Application of artificial intelligence and machine learning techniques in island detection in a smart grid

Electricity grids are getting smarter day by day and simultaneously the importance of distributed generations (DGs) is increasing. The concept of DG has taken the power industry to a new height in terms of improved reliability and continuity. Integration of DGs has reduced dependency on main grid supply as they are sources of energy in a micro-grid and have the capacity to cater to the loads. However, as the penetration of these DGs has increased in energy system, various challenging uncertainties have evolved that are very different in nature from previously known traditional uncertainties. Among various types of modern uncertainties in smart grids, primary one is unintentional islanding of grids. It happens when the DGs continue to feed some portion of the load even after being disconnected from the main grid. Unintentional islanding comes with a large number of threats. Thus, it is very critical to detect the unintentional island incidents. The traditional methods used for island detections involves measurement of various parameters like voltage, frequency, etc. or the effect of injecting some intentional disturbances and analyzing the effects. The island cases are then identified by comparing the measured values with a threshold. However, the setting of the threshold value is difficult because, if the value is less sensitive, the detection will not occur very accurately. However, if a highly sensitive value of threshold is set, it will cause false detection. So an appropriate tool is required to attain high sensitivity and also high accuracy. Thus, the island detection researchers incorporated artificial intelligence and machine learning techniques, which help in fulfilling this task effectively.