Distributed-Swarm: A Real-time Pattern Detection Model based on Density Clustering

The advancement of power technology and the improvement people’s living standards promote expansion grid scale sharp rise in electricity consumption. In system, due to use various sensors, we can collect a large number data (eg. spatial-temporal information electric vehicle charging). Usually, such is generated form stream. analysis mining be widely applied equipment condition monitoring maintenance, user anomaly warning, urban other scenarios. Among them, pattern detection plays key role analysis. Since as charging time-sensitive, it crucial perform real-time systems. However, state-of-the-art methods are built on batch mode. Extending works directly an online environment tends result (1) expensive network cost, (2) high processing latency, (3) low accuracy results. this paper, propose framework for frequent motion distributed environment. Through softmax differentiation function, filtered reduce workload improve performance framework. At same time, concept historical state matrix solve problem that nodes each physical partition not perceive other. Extensive experiments conducted real dataset experimental results show our about 70% faster than baseline methods, which proves huge advantage approach over available solutions literature.