The document compares costs and reliability of different emergency power system topologies for data centers. It finds that systems using integrated paralleling are more reliable at lower cost. While medium voltage increases price, it does not significantly improve reliability. Scalability allows modular reliability improvements. Case studies of four data centers show successful implementations of redundant bus and distributed redundant topologies. In conclusions, standardization, modularity and scalability are driving factors in new emergency power system designs.
1. EMERGENCY POWER
TOPOLOGIES AND EVOLUTION
7X24 Fall Phoenix
Curt Gibson PE ATD
Power Solutions Manager
Generac Industrial Power
(209) 483-3910
Curt.Gibson@Generac.com
2. Introduction
Curt Gibson
Data Center Specialist for Generac
2 – 3 data centers per week
Accredited Tier Designer
Registered Professional Engineer
20 years in engineering of Emergency
Power Systems for ASCO and Generac
2
3. Goals Today
Compare costs and reliability of different topologies
Reliability
Cost
Scalability
Four case studies
University of Utah
Venyu Data Center
C7 Data Centers
Green House Data
Question and Answer
3
4. Typical Data Center Concerns
Energy efficiency
True reliability and availability
Cost efficiency
Return on investment
Scalability
4
5. Recent Trends
Shorter lead times, just in time
Cost competition
Less custom
Modularity
Standard products, off the shelf
Widely available support
5
6. Modular Era
Cost competitive
Future proof
Higher efficiencies
Scalable
Repetitive designs
Economies of scale
Lower cost of ownership
6
7. Paralleling Benefits
Paralleling is easier now
Improves reliability
Lowers cost
Makes systems more scalable
Offers more availability
7
9. Paralleling Controls are Evolving
Smaller digital controllers
Every major manufacturer
(Cat, Cummins, MTU, Kohler, Generac)
Costs are decreasing
Standardization provides reliability and modularity
9
10. Comparison of Concepts
N = 5MW
N+1 during all three construction phases
Co-location data center
Outdoor installation
24 hour tanks
80dB enclosures
Four load buses
480V loads
10
12. Uptime Institute Tier Requirements
TIER REQUIREMENTS Tier I Tier II Tier III Tier IV
Number of Delivery Paths 1 1
1Active
1Passive
2Active
Redundancy N N+1 N+1
N after any
failure
Compartmentalization No No No Yes
Concurrent Maintainability No No Yes Yes
Fault Tolerance No No No Yes
12
Often confused with
“Common Attributes”
Tier IV does not require
2N capacity
Compartmentalization
includes dead lugs
when servicing
Automatic fault
response is typical in
Tier III
13. Reliability Calculations
13
Not “Availability”
Simple equations
Only to the UPS
Probability the load is
fed
Assumed 98% for
generators
Assumed 99.9% for
buses
Stattrek.com calculator
IEEE 3006.7
22. Single Bus, N+2, T2, DR
$850 $600 $600 $2050 99.745%22
Phase 1 Phase 2 Phase 3 Total
Reliability
N+2 System
Tier 2 Compliant
Distributed Redundant
Paralleling on Gen
65KAIC Switchgear
Breaker pair transfer
Day One = 2X1000
Reliability
(N12, n10, P0.98) = 0.99846
0.999
(N4, n3, P0.999) =0.99999
(099846)(0.99999) (0.999)
= 0.99745
F
23. Redundant Bus, N+2, T3, DR
$1000 $600 $600 $2200 99.845%23
Phase 1 Phase 2 Phase 3 Total
Reliability
N+2 System
Tier 3 Compliant
Distributed Redundant
Paralleling on Gen
65KAIC Switchgear
Breaker pair transfer
Day One = 2X1000
Reliability
(N12, n10, P0.98) = 0.99846
1-(1-0.999)(1-0.999)= 0.99999
