Transcript Report of the EPRI-NSF Workshop in Playacar , Mexico

Summary of EPRI-NSF Workshop held in Playacar,
Mexico, April 2002, on
GLOBAL DYNAMIC OPTIMIZATION OF
THE ELECTRIC POWER GRID
By
Ronald G. Harley
Duke Power Company Distinguished Professor
November 3-4, 2003
NSF Workshop, Washington DC, Nov 2003_ R Harley
1
Playacar, Mexico
• Over the next ten years, demand for electric power in the USA is
expected to increase by about 20% while under the current plans the
electric transmission capacity will increase only by 4%.
• Three-day workshop, sponsored by EPRI and NSF, held in Playacar,
Mexico, from April 11 to 13, 2002.
• Theme centered around optimizing the dynamic behavior of the electric
power grid on a global scale in order to produce and transmit more
electric power with existing generators and transmission networks,
without jeopardizing various operating constraints.
• 26 persons from academia, industry, EPRI and NSF attended selected
from the power engineering, control and computational science
communities.
• Six Grand Challenges were considered.
• Conclusions from the workshop are that some technology already
exists, but some power engineers seem to be unaware of this, and in
other cases new technology will have to be developed.
• Some research groups have already developed optimization
algorithms, which should now be demonstrated and validated on a
small pilot electric network.
NSF Workshop, Washington DC, Nov 2003_ R Harley
2
The Grand Challenges
1: How to select the type of control hardware, size it and choose its
location.
2: Integrated network control
3: Should we have centralized or decentralized control; how to
coordinate?
4: What infrastructure hardware will the various implementation strategies
require?
5: A benchmark network model is needed for testing theories
6: Pilot schemes will be needed to prove validity of concepts after
simulation
NSF Workshop, Washington DC, Nov 2003_ R Harley
3
Grand Challenge 1: How to select the type of
control hardware, size it and choose its location.
•
Placement of control devices, such as FACTS devices, phase shifters, tap
changers, switched capacitors.
•
Objectives:
– Transient stability improvement.
– Inter-area oscillation damping.
– Voltage collapse avoidance.
– Subsynchronous resonance mitigation.
•
Tools to select, size, locate and control one or more of these control
devices in a network according to some optimization criteria and avoid dynamic
interaction between these devices and the rest of the network.
•
Global coordination suitable for dynamic control requires knowledge of states
of control devices at speeds much higher than currently in use.
NSF Workshop, Washington DC, Nov 2003_ R Harley
4
Critical Capability Gaps
•
Most research work on control devices has been on single machine and two
machine power systems and do not address the unsolved problems mentioned
above.
•
Global dynamic coordination and control requires data transmission at speeds
higher than existing systems can handle.
•
A suitable variety of proven mathematical tools for optimal sizing and placing of
appropriate control hardware.
•
Power engineers are mostly unaware of the tools available in other disciplines.
•
How to control the generators, the FACTS devices, the switched capacitors and
the transformer tap changers, etc. in a global fashion over a wide area
geographically.
NSF Workshop, Washington DC, Nov 2003_ R Harley
5
Grand Challenge 2: Integrated network
control
• Electric grid is a sprawling network with many operational levels
involving a range of energy sources with many interaction points.
• Modern power system problems are becoming increasingly complex,
diverse and heterogeneous.
• The need for seamless interaction of numerous heterogeneous power
network components represents a formidable challenge, especially for
networks that have traditionally used simple methods of system
optimization and control.
• Mathematical models of such systems are typically derived based on
linear techniques, and wide margins of safety are allowed in order to
ensure stable operation.
• Optimization would have to be subject to boundary conditions such as
dynamic and voltage stability, security, reliability, thermal overloads and
market forces.
NSF Workshop, Washington DC, Nov 2003_ R Harley
6
Critical Capability Gaps
• Current corrective measures for emergency recovery depend on
preprogrammed actions based on local information, and are executed
independently in many control rooms. These should rather be dynamic
actions based on both local and global information, executed in a
coordinated fashion by local or global intelligent agents.
• End-use technologies with adaptability and robustness.
