Hello Nitin,
Lots of good questions here, but you are also asking for what could amount to a small book of answers if tackled comprehensively. I cannot share any commercially developed excel samples at this time in good conscience, but can provide a couple answers to help set you and others in the right direction. Perhaps the tides will shift and I will find myself sufficiently motivated to writing that ?book,? in the future ;-).
Recommended additional reading in response to this series of questions would include this article from the doe2 reference manual: Volume 2: Dictionary > HVAC Components > CURVE-FIT > INPUT-TYPE = DATA > INDEPENDENT-2
* I just want to clear the concept about how software works when we enter the chiller raw data points?
* The software is using raw data points, which reflect how a dependent variable (like COP or Capacity) varies as one or more independent variables (like chilled water temperature or part load ratio) are provided in different combinations.
* Try using the raw data input for a custom linear curve. It should become apparent that this is fundamentally exactly what is happening when you ask excel to create a trendline from a series of data points. The twist is that for custom performance curves with chillers and other equipment, you will find they are often three dimensional instead of just two.
* ? and how many raw data points from the chiller datasheet will be sufficient for software to give appropriate output.
* No easy/fast answers here. Fully answering this prompt could require at least a chapter for the aforementioned ?book.?
* To answer a slightly different prompt: The minimum number of points eQuest requires to derive custom coefficients varies based on the kind of curve you want to make.
* Linear ? 2
* Quadratic ? 3
* Cubic ? you can do this!
* Bi-Linear ? 4
* Bi-Quadratic ? see the pattern?
* I would never recommend a pursuit of the bare minimum number of data points, for reasons answered by addressing your actual prompt, next:
* You want to know, reasonably, the number of raw data points for an ?appropriate? curve regression. We therefore care to achieve ?appropriate? outputs for this analysis, which means there exists some threshold of accuracy expected of the curve/model outputs.
* If I am wrong and the accuracy of the output really isn?t a factor, then I would contend you could save a lot of time by renaming the default library curves and calling them custom.
* If your goal is to create a family of curves that will together closely match lab/field measurements (i.e. always determining an operating efficiency within COP +/- 0.1 for identical tested/measured conditions), then the unfortunate reality is that you will not know how many points are needed until you try and find out whether you have too few. It?s kinda like studying for the P.E. examination ? you don?t get your score so you never will know if you studied too much? only if you studied too little.
* Just as I advise candidates for the P.E. examination, I would in practical terms advise aiming high as a matter of time efficiency. Pursue and secure more information (data points) than the minimum required, up to and even exceeding the maximum number that the eQuest interface permits you to enter (20 per curve ? mind that Excel has no such limit). If/ when that becomes unreasonable (such as working with with a rep who is not practiced in this sort of venture and needs many hours to produce even the minimum data you can work from), then you should sensibly bound and request a finite set of data points to cover the spread of actual operating conditions you expect your model to encounter. If after receipt and first-pass regression efforts the resulting curve is determined to be of insufficient accuracy, I would then assess and articulate the conditions around which more data is needed to better calibrate the curve, then follow up with a second-pass effort to gather new (additional) data points informed by the first pass effort. Rinse and repeat until you achieve an acceptable result.
* Which temperatures (LCHWT & ECWT) should be constant & should not be constant while working on the customized chiller curve using actual part load data available from the chiller datasheet?
* This question as posed is a little flawed: There is no requirement for absolute constance or variance for a given independent or dependent variable, but in specific cases a constant input can get you into trouble.
* If you are holding a dependent variable constant between all of your raw data entries, you are essentially asserting none of the independent variables matter and so there isn?t really even a point to making the performance curve.
* If you are holding one independent variable constant for a curve expecting two independent variables (for example: holding chilled water temperature at 44F for all raw data entry points when regressing coefficients for a chiller capacity curve), you are generating a curve whose outputs cannot be trusted for any conditions other than that constant. In some cases where you might be tempted to ignore an independent variable for good reason, you would be better advised to generate a 2D curve. Shifting to one independent variable from two is a viable approach and consideration for specific performance curves in doe2/eQuest (see EIR-FPLR for example).
* When developing 3D curves, you will typically encounter a series of points with one independent variable held constant, to explore what happens with only the other independent variable changing. It is fine/normal for a series of independent variable inputs to share a common value, just ensure between all raw data points the same variable is ultimately varying within the expected range of operating conditions.
* If you are ever duplicating raw data entry points to meet a minimum required, that is analogous to asking someone to draw a line of with only one correct slope on an X-Y axis, given only two matching points. It is not enough information to complete the task at hand.
Hope this is helpful!
~Nick
[cid:image001.jpg at 01D7216E.83117EC0]
Nick Caton, P.E. (US), BEMP
??? ????, P.E. (US), BEMP
Senior Energy Engineer
Energy Manager, Yokota Airbase
ESS - Energy & Sustainability Services
M JP
M US
Email
+81 . 070 . 3366 . 3317
+1 . 785 . 410 . 3317
nicholas.caton at se.com
?????????????
???????????????
ESS - ?????????????
[cid:image002.png at 01D7216E.83117EC0]
[cid:image003.png at 01D7216E.83117EC0]
[cid:image001.jpg at 01D7216E.83117EC0]
