an ID mechanism operating with the PV system in a grid connected environment, should be capable of disconnecting the PV from the grid in case of grid abnormalities by obeying to specific grid codes. For any condition if the ID mechanism fails to adhere to this specific requirement, the grid abnormality will disperse a large part of generation unit, and also damages the equipment connected in the grid. In light of the above concerns, numerous ID techniques (IDTs) have been introduced in the literature. These techniques are categorized into five classes, based on the input parameters, point of implementation in the system, and the methodology adapted to develop them. This paper is focussed on surveying the various IDTs developed in the literature to identify their advantages and disadvantages. A machine-learning strategy based on signal processing is proposed to overcome the drawbacks of existing IDTs in the current technical literature. The signal processing part is implemented with wavelet transform, and the support vector data descriptor (SVDD) is trained as a machine learning classifier. To assess the operation of the proposed approach, experimental analysis is carried out on a grid-connected PV system in different islanding conditions. The results identified training accuracy, testing accuracy, and an average detection time of less than for all the testing conditions.
AB - Photovoltaic (PV) systems are increasingly assuming a significant share in the power generation capacity in many countries, and their massive integration with existing power grids has resulted in critical concerns for the distribution system operators. Most of these critical concerns arise due to abnormalities at the grid side (undervoltage and short-circuit events), and as a consequence of failing islanding detection (ID). Generally, an ID mechanism operating with the PV system in a grid connected environment, should be capable of disconnecting the PV from the grid in case of grid abnormalities by obeying to specific grid codes. For any condition if the ID mechanism fails to adhere to this specific requirement, the grid abnormality will disperse a large part of generation unit, and also damages the equipment connected in the grid. In light of the above concerns, numerous ID techniques (IDTs) have been introduced in the literature. These techniques are categorized into five classes, based on the input parameters, point of implementation in the system, and the methodology adapted to develop them. This paper is focussed on surveying the various IDTs developed in the literature to identify their advantages and disadvantages. A machine-learning strategy based on signal processing is proposed to overcome the drawbacks of existing IDTs in the current technical literature. The signal processing part is implemented with wavelet transform, and the support vector data descriptor (SVDD) is trained as a machine learning classifier. To assess the operation of the proposed approach, experimental analysis is carried out on a grid-connected PV system in different islanding conditions. The results identified training accuracy, testing accuracy, and an average detection time of less than for all the testing conditions.
KW - Islanding detection
KW - Phasor measurement units
KW - Distribution Generators
KW - Non-Detection Zone
KW - Point of Common Coupling
KW - Support Vector Data Description
U2 - https://doi.org/10.1016/j.rser.2021.111854
DO - https://doi.org/10.1016/j.rser.2021.111854
M3 - Article
SN - 1364-0321
VL - 154
JO - Renewable and Sustainable Energy Reviews
JF - Renewable and Sustainable Energy Reviews
ER -