Presentation is loading. Please wait.

Presentation is loading. Please wait.

Allocation of Unidentified Gas Statement – Interim Report findings 17 th September 2012.

Similar presentations


Presentation on theme: "Allocation of Unidentified Gas Statement – Interim Report findings 17 th September 2012."— Presentation transcript:

1 Allocation of Unidentified Gas Statement – Interim Report findings 17 th September 2012

2 Outline Consumption analysis Theft Analysis Q&A Next Steps

3 Consumption Analysis: Background Estimate of Unidentified Gas total Uses meter read data for both SSP and NDM LSP market sectors Compares allocated and metered demand totals Total UG is the difference between them Calculated figure covers all sources of UG SSP-assigned LSP-assigned

4 Consumption Analysis: Data Used Data provided for time period 01/04/2008 – 31/03/2011 EA LDZ for pilot study Meter read data on MPRN by MPRN basis, EUC included Allocation data (including CSEPs) at EUC level List of MPRNs with no meter reads in time period List of new and lost sites with start/end date MPRN count by EUC for CSEPs Allows MPRN population by EUC to be calculated

5 Consumption Analysis: Meter Read Data LB1 (Lower Bound 1) – the latest meter reading prior to the start of the formula year LB2 (Lower Bound 2) – the earliest meter reading within the formula year UB1 (Upper Bound 1) – the latest meter reading within the formula year UB2 (Upper Bound 2) – the earliest meter reading after the end of the formula year

6 Consumption Analysis: Calculation Rules Select meter reads for consumption calculation using set of rules Full definition in interim report Max time period between LB1 and start of year = 540 days Max time period between end of year and UB2 = 540 days Min distance between selected meter reads = 120 days Min overlap between metered period and relevant year = 60 days

7 Consumption Analysis: Calculation Calculate consumption for metered period Convert to formula year consumption using (volume) ALP Convert to energy using weighted average CV for formula year Check against AQ and reject if more than 5 times smaller or larger Aggregate to EUC level Calculate sample size used for each EUC

8 Consumption Analysis: Final Calculation We now have Actual population for each EUC Sample size in each EUC Aggregate demand for sample for each EUC Multiply sample demand up to cover full population in each EUC Add EUC totals to give overall metered total Subtract from allocated total to give UG Use of AQs for missing sites rejected due to inaccurate results

9 Consumption Analysis: Sample Size Sample size consistent for 2009/10 and 2010/11 Lower in 2011/12 due to lack of UB2 for all meters Formula YearPopulation SizeSample SizeSampling % 2009/101,874,7371,580,05584.3% 2010/111,893,2091,597,24084.4% 2011/121,907,8351,302,06968.3%

10 Consumption Analysis: Unidentified Gas Results Consumptions and UG calculated as described Confidence Interval calculated around best estimate Allocation (GWh) Metered Consumption (GWh) UG (GWh) Best Estimate % Low %High % 2009/10 39,80538,5411,2643.18%2.80%3.56% 2010/11 41,09539,9851,1102.70%2.40%3.00% 2011/12 34,08033,6734071.19%0.73%1.66% Average 2.36%1.98%2.74% Best LowHigh UG Estimate (GWh)803 673933

11 Consumption Analysis: UG using AQ Substitution Average of a similar order Year-to-year variability high and includes negative UG Allocation (GWh) Metered Consumption (GWh) UG (GWh) Best Estimate % 2009/10 39,80538,5351,2703.19% 2010/11 41,09539,9101,1842.88% 2011/12 34,08034,183-103-0.30% Average 1.92%

12 Consumption Analysis: Comparison with 2011 Method Total UG estimates for EA are: 803 GWh (new method) 841 GWh (old method) Confidence Interval of comparable width New method avoids weaknesses and assumptions of old method Therefore recommend that new method is adopted for 2012 AUGS for 2013/14

13 Theft Analysis To consider theft split based on metered + unmetered consumption Also looked at variations of the above to fix split in time to avoid risk of manipulation Further alternative method based on throughput Also examined issue of theft in a year exceeding 73,200 kWh vs AQ Modified original method to include this override for like for like comparison with new method Obtained meter reads associated with theft affected sites Obtained some information regarding customer occupancy changes

