Richland Home Temperature Measurement:

In what is likely to be a moment of nerd weakness, I have started on the Richland version of the home temperature measurement system I had up and running in Wichita. (This is ignoring many, many other things that I have on my to-do list such as:

• Update/repair the digital picture frame (video driver is dead and I'm probably going to update to a Raspberry Pi 3)
• Finish the retaining wall in the back yard
• Debug the Lorenz project so I can take it into work with self-respect
• ...)

Due to the layout of this house, I'm not able to very easily implement this as a wired system and have been hunting around for good wireless options. I decided on using Low Power Lab's Moteino which uses the RFM69 radio. The radio has range should be more than sufficient to cover the range of my house and the Moteino can come with the radio installed and associated libraries make it easy to get the system up and running.

As I found out today: in less than fifteen minutes I had the sample system up and running with one node sending packets and the other receiving them (and acknowledging back to the sender the reception). With another thirty minutes of work I had the remote programming also working. This will be important as I plan on putting some of these nodes in hard-to-reach places and having the ability to wirelessly update their code will be important.

This was the hard part made easy. Now the hard part for me: start thinking through how this system is going to work and get my head around wireless communication paradigms.

Garage Heating

Graph from yesterday's home temperature measurements:

Of particular note are the garage (dark blue) and attic (green) traces. There are three bumps in the graphs that mirror each other and are co-incident; these events are when the garage door opened to let a car in or me out to shovel snow. The two areas are connected by an access hatch that I tend to leave open, allowing air to flow between them easily.  During these bumps in the graph, the garage temperature falls and the attic temperature rises.  The warm air that has formed in the garage is displaced into the attic by the cold air moving in through the open door.  When the door closes, the temperatures in both places begin moving back to their former state.

This leads me to to observations:

1. The air flow through the garage is not very large when all the doors are closed.  The space is far from air-tight but it must be tight enough because the garage warms back up once the doors close.  The cold air from the outside isn't getting in near as well.  And where is that heat that is warming the garage coming from....
2. The insulation between the living space and the garage could be better. The heated house is the only source of energy that could be warming the garage after the door closes.  This has implications for the summer; keeping the garage cool by opening doors and allowing air to flow will hel reduce the cooling demand in the living space.

Daylight Savings Time

This is the effect that the time change had on my home temperature measurement system; look at the green and red traces between 1am and 2am.

Air Conditioner Thermostat Strategy

If you are a frequent reader of my infrequent writing, you'll know that last summer (2011) I installed a system to measure the temperature in and around our house as well as keep track of when the central fan is running (when the air conditioner or furnace is on). Last summer I ran two experiments

1. What is the effect on the attic temperature due to installing additional attic ventilation?
2. What is the effect on the amount of time the air conditioner runs during the day based on whether I turn it off or leave it on while the house is unoccupied?

The data from the first experiment showed about nine degrees of cooling after adding the extra ventilation determined simply by comparing the average temperatures before and after making the change.  The second experiment was far less conclusive and as I said at that time, a more complex statistical analysis would probably be necessary to make a determination.

That was a little over a year ago and I now have another season of data and the desire to jump into the statistics. You science nerds ready to rumble and see where this leads?

I made the choice this past summer to throw in another variable: rather than just measuring the effect of turning the air conditioner off during the day, I also investigated the effect of changing the thermostat set-point.  Every week on Monday morning I would re-program our thermostat with a different set-point and then each day of the week decide whether to bypass the schedule and leave it on all day or let the schedule run, cooling the house in the evening to the set-point I had chosen.  The three set-points I chose were 77, 78, and 79 degrees Farenheit; sadly, the 79 degree data set turned out smaller than I would have liked so I won't be able to use it for this analysis; hopefully by the end of next summer this will not be the case.  

I threw two summers of data into the statistics machine and hoped to answer a few questions:

Question 1 - Is it more energy efficient to turn off your air conditioner during the day (or when the house is unoccupied)?

