Heating system hysteresis and efficiency

[paulca]

I'm upgrading my custom control heating system.  Better put, I have "been" upgrading my custom heating control system and the core architecture needs redone.

While I'm at it, I want to address a problem.  A problem I thought I'd solved, but it resurfaces.

Cycling. 

So my "make it as simple as you can" approach resulted in a design which simply used a minimum expiry time on a heating demand of 5 minutes.  If anything turned the heating on, it would remain on for at least 5 minutes.  This works.  It results in a slightly over temp in zones by 0.5*C or more in some rooms, but it prevents on/off/on/off/on/off cycling every 2 minutes fighting over 0.1*C.

But it still results in 30-40 cycles per day this winter.  I feel I need to find other hysteresis methods.  At the same time I'm trying to get an answer from the boiler manufacturer as to reasonable cycle frequency over the boilers lifespan.

Problems and solutions:
1. Multiple zones request heating at different intervals.  These are processed individually.  So it's open to inefficient demand from individual zones in sequence when the demand could be handled more intelligently so multiple zones can be heated together instead of in turn.  Requires predicting if zones are close to triggering demand while the heating it already on for other zones.  To put it an other way, the living room might demand heating, its radiator comes on, the boiler comes on.  10 minutes later it stops demanding heating and it's radiator and boiler shut off.... but 2 minutes later the bedroom demands heating, as it's radiator was off it didn't get any.  This is a waste.

2. On at 16.9*C and off at 17.1*C margin cycling.  This is caused by the lack of temperature hysteresis control.  Effectively a zero dead-zone hysteresis.  If it's over temp, demand is removed, if it's under demand is applied.  Adding temperature hysteresis would provide a negative trigger (target -0.1*C) and a positive (continue heating until +0.5*C).  So a positive + negative hysteresis.   This is also required for the above to determine if a zone, while not triggering, could use heat as it's close to triggering.  Any zone under it's positive hysteresis limit could be heated as soon as any zone triggers it's lower negative hysteresis limit.

3. Lowering boiler output (flow temp) lowers cycling, but slows warming unused zones up for use.  A heating saving feature is to not heat (as much) zones not in use.  This requires I can heat them up quickly for use.  The solution to this, considering that lower flow temp will lead to less cycling, is dynamic flow temp control.  Just as the boiler itself will modulate back to maintain a flow temp, the flow temp can be modulated back as target temps are reached.  When a zone heated to 15*C comes into service, it can be rapidly heated with the boiler maxed at 85*C but the flow temp back well off as it approaches target of 19*C and then, maybe balance out with the boiler able to supply enough heat to stop the temperatures rising or falling. Another solution to this is predictive heating.  I already have an implementation of fore running schedules for a time in the future, deciding on target temperatures based on gradients.  So if I know in advance a zone will jump up in temp, I can pre-ramp it in advance and not require high flow temps.

Other things I could consider are load sensed, weather aware predictive heating.  Boiler manufacturers seem to do this and aim for long run times.  The challenging thing is how to implement it for multiple zones with varying load patterns.  I don't think my routine is stable enough to predict.

Sorry for rambling.  Having lack of coffee pot conversations in work isolation issues with lock down.

[nctnico]

The solution is simple: make sure the system has a loop between hot and cold (return) at the fartest point from the boiler and let the boiler shutdown when the return water is hot. This means no (or very little) water flows into the radiators (which must have thermostatic control valves). Once the boiler sees the return water is getting cold again, it can switch on again. Since the heat will end up in your house, no heat is lost.

[nali]

But it still results in 30-40 cycles per day this winter.  I feel I need to find other hysteresis methods.  At the same time I'm trying to get an answer from the boiler manufacturer as to reasonable cycle frequency over the boilers lifespan.

Take a look at your boiler's manual and see if it supports chrono-proportional thermostats. These basically PWM the boiler every few minutes as part of a PID control so it would easily handle 30-40 per day if compatible.

