in-rack direct dc powering of servers with solid oxide and … · 2019-04-17 · in-rack direct dc...

Jack Brouwer, Ph.D.Director

March 20, 2019

In-Rack Direct DC Powering of Servers with Solid Oxide and Proton

Exchange Membrane Fuel Cells

US DOE H2@Scale Data Center WorkshopSeattle, WA

© Advanced Power and Energy Program 2019 2/28

Fuel Cell Systems for Data Centers

• eBay’s Data Center in Utah loses $6,000 per second of downtime

• The company’s sustainability mission was in conflict with UT’s electric grid which sources 80% of it’s electricity from coal

Challenges

• 6 MW of fuel cell systems provide primary, onsite, reliable power matched to the operational requirements of the data center

• System provides 100% of electricity demand while drastically reducing carbon footprint

Solution

• Redundant, modular architecture provides highly reliable power

• System architecture replaces large, expensive & polluting backup generators and UPS components

How it works


Microsoft STARK Concept• In-rack Distributed Generation


Microsoft STARK ConceptIn-Rack Distributed Generation• A direct generation method that places fuel cells at the rack level

inches from servers– limits the failure domain to a few dozen servers– Low voltage DC direct connection enabled– Equipment such as power distribution units, high voltage transformers,

expensive switchgear, and AC-DC power supplies in servers could be eliminated

• Hybrid fuel cell systems designed, installed and tested – Use of a 10kW PEMFC stack and system as the distributed power source to

power a server rack – Use of a 2.5 kW SOFC stack and system as the distributed DC power source

to power a server rack


Direct DC Powering of Servers


PEMFC Stack and System Performance

30%

40%

50%

60%

70%

0 2 4 6 8 10 12

Effic

ienc

y

Fuel Cell Power (kW)

35

40

45

50

55

60

0 100 200 300

Volta

ge (V

)

Current (A)

-2

2

6

10

14

0 10 20 30

Pow

er (k

W)

Time (s)

Total Power

Fuel Cell Power

Battery Power

0kW to 9.0kW at t=10

-2

2

6

10

14

0 10 20 30

Pow

er (k

W)

Time (s)

Total PowerFuel Cell PowerBattery Power

4.5kW to 9.0kW at t=10s.


PEMFC Stack and System Server Dynamics

Cold startup time = 90s

t=30, hybrid system turned on

t=90, FC started purging

t=120, FC met the load

-1

0

1

2

3

4

5

6

0 30 60 90 120

Pow

er (k

W)

Time (s)

Total Power

Fuel Cell Power

Battery Power

Hot start


571.4 kWfuel

to the Power Plant

371.4 kWGeneration

Losses

(A) Traditional Data Center (with U.S. Grid Average Efficiency, 2011)

200 kWGenerated

16 kW T&D Losses

184 kW to the Data Center

18 kW UPS

PUE = 1.838Fuel to Server Efficiency = 17.5%

50.8 kW

(B) PEMFC Powered

61 kW to Cooling

5 kW Lighting

100 kW to the Servers

Microsoft STARK Concept

U 838ue to Se e c e cy 5%

338.7 kWNatural Gas

to the PEMFC

50.8 kWConversion

Losses

(B) PEMFC Powered Data Center

287.9 kWHydrogen to the

Fuel Cell

119.9 kW Generation

Losses

7.1 kW to FC BoP

PUE = 1.575Fuel to Server Efficiency = 29.5%

3.4 kW to FC Cooling

157.5 kWto the

Data Center

52.5kW to Cooling100 kW

to the Servers

5 kW to Lighting

NG

NG

Fuel Cell – 29.5%

Grid – 17.5%

Savings: (1) energy conversion losses(2) AC/DC converter elimination(3) reduced cooling


PEMFC Stack and System System Losses

The power outputs of the 12kW in-rack PEMFC system under various external loads. Error bars in the data indicate + one standard deviation from 5 different

measurements.


Direct DC Power Dynamics

• What do we do before ubiquitous hydrogen infrastructure?o Solid Oxide Fuel Cells – natural gas operation (three systems evaluated)

Engen 2500

Natural gas

208VAC

120VAC

Cooling water out

Cooling water in

Reformer water

Exhaust

Forming gas

Condensate

2.5kW

Combustible gas detector

Ethernet cable

Drain

Air


SOFC Stack and System Steady-State Performance

• Characterized I-V relation within the operating range.

• Provide information on overall data center design: bus, power supply, DC/DC.

• Electrical efficiency >52% under standard operating conditions.

0

10

20

30

40

50

60

70

0 5 10 15 20 25 30

Stac

k Vo

ltage

(V)

Current (A)

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0 5 10 15 20 25 30

Stac

k Ef

ficie

ncy

Current (A)


SOFC Stack and System Steady-State Performance

• Characterized the heat rejected at various power outputs.

• Provide information on sizing of the cooling system for data centers.

• Power to Heat Ratio over 1 at full load.

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

kW

Fuel Cell Electrical Output (kW) Fuel Cell Heat Rejection (kW)


SOFC Stack and System Transient Performance

• Characterized ramping behavior of the fuel cell system with controlled ramp loads.

• Ramp rate of 1 A/s achieved.

• No significant power overshoot observed.

• With proper system design the SOFC system could ramp fast.

1.5

1.7

1.9

2.1

2.3

2.5

2.7

0 50 100 150 200

Syst

em P

ower

(kW

)

Time (s)


SOFC Stack and System Transient Performance

• Ramp up and down with various ramp rates were tested.

• System responds immediately to the transient demand perturbation.

• The SOFC system could follow fast load transients.

