A Review of Leyland and Transpower Security of Supply Projections Prepared for: Andrew Smith Manager Energy Markets Information and Services Ministry of Economic Development By Tom Halliburton 25 September, 2002 Summary and Conclusions This study reviews aspects of two reports that examine the adequacy of supply in the New Zealand electricity system. These reports are Transpower’s “System Security Forecast 200102” and a draft of “Electricity Supply and Demand to 2015” to be published by Sinclair, Knight, Merz and the Centre for Advanced Engineering (the Leyland Report). Transpower’s report gives two sets of results: An overall generation shortfall assuming a “1 in 20”1 dry year with a probability of shortfalls from 2006-07 assuming demand growth of 2.2% per annum, or from 200405 for high demand growth of 2.6% per annum. A probability of small shortfalls in supply (less than 0.5%) from 2001-02 onwards in some parts of the country. Leyland’s study, based on what he considers to be an approximately “1 in 20”2 dry year, gives significant generation shortfall from 2003-04. Both reports use years ending 30 June. A statement that shortfall might occur in 2003-04 indicates that shortfall is possible at any time from 1 July 2003 to 30 June 2004. Modeling Approaches Transpower’s two sets of results are from two different modeling approaches - the first is a simple stack model using one year time steps which compares current capacity against demand. Results are presented as figure 2.12 in its report. The second approach outlined in an appendix to its report gives results from a stochastic optimization model (OCCAM) used in conjunction with a transmission system model (SCNPD) for more detailed studies. Leyland uses a more complex stack model, also with one year time steps, implemented on a spreadsheet. This compares total demand over the year with the generation resources available, accounts for some fuel constraints, and includes the HVDC link. Rules are incorporated to estimate how individual plants would be loaded and what fuels would be used. The Leyland model is difficult to review due to its complexity, especially as data are not always separated from the model’s calculations. Drawbacks for the New Zealand system for stack models in general include inability to model the transmission grid, and no modeling of inflow variability, either within or between years (other than by scaling the year’s total inflows), or of hydro storage, or of seasonal patterns in load. Use of Hydro Data Leyland obtains mean hydro data by taking the average of actual outputs over the period 1984 - 2001, whereas better results would be obtained using the full record back to 1931. Transpower inflow data for its stack model are thought to be derived from 72 years of records, but the exact origin of some of the data is unknown within Transpower and therefore the data cannot be readily audited. Transpower believe the Waikato data may be incorrect. There are some weaknesses in inflow data used in both reports. 1 Transpower obtain the 1 in 20 dry year reduction by fitting a normal distribution to data and taking the mean minus one standard deviation, which gives a 16.6% reduction in flows. 2 Leyland considers a 1 in 20 dry year to be a 30% reduction over six months. 314055-1 2 Differences in input assumptions There are four major differences in the assumptions used by the reports reviewed: Dry year inflows - Transpower reduces total hydro energy by 4047 GWh, or 16.6% as representing the lower 5 percentile on a normal distribution. From Ministry records, the March 2002 year generation was about 3300 GWh below mean. Leyland models the effect of a 30% reduction over six months only, by applying this reduction over the full year3. This is not unreasonable given the limitations of the model, but is a much more severe condition than that used by Transpower. Plant factors - Transpower uses a New Plymouth plant factor of 95% which is clearly too high. Its assumed Huntly plant factor of 85% is perhaps reasonable for a dry period, but not for the full 12 months as it has assumed - bearing in mind that even in a dry year, much of the year will consist of apparently normal conditions. Leyland’s maximum plant factor of 66% for Huntly for a normal year is somewhat pessimistic, but he uses 90% plant factor for four months during a dry period. This performance was exceeded in 1992, and accordingly Leyland’s figure seems reasonable. Water rights - Transpower assumes no loss of water rights. Leyland assumes hydro water rights losses of 1030 GWh in a mean year, which would result in approximately 721 GWh loss in a dry year. These are rather high, at least in the short term. New plant - Transpower assumes no new plant commissions, whereas Leyland has three categories of assumptions for new plant (committed, probable and possible). Both studies used similar forecasts for total New Zealand demand, although