Application of EP2 in derivation of SNCWV 23 rd September 2009

1. Application of EP2 in derivation of SNCWV23rd September 2009

2. Application of EP2 data in derivation of SNCWV • DESC meeting to discuss and agree approach for using EP2 data in the derivation of SNCWV from 1st October 2010 • DESC were advised 24th July that additional meeting would be necessary during September to discuss practicalities of how to use EP2 data • xoserve issued DESC members with the Transporters proposed approach for applying EP2 data • xoserve welcomed feedback in advance of DESC meeting in order to address any issues or concerns today • Two shippers have provided feedback – points raised to be covered later in the presentation

3. Summary of material issued 28th August :‘Applying EP2 data to SNCWV’

4. Seasonal Normal Review process • Two key decisions have to be made as part of 5 yearly Seasonal Normal Review: • Is the Composite Weather Variable (CWV) methodology fit for purpose and if so agree the appropriate number of years required to derive the CWV parameters • AGREED - March 2009:Industry accepted CWV formula was robust and number of years used should be 13 (1996/97 to 2008/09) • Produce a Seasonal Normal Composite Weather Variable (SNCWV) which is a reasonable representation of “normal” weather for an LDZ • AGREED - June 2009:Industry accepted SNCWV will be derived using EP2 data, pending implementation of UNC Mod 254

5. Seasonal NormalComposite Weather Variable (SNCWV) • The SNCWV should be a reasonable representation of “normal” weather for an LDZ for each gas day within the currently defined seasonal normal period • The current set of SNCWVs were agreed in 2005 and are based on 17 years of historical data (1987/88 to 2003/04) and cannot be used after 30th September 2010, as per UNC. • The revised SNCWVs are required for the next defined seasonal normal period, namely from 1st October 2010 until 30th September 2015 • Output from EP2 project was selected by industry to be used in the calculation of the SNCWV from Oct 2010 to Sep 2015 • The SNCWV is..... • a daily number which represents normal weather for each future gas day • calculated before gas day based on an industry agreed weather data basis • a value for a day in a gas year BUT is the same for each of the 5 years (including on leap years when they occur) i.e. value for 1st Oct 2010 is the same as 1st Oct 2011.

6. Principles for applying EP2 data to derivation of SNCWV • Key points xoserve needed to consider in applying EP2 data: • Output data from EP2 should be used in derivation of new SNCWV, however… • The SNCWV to be calculated in a manner which means it is fit for its primary purpose - modelling gas demand i.e. modelling parameters are not degraded • Maintain EP2 temperature ‘profile’ i.e. key features such as ‘buchan spells’

7. Weighting for Wind Chill value (l2) defined from relationship between LDZ NDM demand and CWV from 1996/97 to 2008/09 EP2 data not required Weighting for pseudo Seasonal Normal Effective Temperature (1 – l1) defined from relationship between LDZ NDM demand and CWV from 1996/97 to 2008/09 EP2 data not required Weighting for Effective Temp (l1) defined from relationship between LDZ NDM demand and CWV from 1996/97 to 2008/09 EP2 data not required Composite Weather = I1 * E(t) + (1-I1) * S(t) – I2 *WC(t) EP2 Forecast Temp represents Actual Temp when calculating Effective Temp Effective Temperature: Half of Yesterdays Effective Temp + Half of Today's Actual Temp EP2 Average Wind Speed and EP2 Forecast Temp used in Wind Chill Term Wind Chill Formula: Average Wind Speed * Max (0,(14-Actual Temp) Pseudo Seasonal Normal Effective Temperature defined from relationship between LDZ NDM demand and CWV from 1996/97 to 2008/09 EP2 data not required Applying EP2 output data to SNCWV calculation • Shaded boxes indicate where EP2 data output is ‘plugged’ into CWV formula • Note, the various parameters (e.g. cold weather upturn) will also be applied before finally calculating a SNCWV value for the gas day

