Skip Navigation
Search

Doctoral Defense

Monitoring and Control of Large Scale Power Grid Failures with Imperfect Communication

Jose Cordova-Garcia

September 29, 2017
12:00PM
Light Engineering room 250
Advisor: Prof. Xin Wang

Abstract: The wide area monitoring and control network is the core component of the envisioned Smart Grid (SG). A common assumption in the design of SG applications is that such data network behaves perfectly, i.e. no delay, no packet loss. In this dissertation, we examine the effect of communication issues on the power grid operation and present the design of communications-aware schemes for monitoring and control schemes of power line failures.

Recently, high precision data have become available through Phasor Measurement Units (PMUs). Ideally, using phasor data the system-wide status of power lines can be monitored. We argue that the discovery of line failures (outages) can be compromised by the occurrence of different missing data patterns according to the topology of the PMU data network. We propose learning outage subspaces of individual nodes and the adaptive selection of nodes to enable robust failure detection. Additionally, we present a comprehensive design of a power line monitoring network based on wireless sensors in case PMUs are not available.

Upon the discovery of a failure and to prevent failure expansion, i.e. cascading failures, the control center should plan system-wide control actions that must be communicated through the control network. We present a communications-dependent cascade model to describe the impact of delayed control actions on failure control. Based on this model, a control scheme is proposed that incorporates delay information into the optimal control formulation. For the distributed control scenario, we present a control separation scheme based on virtual representation of tie-line variables. The proposed distributed control scheme provides low communication-overhead and allows the operator to set cooperation agreements to meet the global energy demand required.

Finally, we argue that graph-based techniques can be incorporated in the design of optimal control strategies. We propose a failure control algorithm that can efficiently tradeoff between the topological and electrical characteristics revealed by the power grid graph representation to alleviate failures. We show that social characteristics derived from the physical information of the grid can be to reduce the non-monotonic failure effect expansion that is commonly observed in power grids.