Anjan Bose, Washington State University Tom Overbye, University of - - PowerPoint PPT Presentation

Seamless Bulk Electric Grid Management:

A Platform for Designing the Next Generation EMS PSERC Project S-62G

Anjan Bose, Washington State University Tom Overbye, University of Illinois Santiago Grijalva, Georgia Inst. Of Tech. PSERC Public Webinar April 5 2016

1

Project Overview

• The overall research objective is to develop a

flexible platform in order to test the next generation EMS and associated analytics

• Platform should be able to simulate different layers

related to power grid operations including the system itself, its cyber infrastructure, various software applications

• Platform should be flexible to allow different

components to be included

• Need to provide case study example systems

2

EMS Background

• The grid has long been a technology leader

3

Commonwealth Edison Control Room Circa 1920 Utility Control Room, 1960’s

Source: W. Stagg, M. Adibi, M. Laughton, J.E. Van Ness, A.J. Wood, “Thirty Years of Power Industry Computer Applications,” IEEE Computer Applications in Power, April 1994, pp. 43-49

EMS Background

• And we are continually getting smarter!

4

PSE&G Control Center in 1988

Source: J.N. Wrubel, R. Hoffman, “The New Energy Management System at PSE&G,” IEEE Computer Applications in Power, July 1988, pp. 12-15.

ISO New England Control Center

Previous Work

• Previous Work under this project established a

framework and requirements for next-generation Seamless Energy Management Systems.

• Key Requirements included:
• Support for explicit modeling of the effects of

imperfect communications in cyber-control.

• Recognition of the need to manage faster and more

dynamic effects in the system.

• Trends and opportunities of decentralized control.
• Study requires simulation of a cyber-physical

system.

5

Project Coordination

• Project involves coordinated work taking place

at UIUC, WSU and Georgia Tech

6

PMU Level Grid Simulation and Case Development (UIUC) Communication Simulation (WSU) Decentralized Applications including Real-time Control (Georgia Tech)

Seamless Bulk Electric Grid Management (S-62G)

Part 1: PMU Level Grid Simulations and Cases (UIUC)

Thomas Overbye Students and Staff: Frank Borth, Richard Macwan Overbye@illinois.edu

7

PMU Level Simulations

• Traditionally the EMS has

been driven by SCADA data

• Dispatcher Training

Simulators have also used this time frame

• Uniform frequency
• EMS of the future needs to

work in the PMU (transient stability) timeframe, so this is required in the simulation

• EMS is most important during

stressed operation!

8

Image Source: Jay Giri (Alstom Grid), "Control Center EMS Solutions for the Grid of the Future," EPCC, June 2013

Real-time, PMU Level Simulation Environment

• Project leveraged commercial, interactive, real-

time transient stability simulation platform

• Data is

exported in c37.118 format

• Closed-loop

control is also implemented

• Standard

transient stability models are used, including generator

• ver excitation limiters and line relays

9

Case Development

• First case developed for this testing was a 42

bus, 345/138 kV, 11 GW of Generation/Load

• Rather detailed dynamics models were included

allowing for interactive, transient stability simulation

• RTUs were modeled

for each of the substations

• Scenario considered

was a tornado moving by a substation, taking out three 345 kV lines and 500 MW of generation

• Case is public domain

10

Event scenario for 42 bus system

This is described as follows:

1.At 2.0 seconds, the system has a 3-phase ground fault at bus 15. 2.At 2.5 seconds, the line between buses 43 and 15 opens. 3.When control center receives fault data, it sends back control signal to trip the generator at bus 43. 4.The generator trip to keep system stable.

Each PMU data packet will have PDC processing delay when it goes through each substation or PDC, which varies and needed to be considered.

Seamless Bulk Electric Grid Management (S-62G)

Part 2: Communications (WSU)

PSERC Industry-University Meeting May, 2015 Anjan Bose Students and Staff:Yannan Wang, Fan Ye bose@wsu.edu

Communication simulation using NS3

I. Preparation of the communication network

• Communication network based on the power network
• Control center based near existing substation

II. Protocol Stack

• TCP or UDP? Why?
• III. Communication network simulation and results
• NS3 used to calculate time delays in communication
• Output is PMU data plus time delays

Processing preparation

Communication network overlay for Illinois 42-bus system

Processing preparation, cont.

General rules in this case:

1.PMUs are installed at both ends of each transmission line, measuring voltage and current phasors both. 2.The sample rate for each PMU is 60 samples per second. 3.In each substation, phasor data concentrator (PDC) is collecting PMU measurements from PMUs connected to that substation. 4.PDCs are communicating each other over communication channels.

Our work is to compare different communication architectures to minimize communication delays.

Protocol stack

Communication network has 5-layer stacks according to TCP model.

• Application Layer: C37.118 is specifically designed

for PMU messages exchanging.

• Transport Layer: Our choice is UDP but why?
• UDP packet header is lighter thus transmit faster

than TCP.

