When it comes to CW systems for data center cooling, many air conditioning specialists believe that it makes operational and economic sense to do away with water-glycol refrigerant in the data center interior. But more detailed analysis suggests that this theory is the exception, not the rule.
Whether pure water should be used for data center cooling chillers, dispensing with glycol altogether, is an issue of ongoing debate. The basis for this discussion is the fact that using glycol comes with a range of disadvantages:
1. heat transfer is not as effective with a water-glycol mixture.
2. glycol is much more costly than water.
3. larger pumps are required to circulate a water-glycol mixture than for pure water, increasing not only the scale of the project but electricity consumption as well.
Proponents of glycol-free data center interiors argue that using pure water would reduce investment and operating costs and improve cooling capacity. In their view, glycol should only be used where anti-frost requirements render it indispensable: in the pipeline systems leading to the chiller outside the data center.
But closer examination of the facts quickly shows that there are loopholes in this argument, because it omits to say that system separation is a prerequisite for dispensing with glycol. Instead of a single chilled water circuit between the interior air conditioning units and the chiller on the roof of the building, a glycol-free data center interior means splitting the system into two chilled water circuits: an interior circuit with pure water and an outside circuit for Free Cooling, which is still filled with a water-glycol mixture. This system separation takes the heat load from the water circuit and transfers it to a brazed plate heat exchanger in the water-glycol circuit which then conveys the heat from the interior of the building to the outside chiller equipped with a Free Cooling system. Obviously, this uses less glycol than a CW system with only one refrigerant circuit.
But it requires additional components: alongside the heat exchanger it necessitates a pump as well as frost-repelling pipeline heaters for the pure water circuit, and a number of additional components such as special piping and wiring work. So dispensing with glycol doesn't just save money: it also creates extra work. Ultimately there is so much additional effort involved that it negates the savings made by removing glycol from the equation. So in the final analysis, the supposed reduction in investment costs is untenable.
But what about the theory that glycol-free systems are cheaper to operate? A system comparison using the example of a continuously operational data center in Hamburg provides some basic information on this. Operating costs were calculated for an air-cooled chiller with a 700 kW cooling capacity, integrated Free Cooling and input and output temperatures of 18°C and 12°C respectively. The electricity costs were estimated at 15 euro cents per kilowatt hour. Under these conditions, the annual operating costs for the pure glycol system with only one refrigerant circuit were 33,000 euros cheaper. This takes account of the fact that a water-glycol mixture requires more pump power and that the losses in capacity caused by heat transfer must be offset by increased electricity consumption on the fans in the precision air-conditioning unit.
So why is it that the glycol-free system ultimately costs far more to operate?
The key factor is the compact integrated brazed plate heat exchanger between the inside and outside circuits. Firstly, the heat transfer losses that occur in this part of the system increase compressor running time. Secondly, the pressure drop that occurs in the flow through the brazed plate heat exchanger pipes increases the pump power requirement significantly in both the interior and Free Cooling circuits. This additional energy consumption means that at least part of the efficiency benefits of Free Cooling is canceled out. In contrast, the pure glycol system can exploit the benefits of Free Cooling without compromise, and need not compensate for any heat transfer losses on the heat exchanger.
This means that on closer analysis, the argument that glycol-free systems are cheaper to run is equally untenable – at least under conventional pre-existing site conditions such as those in our example. But it can make financial sense to use glycol-free systems at sites where Free Cooling is unfeasible and therefore where permanent compressor operation is required. Although even then, there is still the disadvantage that each additional component increases the statistical probability of a system failure. So whatever the case, system separation entails other risk factors in the data center.