14 Theft – Original vs Alternative Method 1 Original method Apportions theft by calendar year Applies flat profile to allocate theft to calendar year Pre theft AQ used if available otherwise uses current AQ Multi-year thefts are given one market sector classification For comparison modified version of original method with theft override at 73,200 kWh was used Alternative Method Apportion theft by formula year Applies ALPs to allocate theft to formula year Meter reads used to calculate consumption and seasonal adjusted theft added to get sector split Pre/post/current AQ used if consumption calculation fails Multi-year thefts are given individual market sector classifications If theft in a year of occurrence exceeds 73,200 kWh assigns to LSP regardless of AQ/consumption result

15 Metered + Unmetered Consumption Consumption could not be calculated for ~50% of theft affected sites Not surprising as meter read data shows missing reads constant values reads giving rise to negative consumptions etc For consumption failures fallback was to use pre/post AQ method Year of Occurrence Number of Thefts Consumption Calculation Failures Failure Rate 2007200190845% 20082595111943% 20092774164059% 2010158398762%

16 Comparison of Original vs Alternative Method results 200720082009 2010 Original Method Alternative Method 1 Original Method Alternative Method 1 Original Method Alternative Method 1 Original Method Alternative Method 1 Consumption Calculation Successful 11.5%13.1%15.7%13.0%19.4%15.5%15.8%8.9% Consumption Calculation Failed 20.3%26.7%29.6%31.0%25.6%24.1%28.6%26.4% Combined 16.4%19.3%22.5%21.4%23.0%22.6%24.8%20.6% Consumption calculation failure related results are similar as expected Consumption calculation successful results vary and generally give smaller LSP split than successful consumption calculations (potentially different subset of thefts)

17 Theft issues – detected theft estimates The theft estimate is used twice for the size of the theft added to metered consumption to get total when assigning sector The theft period is also estimated and subject to error There are examples where meter reads show potential theft outside of the theft start/end dates If over/under estimation of the theft and theft period is unbalanced then this can affect the overall theft split

18 Thefts issues - unregistered sites Unregistered sites illustrates a key problem with the method Should they be included or not when determining unknown theft split? Can be interpreted in different ways If included, unfair to LSP shippers who will claim that the site would not have stolen gas if it had been their customer If excluded (and part of Shrinkage) may be seen as unfair by predominantly SSP Shippers Anomalies such as this can have a significant impact on the theft split from year to year

19 Theft issues – Shipper influences The theft split can be manipulated by shippers who focus detection on a specific market sector The theft split can also be affected by the level of effort in detection and prevention Detecting theft in a particular sector will increase that market sectors theft portion for unknown theft and acts as a significant disincentive

20 Theft Issues – where AQ is used Where there are customer changes AQs do not always reflect the level of demand for the previous occupant during theft Post theft AQ cannot be assumed to be theft affected – there are cases where the AQ rolls over and adding theft skews the split to LSP There are also cases where it is affected so to not add theft would skew the split to SSP Similarly for pre-theft AQ Customer changes are generally in the SSP sector, although there are some examples of customers changing in the LSP sector with similar effects

21 Variations of alternative method 1 To prevent Shipper influences considered limited detected theft data to end 2011 However, in time the resulting split may be seen as out of date Further variation to use this split and scale it in line with consumption going forward

22 Theft split using throughput Considered split based on NDM LSP vs SSP consumptions Key assumption is theft occurs in proportion to throughput in each sector Final figure here based on projection to 2013/14 (to be revised each year) Sector 2007 200820092010 2011 Final NDM LSP 135.1134.4134.5125.0124.1 SSP 348.7368.4379.4363.2376.2 Total 483.7502.9514.0488.2500.3 LSP % 27.9%26.7%26.2%25.6%24.8% 23.3%

23 Summary Theft split based on throughput (NDM LSP vs SSP) Removes many of the sources of error in the original method Cannot be manipulated by Shippers Less volatile over time Is simple to implement and administer Provides an incentive to detect and prevent theft Recommend theft split to be based on throughput (NDM LSP vs SSP) for 2013/14 onwards

24 Thank you for your attention


Download ppt "Allocation of Unidentified Gas Statement – Interim Report findings 17 th September 2012."

Similar presentations


Ads by Google