Scalability, Resilience, and Complexity Management in Laminar Control of Ultra-Large Scale Systems

Ultra-large scale control systems that are based on layered optimization decomposition and Network Utility Maximization (NUM) have structural properties and operational modes that givesuch control systems useful scalabilityand anti-fragile properties. We denote control architectures based on these principles as Laminar Control and we suggest that the structure so defined has properties such as layer-by-layer segmentation of control signal traffic, abstraction of local state into scalar signal flows, self-similarity of data flow patterns at each layer, and support for islanded operation and re-connection. While the underlying physical networks (such as power systems) and associated communication networks are not necessarily self-similar, the imposition of the Laminar Control paradigm allows the creation such a structure at the application node level. Consequently, instead of having a set of agents with randomly structured logical connections, the optimization nodes become hubs and furthermore,it may be possible to restrict peer-to-peer communication at any level in the hierarchy to delimited specifiable domains. State determination can be similarly partitioned so that domain state need not be shared globally. In combination, these properties can invest distributed control networks with three valuable characteristics: scalability of control system real time network communications, resilience of the logical control network, and complexity bounds. Introduction: Laminar Control Architecture and Data Flows Distributed and hierarchical control methods have been available for decades. In the case of electric power systems, some of these concepts have been used in portions of the grid, but not as an Ultra-Large Scale (ULS) control. By ULS we mean the concept developed at the Software Engineering Institute to describe extremely large, complex systems with the following characteristics: • Decentralized data, development, and control • Inherently conflicting diverse requirements • Continuous (or at least long time scale) evolution and deployment • Heterogeneous, inconsistent, and changing elements • Normal failures (failures are expected as a normal part of operation) In the power grid domain, certain approaches to wide area grid control have employed a single physical variable that presumes to characterize a key aspect of system state. At the transmission and generation level, system frequency is used for this purpose. It is widely used in incremental area balancing via Area Control Error and Automatic Generator Control (AGC).System frequency has also been proposed as the basis for control of large number of responsive loads not owned by the electric utility.At the distribution level, feeder voltage is used in Volt/VAr control systems, both inthose that act centrally and those that are composed of a collection of independent agents.Such methods have enjoyed a degree of success but encounter difficulties in two areas: 1) Such systems can become unstable due to feedback through the grid itself 2) When multiple functions want to use the same infrastructure for possibly competing or even conflicting purposes, as is happening with Peter Feiler, John Goodenough, et al., Ultra‐Large‐Scale Systems The Software Challenge of the Future, Software Engineering Institute, June 2006 NERC Resources Subcommittee, Balancing and Frequency Control, available online: http://www.nerc.com/docs/oc/rs/NERC%20Balancing%20and%20Frequency%20Control%20040520111.pd f PNNL Staff, Grid Friendly Controller Helps balance Energy Supply and Demand, available online: http://readthis.pnl.gov/MarketSource/ReadThis/B3099_not_print_quality.pdf Naveen Venkatesan and S K Solanki, Coordination of Demand Response and Volt/VAR Control Algorithm Using Multi‐Agent System, IEEE 2012 Transmission and Distribution Conference and Exposition, 2012 IEEE PES, Orlando,