In an ideal world a heat pump unit would contain two compressors, a nominal 12kW capacity heat pump having one compressor rated at 8kW and one compressor rated at 4kW. The 8kW compressor providing heating during most of the Winter period, with the 8kW plus 4kW meeting demand during particularly cold spells.

As milder weather conditions are encountered the system would changeover from the 8kW compressor to the 4kW one, thereby allowing a much lower thermal energy output without the possibility of cycling, and probably providing an increase in overall efficiency.

Unfortunately we don't yet live in an ideal world.

I believe that most, if not all, heat pumps control the stopping and starting of the compressor based upon the temperature of the LWT and RWT, such that when either of these temperatures reaches a defined limit, the compressor is started or stopped. This of course is more likely to happen during milder weather conditions.

The 'on' period and the 'off' period is therefore dependent upon the limits set within the controller and the actual temperature measurements produced by the LWT and RWT temperature sensors. One of the major determining factors affecting the frequency of cycling could be the location of the LWT and RWT sensors within the system. If the sensors are located within the outdoor unit they are likely to cool at a more rapid rate than if they were located within the home's thermal envelope, and therefore increase the cycling frequency.

From everything I've seen so far, I think the heat loss calculations are more useful for individual radiator sizing than for overall heating requirements.

Thats precisely the same conclusion as I have reached. 

I have challenged MCS on the method they promote.  Their response was that they are aware of it's deficiencies but 'we don't currently have an alternative'.

I'd already concluded that the higher estimates are excessive, which is why I'm going with the lower of the three figures that I've received.

To my mind, empirical data should always trump theoretical calculations that are based on uncertain assumptions and approximations, but I realise the MCS approach doesn't really allow for this.  From everything I've seen so far, I think the heat loss calculations are more useful for individual radiator sizing than for overall heating requirements.

We got three surveys that came out with heat loss estimates between 8.1 and 13.5kW, hence my uncertainty about the size of the system needed.

My property is 200sq m.  Two MCS surveys each taking 3hrs came out at 16kW.  They ignored fabric upgrades I told them about and double counted room to room loss.  Correcting for these gives 10.5kW. 

Working from my annual gas consumption of about 20MWh would equate to a 7-8kW actual demand which is what I measure on my smart meter (note this translates to about 4kW most of the time ie other than when it's not very very cold).  To prove this I have turned my boiler max output down and downloaded/checked my smart meter readings. 

I can get a spreadsheet calculation to equal 7 5kW by assuming 0.5ach instead of the default values MCS assumes.

So my strong advice it to be very sceptical about the higher figures you have been quoted and if you can, get a measurement.  Oversizing leads to excessive cycling and loss of efficiency and the spreadsheets suffer from gigo in a retrofit where the fabric has been non-uniformly upgraded.

We got three surveys that came out with heat loss estimates between 8.1 and 13.5kW, hence my uncertainty about the size of the system needed.

Thanks again.  I've explored quite a lot of data (and plotted many graphs), but only have daily gas consumption figures from late January onwards (just caught the end of the last cold snap) and don't have anything with negative AOT values.  However, the relationship looks pretty linear (just like yours), which I believe concurs with the theory, so I'm pretty happy extrapolating down to lower temperatures.

Assuming 90% boiler efficiency and calculating an upper 95% prediction limit on the regression, I get a total consumption of 130kWh at -5C.  This is with the heating on for 17 hours (it's currently turned off between midnight and 7am) which equates to an average of 7.6kWh per hour through the day. 

This ought to be well within the capability of the 8.5kW Ecodan.  The 11.2kW unit would obviously be the safer bet, and would be fine at this temperature.  I'm just uncertain about whether there would be a performance penalty in milder weather when it only really needs to be putting out 4-5kW to maintain the temperature.