Hierarchical Control of Microgrid Using IoT and Machine Learning based Islanding Detection /

Abstract

Due to the increase in pollution free renewable energy resources and efficient energy storage systems, the need for a proper Microgrid control structure is increasing. Information communication technology (ICT) plays a key role in the efficient performance of Microgrid. With the development of new technologies like 4G, 5G, and fast sensing devices and embedded systems, the internet of things (IoT) can play a huge role in Power systems. The combination of Machine learning and IoT technology can produce more efficient and accurate results in real time. In this work, IoT based hierarchical control structure for remotely located Distribution Generation (DGs) has been developed along with a Machine learning based island detection system. Microgrid has two operational mode i.e. grid connected mode and islanding mode. During grid connected mode, the voltage and frequency are utility dominant parameters, while Microgrid injects power in the system. In islanded mode, the microgrid acts as standalone and the main function of control is to maintain voltage and frequency into the desired range. In this work for the primary layer, Proportion integral (PI) based current following control algorithm was developed for a grid connected mode, whereas droop controller was used for islanding mode operation of the microgrid. In the secondary layer, to remove deviance of frequency and voltage occurred in the primary layer, IoT based communication system was developed to share voltage and frequency among each DG. The context aware policy was also applied to reduce the amount of data in the communication channel. In tertiary layer, a remote Machine learning algorithm was applied using cloud through which the microgrid shifts to gird connected mode or standalone mode. The data for the training of machine learning model was taken by simulating different islanding scenarios i.e increase in load demand, decrease in load demand, or during faulty conditions. The proposed system was simulated on IEEE- 13 bus system in MATLAB/ SIMULINK. The Microsoft Azure cloud services were used for the IoT and implementation of the Machine learning model.