The goal of this exercise was to compare an estimate of interior temperature change of my home refrigerator to data collected from within the refrigerator over a 24 hour period. The experiment involves learning a valuable technique for (remote) temperature data collection, and serves to illuminate the nuances of a mechanical refrigeration cycle, which can be seen as a small-scale demonstration of how cooling/compressor cycles of a building may work.
I utilized the GE Hotpoint 182-A refrigerator unit in my studio apartment in North Berkeley for testing. The refrigerator is set to 7, on a scale from 1 to 9, where 9 is the coldest. It is located in the small kitchen adjacent to two windows that face east. Due to the relatively high temperature on the day of testing (58-91 degrees F), the windows in the kitchen were left open. At 11:30pm on August 31st 2010, I placed a HOBO temperature logger on the middle shelf of the refrigerator, several inches away from any food items. The logger was left in place until after midnight the following day, set to collect temperature data every 5 minutes. That data was then retrieved via computer from the logger, imported into Excel, and adjusted for daylight savings time. The temperature values for the 24-hour period of September 1st were graphed and overlaid with the graph which I crudely estimated at the start of the exercise.
The temperatures collected range from 39.4 to 51.7 degrees (F), with a mean temperature of 41.5 degrees. There were sharp peaks in temperature at approximately 3am, 6am, 6pm and 9pm. The highest temperature by far occurred around 9pm. And the lowest occurred at about 9am. The dead band (representing the temperature at which the compressor turns on an off) exhibits a range of approximately 1.5 degrees, with a midpoint of about 40.5 degrees.
Observations and Conclusions
The setpoint that I estimated was quite a bit higher than that which was observed, 45 versus 40.5 degrees. I was not aware of the setting of my refrigerator, or how close to freezing refrigerators typically are kept at.
I additionally did not account for the dead band range, assuming that the compressor would be able to produce an almost constant temperature. This range seems fairly narrow, allowing the refrigerator interior to rise just over 1 degree. It is surprising that it takes only about 10 minutes for the temperature to rise the 1.5 degrees from the low to the high end of the range. Additionally it is surprising that it takes an equal amount of time to cool the fridge to the low end of the range. I would have thought a securely closed unit would maintain the cooler temperature for a longer period without rising, and that it would take a shorter period to mechanically cool the unit than it took to warm up again on its own.
Anticipating my data, I also thought there might be some increase in the fridge temperature as the ambient temperature in the room increased, from the time of morning sun on through the day. But the increase outside seems to have resulted in only a fraction of a degree increase inside the fridge.
I predicted peaks in temperature at times when I typically open the fridge for meals and snacks. Though my actually times of accessing the fridge vary a bit, the magnitude of peaks and their correlation to the door being opened seem consistent with my estimate. The one inconsistency is the peak in temperature around 3am. I am uncertain whether this is due to a power saving or defrost mode the refrigerator goes into when the outside temperature drops or at certain intervals of time.
I predicted that peaks in temperature would occur almost instantly when the door was opened, and that the temperature would remain there for some time until it gradually was cooled back down to the original temperature. In actuality, the rise in temp occurred a little more gradually than I predicted. This may partially be due to the cold contents of the fridge still giving off cooler air after the door is opened, thus slowing the overall increase in temperature.
After peaking, the decrease in temperature occurred almost instantly as well, without lag at the highpoint. The decrease appears to be fairly steady and linear, however it is interesting that the dead band cycle begins again before the temperature returns to the original setpoint. It seems to begin at a slightly higher temperature, with the entire dead band cycle slowly decreasing in temperature from there. This may be to conserve the amount of energy needed to cool it all the way down, by supplying shorter bursts of cool air versus a steady flow.
I would be interested to see more detail of the peaks in temperature. Perhaps collecting data every minute would show somewhat more curvature in the graph, and leveling off at the peaks. Additionally, plotting external factors, such as the exact times the door was opened or the ambient temperature in the room would possibly reveal unseen relationships.