But that does seem a little high. I zoned my heating with motorised valves and occasionally suffered from hunting on/off. I cured that by throttling back the radiators, they were too high meaning a surge of heated air was shutting off the 'stat for a few minutes causing a cycle until the room heated up. Looking up how to balance radiators the starting point seemed to be to close off the lockshield valves and crack them open 1/2 turn which seems to work OK.

[NiHaoMike]

1. Multiple zones request heating at different intervals.  These are processed individually.  So it's open to inefficient demand from individual zones in sequence when the demand could be handled more intelligently so multiple zones can be heated together instead of in turn.  Requires predicting if zones are close to triggering demand while the heating it already on for other zones.  To put it an other way, the living room might demand heating, its radiator comes on, the boiler comes on.  10 minutes later it stops demanding heating and it's radiator and boiler shut off.... but 2 minutes later the bedroom demands heating, as it's radiator was off it didn't get any.  This is a waste.

2. On at 16.9*C and off at 17.1*C margin cycling.  This is caused by the lack of temperature hysteresis control.  Effectively a zero dead-zone hysteresis.  If it's over temp, demand is removed, if it's under demand is applied.  Adding temperature hysteresis would provide a negative trigger (target -0.1*C) and a positive (continue heating until +0.5*C).  So a positive + negative hysteresis.   This is also required for the above to determine if a zone, while not triggering, could use heat as it's close to triggering.  Any zone under it's positive hysteresis limit could be heated as soon as any zone triggers it's lower negative hysteresis limit.

Both of those can be solved by making the hysteresis system wide. That is, all the thermostats raise their trigger points by the hysteresis value you set once the boiler turns on and lowers it when the boiler turns off.

[james_s]

Most of our systems here are forced air, so the details are a bit different. Ideally the system should be sized so that it runs as close to continuously as possible on the coldest days and minimizes short cycling, although in practice that is difficult. Historically our heating systems tend to be grossly oversized which is nice in that you can warm up a cold house very quickly but they end up cycling frequently enough to really cut into the efficiency. Some of the fancier gas furnaces have multi-stage gas valves that can help with this.

[Siwastaja]

Boiler burning what? Electricity, oil, gas?

How it's normally done here is, assuming oil or gas burner, the boiler does some 80 degC water with a relatively large hysteresis (like 10 degC) and a few hundreds of liters of buffer; for example my ~20kW oil burner cycle tends to be 10 minutes on, an hour off, boiler temperature oscillating between +75 and +85 degC. It has to be this way, running the boiler too cool causes excess condensation in the (corrosive) exhaust gas.

Then this nearly constant temperature boiler water is shunted into the radiator loop using a 4-way valve, i.e., the radiator loop runs with a circulating pump, and adjustable amount is mixed into the boiler, and back into the radiator loop.

The primary way of control is not feedback at all, it's feedforward! After all, the required heating power is proportional to the heat loss of the house, which is proportional to the outside temperature. So the shunt controller senses the outdoor temperature. For example in my system, when it's 0 degC outside, the shunt controller adjusts the 4-way valve to make the radiator loop hit 40degC. When it's -25 degC outside, the radiator loop would be, say, at 60 degC. The curve can be adjusted, and the only sensor is the outdoor temperature sensor.

Such control is widely used here and it's surprisingly good with absolutely no feedback at all, it's only feedforward. Certain rooms overheating is prevented by radiator thermostat valves, but these are not precision devices, you need to realize they are crappy P controllers. So it's important that the circulating water is not way too hot (i.e., set the feedforward curve right), you want the radiator thermostatic valves to be almost fully open in "normal" state, just on the edge of their controlled range. This way, if you hold a party in one room and 10 people are heating the room up, or you have a fireplace in a room, or sun is shining inside from large windows, then the thermostatic valves react and reduce the flow through that particular radiator.

Such system where thermostatic valves are mostly fully or nearly open, and primary control is through the water temperature, the need for the loop nctnico mentioned is not critically necessary. I don't have such a loop, for example. There never is a situation of thermostat valves going all shut.