This questions has been asked for many decades and the technical term I've found describing the strategy is "thermostat setback".  Much of the research seems focused on using this strategy during the winter for saving energy on heating and the internet is full of opinions.  I haven't looked very hard but I did find one academic paper from 1978 ("Energy Savings Through Thermostat Setbacks" by Nelson and MacArthur) in which the researchers used a computer simulation to try to answer the question.  Their general conclusions support the use of thermostat setback with an unsurprising caveat: the effect of the setback is most noticeable when the degree of the setback is large and the length of the setback is long.  The lower the change in thermostat setpoint and/or the shorter the duration of the change, the less significant the effect. In scenario at our house, both of these conditions are satisfied (roughly): the setback period is at least 8 hours and the change in temperature is high enough that the air conditioner does not run at all when setback.

To do the analysis on the data I had collected, I split the dataset into three parts based on the thermostat setpoint when it ran during the evening.  Each subset contained data showing the daytime state of the thermostat (cooling or not), the evening thermostat setpoint, the 2-hour average peak temperature of the day, and the number of hours the air conditioner ran that day.  I then ran a multi-regression analysis using the air-conditioner run-time as the dependent variable and the outdoor temperature and day-time state as independent variables. (For those of you who don't know, multi-regression analysis tries to determine the mathematical relationship between variables based on a set of data.  More importantly for our purposes, it will also calculate whether a given input variable has a significant impact on the stated output variable. Specifically, it will tell us whether the daytime state of the air conditioner has a statistically significant effect on the air conditioner runtime.)

Answer 1

• Thermostat setpoint = 77'F: Daytime state does have an effect on how long the air conditioner runs for the day.
• Thermostat setpoint = 78'F: Daytime state does not have an effect on how long the air conditioner runs for the day.

It looks like I just happened to stumble across the turning point.  The statistics imply that if I set the thermostat at 78'F, I will not experience longer run time if I just leave the air conditioner on all day rather than turning it off when I leave in the morning.  If I set the thermostat at 77'F and do choose the turn the air conditioner off during the day, the statistical model predicts a reduction in air conditioner runtime of almost 2.5 hours if I choose to do this. 

Question 2 - Does the thermostat setpoint have a significant effect on how long the air conditioner runs for the day?  If so, how much?

I haven't done the research on this one to have an informed opinion so I'm just going to jump to my analysis. Dataset was the same as above but this time was split into two datasets, one in which the AC was running all day and one in which it was off during the day.  I then ran the same statistical analysis to build a model of that would allow me to predict how long the air conditioner would run given the two-hour average peak outdoor temperature and the thermostat setpoint.

Answer 2 - The thermostat setpoint is statistically significant in determining how long the air conditioner will run each day regardless of whether the air conditioner is off or on during the day.

• Air conditioner off during day:  Each degree Farhenheit the thermostat is reduces saves 0.65 hours of air conditioner run time that day.
• Air conditioner on during day: Each degree Farhenheit the thermostat is reduces saves 0.80 hours of air conditioner run time that day.

Question 3 - How much money can be saved by using thermostat setback or increasing the thermostat set?

Answer 3 - I recently was able to measure the power of my air conditioner: 4kW when its running. Let's use the ballpark value of  $0.10/kWh for energy.  This means I'm charged $0.40 for every hour my air conditioner runs.

• Thermostat setback - If I choose to set my thermostat to 77'F and turn it off during the day (rather than leaving it running), I'll save almost 2.5 hours of air conditioner runtime which translates into $1.00 of savings per day.  Over a 30 day month this is $30 in savings.
• Thermostat setpoint increase - We can save $0.26 to $0.32 each day per degree the thermostat is increased.  This doesn't sound like much and over a 30 day month, this is a total reduction in the cooling costs of $8-$10 per degree.

All of the statistical models were linear in nature and we know from Newton's Law of Cooling that the heat loss rate of a house is non-linear; the hotter it is outside, the faster the house heats up. It should take much more cooling effort to keep a house 20 degrees cooler than the outside than just 10 degrees.  A linear model predicts it will take exactly half and this is not correct.  The model predicts the same amount of reduction in air conditioner runtime by moving the thermostat setpoint down one degree Farenheit whether it is 85'F outside or 110'F outside.  A non-linear model would work better here but until I figure out how to make the magic statistical software do this, we'll have to stick to the linear model.

