Updated: Aug 11, 2020
HVAC cooling systems for data centers often have several layers of redundancy through redundant equipment or distribution systems but these systems are useless without a functioning controls system. Electronic components such as sensors can go out of calibration and controllers can fail and at a rate that is usually higher than the mechanical components of the system.
These components can easily become the Achilles Heel of an HVAC system unless careful consideration is made for failures and redundancies in the controls system. Additionally, considerations must be made for the controls system to run during loss of normal power until emergency power systems are fully online. Finally, fully redundant controls systems could be considered.
Sensors for measuring room, outdoor air and water temperatures or humidity are often control system inputs and thus can cause a system to operate improperly if the sensors are providing bad readings to the control system. To address this, two sensors could be used for the same data point with programming such that if the difference in reading between the two sensors exceeds a value, an alarm is generated to alert facility personnel to a faulty sensor. The sensors must be placed in the same location to ensure they are reading the same source. In lieu of two separate sensors, a single sensor could be used but with high and low limits programmed into the system so that if the sensor indicates a value outside of the expected range an alarm could be generated to indicate a faulty sensor.
In most data centers HVAC cooling system mechanical components such as chillers and pumps are fed from an electrical distribution system that is on standby power. This standby power is most commonly achieved with an engine driven generator. Upon loss of normal power, the engine starts and ramps up to meet the load. The period between loss of normal power and the engine coming up to speed can last several seconds which is enough to disrupt power to HVAC equipment and controls. With key equipment such as a chiller, considerations must be made to ensure that automatic restart is provided when power is restored otherwise facility personnel might have to manually restart the equipment, losing valuable time and critical space conditions. Controllers for ATC systems should be provided with UPS’s with several minutes of backup time to bridge the gap between a loss of normal power and establishment of standby power. Furthermore, the ATC systems must be powered from electrical circuits which are fed from panels on standby power.
Modular equipment is popular with data centers as it can provide redundancy at reduced footprints and first costs. This is mostly commonly seen in air cooled chillers that consist of several modules with one or more redundant modules. These chillers often come as a “single” unit controlled by a master controller. Some manufacturers offer options where failure of the controller does not shut down the unit, but rather the chiller will run in its last operating mode prior to controller failure. The configuration and available options for the controller must be understood and factored into the overall resiliency plan for the plant to ensure loss of a single component does not shut down the entire cooling system.
Finally, entirely redundant controls systems should be considered for an HVAC cooling system. This would provide the ultimate level of resiliency albeit at the highest first cost. Like a chilled water distribution system with separate “A” and “B” redundant loops, a complete separate network bus with network level and terminal unit controllers and sensors could be provided. This redundant controls system could be designed such that a single system controls all equipment or if the equipment and associated distribution system is N+1 then each separate system could be provided with its own controls system.
Written by: Chris Wysoczanski, PE, LEED AP
Senior Associate &
Mechanical Group Leader