Thursday, March 29, 2012

Incorrect kW load results from corrupted TLC three phase TOU meter data

[Updated - see end of post for details]

In the last few days I have informed TLC's CEO (John Anderson) of the details of a serious data error that appears likely to affect a large number of customers with three phase supplies and who also have Time of Use meters. These would typically be larger consumers but will include some domestic homes also.

In essence the problem is that the half hour data from certain types of time of use meters used at three phase installations (including at least the EM5100, EM5200, EM5300 models) has been corrupted and when that corrupted data has been processed incorrect kW load values have been produced. Often, perhaps in the majority of cases, I would expect the incorrect results to be higher than they should, hence causing higher payments for the customer.

TLC have not disclosed details of how the new Accommodation/Holiday Home formula has been derived but if this was done using some data from installations with this problem then the formula may well be similarly affected, especially as I expect a higher than average proportion of 3 phase installations within that group. It may affect the dairy shed formula also.

There is no direct evidence that the meters themselves are faulty. Instead it seems likely that the data has been corrupted either during the data download from the meter (a manual procedure carried out after each winter) or possibly in some subsequent processing by TLC or its agents. (Please note carefully at this point that I am no longer contracting to TLC - they informed me that my services were no longer needed last December!)

The reason for suspecting that the problem is not directly within the meter itself is that the LCD display has shown the (approximately) correct time over the last year or two, so the meter clock itself has not obviously been wrong.

I provided John Anderson with a detailed analysis of the fundamental nature of the corruption within the data files. The essence of that is as follows:

1. Some meters (including the models listed above) have sufficient memory to store roughly two years worth of half hour data.

2. If they have been installed for a sufficiently long period of time, this means that the data downloaded after the 2011 winter period also included much or all of the 2010 data, often including dates when daylight saving officially ended in April

3. If the above conditions are true (but perhaps others also) then the half hour data that TLC processed (and also supplied to those customers who requested it) in order to produce the kW load figures that determine the monthly charges from 1 April 2012 was incorrect, as the April 2010 daylight saving transition did not appear in the data (for some reason not yet determined).

4. Later daylight saving transitions did appear. The net result is that the data (from the point where the April 2010 daylight saving transition should have appeared) is one hour advanced from the "correct time".

5. As the process of determining a kW load figure from the data requires reconciling the half hour data from the meter (with its own dates and times derived from the meter clock) against TLC's separately determined records of when load controlling was occurring on the appropriate channel, this means that consumption during the hour before load controlling started will be incorrectly included in those calculations, and similarly, the last hour of each qualifying period of load controlling will be treated as if it was outside of the controlled period and therefore ignored.

6. If there are any situations during the winter period where inclusion of the lead-in hour could result in significantly higher kW load (e.g. from controllable load such as hot water and so on), then the overall result from the average of the top 6 such periods (and charges to the customer for the following year) will also most likely be inflated.

The story of how this problem came to be uncovered is interesting and goes back a long way but I will leave that for later. Suffice to say that the first set of half hour data showing this problem led to a pricing letter from TLC that had the (incorrect!) kW load value at roughly 3 times what it should have been, and for this (large power user) that would have meant roughly $700 extra per month, coincidentally with payments starting on April Fool's Day...

I recommend any customer with an EM5100, EM5200, or EM5300 time of use meter installed (look at the top left on the front for the model number) who has any concerns that they could be affected negatively by this to immediately request a full copy of the raw data used to generate their kW load results from TLC, including the load control timing information that was also used, plus any other relevant information to allow full verification of the kW load calculation. The EM1000 meters like the one at my home only store about 9 months of data and don't seem to be affected but you could ask for your data anyway, just to be sure. There are many other situations in which similar problems could crop up.

If a few other customers hadn't asked for their data then this problem would probably not have been uncovered.

TLC's Elizabeth Anglesey kindly pointed out in a letter to the Ruapehu Press on 25 Jan 2012 that any customer was entitled to ask for their information. See the extract below for what she said:


I'd be very interested to hear from anybody who has asked for and received their data. Clive at xnet dot co dot nz will reach me or else try 07 8954556. At the very least I should be able to help you determine if you are affected by the problem I've described above.


Update: On 2 April I received an email from TLC about this issue. After thanking me for my help the email went on to say:
We have discovered that one of the hand held readers used by the TOU meter readers was found to have faulty/incorrect software.  Fortunately it was not the reader used for the majority of the customers.  We found around 40 customers who kW load needs to be reduced.  All these customers will be notified directly and will have their load level reduced for 2012.
Note that I currently do not know the full number of meters affected. I also do not know if any of the affected installations were included in the data sources used to generate the various formulae. I have asked TLC for further information on both points.

Friday, March 9, 2012

Is a "formula" the same as a licence to print money?

