prosumer-based decentralized cyber-physical architecture ......s. grijalva, m. costley,...
Post on 15-Mar-2021
5 Views
Preview:
TRANSCRIPT
1 1
Ninth Annual Carnegie Mellon Conference on the Electricity IndustryRole of Coordination in Resilient & Fine-Grain Control of Power Grids
February 4-5, 2014
Ninth Annual Carnegie Mellon Conference on the Electricity IndustryRole of Coordination in Resilient & Fine-Grain Control of Power Grids
February 4-5, 2014
Prosumer-Based Decentralized Cyber-Physical Architecture for Green Electricity InternetworksProsumer-Based Decentralized Cyber-Physical Architecture for Green Electricity Internetworks
Santiago GrijalvaDirector, Power System Engineering Center, NREL
Associate Professor, Georgia Institute of Technology (on leave)
Santiago GrijalvaDirector, Power System Engineering Center, NREL
Associate Professor, Georgia Institute of Technology (on leave)
2 2
Smart Grid functionality restores the balance
Hydro power plants
Nuclear Power Plants
Natural Gas Generators
Transmission System
Distribution Substations
Customers
Distributed storage
Solar Farms
Wind Farms
The Emerging GridThe Emerging Grid
Home EnergyStorage
EnergyEfficiency
PHEV
Rooftop Solar
Distributed wind
Commercial Customers
2© 2014 Georgia Institute of Technology
3 3
Core Trends and ParadigmsCore Trends and Paradigms
Domain Trend or Paradigm Change Future System
Objectives • Reliability and Economy + Sustainability Sustainable
Sources • From fossil fuel to renewable• From bulk centralized to distributed• Highly Variable
RenewableDistributedStochastic
Information • Can control entire system through SW• Increased digital control ->• Privacy and cyber-security issues
Cyber-ControlledCyber-Physical
PrivateActors • Consumers can also produce and store
• Consumers seek their own objectives• Massive number of actors and devices
ProsumersSmart
ScalableCarriers • Interdependencies with other systems Integrated
33
Prosumer-Based Decentralized Cyber-Physical Control and Management Architecture for Green Electricity Internetworks
© 2014 Georgia Institute of Technology
4 4
Georgia Tech ARPA-E GENI ProjectGeorgia Tech ARPA-E GENI Project
Investigators:Santiago Grijalva, Magnus Egerstedt, Marilyn Wolf, Shabbir Ahmed.
Objectives: Develop and demonstrate at large-scale:1. A control and management architecture that will allow the
electricity industry to operate with characteristics similar to the internet: Distributed, Flat, Layered, Scalable
2. A distributed services cyber-infrastructure that supports prosumer interaction. This cyber-infrastructure can be understood as an “Electricity Operating System”.
3. A real-time decentralized prosumer frequency controller4. A decentralized prosumer energy scheduler
4© 2014 Georgia Institute of Technology
5 5
Architecture: Prosumer AbstractionArchitecture: Prosumer Abstraction
• A generic model that captures basic functions (produce, consume, store) can be applied to power systems at any scale.
• The fundamental task is power balancing:
• Energy services can be virtualized.
ExternalSupply Energy
Storage
LocalEnergy
Wires
Load 1 Load 2 . . . . Load n
INT G D Loss STO STOP P P P P P
5© 2014 Georgia Institute of Technology
6 6
Architecture: Flat Electricity IndustryArchitecture: Flat Electricity Industry
Interconnection
ISO
Utility
Grid, Building, Home
S. Grijalva, "Multi-Scale, Multi-Dimensional Computational Algorithms for Next Generation Electric Power Grid", DOE Workshop on Computational Needs for Next Generation Electric Power Grid, Cornell University, Ithaca, NY, April 18-20, 2011. 6
• Interactions occur among entities of the same type (prosumers)• Can achieve “flatness”
© 2014 Georgia Institute of Technology
7 7
Architecture: Prosumer InteractionsArchitecture: Prosumer Interactions
• Interactions will leverage recent developments in networked and autonomous control.
S. Grijalva, M. Costley, "Prosumer-Based Smart Grid Architecture Enables a Flat, Sustainable Electricity Industry", IEEE PES Conference on Innovative Smart Grid Technologies (ISGT), Anaheim, California, January 17-19, 2011. 7© 2014 Georgia Institute of Technology
8 8
Architecture: Cyber-Physical Layered ModelArchitecture: Cyber-Physical Layered Model
9 9
Prosumer States
Prosumer Power AgreementProsumer Power Agreement
Given desired power levels, how to agree on power outputs?
An agreement dynamics, incorporating constraints, weights, and trust, is used to agree on the power levels across the network in a decentralized fashion.
ˆ : Desired Power: Agreed Upon Power : Actual Power
i
i
i
ppp
2
1
ˆmin || ||N
i i iipw p p
Ramachandran, Costello, Kingston, Grijalva, Egerstedt. “Distributed Power Allocation in Prosumer Networks”, NecSys, 2012.
