ASHRAE Journal - February 2024 - 44

TECHNICAL FEATURE
hand, the HSP method operates at an operating load
factor of 80%, with 120 fan wall units, including the
spare, actively covering the same load. Although HSP
operation increases the number of active fan walls, VSD
fan power consumption is reduced by approximately
9%, from 1,515 kW to 1,372 kW, due to the difference in
the operating load factor.
With CSP operation of the plant system, one spare
unit is on standby and 5 chillers and cooling towers
are active, with operating load factors of 95% and 87%,
respectively. If the same capacity is operated using
the HSP method, six chillers and cooling towers,
including the spare, are operated, and the operating
load factor is reduced to 79% and 83%, respectively. In
this case, the variable speed chiller has an increased
COP, which reduces input power by about 17% from
3,060 kW to 2,527 kW. Additionally, the cooling tower
fan is controlled by VSD, which reduces the air volume
proportionally to the load factor of the chiller, resulting
in a reduction of input power by about 33% from 328 kW
to 220 kW.
As a result of applying HSP, the operating load factor
of the chiller and cooling tower is reduced, leading to a
decrease in the chilled water circulation pump from 95%
to 79% and the condenser water circulation pump from
87% to 83%, compared to CSP operation. Consequently,
the input power at this stage can be reduced by about
31% from 552 kW to 381 kW for the chilled water pump
and 4% from 1,500 kW to 1,440 kW for the condenser
water pump.
Table 3 presents a comparative analysis of energy
savings for HSP- and CSP-based cooling systems by
component in Data Center A. The savings rate of each
cooling system varies depending on the load factor and
the characteristics and efficiency of the equipment. The
rated input power of the entire data center's cooling
TABLE 3 The comparison of rated cooling power for the two cooling strategies.
COOLING
SYSTEM
Fan Wall Units
Chillers
Cooling Towers
Chilled Water Pumps
ACTIVE UNITS,
EA
LOAD FACTOR,
%
RATED POWER
PER UNIT, kW
system is reduced by approximately 15% by applying HSP
compared to CSP based on the same design capacity.
The typical power usage effectiveness (PUE) is
calculated by dividing the total annual data center
power by the power used by IT equipment.12 However,
to evaluate the energy efficiency of the cooling system,
the peak design PUEcooling is calculated and compared,
which focuses on the rated cooling power. Peak PUEcooling
is calculated by dividing the total rated power of the IT
equipment and cooling system by the rated power of the
IT equipment, as shown in Equation 5.13
(5)
For Data Center A, which has a capacity of 30 MW,
the rated power of IT equipment is 30,000 kWh, and
the rated power use of the CSP-based cooling system is
6,954 kWh. The peak PUEcooling is 1.23, indicating that
about 23% of the IT energy is used for cooling to operate
the data center. On the other hand, the HSP-based
cooling system uses a rated power of 5,940 kWh, which
is 1,014 kWh less power than the CSP-based cooling
system. The peak PUEcooling for HSP-based cooling is
1.20, which means that about 20% of IT energy is used
for cooling.
Discussion
INPUT POWER,
kW
HSP CSP HSP CSP HSP CSP HSP CSP
120 100 80 95 11 15 1,372 1,515
79 95 421 612 2,527 3,060
6
6
6
Condenser Water Pumps 6
5
5
5
5
83 87 37 66 220 328
79 95 64 110 381 552
83 87 240 300 1,440 1,500
Total 5,940 6,955
44
ASHRAE JOURNAL ashrae.org FEBRUARY 2 0 2 4
In data centers, the cooling system is crucial for
ensuring uninterrupted operation and responding
quickly to changes in IT services. In recent years, there
has been an increase in the use of VSD/VFD equipment
(such as chillers, pumps and fans) capable of adjusting
water flow and air volume to accommodate varying IT
loads in the cooling system components of data centers,
which can maximize energy efficiency within a certain
range of operating load factors. This
article proposes an HSP-based cooling
strategy that can effectively address the
redundancy of N + 1 or more redundant
equipment required for uninterrupted
operational reliability and optimize
energy efficiency. Through a case analysis
of a large-scale 30 MW Data Center A,
comparing the operating effects of HSP
and CSP, the input power of the cooling
system of the entire data center is reduced
DIFFERENCE,
%
-9.1
-17.4
-32.9
-30.9
-4.0
-14.6
http://www.ashrae.org
ASHRAE Journal - February 2024

Table of Contents for the Digital Edition of ASHRAE Journal - February 2024
