Temporal, Functional and Spatial Big Data Computing Framework for Large-Scale Smart Grid

With the deployment of monitoring devices, the smart grid is collecting large amounts of energy-related data at an unprecedented speed. The smart grid has become data-driven, which necessitates extracting meaningful data from a large dataset. The traditional approach of data extraction improves the computing efficiency in temporal dimension, but it is made for only one task in the smart grid. Moreover, the existing solutions neglect the geographical distribution of computing capacity in a large-scale smart grid. The future large-scale smart grid will run over the internet of energy where the dataset will be sent to a specific destination along power routers hop-by-hop. Consequently, we design a novel temporal, functional and spatial big data computing framework for large-scale smart grid. In functional dimension, we divide every dataset into sub-groups, each of which has data items shared by different tasks. In spatial dimension, we determine which location the power router should be placed to harvest computing resources used for extracting the sub-group of data items. Our method achieves a promising computing efficiency approaching to the optimal solution with 95 percent convergence ratio, and it saves the in-path bandwidth with 81 percent improvement ratio over benchmarks.

I. Introduction

The data of smart grid phenomenally increases, and almost all of nuclear power stations collect a large amount of data from wireless sensors, distribution network equipment, and even communication objects interacting with consumers. It is quite challenging to process such massive, multi-source and redundant data using existing technologies. In fact, only a partial set of massive data is meaningful to executing one task in the smart grid. Without effectively extracting useful data from huge information, the processing and decision of the corresponding task will be severely delayed. Therefore, data extraction becomes very important to improve the computing efficiency in temporal dimension.