Obviously there is always an option to add feedback to such shunt control, so measure the room temperature and automatically fine tune the curve; my controller from early 2000's has an optional input for such sensor. But the bulk of the control should be based on feedforward, this way the room temperature will be stable against quick variations in outside temperature.

For reference / general interest, this is the manual of the 4-way valve controller I have installed: https://ouman.fi/wp-content/uploads/2018/08/EH-80__manual__en.pdf

[Siwastaja]

The takeaway in my previous post is, having excess power makes tuning control loops more difficult than it needs to be, this is seen in any control system, not just heating

For example, if you have an oven with heating elements that are powerful enough to get it to 500 degC, it's likely quite unstable when set at 50 degC.

Similarly, if your circulating water is hot enough for the coldest winter day, the thermostatic valves and their controller will have hard time achieving stable temperature when they run at say average 10% power.

The solution is always feedforward. Even if you end up designing the best PID thermostat controller ever, adding FF from the "expected" output power (for example, based on the difference between outdoor temperature and indoor setpoint) makes the PID only finetune, the duty cycle or power setting will be higher, and risk of oscillation much smaller.

[paulca]

Interesting ideas.

Clarification, it's a condensing gas combi boiler.  It's oversized by a factor of 2, as is common for these things, the 30kW output is nearly double what the radiators need (8x2kW more or less), but that high output makes for lovely hot mains water pressure showers!  (Which I have set to 42C)

The loop return temp must be below 54*C for the condensing heat salvager to work.  I have all my lock shields tuned to give a 10*C drop between outflow and inflow, although for legacy reasons it's measured in the upstairs/downstairs link pipes (they are easy accessed near a plug socket!).  The radiators where painstakingly tuned using a thermal camera.  The slight annoyance.  The boiler pump speed is so high the lock shields are closed so tight you can hear the water shooting through the valves when the heating is on.  To stop the noise the values need openned much wider and then there is little to no drop across the rad.  I think it's cheap valves they installed TBH, but I can live with the slight "heating" noise.

Anyway

I started with the simplistic approach and lowered the loop flow temp from 75C to 50*C.   The idea being, accept that rooms will warm up slower but the heating will run and modulate itself longer with less cycling.

The result was it couldn't heat the house and the heating was just ON.  The living room target of 20 could not be reached it stalled at 18.  The bedroom barely reached it vacant target of 15*C!  Now I realise leaving it on for days and days might warm the room core up and it might do better.  However this is not the type of heating system I'm after.  This type would be useful in a building which is trying to maintain a constant temp, in all zones, 24/7 like a large public building.

Also the boiler is too oversized to sustain 50*C and it goes into periods of rapid ON, nope!, OFF.... ON, nope!, OFF.  I believe it's "load sensing".  But it turns on the blower fires the boiler up and immediately shuts it off 10 seconds later.  Then periodically it will actually fire up and run for 5 minutes and go back to short cycling itself.  Doesn't sound efficient.  Clearly at 50*C my radiators cannot transfer even it's minimum output to the air.

So I put it back up to 60C for a day.  This morning, based on 24 hour running at 60*C, I used 11 hours of run time and still had 30 demand cycles.  It's colder this morning, 3*C and it's struggling.   I'm putting it back up to 75C to confirm this has the effect I expect.

I wish I had a gas flow meter to really see how much I'm saving or wasting, as I tune things.  Maybe should consider asking how much having one installed is.  Although I'm not sure I trust the "Smart" offerings from china as a gas flow valve, so a better solution would be a proper rated flow meter and make it smart myself somehow.

I'd like to have dynamic flow/loop temp control, the boiler does support it, but its a bit of a "not quite standard" EMS protocol.  An adapter product exists, but it's very DIY and carries warnings about breaking your boiler.

Until I finally go that way, I'm going to make my zone targets slightly more complex to add temperature hysteresis and also implement heating as many zones as needed while the heating has been triggered.

[Siwastaja]

To me, it really sounds like you need a buffer tank which has enough water to run long enough burner cycles, say bare minimum 200-300 liters and 5-10 degC hysteresis, then the burner set to provide the buffer tank with temperature which is the optimum for the burner (regardless of the radiator temperature), which for the condensing burner is less than for classical burner where exhaust condensation is undesired.