8. EP2 Work Package 8 output: • EP2 data output was provided for each weather station in the form of: • Temperature: • (A) - Hourly smoothed average temperatures for the base period of 1971 to 2006 (36 years) • (B) - Hourly climate change temperature increments for the forecast period of 2008 to 2018 relative to the base period averages • (C) - Hourly smoothed average temperatures for the forecast period of 2008 to 2018 (expressed as the sum of items 1 and 2 above) ** • Wind Speeds: • (D) - Hourly smoothed average wind speeds for the base period of 1971 to 2006 (36 years) ** • There were no forecast increments for wind speeds. Note: ** relates to data required for CWV formula

9. Features of EP2 Work Package 8 output • The forecast temperature increments are appropriate to the weather stations in current use BUT were not calculated from the base period average data • The smoothed base period average (1971 to 2006) is a means of representing the predicted climatologies • Base period average for temperature and wind speed was based on data from a variety of weather stations • EP2 forecast temperatures for the gas years in question (2010 to 2014) have been calculated by applying EP2 temperature increments appropriate for each year to the EP2 base period average • The key temperature data output from EP2 is the predicted temperature increments

10. Analysis of SNCWVs using EP2 data only and alternative method • An EP2 CWV has been calculated for the middle year (2012/13) of the five years over which the seasonal normal basis will apply using EP2 forecast temperatures and EP2 wind speed averages (referred to as avg. data stream) • A CWV has been calculated for 2012/13 using an alternative approach relating individual days’ temperatures and wind speeds when computing wind chill (referred to as “daily data”) • An EP2 temperature only profile has also been calculated for 2012/13 (referred to as EP2 temperature) • A comparison can be made of the ensuing changes to the daily average degree days value over the full gas year. • In all cases, the appropriate threshold used for degree days is that for aggregate NDM demand. • The table on the next slide shows the changes to average daily degree days (over the gas year) relative to the current 17 year SNCWV for the various CWV cases and relative to a 17 year SNT for the temperature case.

11. Analysis of SNCWVs using EP2 data only and alternative method

12. Results using EP2 data only and proposed approach in deriving SNCVW • EP2 temperature only “warming” is lower or very similar to the EP2 CWV using average data streams. 12 of 13 LDZs within ±0.1º • Result is surprising - wind chill should have a greater effect. • Indicates that the effect of wind chill not correctly captured when the EP2 average data streams are used to derive CWV. • Also, EP2 temperature only “warming” is actually lower than EP2 CWV with average data streams in 7 (of 13) LDZs. • CWV basis (containing a wind chill term) should be less “warm” than a temperature only basis. In 7 LDZs CWVs basis is more “warm”. • Additional effect of the lag term (effective temperature) with CWV based on average data streams - may be giving a “warmer” CWV than warranted. • EP2 CWV computed from sequential actual daily data gives (as expected) lesser “warming” than EP2 temperature only. (Compare columns 2 and 4). • Suggests that direct use of the EP2 average data streams leads to inappropriate CWV values. Leads to risk of less good gas demand models.

13. Analysis of SNCWVs using EP2 data only and alternative method – WM by month

14. CWV calculated using gas industry history only – Average and Daily approach (1971 to 2006)

15. Impacts of using EP2 data only in deriving SNCVW • The impacts of using EP2 temperature increments added to averaged EP2 base period history is that the CWV will not be appropriately computed i.e. the wind chill effect will be smaller thus producing warmer CWVs than necessary. • The wind chill effect is incorrectly computed (warmer than it should be) all through the year (winter and summer) but to a proportionately greater extent in the summer. • The result is that seasonal normal based ALP profiles for EUCs will be much lower that they should be in the summer and correspondingly peakier in the winter • It is important that any ‘warming’ of the SNCWV is due to the EP2 temperature increments and not the method of calculation.

16. Proposed Approach Summary • Apply EP2-WP8 temperature increments to individual years of gas industry data (36 years) to get 36 different incremented daily temperature streams for each target forecast year (e.g. 2012/13) • Windspeed data will be actual windspeeds for each of the gas industry base period days (no increments specified by EP2-WP8 for wind speed). • Compute 36 different CWVs for each future day and average to a single value. • Smooth computed CWV profile to remove excessive day-to-day variation in CWV profile BUT ensure same area under SNCWV curve and retention of similar bumps and kinks shown in the corresponding EP2-WP8 temperature profile.

