pondering zone thermal mass

Ron Schroeder

Hi Jerry,

"Unfortunatly" meaning no conventional "thermostat" settings from each room. I have to change settings on the central computer. My logging sensors are type K thermocouples and the room temperature sensors and boiler control sensors are precision 10K thermistors that have the same resistance curve as Tekmar. I am using 4 wire kelvin connections to the thermistors that are far away.

jerry scharf_3

the next step in my control system, need ideas

For those who might not remember, I put in a munchkin system with a bunch of diderent types of radiant heat and varying building insulation levels. The system is microzoned, with 13 zones to control. Each zone (approx 1 per room) has a room temp sensor and a proportional globe valve controlling the water flow to the loops. It's a single water temp system. the current status is that all the zones are under simple control and the boiler temp is a simple outdoor reset model run through a "tekmar interface" to the munchkin.

I was hoping to get some other people's thoughts on the next probelm I face. I am running the valves staticly set at just above the flow rates needed to achieve steady state room temperature. Each valve will go on and off 1-2 times a day. This is very comfortable, but time to change the room setting is hours.

So I want to work on a system that when the zone is substantially different than the target, I want to dump a large quantity of heat then back off to the near steady state rates. To do this, I need to come up with a model of the room thermal mass and heat loss, then have a way of learning these based on the control system inputs and responses. I haven't come up with anything that seems to work for either the model on to way to calculate it. I'd love people to toss out ideas.

Some lessons learned:

My acorn nut and socket screw MRT sensor appears to have too much mass. I am builing a comparison sensor, but beleive it adds an artificial lag to the computer's view of the rooms. More tests with other sensor designs are clearly needed.

Python is a nice language to write a control system in. My entire control is running around 500 lines of code (comments? What are those??) The next step here is to start moving away from the current procedural model to a more class based design.

The 1-wire sensors and the Spartan Peripherals globe valves are very nice to work with, despite the low costs. Would still love to find a cheaper solution for the 0-10V analog outputs, but I have what I need for now. Have worked out a nice way to handle all the wires needed to do all this.

Turns out I need more than one outside air sensor, I am seeing effects that the building and trees are having on the measured temp.

Once I get the thermal mass and loss worked out, next comes modelling the solar heat gain from the windows.

Mike T and I are both hoping for an ultrasonic flow meter under the christmas tree. If that fails, I may have to start working on a plan to build a cheap one. I have talked to a bunch of people, and it looks much simpler than the cost of current units would indicate.

thanks,
jerry

Ron Schroeder

Thermal transients

Your answer may be here...
http://www.ornl.gov/sci/roofs+walls/research/detailed_papers/effects/influence.htm#effect

Harold

Would you be willing to share details, particularly about your control hardware and software. I am interested in that sort of thing and would like to fool around with my system once I can get the house finished. I tried and failed with a variable speed pump because of motor noise and would still like to see if that can be implemented later.

Ron Schroeder

Which Spartan valves are you using?

jerry scharf_3

thanks

Greg,

That was just what I was looking for in terms of a model. What I have is fairly simple foam insulated stick frame wall construction. But the approach and the way of breaking down the problem was what I needed. In particular, the understanding of the lag and the inside facing construction factor were just the ticket.

jerry

jerry scharf_3

let me write someting up

I need to write up a document on this. I should have something in the next week or so.

jerry

jerry scharf_3

I am using the syncube series

Ron,

When I went out looking for a control valve, I assumed I would be getting something like a Belimo contoured ball valve. After looking around a bit, I met the folks from Spartan at the 2004 ISH show.

I really like the design. The great part is that the body and the valve are separate, so you can change the valve at little cost and no repiping. They offer a range of valves for each size based on the Cv. I am using the 1.0 Cv valves right now, they seem to give me conrtol autority over about 60% of the actuator range. The control to flow gets a bit touchy right near the closing point, but that's true with just about anything because of the localized head drop as they crack open. You do seem to need to characterize the open point on each valve, but that's a small thing. I haven't tried any of the .5 Cv valves, but they might work better in low flow applications.