Then just shunt (mix) water for the radiator loop, and adjust the mixing based on outside temperature so that the room temperature control sits always somewhere around, say, 70-80%? This way, the room temperature feedback only finetunes the radiator valves. You don't even need an automated controlled, a hand adjustment gives acceptable starting point.

If you do this, you'll completely decouple the burner from the feedback loop, and it can live its own life with long enough operating periods, and the duty cycle adjusts automatically long-term.

For example, I measured my burner running 9 min 15 secs then idling 1 hour 4 min 38 secs then running 9 min 4 secs and so on, which makes duty cycle of 12.5% and average output power of 2010W, when outdoor temperature was -4.5 degC, indoor +21 degC.

[nctnico]

The takeaway in my previous post is, having excess power makes tuning control loops more difficult than it needs to be, this is seen in any control system, not just heating

For example, if you have an oven with heating elements that are powerful enough to get it to 500 degC, it's likely quite unstable when set at 50 degC.

Similarly, if your circulating water is hot enough for the coldest winter day, the thermostatic valves and their controller will have hard time achieving stable temperature when they run at say average 10% power.

The solution is always feedforward. Even if you end up designing the best PID thermostat controller ever, adding FF from the "expected" output power (for example, based on the difference between outdoor temperature and indoor setpoint) makes the PID only finetune, the duty cycle or power setting will be higher, and risk of oscillation much smaller.

That is where an outside temperature sensor comes in which sets the outgoing water temperature of the boiler. In the end the temperature outside determines how much energy is needed.

@Paulca: you are going the wrong way entirely; you are massively overcomplicating things. Just implement the loop system I described earlier (where the boiler shuts off when the return water gets too warm) and use an outside temperature sensor to set the boiler's outgoing water temperature. That is how these kind of heater systems are installed & operate.

[paulca]

That is where an outside temperature sensor comes in which sets the outgoing water temperature of the boiler. In the end the temperature outside determines how much energy is needed.

@Paulca: you are going the wrong way entirely; you are massively overcomplicating things. Just implement the loop system I described earlier (where the boiler shuts off when the return water gets too warm) and use an outside temperature sensor to set the boiler's outgoing water temperature. That is how these kind of heater systems are installed & operate.

Well, I already have the auto shut off when the return water gets too warm.  The boiler does that and more automatically.  If I ask it to sustain a given loop temp, it will, intelligently, modulating and load sensing.  In fact if I purchase and install their (Worchester Bosch) smart "comfort" controls, it will handle the set point temp and monitor the outdoor temp.

The problem with this, is it's expensive and exclusive to their boiler, has a cloud service etc. etc.  Also it doesn't handle multiple zones with individual set points.

I appreciate that one way to run these condensing boilers is to put them on and leave them on at low temp and just let the house build up heat over time.  Lower the loop temp in summer, raise it in winter, or use the outdoor temp sensor for load prediction instead.  That's fine.  But it's not what I currently need or want.  It would cost me much more money to do this.  I live alone in a 3 bedroom house.  It pains me to watch the bedroom get heated all day when nobody is using it.  So I close rooms off and drop their temps on schedules. It also responds to certain events which denote the room is now in use which will expect that room to warm up until I leave. 

The heating equation is heat in versus heat out which is very dependant on the temperature delta.  It costs a lot less to heat a room to 14*C with a 10*C delta to outdoor, than to heat the same room to 20*C with a 16*C delta.  So the approach of keeping the house warm constantly with mild heat, versus heating it up from cold all the time with high heat really depends on how often you use each zone.  I'm only really in one zone at a time.  So it is more efficient for me to lower the deltas.  In fact I may automate that, as I've found out, my bedroom radiator breaks even with heat lose around 14*C delta.  In that the radiator is on constantly and only producing the amount of heat I'm losing due to poor insulation.

But we aren't far off agreement, it just is a little more complex to manage.  Yes, I have an outdoor sensor and as mentioned above, I can calculate the realistically efficient target temps for "vacant" or even sacrifice a degree off the comfort temps if it's particularly cold.

