Scalable real time data management for smart grid

Abstract

This paper presents GridMW, a scalable and reliable data middleware layer for smart grids. Smart grids promise to improve the efficiency of power grid systems and reduce green house emissions through incorporating power generation from renewable sources and shaping demands to match the supply. As a result, power grid will become much more dynamic and require constant adjustments, which requires analysis and decision making applications to improve the efficiency and reliability of smart grid systems. However, these applications rely on large amounts of data gathered from power generation, transmission, and consumption. To this end, millions of sensors, including phasor measurement units (PMU) and smart meters, are being deployed across the smart grid system. Existing data middleware does not have the capability to collect, store, retrieve, and deliver the enormous amount of data from these sensors to analysis and control applications. Most existing data middleware approaches utilize general software systems for flexibility so that the solutions can provide general functionality for a range of applications. However, overheads incurred by generalized APIs cause high latencies and unpredictability in performance, which in turn prevents achieving near real time latencies and high throughput. In our work, by tailoring the system specifically to smart grids, we are able to eliminate much of these overheads while still keeping the implementation effort reasonable. This is achieved by using a log structure inspired architecture to directly access the block device layer, eliminating the indirection incurred by high level file system interfaces. Preliminary results show our system can significantly improve performance compared to traditional systems.