Decentralized Sensor Networking Models and Primitives for Smart Grid

Enhanced measuring/monitoring systems can improve grid security and resilience. This project focused on data use and data management for Smart Grid applications. Specifically, we carried out a statistical analysis of topological and electrical parameters of power grids, to understand what are the key statistical features that characterize the grid and also to generate according to the model samples that have the same statistical properties of actual grids,. This model was the first of its kind. It was meant to generate a variety of scenarios and potentially much larger test cases than the simple IEEE bus systems available for Power Systems Research, allowing researchers to test algorithm meant to advance the security and reliability of the grid on a wide range of conditions. In addition to studying the statistics of the grid parameters we turned our attention to studying the influence of the operating conditions of the grid on cascading failures events. We also defined metrics for electrical centrality of nodes which generalize common measures of importance used in graph theory to spot important elements of networked systems and showed that these locations on the grid are very close to the optimal location for placing PMU measurement devices in Hybrid Power Systems State Estimation, as they improve both the accuracy and the convergence of the non linear least square solution for the state estimation.

We then turned our attention to deriving metrics that gauge the vulnerabilities of the grid that are based on first- and second-order statistics of the line flows, which are coupled through the power flow equations. We developed a stochastic model to study cascading failures in power grids. In our model, the grid state is conditionally Markovian, given a certain line flow. The transition rates of the line from the on to off state depend on the sojourn time of the line flow above the overload threshold of the line switch. We used the statistics of the sojourn time of line flows, derived from a Gaussian model, to obtain the expected active time of the line. These expected times provide a metric of the risk of cascading failures and also the time margin to perform corrective action. In our study, we derived the first- and second-order statistics of the flows from a DC power flow model applied to the load and generation data. In the future, we plan to use flow measurements and sensor data to compute the metric online directly. The results of this project include a model of grid states as conditionally Markovian, given the process of line flows F(t); a model for the line flow statistics, using measurement data on generation/loads; new vulnerability measures, where the metric developed is the expected active time of each line before the advent of a line trip; experiments and validation of the proposed model; and software tools for vulnerability analysis and a simulation package.

Research Demo:

Metrics of Grid Vulnerability to Cascading Failures: Presented by Zhifang Wang during the TCIPG Industry Workshop in November 2011.