• TCP protocol is complicated because it has many

delivering-guaranteed mechanisms such as retransmission, congestion control, flow control.

• Network Layer: IPv4 (more common than IPv6).
• Link Layer: Ethernet.
• Physical Layer: Optical fiber.

Four different communication architectures with delay results

Two main kinds of communication network are presented: star and mesh networks.

• Star network: it defines a network in which all

communication nodes communicate directly one node, in this case, control center.

• Mesh network: it defines a more flexile

communication network in which communication links are along the same or similar right-of-way as the transmission lines.

Four different communication architectures with delay results, cont.

These four communication architectures have mesh and star networks both. They are:

1)Network along with the transmission lines. (Mesh) 2)Network divided by three areas. (Mesh) 3)Centralized structure. (Star) 4)Decentralized structure. (Star)

The following slides describe them one by one, with delay results and demonstration on transient stability followed.

One control center with communication lines along the same right-of-way as the power transmission lines.

Network along with the transmission lines (network type 1)

Three sub-control can help PMU data routing. In each area the communication links along the same right-of-way as the power transmission lines.

Network divided by three areas (network type 2)

One control center through substation 9. Each substation is directly linked to Sub 9.

Centralized structure (network type 3)

Similar to type 2 yet each substation link

• ne sub-control

center in its

• wn area.

Decentralized structure (network type 4)

5 10 15 20 25 30 35 40 45 50 55 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 Bandwidth(Mbps) Delay(ms) Total delay of Type1 to trip the generator at substation43 PDC delay 10ms PDC delay 50ms PDC delay 100ms PDC delay 500ms

Different communication bandwidth considered. When the bandwidth is below 5Mbps, the queuing delay is increasing much.

Type 1 delay results

The communication delays are almost “stable” even the communication transmission rate is below 5Mbps.

Type 2 delay results

5 10 15 20 25 30 35 40 45 50 55 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 Bandwidth(Mbps) Delay(ms) Total delay of Type2 to trip the generator at substation43 PDC delay 10ms PDC delay 50ms PDC delay 100ms PDC delay 500ms

5 10 15 20 25 30 35 40 45 50 55 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 Bandwidth(Mbps) Delay(ms) Total delay of Type3 to trip the generator at substation43 PDC delay 10ms PDC delay 50ms PDC delay 100ms PDC delay 500ms 5 10 15 20 25 30 35 40 45 50 55 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 Bandwidth(Mbps) Delay(ms) Total delay of Type4 to trip the generator at substation43 PDC delay 10ms PDC delay 50ms PDC delay 100ms PDC delay 500ms

Type 3&4 delay results

Similarly to type 1, rate as 5Mps can become a critical rate point, yet their delay values are different.

Communication results demonstration

Three circumstances are considered in power system transient stability: stability, critical stability and instability. This system has the critical delay point as 800ms approximately.

We examine the situation (case 1) in the first place where the generator is tripped in a very short time (<300ms). The rotor angle performances of two generators in substation 43 are shown here.

Critically stable case in which the generator is tripped after around

• 800ms. The maximum degree for
• scillation is roughly 122 degree,

which is greater than case 1. Instability case in which the generator is tripped after a long time (>800ms). The system goes unstable even the generator is tripped.

Conclusions

In our work four different communication architectures are compared and studied. It’s hard to say which one is the best and which one is the worst. For those applications which don’t have strict latency requirements, type 3, type 4 even type 1 might be a good choice. However, for those like real-time PMU-based applications, type 2 might meet the demands.

Seamless Bulk Electric Grid Management (S-62G)

Part 3: Decentralized Applications (Georgia Tech)

PSERC Industry-University Meeting May, 2015 Santiago Grijalva Students and Staff: Leilei Xiong sgrijalva@ece.gatech.edu

29

Motivation

30

• We need to integrate into the grid:
• Large amounts of highly variable and spatially

distributed renewable energy.

• Need much faster, better, tighter coordination

across subsystems: ISO, utilities, microgrids, etc.

Motivation

• Various part of the grids are operated by EMS and

DMS systems.

• What happens in a region of the grid affects other

regions.

• Decentralized coordination issues must be addressed.
• A dynamics co-simulator could be used to test

decentralized control applications including the effect of imperfect communication and delays.

• Use Cases:
• ISO to ISO coordination within an interconnection
• Distribution Utilities to ISO coordination

31

Example: Decentralized Power Agreement

• Assume that a region needs to change their

reference production (net interchange) due to:

• Power plant emergency
• Contingency
• Variation in renewables
• Each region has:
• Can the regions agree on the needed levels of

interchange to balance the system in a decentralized manner?

32

: Desired Power (ex-ante) : Agreed-upon Power (algorithmic) : Actual Power (ex-post)

i i i

p p p  

Decentralized Power Agreement

• Given desired power levels we want to:
• An agreement dynamics, incorporating

constraints, weights, and trust, is used by regions to agree on the power levels across the system in a decentralized fashion.

• How fast can an agreement be reached?

