Tech Briefs

Communication network-based power measurement and control approaches can impact power networks in the face of weapons of mass destruction-induced cascading failures.

Emerging distributed power grid systems will integrate tightly coupled networks, namely a power grid and a communication network, to provide distributed power monitoring and control. While distributing power monitoring and control away from a central site enhances the robustness of the power grid to multiple, dispersed failures and attacks, such communication-based distributed control schemes can introduce complex cascading failure scenarios in the face of large-scale weapons of mass destruction (WMD) attacks.

The objective of this project is to model and analyze the interactions between coupled communication networks and power distribution grids so that potential cascading failures in the composite complex bi-infrastructure network can be examined, analyzed, and avoided. This work is motivated by fundamental challenges presented by potential distributed power grid topologies and structures currently under discussion in the power community. Using a combination of graph-theoretic dynamical modeling of cascading failures, optimal distributed control algorithm design, and distributed estimation techniques, it is possible to see how various communication network-based power measurement and control approaches impact the robustness and efficiency of power networks in the face of WMD-induced cascading failures.

Investigation of cascading failure mechanisms has shown that load shedding can be an effective method of reducing the likelihood of cascading failure, if implemented in a timely manner. However, currently load shedding is generally implemented by an operator manually, following a request from a system operator. Moreover, requests are subject to human judgment, and sometimes fail to be implemented correctly, in a timely manner, and/or in their entirety. An automated response, making use of energy storage, would have two benefits: first, its implementation would not depend on human judgment, and second, its impact on end users would be less severe. For example, storage units (thermal storage, community electric storage, utility-scale electric storage) may be deployed to produce a certain level of load shed, rather than resorting to rolling blackouts or brownouts.

There are a number of challenges in the implementation of automated, fastacting load shedding strategies for WMD protection. First, the mechanism must be compatible with “normal” grid operation mechanisms. Second, the mechanism must be compatible with local regulations. For example, control by real-time pricing (RTP) may be possible in some regions, but not others. Where RTP is not available, other control mechanisms must be adopted. Finally, the control strategies must be robust – for example, high-bandwidth communications requirements may pose problems because, in the case of WMD attack, communications networks may become congested or even inoperable. Two strategies for load control were considered – RTP and stochastic.

With RTP control, the cost of power is a function of the residual capacity of the system. This is in line with industry practices, and is also compatible with WMD response, since a WMD attack would automatically result in reduced capacity. The customers use automated response to activate their storage devices. Thermal load is either met by depleting storage or by purchasing power from the network, depending on an internal power price threshold. The threshold for meeting thermal load, in turn, is set based on the state of charge of the storage – a high threshold when state of charge is low, and vice versa. The threshold for charging is set similarly.

Stochastic control (SC) could be used in regions where RTP is not allowed by regulation. In this case, the utility or some control agency would send a signal to all customers, between 0 and unity, representing the probability of activating a distributed resource. In turn, each resource generates a random number, which is compared with the utility signal. Based on the result of the comparison, the device is either activated or not. With stochastic control, loads can choose to op-in, and be financially compensated as a result.

Results have shown that both RTP and SC are effective means of controlling the total load on the grid, and can respond effectively to both renewable energy intermittency and load shedding requests. Moreover, the bandwidth for communication is very low – only requiring a common signal to be broadcast at regular intervals, typically on the order of minutes. These same mechanisms could be used for enhanced readiness to WMD threats. For example, it is possible to increase the storage SOC in cases of enhanced threat, simply by modifying the RTP or the stochastic signal.

Simulations were performed to reveal the effects of the control/communication system parameters on the cascading failures in power grids. Simulations have shown that vulnerabilities in the control/ communication systems can increase the probability of large cascading failures initiated by small disturbances over the power grid. In addition, the topological location and characteristics of the failures affect the cascading behaviors in the power grid. Since inhibition and clustering can be important attributes of WMD attacks, which exhibit a great level of spatial correlation, the effects of spatial inhibition and clustering among the uncontrollable load buses were studied. These effects can be important due to the nature of certain disaster events that may affect the power grid and their control/communication systems.

This work was done by Majeed M. Hayat, Andrea Mammoli, Yasamin Mostofi, and Patrick Bridges of the University of New Mexico for the Defense Threat Reduction Agency. DTRA-0002

This Brief includes a Technical Support Package (TSP).

Cascading Failures in Coupled Distributed Power Grids and Communication Networks (reference DTRA-0002) is currently available for download from the TSP library.

Please Login at the top of the page to download.