Islanding detection in grid-connected photovoltaic distributed generation using inverter DC-Link voltage

• Main Author: Kumo, Mohammed Saidu
• Format: Thesis
• Language: English
• Published: 2017
• Online Access: http://psasir.upm.edu.my/id/eprint/71167/1/FK%202017%2046%20-%20IR.pdf; http://psasir.upm.edu.my/id/eprint/71167/

There is an increase in the spread of Distributed Generation (DG) in the form of solar photovoltaic (PV), wind turbines, fuel cells, etc. as renewable energy resources, giving numerous advantages if connected to the existing electric grid system. However, their integration into the grid introduces certain problems to the conventional distribution system, of which islanding detection is the most important. Islanding a situation in which a DG powers its local load while in the absence of the grid supply. The occurrence of islanding causes numerous problems to the DG, the grid and the maintenance personnel. Therefore, its occurrence must be detected within two seconds. The aim of this thesis is to study the viability of using the inverter DC-Link voltage as a parameter for passive islanding detection. The most significant shortcoming of passive islanding detection methods is the presence of large non-detection zone (NDZ), which is a region of power mismatch between the DG and the local load where islanding cannot be timely detected. For the study, a detailed model of 100 kW, 480V, gridconnected PV DG is implemented in MATLAB/Simulink. Then the response of DCLink voltage to system load variations in islanding and grid-connected modes were studied. Furthermore, its responses to islanding on three inverter interface controllers, the constant power controller (CPC), the constant current controller (CCC) and the open-loop controller (OLC) were evaluated. The NDZ of the DC-Link voltage was determined using the UL 1741 test conditions on the IEEE 1547 anti-islanding (AI) test circuit. The effectiveness of any AI method depends on its NDZ, therefore the NDZ of DC-Link voltage was improved using the Detrending Algorithm. The effect of nonislanding grid-side faults on DC-Link voltage was equally examined. The system performance is verified with the MATLAB time-domain simulations. DC-Link voltage was found to be viable for passive islanding detection with an NDZ of +20%. The NDZ is improved to ± 1.0% by detrending the DC-Link voltage, which is a novel achievement. An AI detection system using DC-Link voltage and detrended DC-Link voltage as inputs was able to detect the occurence of islanding within 33 ms against the 2 seconds required by the standards. Detrended DC-Link voltage responds to each nonislanding event distinctively. To sum it up, DC-Link voltage is viable for being a parameter for passive islanding detection as it is very fast in detecting islanding, discriminative from non-islanding faults and has almost zero NDZ. The fact is validated in comparison with work done with wavelet analysis based on a neuro-fuzzy system.