How can I compare chillers for data center applications?
Every planner/consultant or customer is familiar with this problem. Once the decision to purchase a system has been made, the tendering process begins and you are confronted with the major task of weighing up which product is most suitable for this project.
This article deals with the question of what technical data can be used to compare equipment, and how we can question or verify the credibility of this data relatively easily.
Before we get started, we should first be clear about what our main focus is. What are the most important factors for my project? Is the focus on capital expenditure (CapEx), operating expenditure (OpEx), noise levels, or the easiest possible integration in an existing system? Comparing capital expenditure is relatively straightforward. However, it's very important to ensure that both machines can feature the same equipment. Does the standard version of a chiller have something to offer that is only available as an optional extra in the other model?
Where integration is concerned, the key aspect is collaboration with the manufacturer, and how flexible they are. Here, comparisons are already getting more difficult. However, it's clear that some manufacturers are more flexible than others. In this context, flexibility means far more than simply adding to the standard range of options; it could involve a larger compressor with the same footprint, the adaptation of load entry points, specific electrical requirements, and a great deal more.
Once we have gathered documentation for the same basic configuration from two or more manufacturers, the great comparison of technical data can commence. Defined KPIs such as EER and ESEER are a popular means of comparison. But how meaningful are these really? To gain more clarity, let's first define what these two values are actually about. First of all, here is an explanation of how they are calculated:
Energy Efficiency Ratio (EER)
EER is the ratio of cooling capacity to power consumption. This value should be as high as possible (i.e. not much energy is required to produce the desired cooling capacity).
|Comparison is possible with the same operating point.||Only one operating point is considered.|
|Does not take account of efficiency throughout the year.|
|Imprecise, because not all manufacturers use the same consumers for their calculations (e.g. pump).|
European Standard Energy Efficiency Ratio (ESEER) = 0.03 A+0.33 B+0.41 C+0.23D
whereby A-D stand for the following design parameters:
|Operating point||EER factor||Outside temperature|
|A- 100 %||0.03||35 °C|
|B- 75 %||0.33||30 °C|
|C- 50 %||0.41||25 °C|
|D- 25 %||0.23||20 °C|
With the ESEER, it is assumed that cooling capacity decreases as the outside temperature falls, as there is less need for cooling.
|Comparison is possible for several operating points.||Cooling capacity constant in the data center.|
|The above assumptions do not always apply.|
|Water temperatures are not project-specific, but predefined.|
Both these KPIs only provide limited information on the efficiency of a chiller in a data center. One alternative would be a comparison of operating expenditure over the year as a whole, including all project-specific data. This method is based on a weather profile at the respective location, so it takes into account behavior over a broad temperature range, and all operating modes (DX, MIX, FC). When comparing operating expenditure, it's important to ensure that calculations are based on the same temperature profile. Otherwise, significant discrepancies may occur. A lot of operating hours at cold temperatures considerably improve the statistics.
Can I to rely on the technical data, or can I shed more light on them? Occasionally, we need to question the technical data. A closer look makes sense whenever the difference in price is considerable and the technical data are very similar.
Which components are installed?
First of all, we need to verify which components from which manufacturers have been installed. There will be a lot of overlap. But even here, it's worth putting this under the microscope. Why? Because a chiller is a complex system, not a single component. So what matters is harmonious interaction between the individual components. Take the compressor, for example: even if the same compressor appears to be installed in both chillers, the important thing is its environment. Every chiller has an evaporator and a condenser, as well as the compressor.
So what's the difference? The evaporation temperature and condensing temperature are the reference variables for the energy consumption and operating behavior of the compressor. So as you can see, there is a direct influence. The evaporation temperature should be as high as possible, while the condensing temperature needs to be as low as possible. The difference between the evaporation and condensing temperature is the path, so to speak, that the refrigerant has to take, with the aid of the compressor. Every kelvin that can be saved here translates as energy savings of 3-5 % for the compressor. But how do we obtain a low condensing temperature and a high evaporation temperature? Both depend on the design and the material. For both the evaporator and the condenser,
Q = A ⋅ k ⋅ΔT
A is the surface area in m2
k is the heat transfer coefficient
ΔT is the difference in temperature between the chilled water inlet and the chilled water outlet temperature.