(N4, n3, 0.999)= 0.99999
=0.99845
G
24. Redundant Bus, N+4, T4, PR
$1040 $540 $790 $2370 99.998%24
Phase 1 Phase 2 Phase 3 Total
Reliability
N+4 System
Tier 4 Compliant
Parallel Redundant
Paralleling on Gen
150KAIC Switchgear
Breaker pair transfer
Day One = 2X1000
Reliability
(N14, n10, P0.98) = 0.99969
1 – (1-0.999)(1-0.999)= 0.99999
0.99999 x 0.99999= 0.99998
H
25. Topology Phase
1
Phase 2 Phase 3 Total
(Thousand)
Reliability
98%, 99.9%
Reliability
94%, 99%
A Medium Voltage, 2N, PR, T4 $2250 $750 $750 $3750 99.994% 99.91%
B Medium Voltage, N+1, DR, T2 $2000 $650 $650 $3300 99.645% 97.02%
C Parallel Redundant, LV, 2N, PR, T3 $1700 $600 $600 $2900 99.994% 99.91%
D Distributed Redundant, LV, N+1,DR, T3 $1200 $600 $600 $2400 99.742% 97.94%
E Block Redundant, LV, N+1, BR, T3 $1100 $550 $550 $2200 99.742% 96.76%
F Single Bus, LV, N+1, DR, T2 $850 $600 $600 $2050 99.745% 95.87%
G Redundant Bus, LV, N+2, DR, T3 $1000 $600 $600 $2200 99.845% 96.83%
H Redundant Bus. LV, N+4, PR, T4 $1040 $540 $790 $2370 99.998% 99.89%
Comparison of Options
25
26. Conclusions
26
Systems that used integrated paralleling were
more reliable at lower costs
Availability was improved by redundant buses,
but reliability was only slightly affected
Medium Voltage increased the price but not
reliability
Scalability offers modular reliability
improvements
27. University of Utah
11 MW design, but only 5MW installed
20,000 Gallon fuel
27
29. Outcomes and Conclusions
Power usage is not as high as forecasted
Provisions for future growth are available
Bi-fuel may have improved fuel storage risk
Smooth project execution
Video
29
32. Factors and Decisions
Redundant buses allow concurrent
maintenance
Automatic transfer is fault tolerant
All generators on line in 8 seconds
Second floor services and high efficiency
transformers are other innovative aspects
Video
32
35. Outcomes and Conclusions
Innovation is a key to success at C7
Power usage in the facility is lower than
expected
Energy efficient servers and cooling may
cause the expansion at this facility to never be
built
Other buildings are receiving the same units
Future large expansion is planned with this
topology
35
38. Cheyenne Wyoming
5MW total final
N + 1
2MW first phase
Single bus
Distributed Redundant
Outcomes and Conclusions
38
39. Topology Phase
1
Phase 2 Phase 3 Total
(Thousand)
Reliability
98%, 99.9%
Reliability
94%, 99%
A Medium Voltage, 2N, PR, T4 $2250 $750 $750 $3750 99.994% 99.91%
B Medium Voltage, N+1, DR, T2 $2000 $650 $650 $3300 99.645% 97.02%
C Parallel Redundant, LV, 2N, PR, T3 $1700 $600 $600 $2900 99.994% 99.91%
D Distributed Redundant, LV, N+1,DR, T3 $1200 $600 $600 $2400 99.742% 97.94%
E Block Redundant, LV, N+1, BR, T3 $1100 $550 $550 $2200 99.742% 96.76%
F Single Bus, LV, N+1, DR, T2 $850 $600 $600 $2050 99.745% 95.87%
G Redundant Bus, LV, N+2, DR, T3 $1000 $600 $600 $2200 99.845% 96.83%
H Redundant Bus. LV, N+4, PR, T4 $1040 $540 $790 $2370 99.998% 99.89%
Comparison of Options
39
40. Observations and Conclusions
Short lead-times, lower cost, and risk reduction
are driving decisions.
New technology is available and being
adopted
Standard components are favored over
custom
Scalability is more important than before
Many solutions exist to accomplish goals
40
Welcome
This break out session will cover some advances in emergency power technology, and the options available for the data center designer or builder.
