IGridEdgeDrone: Hybrid Mobility Aware Intelligent Load Forecasting by Edge Enabled Internet of Drone Things for Smart Grid Networks

Abstract

With the growing prevalence of Internet connectivity in the civilized world, smart grid technology has become more practically relevant to implement. The smart electric grid is more than just a generation and transmission infrastructure. Modernizing such electric grids to automate the process of tracking the electricity consumption at multiple locations, while intelligently managing the supply is an exciting transformation, which offers both challenges as well as opportunities. Further, it must consider the change in prices with demand throughout the day. Advanced communication policy and intelligent sensing mechanism and decision making must be adapted to collect, monitor, and analyze real-time information, performing automatic metering, home automation, and inter-grid communication within vast geographical distance. In this paper, two crucial issues in the domain of smart grid management and communication have been addressed and the potential solution is provided. Firstly Delay Tolerant Network assisted the Internet of Drone Things based communication paradigm has been modeled for smart-grid communication in intermittent connective smart grid networks. The hybrid cluster-based 3D mobility has been engineered to that pursue the information sharing and offloading within IoT based cloud infrastructure. The routing mechanism results in 97% message delivery in 5 MB buffer size and about 8.5 × 106 J data transmission energy dissipation. In the latter half of this work, a load forecasting strategy is proposed, combining the mathematically robust gradient boosting strategies and a popular Deep Learning methodology, termed the Long Short-Term Memory approach. A hybrid architecture is developed for enhanced prediction in the presence of noise and faulty transmission of data from the physical layer. Further, the proposed model is also capable of generalization to a variegated set of data and produces forecast results with 0.167 MSE score, 7.231 MAE score, and 4.9% resource utilization which are better than the conventional frameworks on resource-constrained edge computing platforms.