Sunday, July 8, 2012

More on Thermal Modelling

7 July 2012: While Serena Williams has won her fifth Wimbledon, and the unseeded Jonathan Marray becomes the first Brit to win a Wimbledon Final for 76 years (upstaging Andy Murray's noble efforts by one day), I've had a share of my attention on continuing to improve this thermal modelling project using GDL. It has gone really well, and I can now keep the model and graph going, in parallel with my meter readings during the year - it will grow, day by day. Also, this is undoubtedly the stuff for my next WSSET and CIBSE paper, there's been a tremendous amount of thought put into it, and I am glad of the opportunity to explain it all.

Screen capture of some of the working windows

The Algorithm
The dynamic simulation takes every single daily meter reading from September 2009 to now, and calculates the thermal energy extracted from the earth by the GSHP, the solar energy injected into it by the Sunbox and Tubes, and the heat brought in from the infinite surrounding earth. From this, it calculates the theoretical volume of the thermal energy bulb.
   The earth energy extracted from the borehole is done by considering the electric power consumption, making an assumption about the COP (coefficient of performance), and thus deriving the amount of energy drawn from the borehole. If the COP is 3, it should draw 2 kWh for every 1 kWh it uses. (Correct me if I am wrong, please.)
   The Sunbox energy put down is read from the meter, but in reality there must be some system losses (e.g. heat from pipes, heat lost going through water in the manifold). But this is fuzzy assumption, about as fuzzy as a guess of the COP, so I can let the user enter a parameter for system loss, and see what that does to the curve.
   In Winter, as the energy level in the earth reduces, it creates a delta-T between itself and the infinite surroundings and pulls thermal energy from around proportional to a fraction of the square of the distance. In Summer, if the solar panels put down too much energy in the summer, some of this will migrate outwards and be lost. The algorithm takes care of both of these cases.

Could this be done with a spreadsheet?
Now that I have solved it in GDL, it is tempting to go back to Googledocs and develop a spreadsheet version of it using the same algorithm. The use of parameters is perfectly possible in a spreadsheet. I have reached the limit of what Googledocs can manage, and I have had to cut my spreadsheet down and move it to a new document. There is a theoretical maximum number of formula based calculations it can do, and I have passed that point.

Chart showing thermal model WITH SUNBOX above, and WITHOUT Sunbox below

Explaining the graph
This graph very accurately represents the weather patterns we have had during the last three years. Data from meters is recorded at daily intervals, apart from the odd holiday or weekend away - so it is following the meter readings with over 300 points per annum. If you enlarge the diagram, the dates are printed at fortnightly intervals, and the year points are marked by vertical lines. The top diagram is the actual pattern based on real data over three years. The lower diagram is the same pattern, with the Sunboxes recording zero.

Here are some weather notes, explaining the curve shape.
  1. The light vertical line in 2010 is the date of installation of the Sunbox, and the twin vertical lines in 2012 are the month in which I had no Sunbox and no Tubes. Once they were installed in early May 2012, the line grows upwards, satisfactorily. Despite the rainy weather, the heat pump has not been overworked. Solar energy put down is reduced, but so is the energy withdrawn.
  2. The winter of 2009 shows a steep dive because the heat pump was quite overworked, with severe winter conditions in Spring 2010, a regular resort to 'additional heat', and slightly less insulation and air tightness than we had later. 
  3. 2010 also had a short summer and a dive into a very cold winter, so the energy level fell, but not as much as it would have done without the first Sunbox. 
  4. 2011 was a year with no winter at either end, so we achieved record high figures for PV and Sunbox solar capture, for reduced GSHP and house consumption.
  5. There was almost no winter from 2011 into 2012, but suddenly in February 2012, we had a period of snow and cold, followed by nearly 6 months of unseasonal low temperatures and the highest rainfall statistics in weather history. The two short one-week heatwaves of March and can be seen as blips in the curve. 
Note, that the curve does not represent actual energy levels, precisely. There are too many unknowns. However, it represents the Pattern. The shape of the graph is the same even with a few parameter changes - the shape is based on the data, and this is influenced most strongly by the weather patterns.
  What does the Horizontal line represent? it is user definable, for example, in this one, it is at the line of the lowest energy level, at the start of March 2010 after a prolonged very cold spell and no Sunbox. It can be moved to any height, and the polygon is re-formed for fit it.

Does this prove anything?
Well, it demonstrates the general hypothesis with which I set out to do this project, proving that there is truth in it. Now it is a theory!
   After first installation, the GSHP appears to do quite well, initially. After the first Winter, it seems to perform less well and uses more electricity, is on for longer hours. But after a while, it rapidly settles the same pattern of curve, but at a lower energy level - how far precisely is dependent on the size of the house, the depth of the borehole, the variations of the summer or winter, but following much the same shape - steep winter fall-off, long slow climb out in summer.
   With solar augmentation, the borehole is refreshed every summer, and seems to settle into the same curvy pattern, but at a higher energy level - this means that the COP of the GHSP will be improved because the COP improves if its energy source is warmer.
The proof of the pudding has been in the eating, whereby the ground temperatures reflect this curve. However, they level out at just below 14ÂșC because after that the energy bulb is widening, not getting warmer. This diagram finally demonstrates what is really happening.

The three year charging cycles overlaid onto a single diagram.

Simulating COP changes  
The lower graph (Sunbox off) is still influenced by the Sunbox even if it is turned off because the consumption of the GSHP is reduced by the SB and I can't change the readings, but perhaps I can tweak the COP factor. The GSHP uses vastly less power as a result of its improved COP. Maybe I should alter the model to make the GSHP use more electricity and draw less from the ground if the Sunbox is turned off. So in later versions, I can introduce a worsened COP - ironically, I found that this draws less heat from the ground because the GSHP burns more power electrically. 

Additional Heat effect
Another thing I cannot compensate for is that in the early days, until March 2010, the GSHP used to resort to using 'additional heat' (AH) if it couldn't get enough heat from the ground. This explains why the curve in the winter of 2009 descends so deeply. This AH was done for pasteurisation, but it would do it frequently whenever it wasn't getting energy from the earth quickly enough - an average of 220 hours per annum in its first two years. This increases the consumption by 4 kW immediately, and is equivalent to adding 8-900 kWhs per annum to the total consumption. I am not sure whether to tinker with the model and introduce a modifier for the readings from Oct 2009 to March 2010, but have decided to leave it as it is, and explain it here.

Complications of the 2012 additions to system
I thought you might ask this question! I bought a new meter for the Sunbox, and moved the old energy meter to the heat exchanger for the tubes. So I compile the data in a spreadsheet first and copy and paste it to the data file for the model. the spreadsheet can bring in the meter readings, and apply the correct additions and subtractions, based on the reading taken just before the meters were swapped. If you look at my Peveril-Metering spreadsheet, you can see the page on which the data is prepared before being read by the GDL model.

Any comments, please leave below!

No comments:

Post a Comment

Comments will be moderated before showing. Please make them relevant to the subject of the posting. Comments which advertise commercial products will usually be deleted.

Popular Posts