This formula demonstrates that both the material and the size/area are decisive factors. In this case, size matters. Consequently, a glance at the technical drawing of the chiller can already help us to determine whether the data are plausible. Can the chiller with a smaller condenser surface area really have a lower condensing temperature and better energy efficiency? As a rule, the answer is NO.
The same rule applies to Free Cooling. The larger my Free Cooling surface – i.e. the Free Cooling coil – the sooner the system can switch to Free Cooling mode. Here, too, size matters, and plausibility can be checked simply by taking a look at the drawing.
The pressure drop
The datasheets contain yet another value that can often only be compared with difficulty: the pressure drop. The pressure drop determines how large the chilled water pump has to be. If the overall pressure drops are much higher, the chilled water pump may have to be bigger. Even if the pump only accounts for approx. 10 % of energy expenditure in a CW system, when viewed over lifecycle large savings can be achieved. Why are the values often difficult to compare? Because not all manufacturers calculate their data on the same basis. For instance, manufacturer A may only state pressure drops over the evaporator, while manufacturer B specifies the total pressure drops over the entire chiller, including chilled water piping. Here, it pays to exercise caution.
In addition to energy efficiency, the subject of noise is also becoming ever more relevant (see also our blog post on Low-noise data center cooling).
Here, too, a simple comparison can reveal whether the data are correct. Manufacturer A and manufacturer B provide different noise data. Which value are they comparing? The sound pressure, or the sound power level? What's the difference? Sound pressure depends to a large extent on the acoustic properties of the environment. Furthermore, this begs the question of how and under what conditions these measurements took place. As we can see, a genuine comparison is not possible on this basis. Since the sound power level is not dependent on the acoustic properties of the environment, it is a characteristic specific to the equipment in question, and is therefore the only admissible value that should be used for a serious comparison.
If these values are compared, it is relatively simple to check whether the data are realistic. For chillers, in most cases it is the fan noise that dominates. Therefore, we need to take a closer look at the fan and its associated data. What is the fan diameter? And how many revolutions a minute does it need to produce a given airflow? A smaller fan is hardly likely to deliver the same airflow with lower power consumption and less noise.
Is the fan an AC or an EC model? If it's an EC fan, we have to ask what its speed is at the operating point. If it runs at full load, it doesn't offer any advantage over an AC fan, as it achieves its greatest savings in partial load mode.
Last but not least ...
The integration of the chiller in a system also has to be considered. As mentioned above, it is the flexibility of the manufacturer that counts here. However, optional extras and operating limits may also play a role. If too many electrical extras are needed, an external switch gear cabinet is frequently required. This is a vital consideration, because this also increases both CapEx (which is not directly attributable to the chiller) and the footprint, for at the end of the day space needs to be found for this external cabinet. If everything fits inside the chiller's switch gear cabinet, installation is vastly simplified.
Another aspect in this connection is the behavior of the chiller during and immediately after a power blackout – the worst-case scenario for every data center operator! How long does the chiller need to return to generating 100 % cooling capacity? How flexible is it where switching between two networks is concerned? How quickly can it be switched? How high can the water temperature be for it to restart without problem? Where are the operating limits? Can I save buffer tank expenditure through a broad range of application? In the best case, the tank can be smaller, because the chiller can start without problem despite high water temperatures.
So what should I do?
- Use operating expenditure based on the same weather profile to compare energy efficiency.
- Check the plausibility of the technical data.
- Peruse drawings and consult them for your comparison.
- Always take account of the system as a whole, and all the various influencing factors.