The cost of these weren't much more than the cost of a good 24V on/off valve. This made the cost of the 0-10V control point to be a substantial part of the cost increase over a simple relay/zone valve setup.

BTW, what do you do at Brookhaven?

jerry

Ron Schroeder

\"thermal capacity\" \"gypsum\"

Jerry,
I googled the above words to find that article. I was going to calculate the thermal mass of the air and compare it to the sheetrock for you. Seems like you need the resistance from the air to the sheetrock(plaster) and the resistance from the sheetrock to the outside. The delta of the sheetrock will be less than the delta of the inside air. I did not try to understand the article...

What are you using for a controller/computer? I would like to control post purge times for my indirect and BB zones based on the boiler temperature.

Mike T., Swampeast MO

Do I have this correct?

In "normal" operation all of the valves operate with a fixed degree of opening set just a bit higher than actually needed? So you get very long heating cycles in each zone, but they still shut down completely once or twice a day.

Am I correct that the supply curve also is set to be just barely adequate? And that the "normal" setting of most or all of the zone valves is nearly wide open?

If so, then I believe the only way that you'll be able to raise temp in any space fairly rapidly is by increasing the supply temperature rather significantly during the heat-up phase. Once you do this all the rest of the zones will either have to be "more closed" or operate with significantly faster "on-off" cycles.

If these assumptions are correct, then I honestly have little idea on how to determine 1) the amount of supply temp boost and 2) the degree of closing required for each of the rest of the zones. 2) in particular will get EXTREMELY hairy if you try to do it mathematically and you're almost certain to need very accurate flow measurements. Am also assuming that you're not using a dedicated circulator for each zone. If so, do be aware that changing the flow in even one zone significantly will affect the flow in all of the rest. This is what I REALLY like about TRVs or FHVs as such compensation is inherent to their operation.

----------------------------------------------------

I did a "Euro-cave" response test last night. These observations might help you a bit.

I raised all of the TRVs in unoccupied spaces to the nearly the same setting as the occupied. This called for about a 7°F increase in temp. Outside temp was in the low teens and falling rapidly on a clear, night with nearly calm winds.

Prior to raising the TRVs, boiler running constantly for about 6 hours at or very near minimum modulation (25 mbh input). Boiler target and temp were nailed at 97°. Supply temp I measure (3' away on surface of 1" iron pipe) was very close and quite steady around 96°. Measured system delta-t was very steady at 13°.

As soon as the TRV settings were increased at about 9:30 p.m., measured supply and return temps plummeted by 8°. Measured supply temp was now 88°. Within 10 minutes all of the radiators whose setting had been increased were fully warmed (a number were virtually ambient). Boiler input had increased to about 31 mbh and was very steady (clocked 3 times). Boiler temp was again nailed on target at 97°.

Measured supply temp rapidly rose about 1°F to 89° then began a EXTREMELY slow climb. Return temp rapidly increased by 3° giving an 11° system delta-t. The 11° delta-t stayed completely steady as the measured supply temp SLOWLY climbed.

By the time the sun came up this morning and the outside temp began to climb, measured supply temp had only increased 1°. The spaces with the increased TRV setting had also increased by about 2°. The occupied spaces FELL by about 2°. Temp in those spaces had leveled off and had just begun to increase when I ended the experiment at about 10:30 a.m. this morning.

Had the outside temp stayed steady and the experiment continued, I believe it would have taken at least a day-and-a-half to achieve the increased temp setting in the unoccupied spaces (about 60% of the house).

Interesting observations:

With my barely adequate supply curve and indoor temps being maintained, the "buffer effect" in the big gravity mains has nearly disappeared. Measured supply temp is VERY close to the actual boiler temp.

When I began to raise space temp, the "buffer effect" immediatley reappeared. Despite a BIG call for heat, the boiler only increased input by about 6 mbh. The vast majority of the increased boiler energy was going directly to the radiation with almost nothing left to "recharge" the buffer. The ONLY way I could have increased response time would be to increase the heating curve. Had the curve been increased the "buffer" would have been rapidly replenished, but not all the way to the new, higher boiler temp. It would only have increased to the level before the experiment began because such level is adequate to maintain the desired settings.

