cooling summit - dr robert tozer present
TRANSCRIPT
Case study: Energy efficiency improvements in a legacy data centre Robert Tozer Director – Opera>onal Intelligence
• First improve air management – Air supply temperature to servers
• Lower temperature • Reduce range (also control CRAH on supply air temperature)
• Then save energy: – Increase temperature set points (chilled water, air):
• Refrigera>on energy (and free cooling)
– Reduce air flow (to what is required) • Fan energy (cube law, VFDs)
Objec>ves
Air management: improvements • Reduce Bypass flow
– Close cable penetra>ons
– Locate air grilles in front of IT equipment
• Reduce Recircula>on flow
– Use blanking plates, improve rack layout
• Reduce Nega>ve Pressure flow (< air velocity)
• Supply air control (not return air control)
• Physical Containment (cold aisle, hot aisle or rack)
Background
0.3 1.0 0.4 = $1.7m
ANNUAL SAVINGS per year (1 site)
Cooled by 53 CRAH units
3,500m2 data hall area
Current IT load
1,800 kW
Tier 3+
2N Built in 2007 in London
Incremental changes in a
LIVE ENVIRONMENT
Energy op>misa>on programme 1. Energy assessment and data hall air temperature survey 2. Outdoor fresh air AHU dew point control 3. Air management improvements 4. Reduce fan speeds and control on supply air 5. Increase in data hall air set points 6. Increase chilled water set points 7. Free cooling feasibility study 8. Installa>on of free cooling modifica>on
PUE reduc>on 1.49
2.3
$3m
$4.7m Annual energy
saving
0 1
1
Flow Performance 1-‐BP
Thermal Perform
ance
1-‐R
Af=0.25
Af=
4
BP = 80% R = 20% Af = 4 Server intake: 15 and 27°C (upper limit ASHRAE recommended range) Air management a priority
Energy assessment/air temperature survey
50% IT Load
Data cen
tre load re
la>ve to IT
0 25% 75% 100%
100%
75%
50%
25%
0
125%
150%
175%
200%
IT PU
E = 8
C
Original
Ini>al Load
Improved
Scalability
Air management improvements
15
Step 2: Air Performance � Measurement and Solution Air was recirculating around the end of the cold aisle and straight back into ����������� �������� �� ����������into the top of the racks. The installation of temp doors resulted in a reduction of approx. 5°C, increasing height of aisle by about 4°C, net result being an increase in the cooling effectiveness of the aisle � Temp solution mocked up to prove concept with a permamanant end of aisle solution then deployed. Recirculation coming over the top of the racks was combated by installation of flame retardant curtains to increase the height of cold aisle. Effective blanking within the racks also demonstrated significant improvements Total costs of deployment recovered in less than a year.
15
Step 2: Air Performance � Measurement and Solution Air was recirculating around the end of the cold aisle and straight back into ����������� �������� �� ����������into the top of the racks. The installation of temp doors resulted in a reduction of approx. 5°C, increasing height of aisle by about 4°C, net result being an increase in the cooling effectiveness of the aisle � Temp solution mocked up to prove concept with a permamanant end of aisle solution then deployed. Recirculation coming over the top of the racks was combated by installation of flame retardant curtains to increase the height of cold aisle. Effective blanking within the racks also demonstrated significant improvements Total costs of deployment recovered in less than a year.
Blanking plates Sealing floor gaps Revised floor grille loca>ons
Proof of concept: Hot and cold air
stream segrega>on
5C reduc>on 4C reduc>on
1 year payback: deployed throughout data hall
CRAH fan speeds and control
100%
60%
Original fan speed
Op>mised fan speed
Work in Progress: 40% with Dynamic Control
Fan speed, flow rate and power
20 Fan applicat ion guide
the impact that they have on the power requirement of thesystem overall.
In essence, if the load/speed of a fan never alters and isperfectly matched to its required output once installed,then the addition of a control can only reduce the systemefficiency. H owever, these conditions are rarely, if ever,encountered.
T he methods of duty control commonly employed areconsidered in the following sections.
5.2.1 Alteration of the fan rotational speed
See F igure 5.1. For a fan installed in a typical buildingventilation system, the power absorbed is approximatelyproportional to the cube of fan speed. Note that other(typically process) load types may have different charac-teristics.
