Modeling uncertainty for middleware-based streaming power grid applications

The power grid is incorporating high throughput sensor devices into power distribution networks. The future power grid needs to guarantee accuracy and responsiveness of applications that consume data from multiple sensor streams. The end-to-end performance and overall scalability of cyber-physical energy applications depend on the middleware's ability to handle multi-source sensor data, which exhibits uncertain behavior under highly variable numbers of sensors and middleware topologies. In this paper, we present a parametric approach to model middleware uncertainty and to analyze its effect on distributed power applications. The models encapsulate the entire data flow paths from sensor devices, through network and middleware components to the power application nodes that utilize sensor data streams. Using the Ptolemy II framework for modeling and simulation, we generate Monte Carlo samples of uncertain parameters that are used to generate parameterized middleware models that are used in end-to-end Discrete-Event(DE) system simulation simulation. The simulation results are further analyzed using regression methods to reveal the parameters that are influential in the limiting middleware behavior to achieve temporal requirements of the power applications.