Here are a few graphs, showing the way the standard domestic/commercial formulae used by TLC have changed over the last few years. Oddly enough TLC seems to find that every year, without fail, those at the very bottom of the power consumption ladder just keep generating more and more load in peak times, and so must pay more. At about 6 kWh per day or 550 kW over 92 days of uncontrolled consumption (so typically not including water heating) the formula now spits out a 10% higher kW load estimate than it did last year. Add in the notified "price increases" and your bill will be increasing by quite a bit more than that even though your consumption (and most likely true kW load) has not changed.

Or is it all just a way to beat these customers up until they switch to using a time of use (demand) meter?



First, a close up of the lower end of the range, roughly corresponding to those customers who may be eligible for the low user plan. TLC has (roughly, somehow) determined that an "average" kW load value for domestic homes is 2.2 kW although you'd think this needs to be adjusted upwards with the every increasing load estimates.



This version shows a large range of the input values, roughly corresponding to what you might expect from most domestic dwellings. Note that TLC's formulae often turn into straight lines once you get above a certain level of consumption.
This is a very "wide angle" view of the same formulae again. The 2009 formula predicted much kW load figures than those from later years at very high levels of consumption. All these formulae are derived from a small sample of domestic homes (and all from the northern area) so it's quite possible the estimates at the very high levels (which would also typically come from commercial and industrial users of one kind of another) to be even more inaccurate. Note also that up to 2012, accommodation and holiday homes were subject to these formulae also but in 2012 TLC introduce a new formula just for that group.
This shows the very large and painful change TLC have imposed (at rather short notice I might add) on accommodation and holiday home customers (top line) as compared to the formula applied to them (and others) last year (lower, red line).

What is missing from all of this is a complete, detailed, correct description from TLC on how the formula are derived, how (in)accurate they may be, what data was used as the basis for them and so on. The graphs below comes from an earlier version TLC's 2010 methodology but you won't be able to find it on their site any more...


This "shows" how the non-dairy formula TLC used over 2010/2011 was derived.  The red data points are 92 day consumption figures versus peak demand (apparently the single highest 3 hour period "while load controlling" - although this was not data that I processed) based on the smallish sample of homes ALL drawn only from the northern region. Count the dots - I get about 110 which is a miserably low sample size to set the rates for the other (roughly) 20,000 customers. Anyway, the black line is "the formula", derived by finding some kind of "best fit" to the red data points and then expressing that mathematically.

In other words, the line in this final graph shows where the formula behind the blue line in the first three graphs above came from. I wonder what they've used to come up with the 2012 formulae? Looking at the very top graph you'll see they top line is remarkably similar in general shape to the next one down (for 2011) so I think we rest assured the same old tired sample from up north is still influencing everything.

The 2012 Formulae - direct from The Lines Company

Copied and pasted from an email from TLC, below you can find the new 2012 formulae. TLC haven't published this info on their site as of now but no doubt will at some point - please use their information if there is a difference.

A key detail to remember is that the input "consumption" value needs to be the total uncontrolled units (kWh) used over a 92 day period, or at least an estimate of that based on readings taken from some appropriate period from 2011 (June through September for all but dairy farms which instead are analysed using the September through December period).

For a completely uncontrolled installation or meter that has no controllable load attached, use all of the recorded consumption. For "mixed meters" it seems TLC are still using 65% of the total units in the "southern areas" and 75% for Te Kuiti north. For the "Limited Off Peak meters" 55% of the actual consumption recorded is used. Night time consumption is also disregarded if metered separately.

All of these factors and adjustments as above can compound the inherent inaccuracies in the formulae as these percentages are just some kind of estimate that TLC has never properly explained or supported with detailed data.

These formulae clearly also come direct from a spreadsheet (e.g. Excel) so interpret them in that context (* is multiplication, / is division, ^ is exponentiation, IF is the "IF function", etc.)

Standard

    IF(consumption<750,(0.1391*consumption^0.4931)/2*1.0161,((0.001559*consumption)+2.476)/2*1.0161)


Accommodation

    (0.04354*consumption^0.7329/2)*1.0423


Dairy

    (0.0009*consumption+3.9795)*1.0097

Wednesday, March 7, 2012

Those on the red line, see red!

TLC have created a whole new storm after sending out their most recent pricing letters about a week ago. One particularly significant change is that the new 2012 formulae (now masquerading as "profiles") include a new entry to be applied to accommodation businesses (e.g. motels) and holiday homes. In summary, TLC customers lumped (how exactly?) into this category can expect to have new charges based on a kW load figure that will often be about twice as high as that used a year ago even when their average winter daily consumption is essentially  unchanged. Here's the graph, red is the new profile.


Data for the graph was sourced from TLC's updated online calculators which I found earlier today. On the same pages TLC says that a minimum kW load of 2.5 will apply to those in Ohakune, and 2.2 elsewhere. This will roughly correspond to the point where the red line starts at the bottom left but for now I haven't shown that "little detail". What it means though is that the low usage holiday home owners still have to pay charges every month of the coming year based on a "kW load" of (at least) 2.2 kW - even if they actually used zero power for the whole winter!