2 1

ˆ min

N i i i i

w p p

=

−

∑



Ramachandran, Costello, Kingston, Grijalva, Egerstedt. Distributed Power Allocation in Prosumer Networks, NecSys, 2012.

33

Decentralized Agreement: Convergence

• Rate of converge depends on topology and connectivity
• Second eigenvalue of the communication graph Laplacian.

34

Example: Decentralized Power Agreement for Frequency Control

• Current frequency response is proportional to

frequency deviation with respect to nominal.

• Current frequency regulation is unilateral and

may cause inter-area oscillations.

• Question:
• Can approaches be developed that can start

correcting imbalances before frequency deviates?

• Can agreement help drive the frequency closer to

nominal after an imbalance?

35

Small Case Simulation

• 42 bus system with dynamics.

400 MW 505 MW 2238 MW 1650 MW 234 MW 55 Mvar 234 MW 45 Mvar 94 MW 30 Mvar 267 MW

• 50 Mvar

267 MW

• 50 Mvar

267 MW

• 50 Mvar

268 MW 128 Mvar 268 MW 128 Mvar 240 MW 110 Mvar 200 MW 82 Mvar 150 MW 30 Mvar 207 MW 54 Mvar 205 MW 65 Mvar 200 MW 45 Mvar 200 MW 45 Mvar 158 MW 43 Mvar 158 MW 43 Mvar 240 MW 0 Mvar 240 MW 0 Mvar 160 MW 32 Mvar 160 MW 27 Mvar 186 MW 56 Mvar 202 MW 32 Mvar 190 MW 42 Mvar 201 MW 52 Mvar 201 MW 62 Mvar 175 MW 32 Mvar 156 MW 23 Mvar 176 MW 15 Mvar 212 MW 30 Mvar 140 MW 33 Mvar 212 MW 30 Mvar 132 MW 15 Mvar 94 MW 35 Mvar 267 MW 0 Mvar 267 MW 0 Mvar 267 MW 0 Mvar 210 MW 45 Mvar 185 MW 33 Mvar 112 MW 40 Mvar 300 MW 60 Mvar 95 MW 23 Mvar 75 MW 15 Mvar 198 MW 35 Mvar 193 MW 30 Mvar 161 MW 21 Mvar 135 MW 20 Mvar 140 MW 20 Mvar 88 MW

• 49 Mvar

130 MW 45 Mvar 128 MW 28 Mvar 49% 63% 31% 57% 54% 83% 23% 58% 47% 31% 42% 28% 33% 48% 29% 45% 25% 60% 68% 75%

75% 31%

78% 52% 54% 55%

67%

1100 MW 178 MW 162 MW 177 MW 77 MW 55% 1520 MW 250 MW 50 Mvar

Hickory138 E lm138 Lark138 Monarch138 Willow138 S avoy138 Homer138 O wl138 Walnut138 Parkway138 S pruce138 Ash138 Peach138 Rose138 S teel138

123 Mvar 73 Mvar 100 Mvar 124 Mvar 117 Mvar 116 Mvar

56% 52%

B adger D

• lphin

Viking B ear S idney Valley H awk

Illini P rairie Tiger Lake R am Apple G rafton O ak Lion

1520 MW 41% 200 MW 40 Mvar 200 MW 45 Mvar

84%

190 MW 63 Mvar 200 MW 505 MW 63% 57 Mvar

E agle

57%

500 MW

36

Small Case Simulation

• System is partitioned into regions (e.g. control areas)
• No central agent, each region talks with neighbors.

~1400MW Generation Disconnected Communication Links

37

Small Case Simulation

• At time t a generator is

disconnected in region i causing an imbalance equal to ∆Pi.

• As soon as the generator trips,

power agreement is invoked to match the imbalance.

• Regions reach agreement in a

few iterations and take actions immediately.

• Performance depends on

communication delays.

38

2 3 4 5 6 7 8 9 10 11 12

• 1400
• 1200
• 1000
• 800
• 600
• 400
• 200

200 400

Iterations

Regions determine (agree on) the power needed to compensate for generator trip

1, N j i j j i

P P

= ≠

∆ = ∆

∑

Frequency Response Performance

39

Base Case (conventional response) Line Trip Generator Trip at t=25 sec Total time to agreement including delays (sec)

sec 0.1 5 2 1 0.5

Note: Frequency approaches the nominal because imbalance is known.

Summary

• Decentralized coordination methods can be of

service in achieving scalable grid coordination.

• With appropriate communication system and

small delays, fast applications such as frequency response can be improved.

• A co-simulation approach supports investigation
• f dynamic performance of decentralized control

methods under imperfect communications.

40

Next Steps

• Development of larger cases for demonstrating

additional and more realistic scenarios, and more closed-loop control

• Realistic modeling of communication delays using

statistical simulator (WSU)

• More streamlined closed loop control
• Prototyping and assessment of improved

visualizations utilizing the c37.118 data

41

Thank You! Questions?

bose@wsu.edu