Loop temp control is something I want to explore, which I would have driven off the current->target delta and outdoor->target deltas.  It's risking a 3rd party interface to the boilers EMS bus that's daunting.

[Zucca]

Read about passive homes and how the human body feels cold or hot.
Basically air temperature is only an indirect parameter for your body temperature perception.

Buy a thermal camera (even a cheap one) and check your home for thermal bridges.
Then check humidity...

[paulca]

Here is an imaging showing the last 6 hours.

In my view it shows the system working as intended.  Unused rooms are left to be cool, used rooms are kept warm.  The office is easy to heat, as I'm usually in it and the PC provides a constant 100-300W of heat in addition to 100W of body heat.  It's also the best insulated from the outside.

Active overrides are in place trying to lower the demand coming from idle rooms even more.

[james_s]

I wish I had a gas flow meter to really see how much I'm saving or wasting, as I tune things.  Maybe should consider asking how much having one installed is.  Although I'm not sure I trust the "Smart" offerings from china as a gas flow valve, so a better solution would be a proper rated flow meter and make it smart myself somehow.

I assume you must have a gas flow meter somewhere, otherwise the utility would not be able to bill you accurately. Is that not located somewhere you can access? It's pretty easy to turn off any other gas appliances and then read the consumption over a specific interval on the meter. My gas meter is electronically monitored so I can log into my utility site and see a graph of daily consumption in almost real time.

[paulca]

The issue is the frequency of the demands.

or... as you might notice I have insulation issues with the bedroom in particular.  This will be addressed in the Spring with the demand for insulation and heating engineers falls.

EDIT: To avoid confusion the timescale on this graph is 24 hours, the last one is only 6 hours.

[paulca]

I assume you must have a gas flow meter somewhere, otherwise the utility would not be able to bill you accurately. Is that not located somewhere you can access? It's pretty easy to turn off any other gas appliances and then read the consumption over a specific interval on the meter. My gas meter is electronically monitored so I can log into my utility site and see a graph of daily consumption in almost real time.

Of course.  Yes.  I meant an automated, data logging method.

I can, as you rightly point out, just read the meter myself.  Although it's mechanical.

However any realistic period I measure it over will ultimately be skewed by current weather making it hard to make a determination....  a constantly logged data set would make that easier.

[james_s]

Of course.  Yes.  I meant an automated, data logging method.

I can, as you rightly point out, just read the meter myself.  Although it's mechanical.

However any realistic period I measure it over will ultimately be skewed by current weather making it hard to make a determination....  a constantly logged data set would make that easier.

Does it have any sort of mechanism that you could read electronically from the outside, say a reflective opto sensor "watching" a dial pointer go by? A lot of ours have some type of mechanical indicator that rotates so that it's easy to see if gas is flowing, even the old ones that are purely mechanical and manually read. I have no idea if gas meters worldwide are anything similar.

[paulca]

I haven't looked.  TBH.  It might have a standard pulse LED which I think is standard and there is software available to monitor those.  If not, there are projects to OCR Optical Character Recognition the digits off the meter. 

it just comes down to risk versus reward.

Looking closely at the data over the winter, I have serious insulation issues with the bedroom and it's not a surprise.   I insulated the main roof above last year.  This has made the roof space a ventilated cold space.  The trouble is that same cold space borders directly onto the bedroom wall and the bedroom floor/livingroom ceiling.  Causing the gap between the two rooms to be a cold space with no insulation.  I intend to address this.

Asides, it's an interesting test of the system under load.  The priority is to fix the insulation.  Monitoring the gas meter is second fiddle.

[penfold]

What you pose is a very interesting control problem. Based on the amount of swearing I heard when sharing a PhD office with someone tacking this very problem, it's very 'interesting' indeed.

