Real-time disturbance detection in LV islanded microgrid

The rate of change of frequency following any disturbance in LV islanded microgrids is not only relatively high compared to conventional system, but also becomes location specific, due to their high R/X ratio lines. This makes the disturbance detection challenging. Conventional power imbalance calculation methods are not applicable due to variation of inertia with time. This paper proposes a novel real-time disturbance detection technique, independent to the system inertia. Three different Support Vector Machine regression models are created with three different feature selections. Results show that, some local parameters such as voltages at different locations should also be included with the frequency features for the accurate disturbance detection. This method can detect the amount of load disturbance within 0.5 second following the disturbance. The proposed algorithm is tested with an unseen test data. A multi-source islanded microgrid is used to collect the data, which is modelled in MATLAB/SIMULINK.