Sunteți pe pagina 1din 1

Architecture of a Smart Microgrid Distributed Operating System

Bruce McMillin, Ravi Akella and Derek Ditch Computer Science Missouri University of Science and Technology
Future Renewable Electric Energy Delivery and Management (FREEDM)
The FREEDM distribution system may be viewed as a smart microgrid with the goals of power and energy management and reliability enhancement.

Gerald Heydt Electrical Engineering Arizona State University

Ziang Zhang and Mo-Yuen Chow Electrical & Computer Engineering North Carolina State University

Features of FREEDM System


The FREEDM microgrid is organized into Intelligent Energy Management (IEM) nodes, each consisting of advanced power electric technologies of Solid State Transformer (SST), Distributed Renewable Energy Resource (DRER), and Distributed Energy Storage Device (DESD).

Need for Distributed Operating System

The Distributed Grid Intelligence (DGI) is a major computational aspect of the FREEDM microgrid with a portion of DGI running on a computer embedded in each IEM node.

The DGI is a collection of distributed algorithms that manage power and computational resources within the microgrid in a distributed fashion. DGI contains a broker that integrates plug-in software modules that embodies the functions of configuration management, power management, and fault detection and reconfiguration.

Broker
The broker runs as a process that manages individual POSIX threads that instantiating each software module. Listens for incoming connections and provides mechanism to: Maintain TCP connections among IEMs Dispatch incoming messages to appropriate software modules and send messages based on connections
POSIX BROKER Creates Unique IDs (UUID) for each node

Group Management
Manages group membership of IEM nodes by determining the neighbors Handles network partitions or failure of node(s) (through Reorganization) Elects a leader of the group which has special group information to be used by other modules or a new node that joins the group

Power Balancing
Balances the power on the microgrid in a way to meet the net demand/supply through negotiation among peer IEM nodes to control individual SSTs to add or subtract power to / from a shared power interconnection bus A distributed version of Locational Marginal Pricing (D-LMP) is integrated with this module to optimize energy storage, renewable energy controls, line loading management in the distribution network, and power usage

Different classes of the Broker Implementation

Sets up all software modules CBROKER Register each software module with broker -This registers XML portion of ptree (message structure) for this module -Registers callbacks to handle these sections of a ptree

A new node forms a new group with itself as leader

Election between the leaders of subgroups to merge into a single group

CONNECTION MANAGER Maintains an iterator list for all connection objects

Listen on network port for incoming connections CMESSAGE CCONNECTION Message is parsed into ptree Does XML conversion /parsing on the incoming/ outgoing messages

CDISPATCHER Deliver the submessages associated with every incoming message

Initiate connection object Handle Send/Receive, Read/Write

Network partition due to failures leads to election within subgroups

Leader node Failed node New node Member node

Incremental Cost Consensus


At higher levels of penetration, renewable energy controls could have an impact on the transmission LMPs at the supply points. The consensus concept for such cases assures that the D-LMPs are calculated across the distribution network with higher sampling rate. An appropriate consensus algorithm can guarantee that all the consensus variables converge to a common value, , asymptotically within the microgrid.
The figure shows the distributed control consensus network: the local controller will update its own based on its neighbor's . Leader generator (G1) is selected to control the increase or decrease of the group , depending on the sum of total power generation and the actual load .

State Collection
The Chandy-Lamport state collection algorithm is used to collect a logically-consistent state (one in which causality between actions is preserved) of the distributed system It employs correctness predicates to determine correct/faulty behavior (such as for fault diagnosis).

Conclusions and Future Work

Distributed This

microgrids require more management than classical distribution systems. work describes distributed microgrid. an architecture for a

This architecture embraces key cyber and physical


components to perform both economic dispatch and system management.

Fault Management
DGI detects internal software and hardware faults and reports to the Integrated Fault Management (IFM) module. If a fault is detected, the consensus system and group manager are contacted to initiate reconfiguration around the failed component.

To

implement the system requires work in distributed algorithms, power systems, and power system economics.

Acknowledgements
This work was supported in part by NSF MRI award CNS-0420869, NSF CSR award CCF-0614633, NSF ERC award EEC-08212121.

S-ar putea să vă placă și