• Real time techniques to detect dynamic stability margins, and predict
dynamic voltage collapse.
• Multilayered intelligent agents capable of carrying out the following tasks
during dynamically changing local and global conditions:
– Control, Adapt, Optimize, Communicate, Negotiate, Advise humans.
• Power engineers are currently not using multi-objective optimization
techniques for both continuous and hybrid systems, and which have
become available in the last decade. Revise education.
NSF Workshop, Washington DC, Nov 2003_ R Harley
7
Grand Challenge 3: Should we have
centralized or decentralized control; how to
coordinate?
Completely Decentralized Control contributes to fragmented decisions and
catastrophes.
Centralized Global Control could include some decentralized control, and execute
through three layers:
– Reactive Layer: This layer of agents in every local subsystem performs
preprogrammed self-healing actions that require an immediate response.
– Coordination Layer: The agents in this layer include heuristic knowledge to
identify which triggering event from the reactive layer is urgent, important, or
resource consuming.
– Deliberative Layer: This layer consists of cognitive agents that have goals.
The goals of agents in this layer are for example, reliability, economics,
robustness, and self-healing.
•
With system information available to a local intelligent controller, it could take
appropriate actions to optimize local and even some global variables.
NSF Workshop, Washington DC, Nov 2003_ R Harley
8
Critical Capability Gaps
• Centralized control requires system wide variables and currently no
infrastructure exists for this to happen at fast enough speeds of data
transmission.
• The intelligent agents should be embedded with tools like game theory
for example, with computational and resource bounds that can allow
them to carry out automated negotiation and risk management.
•
Self-healing techniques inherent in local and global agents need to be available.
NSF Workshop, Washington DC, Nov 2003_ R Harley
9
Grand Challenge 4: What infrastructure hardware
will various implementation strategies require?
• Centralized control depends on global information of
the power system.
• Power grid dynamics are changing continually.
• Speed and reliability of communication networks,
sensor networks and intelligent system identification.
• Power grid is vulnerable to physical and cyber
disruption which further underlines the need for fast
reliable communication and sensor networks.
NSF Workshop, Washington DC, Nov 2003_ R Harley
10
Critical Capability Gaps
• Present SCADA systems are usually refreshed at a rate of about once
every second. This is sufficient for slow steady state controls, but is
inadequate for dynamic control especially when higher bandwidths are
needed (10 – 100 Hz).
• Intelligent sensors and actuators, and techniques for verification and
validation are required for system identification. Some existing
techniques can be used in automatically providing updates and relayed
via wireless or other networks to control centers to carry out global
optimal control of the electric power grid.
• Tools for optimal selection and placement of sensors.
• Real time wide-area sensing will open issues on who accesses such
information while maintaining security.
• Some information and encryption technology agents are required.
NSF Workshop, Washington DC, Nov 2003_ R Harley
11
Grand Challenge 5: A benchmark network
model is needed for testing theories
• Currently, no suitable large benchmark networks exist for studies
involving controllers.
• In order to compare and evaluate the potentials of different control and
optimization approaches and devices to solving similar problems, a
benchmark network with standard traditional controllers is needed.
• These benchmark systems need to be accepted and widely used by
researchers and industries.
NSF Workshop, Washington DC, Nov 2003_ R Harley
12
Grand Challenge 6: Pilot schemes will be
needed to prove validity of concepts after
simulation
• There is a wide gap between simulation studies and real time practical
implementations. Some laboratory scale tests have been conducted,
but a pilot scheme on a practical plant is needed.
• Most US universities no longer have power programs, and most of the
power programs do not have a power systems lab containing a number
of synchronous generators, transmission line simulators, etc.
NSF Workshop, Washington DC, Nov 2003_ R Harley
13
Summary
• Some form of automated coordinated wide area
control could possibly have averted the recent black
out.
• Wide area control could also be used to provide
Global Dynamic Optimization of the Electric Power
Grid, by optimizing selected system variables during
steady state conditions.
• FINISH
NSF Workshop, Washington DC, Nov 2003_ R Harley
14