That's what I've go for now.  Until next summer when I've got more data, this is what I know.

Updated Home Temperature System

I've been having a problem with my home temperature system for a while and the issue really manifested itself over this winter.  The some of the sensors, some of the time seemed to head off the deep end and this winter, the outdoor sensor took a turn for the worse.  Here's what the system recorded on February 11, 2012.  Note the red outdoor trace headed off to Saudi Arabia.  Over the past week or two the trace has literally been off the charts reading over 140 'F all day long.

Based on the the little trouble-shooting I had done to try to fix this problem it seemed that the problem was rooted in the type of sensor I was using: semi-passive and analog.  The device is cheap and very easy to use but the output is weak and lacks the ability to consistently drive the relatively long cable runs (~50 feet) from sensor to processor.  Most of the sensors seem to work most of the time but I was having enough trouble that it seemed time to make a change.

My plan was simple: put an amplifier out by each sensor to provide enough drive to be immune from the unidentified craziness that seemed to come and go.  I found a candidate part and when it arrived this past Saturday I prototyped a new little sensor board Saturday evening and built and installed the rest on Sunday.  As you can see from the results below, things seem to be working much better.

Besides an increased reliability, the amplified sensors allow me to greatly increase the performance of the system overall.  The number of samples I average per log entry went from 35 to 1000 (take that central limit thereom!) and the time between log entries went from 3.5 minutes to 1 minute.  I could have easily pushed to system more in both direction but for now this seems more than adequate.

Before installing the sensors I tried to calibrate them with little success.  I placed a bad of crushed ice and water over the sensors with the hopes that all the sensors would reach the freezing point and I could compare their outputs.  After ten minutes of this it was obvious to me that this wasn't getting the sensors cold enough and so I settled for the next best thing.  Once the sensors reached room temperature again I simply compared their outputs and added an offset in the processing code to each one, bringing them all into rough agreement.  All the sensors now read the same temperature but there is no guarantee that this temperature is the correct one.  All six sensors are now consistent but not necessarily accurate in absolute terms.

I'm very happy with the results so far. The system is much less noisy and the higher update rate is nice; its like my temperature system is brand new again.

Nerd Details
The TMP36 is a nice little sensor but lacks drive; the output is high impedance and so it seems susceptible to static build-up and noise over these long cable runs I have.  Additionally, to measure multiple sensors on the Arduino, the high impedance forces long delays between read and multiple reads per sensors to get things accurate.  This limited the rate as which I could collect data and even then things didn't work so well, as can be seen above.

The op-amp I used is the MCP6002 which is a low-power, rail-to-rail op-amp. I did a quick characterization of the op-amp as a voltage follower and it followed the input 1:1 from 0.1 to 5V, more than enough range for the expected output of the TMP36 (and much better than the other op-amps I tried, see below).  Doing a little algebra revealed I could probably even amplify the signal a bit to push my nominal values a little bit out of the weeds (77'F is 0.75V).  Instead of configuring the op-amp as a follower I set it up as a non-inverting amplifier with a total gain of 2.82.  To prevent oscillation in driving my long twisted pairs and the capacitance they tend to create I used this table to make an informed guess and added a 470 ohm resistor in series with the output.  Finally, to make installing the sensor much easier, I used a three pin terminal block as the connector.

The changes to the Arduino code were trivial.  I got rid of the extra analogRead(), reduced the delay between reads down to 10ms (I tested at 5ms and it seemed to work just as well), changed the number of reads from 35 to 1000, changed a couple data types to accomodate the much larger values I'd be generating in the running total, and added a unique offset that is applied to each final averaged sensor value based on my quick and dirty "calibration" I had done at my desk.  Oh, and I divided the final converted temperature by the 2.82 gain factor the op-amp introduced.

Results from Measuring Household Temperatures for Most of the Summer

The temperature measurement system I installed in late June has been running for several months now with only minor problems and with the highs now in the 80s, two of the experiments I've been running over the summer have come to a close and I'd like to share the results.