The big problem when tacking this kind of optimisation problem in a domestic setting is trying to work out exactly what 'efficient' means. In a large office it's quite simple, there's one room, one temperature and the target is minimum cost for most consistent temperature. But for a house with more rooms than people and a tendency for said people to want to move from one to the other and expect it to be at the same temperature its a very different problem, there's always a relatively large thermal mass in the fabric of the building compared to the amount of air so its very easy for rooms to 'feel' much colder than they are and there's a lot of thermal mass to heat up before the 'comfort' rating builds up.. then its very inefficient to keep one room warm whilst there's nobody in it and that becomes something you need to factor into your 'efficiency'.

If your gas meter is an electronic one, you'll probably find it pings a packet of data from an IR LED every second or so, the electric meter I have does that and it contains the current reading and other status info... could be useful.

[nctnico]

What you pose is a very interesting control problem. Based on the amount of swearing I heard when sharing a PhD office with someone tacking this very problem, it's very 'interesting' indeed.

The big problem when tacking this kind of optimisation problem in a domestic setting is trying to work out exactly what 'efficient' means. In a large office it's quite simple, there's one room, one temperature and the target is minimum cost for most consistent temperature. But for a house with more rooms than people and a tendency for said people to want to move from one to the other and expect it to be at the same temperature its a very different problem, there's always a relatively large thermal mass in the fabric of the building compared to the amount of air so its very easy for rooms to 'feel' much colder than they are and there's a lot of thermal mass to heat up before the 'comfort' rating builds up.

Not just that. Keeping one room cold means the room at the other side of the wall (or on the next floor) will also feel colder. In the end you don't want large temperature differences in a relatively small building like a home. You can't really treat rooms like single islands because they are physically connected and heat will transfer from one to the other.

Floor heating is something to consider though (especially when mounted directly under the tiles and not covered by several cm of cement) because it brings the heat source closer to the persons in the room and the heat is distributed more evenly. A commonly found claim is that floor heating saves about 15%.

[Siwastaja]

Small temperature differences between rooms may be an OK micro-optimization, but keeping some rooms cold requires adding thermal insulation between the rooms to actually save anything.

Instead, use the same money and effort to put that insulation on the outer envelope of the house -> now you can just keep all the rooms warm with basically the same cost and stop thinking about it.

Really, insulating pays off pretty quickly. You may want to get a thermal camera to look at the discontinuities, to find where the problem is. Most insulation materials also require measures to stop air flowing through them so basically you need to make those vapor proof plastic film / paper / Housewrap / whatever seams completely tight, even small gaps let the wind in your insulation compromising its thermal conductivity value. The material stopping the air flow should be water vapor proof on the inside (typically plastic), but even more importantly, the outside material must let the water vapor through.

[tom66]

I wish I had a gas flow meter to really see how much I'm saving or wasting, as I tune things.  Maybe should consider asking how much having one installed is.  Although I'm not sure I trust the "Smart" offerings from china as a gas flow valve, so a better solution would be a proper rated flow meter and make it smart myself somehow.

I'd like to have dynamic flow/loop temp control, the boiler does support it, but its a bit of a "not quite standard" EMS protocol.  An adapter product exists, but it's very DIY and carries warnings about breaking your boiler.

Until I finally go that way, I'm going to make my zone targets slightly more complex to add temperature hysteresis and also implement heating as many zones as needed while the heating has been triggered.

I have a gas smart meter and get half-hourly consumption figures.  No price variability on gas so no benefit to shifting usage like electricity but could help you gather an idea. Most suppliers will put them in for free.

I think I do need to optimise this home, in winter we are using cca. 50kWh per day of gas.  3 bedroom detached new build.

[nctnico]

Small temperature differences between rooms may be an OK micro-optimization, but keeping some rooms cold requires adding thermal insulation between the rooms to actually save anything.

Instead, use the same money and effort to put that insulation on the outer envelope of the house -> now you can just keep all the rooms warm with basically the same cost and stop thinking about it.

Really, insulating pays off pretty quickly. You may want to get a thermal camera to look at the discontinuities, to find where the problem is. Most insulation materials also require measures to stop air flowing through them so basically you need to make those vapor proof plastic film / paper / Housewrap / whatever seams completely tight, even small gaps let the wind in your insulation compromising its thermal conductivity value. The material stopping the air flow should be water vapor proof on the inside (typically plastic), but even more importantly, the outside material must let the water vapor through.

