For those who have too much time on their hands - consider
a set-up like this (view with fixed-width font):
Thermostate Dim Reservoir
switch
+----+ +-----+ : :
| | | |-------------z :
Power supply ======| |====| | : z :
| | | |-------------z :
+----+ +-----+ +-------o o--------- T probe
|________________| .........
We have a heater (the vertically stacked Z's) in
a reservoir. The heater is controlled by an on/off
thermostate, with tight hysteresis settings. The
thermo sensor of the thermostate monitors the
temperature in the reservoir. The output of the
thermostate is dimmed by a dim switch, where the
setting is known but not the actual factor for
power reduction.
None of the parameters associated with the power
supply / thermostate / dim switch / heater are known
or logged.
The one log one has, is a thermo probe independent
of everything else, that samples and stores the
temperature in the reservoir at a rate of 4 samples
per minute.
Q: How does one get an estimate of power consumption
trends in the heater, from this set-up?
Rune
Puzzle: How to extract desired info from data?
Started by ●June 22, 2011
Reply by ●June 22, 20112011-06-22
On Jun 22, 4:28�am, Rune Allnor <all...@tele.ntnu.no> wrote:> For those who have too much time on their hands - consider > a set-up like this (view with fixed-width font): > > � � � � � � � � Thermostate � Dim � � � � � � Reservoir > � � � � � � � � � � � � � � �switch > � � � � � � � � � �+----+ � �+-----+ � � � � : � � � : > � � � � � � � � � �| � �| � �| � � |-------------z � : > Power supply ======| � �|====| � � | � � � � : � z � : > � � � � � � � � � �| � �| � �| � � |-------------z � : > � � � � � � � � � �+----+ � �+-----+ � +-------o � o--------- T probe > � � � � � � � � � � � |________________| � � ......... > > We have a heater (the vertically stacked Z's) in > a reservoir. The heater is controlled by an on/off > thermostate, with tight hysteresis settings. The > thermo sensor of the thermostate monitors the > temperature in the reservoir. The output of the > thermostate is dimmed by a dim switch, where the > setting is known but not the actual factor for > power reduction. > > None of the parameters associated with the power > supply / thermostate / dim switch / heater are known > or logged. > > The one log one has, is a thermo probe independent > of everything else, that samples and stores the > temperature in the reservoir at a rate of 4 samples > per minute. > > Q: How does one get an estimate of power consumption > � �trends in the heater, from this set-up? > > RuneDo you have information on the mass and thermal properties of what's in the reservoir? How about the thermal interface between the reservoir and the ambient environment? If you know the temperature of the ambient environment and can guess at how much thermal energy is being leeched out over time, you should be able to estimate how much energy was put into the system by the heater over time. Jason
Reply by ●June 22, 20112011-06-22
On Jun 22, 4:28�am, Rune Allnor <all...@tele.ntnu.no> wrote:> For those who have too much time on their hands - consider > a set-up like this (view with fixed-width font): > > � � � � � � � � Thermostate � Dim � � � � � � Reservoir > � � � � � � � � � � � � � � �switch > � � � � � � � � � �+----+ � �+-----+ � � � � : � � � : > � � � � � � � � � �| � �| � �| � � |-------------z � : > Power supply ======| � �|====| � � | � � � � : � z � : > � � � � � � � � � �| � �| � �| � � |-------------z � : > � � � � � � � � � �+----+ � �+-----+ � +-------o � o--------- T probe > � � � � � � � � � � � |________________| � � ......... > > We have a heater (the vertically stacked Z's) in > a reservoir. The heater is controlled by an on/off > thermostate, with tight hysteresis settings. The > thermo sensor of the thermostate monitors the > temperature in the reservoir. The output of the > thermostate is dimmed by a dim switch, where the > setting is known but not the actual factor for > power reduction. > > None of the parameters associated with the power > supply / thermostate / dim switch / heater are known > or logged. > > The one log one has, is a thermo probe independent > of everything else, that samples and stores the > temperature in the reservoir at a rate of 4 samples > per minute. > > Q: How does one get an estimate of power consumption > � �trends in the heater, from this set-up?Do you at least know the thermostat's hysteresis band? Is the sample rate high enough to capture the relevant information? I don't have much patience for this type of puzzle in the abstract. (There are times when it is a practical necessity.) If I wanted to know the energy consumption, I would install a watt-hour meter. Jerry -- Engineering is the art of making what you want from things you can get.