Let's do the easy one first: attic temperature.  A month or so after I got the temperature system going we could see that our attic was getting quite warm on those hot days, often above 130'F.  I talked with my wife about it, we did some reading, and decided that adding a bit more ventilation would be good idea.  The cost would be minimal and we hoped it would lower the difference between the outside air and the attic.  Thankfully it did; the results are below.

The difference between the old and new ventilation is pretty clear: adding the ventilation did lower the temperature difference between the outside and attic air. The statistics show a nine degree cooling in the attic by adding the extra ventilation.  (For you statistics nerds, the sample size is ~25 for both conditions, and standard deviation is ~5 degrees.)

The other experiment I ran was based on a conversation I had at a cook-out over the summer.  One of the gals there said her father was a HVAC guy and that he recommended keeping a constant set-point during the summer; that is, don't turn the AC off when the house is unoccupied.  It was hard for me to believe that this would use the same amount of electricity as a more "conservative" approach of turning it off when gone but I realized I didn't actually have anything more than opinion to back up my assertion.  Time to do some science.

I decided to test this theory and added a sensor that would show me when the central fan in our house was running.  The fan only kicks on when the air conditioning (or furnace) are on and so this allowed me to measure how long the air-conditioner ran during a given 24-hour period.  I semi-randomly changed the programming on our thermostat to either hold a constant temperature all day or to turn the AC back shortly before we got home from work and school.  Here's an example of each:

The yellow line at the bottom is the state of the fan: 55'F is on, 15'F is off.  For a day when the AC was on a schedule, it turned off around 7am and would come back on around 2:30pm in an attempt to get the house down the temperature by 5pm when we came got home.  You can see the purple (kitchen) and cyan (hallway) lines rise throughout the day and then when the AC comes on in the afternoon, begin descending.

When we kept the thermostat constant all day, the AC cycles to keep the temperature in the house at the thermostat set point. The purple and cyan lines stay at an even value throughout the day.  (You'll notice the basement sensor is relatively unaffected by the AC.  This is why everybody should have a basement if they live in a place that gets hot.  Basements are the best.)

After a summer of running both cases, here are the results.

The results are much more mixed than I would expect.  I think to make any good conclusions a statistical linear regression would need to be done; I haven't done that yet and probably won't ever get around to it.  Its clear that the run-time of the AC is strongly related to the peak temperature for the day.  This should be no surprise to anybody.  It is less clear which thermostat schedule uses less energy.  For the very hot days (> ~105'F) you could make a pretty good case that turning the AC off when you're gone at work will save some energy.  For days when the highs are less than 100'F, though, it seems there is very little difference between the two cases.

These results are surprising to me.  I would have expected that keeping the AC off for seven or eight hours a day would cause it to run less in the grand scheme of things.  It may but the difference isn't huge. I may try repeating this experiment next summer, just to see how it turns out.  I guess the good news is that if you're home all day with kids or work you don't have to feel too guilty about having the air-conditioner running the whole time; its not killing your bill much worse than the rest of us.

Basement Sensor

My previously mentioned tutor and nerd mentor who enabled me to get started on this project took one look at my code and found the mistake that was keeping the basement sensor from working properly. Making its first-time ever screen appearance, I give you: the Basement Temperature (brown line).

The next problem to solve: why all the "noise" in the data during the second half of the day? We have a few guesses but don't have any firm convictions at this point but we're assuming that something is happening to drive a single measurement to an unreasonable level. The proposed fix is to throw out all unrealistic measurements either based on absolute limits (no temperatures greater than 200'F and less than -30'F) or a relative limit (no temperature change more than 50 degrees away from the previous measurement). This will complicate the code slightly but, hey, that's what nerd friends are for.

Whole-House Fan and Attic Temperature

A year or so ago we installed an Airscape whole-house fan. The fan is supposed to provide cooling to the house in two ways:

1. Pulling cool air from the outside into the living area, replacing the air in the house as well as cooling the interior structure of the house.
2. Displacing the highly heated air in the attic with cooler air from the outside (via the house).

We've been very happy with the cooling the fan provides to the living area of the house but we've had to simply assume that the fan was adequately cooling the attic.

Until now.