I believe the findings being presented will be interesting in real world comparisons, and trends
My name is Curt Gibson, and I am a data center specialist with Generac
I see many data center projects each month from all over the USA
My data center experience is a byproduct of engineering roles with manufacturers of products for this industry.
I also spent a few years as a Facilities Manager for Sun Microsystems in Milpitas
grateful for this opportunity to present,
How to design a system
This presentation will provide a comparison of costs versus reliability for a eight topologies
We will then review four case studies showing how the systems were actually built
The results show some interesting preferences and biases when the data center is actually built
Concerns in the design of a data center can be grouped into areas
These are pretty consistent in general
Emerson Network Power provided this study last year,
Focusing on Power, I have listed some key concerns related to this presentation
Over the last 15 years I have seen trends that point in a certain direction
Custom products were needed in the early days.
I expect this to continue and see only increasing pressure to become more efficient
Efficiency by removing labor and risk while increasing return on capital
The modular era is here, due to this trend in efficiency.
I stole the graphic on the right from Compass Data Centers
We see modularity in UPS’s, HVAC, switchgear, and now also in generators
There are many benefits to modularity.
Modularity in generators is provided by the ability to parallel the generators
Paralleling includes controls and switching.
Of the technologies mentioned, I would like to focus on integrated paralleling
It brings many benefits, and seems that many people don’t know that it is available, and very reliable.
There are some ways to use the capability to your advantage which may not be obvious.
When we talk about paralleling generators, often this is what engineers think of large generators and paralleleing switchgear
Paralleling has traditionally been used when the generator was not large enough for the load
Implementing a paralleling system was done by using PLC’s, synchronizers, load sharing modules, relays, and lots of custom engineering
There are many risks and costs involved in this way to provide paralleling
There are some great benefits by allowing all your sources to feed all your loads. It is more reliable.
Technology is advancing and making life easier.
Every major manufacturer
On the left is a Generac motherboard, with includes synchronizing and load sharing controls.
The next is a Woodward EGCP2, which was intended as a similar single board paralleling system
The Horner ACM500 has all the processing to sync and share
And the Beaglebone black is
Today we are going to all go through the process of designing a project
We will compare the different topologies
This is the list of the requirements for our hypothetical project.
These requirements are very common for a data center
There are many factors to consider, but for this comparison we would like to keep the components in a the real world.
Here is a list of guidelines that show the commonly available sizes of components. This is a very important One of the slide
For the exercise we need to make sure we comply with basic tier compliance guideline.
Despite no plans to apply for certification, engineers typically use the Uptime Institute Tiers as a guideline.
These are just a few of the requirements.
Reliability calculations are based on these simple equations, and we are only evaluating the emergency power system
Reliability is the probability that the load will be fed in the event of an outage.
The component probabilities used are estimates
98 % reliability for a generator to start and carry load is higher than average, but data centers tend to be more maintained than average
Binomial distribution calculations are provided by Stattrek.com’s online calculator.
b(x; n, P) = { n! / [ x! (n - x)! ] } * Px * (1 - P)n - x
Availability is the brother to Reliability, but considers the time the unit is available for use. My farrarri is very reliable, it always starts, but it is in the shop most of the time, so it is not available.
Steve Fairfax has presented data that show generators to be remarkably un-reliable, but of course he is not talking about your data center, or is he?
I’d like to mention that I did receive help on the calculations from Bob Scheugar with HP-EYP in Los Angeles. He has done remarkable work on IEEE 3006.7 which deals with emergency power reliability.
This is the first of eight topologies to be considered
We will discuss only the blue items
Substation transformers are also not included to keep everything fair
No problems with AIC or bus size
First day 2X2500kW
Will save some money, but give up a little reliability
Not concurrently maintainable, so not T3
Low Voltage
200kAIC, 4000A main bus
First day; 2X2500kW