Reply by ●June 22, 20112011-06-22
On Wed, 22 Jun 2011 01:28:58 -0700, Rune Allnor wrote:> For those who have too much time on their hands - consider a set-up like > this (view with fixed-width font): > > > Thermostate Dim Reservoir > switch > +----+ +-----+ : : > | | | |-------------z : > Power supply ======| |====| | : z : > | | | |-------------z : > +----+ +-----+ +-------o o--------- T probe > |________________| ......... > > > We have a heater (the vertically stacked Z's) in a reservoir. The heater > is controlled by an on/off thermostate, with tight hysteresis settings. > The thermo sensor of the thermostate monitors the temperature in the > reservoir. The output of the thermostate is dimmed by a dim switch, > where the setting is known but not the actual factor for power > reduction. > > None of the parameters associated with the power supply / thermostate / > dim switch / heater are known or logged. > > The one log one has, is a thermo probe independent of everything else, > that samples and stores the temperature in the reservoir at a rate of 4 > samples per minute. > > Q: How does one get an estimate of power consumption > trends in the heater, from this set-up? > > RuneIf your thermostat cycle times (both on and off) are long compared to your sample rate and dim switch output is constant, then rising temp ~= heater on (with some unknown but unimportant delay) and falling temp ~= heater off; power consumption roughly proportional to percentage on time. If dim switch output can vary then possibly measure off time duration only, power roughly proportional to inverse of off time, or factor in rate of temperature rise during heater on time, probably adding average cooling rate during off time, which presumably continues while heater is on, reducing the apparent heat input bellow actual heat input during on time. If thermostat setting and/or dead band is adjustable you should be able to detect that also with your single sensor and factor them in also.
Reply by ●June 23, 20112011-06-23
On Jun 22, 2:16�pm, Jason <cincy...@gmail.com> wrote:> On Jun 22, 4:28�am, Rune Allnor <all...@tele.ntnu.no> wrote: > > > > > > > For those who have too much time on their hands - consider > > a set-up like this (view with fixed-width font): > > > � � � � � � � � Thermostate � Dim � � � � � � Reservoir > > � � � � � � � � � � � � � � �switch > > � � � � � � � � � �+----+ � �+-----+ � � � � : � � � : > > � � � � � � � � � �| � �| � �| � � |-------------z � : > > Power supply ======| � �|====| � � | � � � � : � z � : > > � � � � � � � � � �| � �| � �| � � |-------------z � : > > � � � � � � � � � �+----+ � �+-----+ � +-------o � o--------- T probe > > � � � � � � � � � � � |________________| � � ......... > > > We have a heater (the vertically stacked Z's) in > > a reservoir. The heater is controlled by an on/off > > thermostate, with tight hysteresis settings. The > > thermo sensor of the thermostate monitors the > > temperature in the reservoir. The output of the > > thermostate is dimmed by a dim switch, where the > > setting is known but not the actual factor for > > power reduction. > > > None of the parameters associated with the power > > supply / thermostate / dim switch / heater are known > > or logged. > > > The one log one has, is a thermo probe independent > > of everything else, that samples and stores the > > temperature in the reservoir at a rate of 4 samples > > per minute. > > > Q: How does one get an estimate of power consumption > > � �trends in the heater, from this set-up? > > > Rune > > Do you have information on the mass and thermal properties of what's > in the reservoir? How about the thermal interface between the > reservoir and the ambient environment? If you know the temperature of > the ambient environment and can guess at how much thermal energy is > being leeched out over time, you should be able to estimate how much > energy was put into the system by the heater over time.Nothing that isn't stated is known. Rune
Reply by ●June 23, 20112011-06-23
On Jun 22, 3:00�pm, Jerry Avins <j...@ieee.org> wrote:> On Jun 22, 4:28�am, Rune Allnor <all...@tele.ntnu.no> wrote: > > > > > > > For those who have too much time on their hands - consider > > a set-up like this (view with fixed-width font): > > > � � � � � � � � Thermostate � Dim � � � � � � Reservoir > > � � � � � � � � � � � � � � �switch > > � � � � � � � � � �+----+ � �+-----+ � � � � : � � � : > > � � � � � � � � � �| � �| � �| � � |-------------z � : > > Power supply ======| � �|====| � � | � � � � : � z � : > > � � � � � � � � � �| � �| � �| � � |-------------z � : > > � � � � � � � � � �+----+ � �+-----+ � +-------o � o--------- T probe > > � � � � � � � � � � � |________________| � � ......... > > > We have a heater (the vertically stacked Z's) in > > a reservoir. The heater is controlled by an on/off > > thermostate, with tight hysteresis settings. The > > thermo sensor of the thermostate monitors the > > temperature in the reservoir. The output of the > > thermostate is dimmed by a dim switch, where the > > setting is known but not the actual factor for > > power reduction. > > > None of the parameters associated with the power > > supply / thermostate / dim switch / heater are known > > or logged. > > > The one log one has, is a thermo probe independent > > of everything else, that samples and stores the > > temperature in the reservoir at a rate of 4 samples > > per minute. > > > Q: How does one get an estimate of power consumption > > � �trends in the heater, from this set-up? > > Do you at least know the thermostat's hysteresis band?Nothing except what was stated. The hysteresis switching times are visible in the ligged T-data.