With the installation of my super-nerdy temperature measurement system, we now have proof that the fan is doing its job. Last night we turned it on as we we're going to bed, knowing the overnight lows would be cool enough to provide benefit. Looking at the graph, you can see around 10pm when the fan turns on the attic temperature (green) drops pretty quickly.

Nice the see our assumptions were correct.

Nerd Project: Household Temperatures

I spent most of Saturday crawling around the attic and drilling small holes in the ceiling to complete a project I've been batting around for over a year now. Since we put in our whole-house fan (about a year and a half ago) I've been curious to see what effect the fan would have in reducing the temperature in our attic. This got me thinking about temperature regulation in our house in general: the basement having much smaller changes in temperature during both the winter and the summer, the temperature in the two bedrooms we've don't used and have closed off, how much cooking in the kitchen heats the house, et cetera.

Enabled by a friend of mine who loaned me a critical piece of hardware (because he's even nerdier than me and had extras just laying around his house), I built a little system that measures the temperature in six locations around our house throughout the day. This collection of little programs creates an internal webpage that shows the current temperature for all six locations in the house and every morning creates a graph of the previous day's data and adds a link to that graph on the webpage.

Here's an example from yesterday.

A few items of note:

• The dark blue line for our garage shows a little ramp starting around 5:30am. This is when we started the clothes drier which vents into our garage and thus, warms it up. My wife has been interested to see how pronounced this effect is and whether we need to try to modify the venting so it dumps the air outside.
• The spike in the purple line a little after 6pm is dinner being cooked. Again, another wife-requested measurement.
• You can see attic (green) gets very hot during the day, hotter than the outside temperature. It was too hot yesterday to run the whole-house fan so there was no circulation in the attic. This data seems to suggest that getting some kind of attic fan that ventilates the attic better throughout the day may help in keeping the house cooler. We've got a fair amount of insulation but with the temperature knocking around 130'F during the peak of the day when the outside air is barely at 90'F, it seems like our attic could be acting as a heat source and some of that heat is sure to be leaking its ways back into our house.
• Relatedly, the garage is definitely getting warmer than the outside our throughout the day as well. More insulation between the garage and the house would help but an easier solution may be to open the garage doors to allow the air to ventilate. There are plenty of hot days left in the summer to try this.

This system has been running for a few days now with only minor hitches. The biggest bug is that I something is wrong the basement measurements. I know the sensor is good as it was the first (and easiest) one to install and I used it as a proof-of-concept. I've been talking with my nerd-enabling friend and we've got a few ideas I'm going to pursue. I'm also a bit perplexed at how noisy the data is at times. The indoor data (kitchen and hallway) seem very smooth but the rest vary much more than I would expect. Maybe its not noise and the temperatures do vary that much.

Aside from adding a few more sensors (closed-off rooms, maybe the bedroom and the living room), I would also like to add sensors that detect when the whole-house fan is running and when the air-conditioning/furnace fan is running. I hope to get those last two in sometime this summer but I need to figure out the best/easiest way to do it.

I'll write another post soon in a few days detailing the specifics of how the system is put together for any fellow nerd out there who is interested.

Attic Fan

This past spring, in an attempt to save ourselves some money, hassle, and discomfort, we decided to purchase and install an attic fan. Katie had experience with one growing up and insisted that if we were going to do this it had to be quiet so we ended up on one provided by Airscape. As with my home projects, the install took longer than I thought but once it was in and running we were loving it. Instead of messing with multiple box fans that didn't seem to work very well we open a few windows and flip a switch. The fan pulls in the cool air from outside and shoves the hot air in the house and attic outside.

The fan ceases to be useful, though, when things don't really cool down at night very much. We knew that the fan wouldn't be able to be used all summer but based on historical data on the temperatures here in town, it seemed it would still be worth it the rest of the year. The summer heat has seemed to hit a bit early this year and already the night-time lows are 78'F at our house; there's no sense in exchanging one set of hot air with another so the fan is staying off.

Other than this not so small complication, we have been big fans of the fan. For people who live in places where the night-time lows are more reasonable, I can whole-heartedly endorse the product. Where I grew up, we didn't have air-conditioning and this worked out fine most of the year but a fan like this would have made all the difference and saved my father from same box-fan routine we used to do here.