ˆ( 1) ( ) [ ( ) ]
ˆ ˆ( 1) ( ) ( ( ) ( )) ( ( ) ( )) ( )i i k i
i i i i k j k i i jj N k N N j N
q t q t t LWLq t WLp
q t q t t w q t q t w q t q t w p p
Converges asymptotically to the global minimizer (exponentially in the algebraic connectivity of the network)
i
i ijj N
p p
9© 2014 Georgia Institute of Technology
10 10
Malicious/Malfunctioning ProsumerMalicious/Malfunctioning Prosumer
• Linearize dynamics:
• On state-space form:
• Project the (asymptotically stable) dynamics and evaluate z “after a while” and solve for d:
• Can evaluate x in a distributed way:
??Unstable!
© 2014 Georgia Institute of Technology
11 11
Decentralized ControlDecentralized Control
N. Ainsworth, M. Costley, J. Thomas, M. Jezierny, S. Grijalva: "Versatile Autonomous Smartgrid Testbed (VAST): A Flexible, Reconfigurable Testbed for Research on Autonomous Control for Critical Electricity Grids", in Proceedings of the North American Power Symposium, Champaign, IL, September 9-11, 2012.
Hardware Demonstrations
11© 2014 Georgia Institute of Technology
12 12
Decentralized Frequency RegulationDecentralized Frequency Regulation
• Model Predictive Control solvable in a distributed manner.• Each prosumer solves its sub-problem and shares its
optimal control strategy with its neighbors.
• Prosumer i strategy:
1
2 2
,..., 1
min ( 1)n
n
i i c i iu u i
p x t ru
. . : ( 1) ( ) ( )
( ) ( )i c ii i c ii i c
ij j c ij j cj N
s t x t a x t b u t
a x t b u t
21 2arg min (
i
ki i i i ii i ii i ij j ij jj Nu ru p a x b u a x b u
© 2014 Georgia Institute of Technology
13 13
Decentralized Energy SchedulingDecentralized Energy Scheduling
• 1,500+ Generator case• Full UC realistic data.
© 2014 Georgia Institute of Technology
14 14
Further Elements for De-centralized Control
Further Elements for De-centralized Control
Requirement TheorySynchronization/ stabilization
Decentralized controller for nonlinear coupled oscillators
Effect of losses Bi-directed graphs and non-odd functionsVoltage Stability Locally enforceable necessary conditionTransient Stability Locally enforceable sufficient conditionFlow Constraints Distributed enforcement of coupling constraints
14© 2014 Georgia Institute of Technology
15 15
Decentralized Computing InfrastructureDecentralized Computing Infrastructure
15
M.U. Tariq, S. Grijalva, M. Wolf, "Towards a Distributed, Service-Oriented Control Infrastructure for Smart Grid", ACM/IEEE Second International Conference on Cyber-Physical Systems, Chicago, IL, April 11-14, 2011.
© 2014 Georgia Institute of Technology
16 16
System Control
.
Device Local Control
Power System Devices
Demonstration SystemDemonstration System
OSI SoftDatabase
ISO/Utility Data
Data Organizationand Processing
Electricity Operating System
Scenario Generator
…
Prosumer Grid Data Instance
Market
Communication Network Simulator
CPSConditions
Network Conditions
System Conditions
Market Conditions
Grid Conditions
Prosumer Computational Node InstanceEstimator, SA, Scheduler, etc.
Simulation Testbed Prosumer Data Instance
16© 2014 Georgia Institute of Technology
17 17
Decentralized Control: Large-Scale Systems
Decentralized Control: Large-Scale Systems
• 20 Self-Optimizing Regions in a Large-Scale RTO System.• Non-fictitious Tie-Line Bus LMP Difference Convergence
obtained using Distributed Optimization.
17
M. Costley, S. Grijalva, “Efficient Distributed OPF for De-centralized Power System Operations and Electricity Markets”, IEEE PES Conference on Innovative Smart Grid Technologies, Washington D.C., January 17-19, 2012.
© 2014 Georgia Institute of Technology
18 18
Decentralized Control: Large-Scale Systems
Decentralized Control: Large-Scale Systems
AE
AEP
PJM
PEPCO
DAY
CE
PENELEC
JCP&L
EKPC DVP
PSE&G
PPL
PECO
DEO&K
RECO
DP&L
BGE
AP
UGI
METED
DLCO
IPRV
FE
98%A
M W
© 2014 Georgia Institute of Technology
19 19
Multi-Scale Prosumer BehaviorMulti-Scale Prosumer Behavior
• Prosumer exposes standardized services– Energy balancing– Frequency regulation– Reserve– Sensing and Information– Forecasting– Security
19© 2014 Georgia Institute of Technology
20 20
Decentralized Building ControlDecentralized Building Control
LMP $/MWh Node
Iteration
a) Building prosumer interaction at transformer connection point for a constrained grid.
b) Voltage contouring for PV variability
20© 2014 Georgia Institute of Technology
21 21
SummarySummary
• Future electricity systems will consist of billions of smart devices and millions of interconnected decision makers.
• Only a decentralized control and management architecture will be able to support the objectives and requirements of the future electricity industry.
THANKS!
© 2014 Georgia Institute of Technology
top related