In practice, the varying efficiency of motor and drivesystems with speed may modify th is ‘perfect’ controlcharacteristic.
Although larger fans may be driven by a wide variety ofprime movers, the majority of fans are driven by electricmotors.
In outputs above 200 W, these are typically AC inductionmotors where the rotational speed is dependant on supplyfrequency and internal construction and having efficien-cies that vary from 60 to 90%+ , over the range to 75 kW.In smaller sizes, where AC motor efficiency may be as lowas 20%, recent developments in DC motor technology haveseen these used with the benefits of lower losses and morestable speed controllability.
Where infrequent duty changes are needed, the use of beltdriven fans, with high efficiency drive systems is oftenused for decreasing fan speed and operating at speedsother than those provide by standard motors (note thatany increase in speed generally requires reference to thefan manufacturer).
Belt/pulley changing requires a degree of skill, and someconsiderable downtime, these factors being the principaldisadvantages together with the power loss inherent in thedrive.
More frequent speed changes with fixed ratios (usually of1:2 or 1:1.5) may be achieved at relatively low cost withmulti-speed AC motors and basic electrical switchgearunder automatic (e.g. time based) or manual control.
For smaller induction motors (up to around 2.2 kWoutput) the possibility exists for variable speed operationusing a fixed frequency, variable voltage speed controller,of either a transformer or solid state (triac) based type.The latter tend to be smaller and lower cost but may causemotor noise due to their non-sinusoidal output waveform.
A degree of two-speed operation can be achieved in thisoutput range by using a star/delta wiring changeover,which produces an effective voltage control.
The reason for the stated limitation in motor output is thetemperature increase in the motor due both to increasedlosses and the reduced ability to reject heat as speeddecreases.
The motor and/or control may fail if they are not correctlymatched. Typically a peak in motor current occurs ataround 65% of maximum speed and must be allowed for.Specifiers are strongly advised to seek the approval of thefan manufacturer for any speed control device used, and toensure that all products comply with the latest legislativerequirements.
Undoubtedly, one of the most significant technologicaladvances in the past 30 years has been the availability ofcost effective variable frequency speed control drives(often known as ‘inverters’), in output ratings from 150 Wto several thousand kilowatts.
By varying the effective supply frequency to the motor,speed reductions down to 20% may be achieved withoutsignificant de-rating. T his technique is applicable to awidely available range of suitable motors.
T hese products have made practical a wide range ofsophisticated control possibilities and the ability tointerface easily with building management systems.
The power conversion efficiency of such drives is typicallyabove 96%, although the imperfect output waveform mayreduce peak motor efficiency by 1 or 2 %. Motor efficiencymay itself reduce significantly at speeds below 75%.
There are many manufacturers of these devices, offeringmany advanced features. A limitation on their use hasbeen the lack of experience of H VAC installers with thetechnology, particularly with regard to electromagneticemissions and immunity.Figure 5.1 Flow control by speed regulation
1
2
3
75% full speed
Design operatingpoint
Fan outputcharacterist ics
Fan powercharacterist icat full speed
50% full speed
Full speed
75% full speed
50% full speed
Volume flowrate, qv
Fan
stat
ic pr
essu
re, p
sFa
n sh
aft p
ower
, Pr
Pr1
ps2
ps1
ps3
qv3 qv2 qv1
Pr2
Pr3
Fan Applica*on Guide CIBSE TM42: 2006
0.3
1.0
0.4
$million savings/yr
COP chilled water + chiller delta T
15-‐17/24°C
6/12°C
22°C Supply Air
24°C Return Air
Chilled
water
tempe
rature se
t po
ints
Cooling un
it set
points FR
OM
FROM
TO
TO
Set points
Opera>ng envelope for humidity control widened in line with ASHRAE recommenda>ons
Ven>la>on & addi>onal chw temp
Small number of hours above 15°C dew point
0.3
1.0
0.4
$million savings/yr
ven>la>on extract fan Ven>la>on flow rates
50% RH 5.5°C
dew point
Psychrometric chart
!
