Smart grid data analytics for decision support

As electric grid sensor data originating from several sensors such as the phasor measurement units (PMUs), intelligent relays, and new installation of smart meters, Plug-in Hybrid Electric Vehicles (PHEV) or Gridable Vehicles (GV), are exponentially growing, the data analytic platform for Smart Grid has huge potential (generation, transmission or distribution) and can play a significant role in the decision making process for meaningful data interpretation to act promptly or automate the grid process to avoid any failures or instability in the grid. This paper focuses on identifying the variables of interest that are important in the electric grid embedded in distributed real time data engines which will help decision support process for system operators. More specifically, the applicability and performance of M5 model and J 48 decision tree machine learning technique is investigated using the real electric grid data. We have presented how decision tree model such as M5P can support system operators in making effective decision in the Smart Grid. Two sets of test data are used in this paper; the first data set is taken from a 10 unit commitment with 50000 Gridable Vehicle and the latter analyzes a weekly New York City (NYC) demand data from NYISO.