OK in theory. The downside of such measures is that you'll also need a ventilation system otherwise you get a large amount of moisture inside the house. This can get really expensive quickly because of the remodelling needed to install the pipes. In the end you can only add a limited amount of extra insulation to an old house (from a financial perspective).

[paulca]

I seems to be a balancing point thing.

By example.  If I keep my bedroom at 16C all day it uses about 10-15 minutes of heating an hour when it's 3C outside.

Before lockdown and WFH became the norm, I would have been out all day for 9-10 hours and let it cool to 14C, which only pinged the heating every few hours.

However, 14C is too low a temp to ramp up from quickly to 17/18C when the room becomes in-use.  The heating responds within a minute, gets to temp within 5 and warms the room pretty quickly, but... the furniture is still 14C, which makes the room feel colder.

This switch between 14C and 16C is triggered by me being home.  As I'm pretty much always home now, modifying the vacant target from 16C down to 14C isn't the best option.  I think I have to accept it and adjust this schedule based on time instead of "presence".  Let the bedroom be 14C during the part of the day I am least likely to use it and warm it to 16C during the time I'm likely to use it.  Like evenings and mornings.

Last year I only had a 3 programme day timer. At this time of year I typically had 2 hours morning, 3 hours evening, 1 hour bedtime, usually 2 or 3 additional 1 hour boosts at the weekend or advancing schedules.  Probably 6-8 hours a day with all radiators on (with TRVs).

While I have seen run times as high as 11 hours with the new system, that was what started to prompt this tuning process.  That said, 11 hours run time with 2-3 radiators closed most of it.

I have brought that down to 8 hours with a few tweaks but I think I can get it lower and still retain comfort.

Insulating the bedroom will make a huge difference too.  The installation job on the upper roof insulation failed to realise, as my house is a chalet-bungalow, I have a lower roof space too which surrounds the bedroom floor and living room/hallway ceiling.  Venting the upper roof has also vented the lower part, which has no insulation.  Making those walls freezing cold.  I will have to get them to cut into that inaccessible roof space to insulate it.  Probably costing me a grand or more

[GlennSprigg]

I haven't really worked with 'domestic' systems, but used to be responsible for certain large industrial and
commercial multi-story control systems.  Although (as some people mentioned) P.I.D. (Proportional, Integral
& Derivative) controllers can 'smooth' things out and make cooling/heating valves more efficient, 'Hysteresis'
is most important in that regard when it comes to energy saving!  (More on that shortly)...

However, the main problem is Common Heating/Cooling calls!!  We can't have a situation where just one
sensor/setpoint in just one zone/room/floor, (which may have say a warm computer too close to it!), for being
responsible for starting say a whole buildings Chiller-Plant or Boiler. That requires a pre-programmed number
of sensors, and pre-programmed time delays at times. Otherwise, arses are kicked!! 

Regarding Hysteresis, that is a two-pronged attack!!  One is the cycling on/off Range, either heating or cooling,
but the other is 'Time Of Year Perception' !!  I don't know what other colder countries do, but here in Australia,
'experts' have determined that in our Winter, let's say it's say a 'freezing' (haha..) 14-deg-C outside, then the
general public/workers actually feel comfortable with a cycled range inside, of say 18-21 deg-C. And when it is
say 34-deg-C outside, people are comfortable/happy when the cycled range inside is say 25-28 deg-C inside!! 
The happy 'Norm' here, is calculated to be 23-deg-C, but maintaining that will cost the owner Heaps! 

I'lll throw in one other factor, which is where a 'Company' is the sole multi-story owner/occupier, and where the
building is almost totally sub-letted to numerous other paying tenants!!! If it is a case of the 'latter' here, then the
owners have the control of all the metering, and re-sell the power to all the tenants, making a huge profit there!! 
You try to talk about energy management to them, and they just laugh!! (True!).