----------------------------------

Perhaps this will help you understand your situation Jerry--that is if it's operating similarly to how I imagine.

Maybe this goes without saying, but when you've optimized a modulating/condensing boiler system for efficiency by nearly perfectly providing just the energy required to maintain your desired conditions you will have to increase supply temperature and accept [what appears to be significant] inefficiency in order to raise space temp with any reasonable measure of speed.

Will be glad to attempt to answer any questions. If you post some details about how your control system is hooked up (and how those valves are operated) I'll try to help you come up with a program strategy to preserve as much efficiency as possible while still giving you "headroom" when you want it.

Ron Schroeder

Hi Jerry,

Your system sounds similar to mine (15 zones in a 24' x28' house)except I don't have proportional valves on most of the zones. My reset control is on a PC. The baseline boiler temperature is based on the whole house load but I am using a PID computation based on which ever zone that has the greatest differance between it's set value and its actual temperature to raise the boiler temperature above the baseline curve. Unfortunatly, this requires a room temperature sensor in each zone and the setback is controlled and programed from the PC rather than from setback thermostats on each zone. The Tekmar TN4 system might be able to do this kind of control but I haven't looked to closely at it.

I am in Instrumentation at Brookhaven but have been in HVAC and energy since the '70s.

Ron

Constantin

Have you considered Simulink?

... your local technical college may have a copy for you to play with, it's a great way to figure out what the best coefficients are to mimize the time needed to achieve a certain state w/o overshooting.

My recollections of state-space are pretty fuzzy at this point, but I remember programming controllers for these kinds of issues back in college. While the LaPlace equations were very elegant, computers look for a different kind of input... but you knew that already, so never mind.

Practically speaking, I would consider programming the globe valves to operate like TRVs all the time. Then, if there is a significant ΔT in any zone, the boiler supply temp should shoot up proportionally to the additional temp needed.

Since all the zones act like TRVs, all non-affected zones screw down a notch or two as the supply temp goes up, while the affected zone opens 100%. Between the added flow and the higher supply temps, your zone should come up to speed pretty quickly.

Given the amount of instrumentation you're willing to throw at this problem, one interesting thing to note is the return temp coming out of each zone in addition to the temp of the zone. If you could establish the flow rates at the various "openings" of the valves you're using, you could use that info, open-loop style, to predict how much heat each zone is consuming.

What makes this approach perhaps more interesting is that you can get away from temperature and flow and get up to BTUs in vs. BTUs out. with the air-sensor being the arbiter whether you're making the grade or not.

jerry scharf_3

Slap my hands

if I ever built what you described, a complete lack of control authority.

Mike

Every zone has control authority. There is the highest heat loss zone running at 80% of control span, and everything else goes down from there. Many of them are running under 20% of the control span (not the same as the voltage range.)

This is a general disagreement I have with setting the reset curve at exactly makeup level. It's kind of the same thing as using a single burn to have a rocket change orbits, it's the most efficient but also has many bad side effects. Watch how NASA does it, a burn to speed up and a second burn to slow into the new orbit. They maintain control authority throughout the entire process.

So the outdoor reset is currently set up to heat the big loss room at the 80% setting. Everything else is set up with the proportional valave to run just over the make up rate. Since everything is radiant surface, the relative heating reaction is fairly close between zones.

What you describe is in your situation is exactly what I don't want, the system to take forever to change temperatures. In my book, the fact that the boiler didn't increase output and have TRVs adjust as necessary is a control failure.

BTW, this issue is one that I have heard several Europeans complain about. Without any prompting they told me how they hated the hours of waiting to get a room a bit warmer.