Key decisions • Choice of free cooling system
– Choice of an evapora>ve system – No. of heat transfer processes
• Area of heat exchangers – CAPEX – OPEX Q = A . U. LMTD
Q = cooling (kW) A = area (m2) U = heat transfer coefficient (kW / K m2) LMTD = log mean temperature difference (K) Given fixed Q and U then LMTD α 1/A
T
Air Data Hall 23°C
12°C
6°C
Chilled water
Total Appr 27 K
BP
31°C T
27°C
Total Appr 12K
BP
22°C R
23°C
R
-‐4°C
Free cooling Outside air
2°C
23°C
21°C
31°C
22°C
18°C
15°C
11°C
Heat exchange area Before
Heat exchange area Arer
Free cooling feasibility study
Chillers CRAH units
Buffer tanks
Primary circuit Secondary circuit
Dry coolers / Cooling towers
Dry coolers
Capacity
Approach temperature
CRAH cooling coils
Free cooling feasibility study – best op>on
!
Predic>on:
Free cooling feasibility study
Installa>on of free cooling with dry coolers
0.3
1.0
0.4
$million savings/yr
Year 1
Fiscally posi>ve
3 years
ROI less than
0 1
1
Flow Performance 1-‐BP
Thermal Perform
ance
1-‐R
Af=0.25
Af=
4
ORIGINAL: BP = 80% R = 20% Af = 4
IMPROVED: BP = 42% R = 9% Af = 1.57
Air performance arer improvements
7
PUE Benchmarking: Continued Year-On-Year Improvement
3.38
1.98
1.71 1.68 1.69 1.61
1.48
1.300
1.800
2.300
2.800
3.300
CDC Q1 YonY Performance 2.87
1.96
1.74 1.70 1.70 1.66
1.53
1.300
1.500
1.700
1.900
2.100
2.300
2.500
2.700
2.900
CDC Q2 YonY Performance
2.58
2.06
1.72 1.78
1.68 1.74
1.54
1.300
1.500
1.700
1.900
2.100
2.300
2.500
2.700
CDC Q3 YonY Performance
2.36
1.84
1.67 1.63 1.62
1.65
1.47
1.300
1.500
1.700
1.900
2.100
2.300
2.500
CDC Q4 YonY Performance
A) ������������������������� ���� ����������� ������������ B) Reduction in PUE directly attributed to energy efficiency initiative
A B
0.3 1.0 0.4 = $1.7m
ANNUAL SAVINGS per year
1.41 Instantaneous PUE 1.49 Es>mated annual PUE
35% Overall reduc>on
Conclusions
Dec 1st @ 11:00 2050KW of Heat RejecLon for 10KW electrical energy. COP of 205 @ 7.6 degC ambient vs ~COP 3 from Chillers
Annual PUE Trend YTD Note : DAC Turned on in Dec 13
DAC ON
Conclusions
Conclusions
16
Step 3,4,5 : Changing Set Points, Reducing Fan & Pump Speeds PUE � Improvement of 34% IT Load (69%) � No Change for evaluation UPS Transformation Losses (7.5%) � We have rotary machines that are operating at 90% efficiency so
a fixed loss at present, albeit we matched capacity to load where possible
Lighting and ancillary loads (10%) � Reducing Lighting footprint and run hours, re-educating users, ��������������������� ���
Secondary Pumps (1%) � To re-balance CHW flow rate to cooling load, saves pumping ���
Data Hall Air Movement (7%) � Achieve better control of air distribution, re-balance air flow rate to ����������������������� ��������������� ���
� Change CRAC control Strategy to supply air, decrease fan speed (cube law benefits)
Chiller and Primary Pumps (7.5%) � Raise temperature of CHW to improve COP of the chillers. Re-
balance CHW flow rate to match secondary load, saves pumping ���
� Adjust humidity control window, reduce humidifier hrs � Install Free Cooling benefit from much higher COP � Ability to turn off primary pumps when in 100% free cooling mode
Initiative to date : PUE ~1.49
Pre Initiative : PUE ~2.29
Indicative Graph showing Impact to PUE of Raising Chilled Water Temperatures
UPS Op>
misa
>on
Con>
nued
increase
Data hall supply temperature
26C
Chilled water temperature
21/28C
Free cooling hours
2-‐3%
Further savings
Our workshops are accredited by CIBSE
RobertTozer@dc-‐oi.com www.dc-‐oi.com
www.linkedin.com/company/opera>onal-‐intelligence-‐ltd @OI_RobertTozer
Ques>ons?