10

15

20

25

30

0 100 200 300 400 500 600 700 800 900

Curr

ent (

A)

Time (s)

Current Applied to the Stacks

1.5

2

2.5

3

0 100 200 300 400 500 600 700 800 900

Pow

er (k

W)

Time (s)

Power Output of the System


SOFC Stack and System Cycling Performance

• After over 1000 hours of dynamic operation, slight voltage deviations were observed.


SOFC Stack and System Cycling Performance

• After over 1000 hours of dynamic operation, Negligible power output degradation observed.


Goal Must Be: 100% Zero Emissions

Envision this future, invest in its evolution

• ALL primary energy from sun, wind, wave, …

• Use ONLY zero emissions electrochemical energy conversion to complemento Batterieso Electrolyzerso Fuel cells

• Use ONLY zero emissions energy carrierso Hydrogeno Renewable gases & liquids


Hydrogen Energy Storage Dynamics

Oct. 2 Oct. 3 Oct. 4 Oct. 5 Oct. 6 Oct. 7 Oct. 8

• Load shifting from high wind days to low wind days• Hydrogen stored in adjacent salt cavern

Maton, J.P., Zhao, L., Brouwer, J., Int’l Journal of Hydrogen Energy, Vol. 38, pp. 7867-7880, 2013

• Compressed Hydrogen Storage complements Wind & City Power Demand Dynamics in (Texas)


Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec30

45

60

75

90

105

120

135

150

165

Pre

ssure

(B

ar)

Maximum Pressure

Minimum Pressure

280

290

300

310

320

330

340

350

360

370

Tem

pera

ture

(K

)

PressureTemperature

Hydrogen Energy Storage Dynamics

• Weekly storage and seasonal storage possible with hydrogen and fuel cells/electrolyzers – all zero emissions!

Weekly Seasonal

30

45

60

75

90

105

120

135

150

165

Pre

ssure

(B

ar)

Maximum Pressure

Minimum Pressure

280

290

300

310

320

330

340

350

360

370

Tem

pera

ture

(K

)

PressureTemperature

Oct. 2 Oct. 3 Oct. 4 Oct. 5 Oct. 6 Oct. 7 Oct. 8

Maton, J.P., Zhao, L., Brouwer, J., Int’l Journal of Hydrogen Energy, Vol. 38, pp. 7867-7880, 2013

But what can we do if we don’t have a salt cavern?


• Recent 1-Year Simulation of 100% Renewable Grid in CAo Wind dominant case (37 GW solar capacity, 80 GW wind capacity)

Why we need H2: Amount of Storage Required

Deficit

Surplus

21 millionEVs

CurrentH2 Storage

PumpedHydro

§ 21 million = total CA registered light duty vehicles; Nissan Leaf battery

*Using existing natural gas resources for hydrogen storage


Why We Need H2: World Grid Energy Storage Need

Simulate meeting of TOTAL world electricity demand w/ Solar & Wind• How much storage is needed?

[Nuria Tirado, M.S. Thesis, 2018]

• Batteries needed, but, they cannot do it all!o Li req’d = 3,144 Mt Co req’d = 25,815 Mto Massive cost (connected power & energy scaling)o Self discharge (measured performance in utility applications)

19,981 TWh


Lithium• Resources: 53 Mt (USGS)

o Economic feasibility

Cobalt• Terrestrial resources: 25 Mt (USGS)• Ocean-floor resources: >120 Mt (USGS)• 40% of cobalt comes from the Democratic Republic of the Congo

o Improvements needed in cobalt extraction

Why We Need H2: Lithium-ion Batteries

– Size of the deposit

– Lithium content

– Content of other elements

– Processes used for purification

– Better geologic models

– Improved extraction methods

– Development of recovery processes

– Technological advances (deep-sea)

Figure extracted from USGS.

3,144 Mt req’d

25,815 Mt req’d


Why We Need H2: Lithium-Ion BatteriesRound-Trip Efficiency (>90% in Laboratory Testing)• Measured battery system performance in Utility Applications

• Self-Discharge (the main culprit), plus cooling, transforming, inverting/converting, and other balance of plant

From: 2017 SGIP Advanced Energy Storage Impacts, Itron, E3

Average RTE ~60%


Brief Gedanken experiment• First mix up to X% – tremendous boon to grid renewables• Then piecewise conversion to pure hydrogen

13

7

5

Solar Here!

Close valvesHere!


1L=75 mile

D=30”Line 2036

Cactus City Compressor Station

Min Pressure= 475 psigMax Pressure= 850 psig

L=150 mileD=30”

Line 2036

2

3L=100mile

D=30”Line 2225/1941

North Palm SpringCompressor Station


L=105 mileD=30”

Line 1941

4

5L=124.5 mile

D=30”Line 1918

NewberryCompressor Station


L=118 mileD=30”

Line 1921

6

7L=115.5 mile

D=34”Line 1917

NewberryCompressor Station


L=117 mileD=36”

Line 1922

8

El Paso

Needles

Topock

Blythe

Reference for pipe and compressor: stationhttps://www.arcgis.com/home/webmap/viewer.html?webmap=f8b54b821642463b8dc0becb2711093a

Pressure and Flow Dynamics• With renewable gas injection at border (in desert)


Pressure and Flow Dynamics• 40% of all electric demand – 20 sq. miles of solar, only gas system

use for H2 storage and T&D


Brief Gedanken experiment• Piecewise conversion of gas system to pure hydrogen

13

7

5

Solar Here!

Close valvesHere!


Thank You!

in-rack direct dc powering of servers with solid oxide and … · 2019-04-17 · in-rack direct dc...

Documents