> Is the sample > rate high enough to capture the relevant information?The sample rate adequate.> I don't have > much patience for this type of puzzle in the abstract. (There are > times when it is a practical necessity.)I solved this last week.> If I wanted to know the > energy consumption, I would install a watt-hour meter.Can't be done with historical data. Rune
Reply by ●June 23, 20112011-06-23
On Jun 22, 4:38�pm, Glen Walpert <g...@null.void> wrote:> On Wed, 22 Jun 2011 01:28:58 -0700, Rune Allnor wrote: > > For those who have too much time on their hands - consider a set-up like > > this (view with fixed-width font): > > > � � � � � � � � Thermostate � Dim � � � � � � Reservoir > > � � � � � � � � � � � � � � �switch > > � � � � � � � � � �+----+ � �+-----+ � � � � : � � � : > > � � � � � � � � � �| � �| � �| � � |-------------z � : > > Power supply ======| � �|====| � � | � � � � : � z � : > > � � � � � � � � � �| � �| � �| � � |-------------z � : > > � � � � � � � � � �+----+ � �+-----+ � +-------o � o--------- T probe > > � � � � � � � � � � � |________________| � � ......... > > > We have a heater (the vertically stacked Z's) in a reservoir. The heater > > is controlled by an on/off thermostate, with tight hysteresis settings. > > The thermo sensor of the thermostate monitors the temperature in the > > reservoir. The output of the thermostate is dimmed by a dim switch, > > where the setting is known but not the actual factor for power > > reduction. > > > None of the parameters associated with the power supply / thermostate / > > dim switch / heater are known or logged. > > > The one log one has, is a thermo probe independent of everything else, > > that samples and stores the temperature in the reservoir at a rate of 4 > > samples per minute. > > > Q: How does one get an estimate of power consumption > > � �trends in the heater, from this set-up? > > > Rune > > If your thermostat cycle times (both on and off) are long compared to > your sample rate and dim switch output is constant, then rising temp ~= > heater on (with some unknown but unimportant delay) and falling temp ~= > heater off; power consumption roughly proportional to percentage on > time. �If dim switch output can vary then possibly measure off time > duration only, power roughly proportional to inverse of off time, or > factor in rate of temperature rise during heater on time, probably adding > average cooling rate during off time, which presumably continues while > heater is on, reducing the apparent heat input bellow actual heat input > during on time. �If thermostat setting and/or dead band is adjustable you > should be able to detect that also with your single sensor and factor > them in also.More or less correct. What I did was: - Detect thermostat cycles from T readings - Count the number of rise points and fall points within each cycle, to get thermostat on percentage - Count the number of measurements per thermostate cycle - Multiply the two to get average energy contribution / cycle This works well enough when dim switch and thermostate settings are left alone for the duration of the measurement. Rune
Reply by ●June 23, 20112011-06-23
Rune Allnor <allnor@tele.ntnu.no> wrote:> On Jun 22, 3:00�pm, Jerry Avins <j...@ieee.org> wrote: >> On Jun 22, 4:28�am, Rune Allnor <all...@tele.ntnu.no> wrote:(snip)>> > We have a heater (the vertically stacked Z's) in >> > a reservoir. The heater is controlled by an on/off >> > thermostate, with tight hysteresis settings. The >> > thermo sensor of the thermostate monitors the >> > temperature in the reservoir. The output of the >> > thermostate is dimmed by a dim switch, where the >> > setting is known but not the actual factor for >> > power reduction.(snip)>> Do you at least know the thermostat's hysteresis band?> Nothing except what was stated. The hysteresis switching > times are visible in the ligged T-data.The times are visible, but how much is the temperature difference? You can use that to back calibrate the historical data. (snip)>> If I wanted to know the >> energy consumption, I would install a watt-hour meter.> Can't be done with historical data.Again, it can be used to back calibrate. Well, a wattmeter is probably better, but watt-hour works, too. If the thermal mass is constant, you can compute it from current measurements of power and hysteresis, and on/off times, then use that to figure out the historical data. -- glen
Reply by ●June 23, 20112011-06-23
