Chiller Staging on Measured Capacity
/I have a thought for chiller staging by capacity. Why not measure a chiller's capacity every time another chiller stages up? After all, that's as much as the chiller has just proven it can do after accounting for whatever shape it's in.
Note: This approach appears to be suited best to positive displacement (PD) machines such as screws and scrolls, though a modified version for centrifugal machines also appears feasible.
Background
Staging up, of course, is necessary - running enough chillers to meet the load is essential. Staging down is also necessary - to keep chillers within acceptable operation to avoid surge, low flow, and/or poor performance. Staging up can be simple, as waiting a few minutes after failure to meet set point.
Staging down is much tougher, because knowing the "right" time to do it depends on the chiller characteristics and current operation. Commonly, people set a chiller motor current low limit, as %FLA, when a chiller will stage off after a timer is satisfied. However, the correlation of current to load changes drastically with changing condenser conditions. Since we should be taking advantage of that "relief" in the condenser generally, FLA isn't the first choice any more. We have moved into a new paradigm, that of staging based on measured load vs. capacity. This means we stage a chiller off when doing so will improve efficiency AND the remaining chillers have capacity to meet the load.
Exploration of Concept
Measured thermal load is straightforward. Though the accuracy and the placement of temperature and flow sensors can be an issue, let's assume we have this reasonably accurate (see Notes below for more on that).
What about capacity, which isn't always exactly as per the design, the nameplate, or even the manufacturer's published data? Let's ignore the variable capacity of a centrifugal machine vs lift, for this post. For a PD machine, capacity is a function of at least:
- Suction condition, which is reasonably constant due to generally stable set points for leaving chilled water.
- Bypass, in the mechanical imperfections of the machine. This can change with wear and oil maintenance issues, though admittedly I haven't studied out or measured how much.
- Refrigerant charge, which we know does change with leaks and is affected by incorrect charge.
- Heat Exchanger Condition, which changes over time especially on the condenser
- Capacity control, which for now we'll assume is always at 100% speed/slide valve/load valves etc.
With several potential drift factors above, why not just take the capacity of the chiller to be whatever it most recently delivered, in the last X minutes before a lag chiller was staged up? This has benefits:
- It gives a more accurate picture of the proper stage-down, which is the point of this post... possibly avoiding short-cycling (when capacity is lower than expected) or avoiding missed savings when capacity is higher than expected.
- It offers a chance to observe changes in capacity, as a check on maintenance condition.
Over time, we can then establish a chiller's capacity as a function of lift conditions. If that curve changes, we can also flag it as a service issue.
Since this is a new thought, I welcome comments and questions - please feel free to poke holes in my idea.
Notes, By the Way...
- This same discussion can apply to refrigeration compressors that don't even make chilled water, for flooded or DX systems.
- The accuracy of the measurement of flow and temperature is less of an issue here. Why? The flow and temperature sensors are used in both the load measurement and in the capacity measurement, so errors are systematically neutralized - which sounds more awesome than offset.
- I've left this open for centrifugals, as they are more complex. Still, think of this - last known good full-load capacity as a function of lift could come in mighty handy when staging single-stage or VFD chillers that are best staging up well before full load; the issue there is if you keep them out of full load, you'll never know... maybe that's for a later post.
Agitation flow is less than process-required flow in this case; the process only occurs during roughly (10) 90-second dips per day. The remaining time can have reduced pump speed.