17. Shipper feedback on proposed approach

18. Shipper Feedback: E.On and EDF Energy • E.On Feedback • Would like to check with Met Office that base period averages (1971 to 2006) were not used in derivation of temperature increments • Using gas industry history data removes ‘transparency’ of process • Clarification sought on calculations shown on slide 34 • Numbers used in examples misleading – slides 31, 37 and 38 • Happy that smoothing is necessary if proposed approach is to be used • EDF Energy Feedback • Happy with “95%” of proposed approach • Believe using same temperature increment across all 36 years is inappropriate • Acknowledged requirement for smoothing, however asked if a moving average approach had been considered

19. Transporters’ response to E.On Comments • Transporters’ response: • xoserve are happy that base period averages were not used to derive temperature increments. EP2-WP8 final report sections 3,4 and 5 confirm this • Further detail behind slide 34 has been provided to DESC which we hope assists in your understanding of the analysis • Monthly degree day analysis of differences between EP2 average and daily approach • Example of two CWVs for WM (calculated using gas industry data only) were also provided in order to show how ‘average’ approach results in a warmer CWV compared to ‘daily’ approach • Examples used in slides 31,37 and 38 were included in presentation primarily to show the differences between the two approaches, however in hindsight xoserve accept that the numbers used were misleading

20. Transporters’ response to EDF comments:Increments • Transporters’ response: • EP2 increments are values for each hour of the day added to hourly values of temperature • In order to derive increments for each day of 2012/13 for each of the sets of days of the 36 years (1971 to 2006) further information is required • Increments were derived first from modelling 10 UK climate districts and did not include the specific 1971-2006 average base period data • Increments were not derived from going from a baseline to a future date and so increment values appropriate from 1971 to 2012/13, from 1972 to 2012/13 etc do not exist • For each gas day (e.g. 1st April) we want to go from 36 individual gas industry history date data points to the future date (2012/13) • For simplicity and lack of additional information xoserve have proposed the increment is the same for going from each of the start points • Without additional global modelling work being done we have existing EP2 increments, gas industry daily history and EP2 averaged history as the only data available • Timescales dictate that we cannot manipulate any of the available data or indeed wait for further ‘fresh’ output from the Met Office

21. Transporters response to EDF comments:Smoothing Method • Transporters’ response: • Any smoothing method chosen should (i) retain EP2 Temperature profile and (ii) ensure the SNCWV does not incorporate any ‘warming’ or ‘cooling’ as a result of applying smoothing • xoserve considered moving averages, however this was considered not as statistically sound as the chosen method • Final smoothing method chosen – “lo-ess” method is a statistically robust method and the results from which are optimum for this purpose. • Final smoothed SNCWV profile compares well with the EP2 Temperature profile i.e. retains kinks and bumps • The smoothed and unsmoothed profiles envelope nearly identical areas (0.0002% difference) i.e smoothing has not changed the extent of annual warming

22. Demand Estimation Timetable incl. SN Review • 23rd Sept ’09 • DESC meeting to discuss application of EP2 approach • October ’09 to December ’09 • Calculate revised CWV definitions • Calculate revised SNCWVs • DESC 22nd Dec ’09: Final SNCWV values presented • DESC November 10th 2009 – London • Re-evaluation of NDM Sampling Sizes • Re-evaluation of EUC definitions & Demand Model Performance: SF & WCF • Re-evaluation of Model smoothing methodology • Review of demand attribution to EUC models newly with/without cut offs in 2008/09 • Review of consultation process for annual proposals • January ’10 to Feb’10 • Re-define all ALPs and DAFs on new basis (effectively re-running Spring analysis) • Set of revised WAALPs available for system to start calculating AQs effective from 1st October 2010 • Jan DESC: • Re-evaluation: EUC definitions & Demand Model Performance: RV & NDM sample Strands • Approach for Spring 2010 Analysis • March ’10 onwards • WAALPs used in AQ Review