On Jun 23, 9:14�am, glen herrmannsfeldt <g...@ugcs.caltech.edu> wrote:> Rune Allnor <all...@tele.ntnu.no> wrote: > > On Jun 22, 3:00�pm, Jerry Avins <j...@ieee.org> wrote: > >> On Jun 22, 4:28�am, Rune Allnor <all...@tele.ntnu.no> wrote: > > (snip) > > >> > We have a heater (the vertically stacked Z's) in > >> > a reservoir. The heater is controlled by an on/off > >> > thermostate, with tight hysteresis settings. The > >> > thermo sensor of the thermostate monitors the > >> > temperature in the reservoir. The output of the > >> > thermostate is dimmed by a dim switch, where the > >> > setting is known but not the actual factor for > >> > power reduction. > > (snip) > > >> Do you at least know the thermostat's hysteresis band? > > Nothing except what was stated. The hysteresis switching > > times are visible in the ligged T-data. > > The times are visible, but how much is the temperature difference?Small. 0.5 C or less.> You can use that to back calibrate the historical data. > > (snip) > > >> If I wanted to know the > >> energy consumption, I would install a watt-hour meter. > > Can't be done with historical data. > > Again, it can be used to back calibrate. � Well, a wattmeter is > probably better, but watt-hour works, too. � > > If the thermal mass is constant, you can compute it from > current measurements of power and hysteresis, and on/off times, > then use that to figure out the historical data.That might have worked if one logged the T sensor that fed the thermostate. We don't. We log an auxillary probe, located at a different position in the reservoir, meaning it is exposed to different ambient temperatures and heat flows. The only properties of those readings that map 1:1 to the thermostate, are the switch times. Rune
Reply by ●June 23, 20112011-06-23
Hi Rune,>On Jun 22, 4:38=A0pm, Glen Walpert <g...@null.void> wrote: >> On Wed, 22 Jun 2011 01:28:58 -0700, Rune Allnor wrote: >> > For those who have too much time on their hands - consider a set-uplik=>e >> > this (view with fixed-width font): >> >> > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 Thermostate =A0 Dim =A0 =A0 =A0 =A0=A0=> =A0 Reservoir >> > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0switch >> > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0+----+ =A0 =A0+-----+ =A0 =A0=>=A0 =A0 : =A0 =A0 =A0 : >> > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0| =A0 =A0| =A0 =A0| =A0 =A0|---=>----------z =A0 : >> > Power supply =3D=3D=3D=3D=3D=3D| =A0 =A0|=3D=3D=3D=3D| =A0 =A0 | =A0=>=A0 =A0 =A0 : =A0 z =A0 : >> > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0| =A0 =A0| =A0 =A0| =A0 =A0|---=>----------z =A0 : >> > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0+----+ =A0 =A0+-----+ =A0+-----=>--o =A0 o--------- T probe >> > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 |________________| =A0 =A0=>......... >> >> > We have a heater (the vertically stacked Z's) in a reservoir. Theheate=>r >> > is controlled by an on/off thermostate, with tight hysteresissettings.>> > The thermo sensor of the thermostate monitors the temperature in the >> > reservoir. The output of the thermostate is dimmed by a dim switch, >> > where the setting is known but not the actual factor for power >> > reduction. >> >> > None of the parameters associated with the power supply / thermostate/>> > dim switch / heater are known or logged. >> >> > The one log one has, is a thermo probe independent of everythingelse,>> > that samples and stores the temperature in the reservoir at a rate of4>> > samples per minute. >> >> > Q: How does one get an estimate of power consumption >> > =A0 =A0trends in the heater, from this set-up? >> >> > Rune >> >> If your thermostat cycle times (both on and off) are long compared to >> your sample rate and dim switch output is constant, then rising temp~=3D>> heater on (with some unknown but unimportant delay) and falling temp~=3D>> heater off; power consumption roughly proportional to percentage on >> time. =A0If dim switch output can vary then possibly measure off time >> duration only, power roughly proportional to inverse of off time, or >> factor in rate of temperature rise during heater on time, probablyadding>> average cooling rate during off time, which presumably continues while >> heater is on, reducing the apparent heat input bellow actual heat input >> during on time. =A0If thermostat setting and/or dead band is adjustabley=>ou >> should be able to detect that also with your single sensor and factor >> them in also. > >More or less correct. > >What I did was: > >- Detect thermostat cycles from T readings >- Count the number of rise points and fall points within > each cycle, to get thermostat on percentage >- Count the number of measurements per thermostate cycle >- Multiply the two to get average energy contribution / cycle > >This works well enough when dim switch and thermostate >settings are left alone for the duration of the measurement. > >RuneYou have been purposefully misleading. You originally said the hysteresis band was tight, and no further information was known. Your solution is specifically based on knowing that the hysteresis band is fairly wide, such that a slow 4 samples per minute log can track the thermostat cutting in and out, and get a rough estimation of its on off ratio. Steve