The goal of my project is to find the maximum efficiency available while producing occupant comfort. Sacrificing comfort for efficeincy will not fly in this situation. If I need to increase the supply and return temps for a while to reach comfort levels faster, that's ok.

jerry

Weezbo

i like the concept.

this is off on another tangent buh have you done any slabs with two layers of pex, one closer to the surface,and zoned them for more rapid responce time? the reason that i ask is because i find my 125 foot 3/8th slab with a thin over pour to ramp like you thought Satan just walked into the room .

Mike T., Swampeast MO

To change that situation all I have to do is increase the reset curve--then response is nearly as fast as forced air and the boiler significantly increases output by opening even a single TRV. Temperature maintenance isn't affected but efficiency seems to take a big hit.

Two ends of the same pole but it's hard to tip that pole automatically...

Still believe you'll have to increase the supply curve if you want to decrease response time for a call to raise heat.

Have you clocked the boiler when you increase the setting in a single zone? Does it increase fire significantly?

What I think you need to do is to detect significant differences between actual and desired temp in any zone--say 3° or more. Then boost the supply temp more greatly the further the difference between actual and desired. As long as you bring the supply temp back down as the difference decreases you shouldn't have overshoot. If you want to get really finicky then determine the boost rate based on the proportionate share of heat loss of each zone.

The Buderus mod-con is attempting to do this with the RC-10. But you notice it can only be done in one zone. Trying to do it with a large number of zones sounds like a mathematical nightmare.

Return temperature also offers a route to do this. Do you get a fast, noticeable drop in return temp if you increase the t-stat setting in a single zone?

Ron Schroeder

You could possably move the curve up by some fixed number of degrees, like 10 or so degrees, for a fixed amount of time whenever any zone comes out of set back.

Sometimes you can get 99% of the benifit from just a simple control change.

jerry scharf_3

Unfortunately??

You said unfortunately it requires a temp sensor in each room.???

Of course it does. But there is also all the other sensors you need. Humidity sensors in all wet areas, absolute pressure sensors to control make up air, boiler control sensors so you can manipulate the modulation rather than the set point and more.

The way I get around this is to use Maxim 18B20 1-wire temp sensors. They are under $3 each, have a -40 to +85C range thermister, laser trimmed compensation, 12 bit A/D, unique serial number and serial communications in a 3 pin TO92 plastic package. The computer interface runs from $40-150, and you can hang upwards of 1000' of signal cable and a whole bunch of sensors off a single bus. I have one of the more expensive ones with 4 signal busses off a single serial port. I use Krone Series 2 telecom blocks and cat5 wire to pull it all together and keep it neat. If you aren't using these at work, you should seriously look into it.

I bet you even have Platinum RTDs available to calibrate the units (Me, jelous?)

I'll attach some pictures to this tomorrow.

jerry

jerry scharf_3

that's a future phase

Ron,

There is a 1-wire interface for either 16x4 or 20x2 LCD characeter displays, with three buttons as well. These will allow someone to see the current temp, and modify the setpoint. They won't see anything like set it to 72, instead they will be able to ask it to be warmer or colder, and the system will look at various control responses to the input.

I have to get two more levels of control in place before I play with the controls.

The first is to get the offered heat to take into account the room load and the difference in temp and drive the zone to reach a compromise between best economy and fastest response.

The second is to model the passive solar gains and include this into the zone control logic.

Once those two are in place and tested, I should have a solid model of the system. Then I can then start taking external control inputs. I may or may not get this done before the meat of the heating season is behind me and I get to start thinking about the cooling side of life.

For me the cost of either doing my own conpensation logic, A/D and serial interface or buying them off the shelf is way beyond the $3 for a 1-wire sensor. Since I have a mod/con boiler in a low temp setup, there is nothing that exceeds the 85C limit of the 1-wire parts.

jerry

Mike T., Swampeast MO

THANK YOU JERRY!!!

This post was the final piece of a puzzle.

Information regarding zone thermal mass is already contained in a TRVd system and now I know how to use it!!!

EMT reset

Energy, M****, Time

Self-adjusting, self-correcting, understanding of the desires of the occupants while delivering the highest possible fuel efficiency. Yes, I'm 100% serious.