Leyland’s forecast for South Island growth is higher and seems to be too high. For New Zealand as a whole Leyland uses growth averaging 1.72% per annum, while Transpower uses 2.01% per annum for its medium growth scenario. Its medium growth scenario is used throughout this analysis. Actual demand for the year ending 31 March 2002 was less than that for the preceding year, because the 2001 winter supply situation suppressed demand. The figures used in both reports for 2001-02 demand are consistent with the growth that would otherwise have been expected based on 2000-01 data. Summary of Output Results The reports estimate shortfalls in supply as follows: Table1: Year Shortfalls Begin Mean Year Dry Year Leyland Committed projects only 2009-10 2003-04(2) Transpower (1) No new plant 2011-12 2006-07 (1) Results from Transpower simple stack model (2) Leyland shows insignificant 4 GWh dry year shortfall 2002-03, 1609 GWh 2003-04. 3 The aim of the model is to determine when supply / demand mismatches may occur. Generally the effect of such mismatches will be more severe in winter months. Given that the model uses one year time steps, it applies inflow reductions over full years so as to avoid under representing potential low hydro situations. 314055-1 3 Table 2 below quantifies factors causing the differences in Table1 dates, for data applying to a “1 in 20” dry year using parameters applying in 2001-02. Table 2: Main Contributions to Lower Leyland Supply & Demand in a Dry Year Lower thermal plant factors Approximate hydro inflow water right loss in dry year Approximate hydro potential energy loss, other causes GWh(1) 2155 804 (2) 3057 (1) GWh differences apply to the 2001-02 year. (2) In a mean year the lower hydro inflow potential attributable to water rights loss is 1030 GWh. The above draws on results from the Transpower stack model. The results from the OCCAM and SCNPD network modeling system indicate potential supply difficulties from now on, although it is difficult to interpret these results and to compare them with those of its stack model. It is unclear whether the short term supply difficulties stem solely from transmission capacity limits or whether generation capacity limits contribute. The differences in results from the OCCAM / SCNPD system are due to its more detailed representation of the actual power system. Additional features represented include inflow uncertainty, reservoir operations, the transmission system and decision making without perfect foresight. Conclusions Transpower’s simple stack model suggests a limited security risk until 2006-07 (or 2004-05 for the high demand growth scenario). There are, however, indications that Transpower’s model is overly optimistic in respect of plant factors, and as a result it may understate the risk. Leyland’s more complex stack model shows significant security risk as early as 2003-04. Leyland does, however, make rather pessimistic assumptions about loss of water rights. Without the loss of water rights Leyland would generate a reduced shortfall in 2003-04 which disappears in 2005-06 (due to new plant commissioning) but still exists in the following years. There remain significant differences between the Transpower and Leyland models in the way they treat dry year hydro inflows. It is not possible to form an opinion as to which is a more accurate representation of reality in respect of these inflows. Therefore, it is not possible to form a firm judgment as to the real security risk situation. Both Transpower and Leyland concede the limitations of their models. Further modeling, using a more comprehensive model is needed to provide a sound assessment of the security risk. Given the modeling that has already been done and the data available, a much sounder assessment of the situation could probably be produced within a time-span of 2-3 months. 314055-1 4 Contents Objective .................................................................................................................................... 6 Modeling Methodologies and Conclusions of Reports .............................................................. 6 Demand Forecasts ...................................................................................................................... 8 Generation ................................................................................................................................ 11 Large Hydro Schemes .................................................................................................. 11 Large Thermal Stations ................................................................................................ 12 Fuel Supplies ............................................................................................................................ 14 Dry Year Shortfall Sensitivity in Leyland Report ................................................................... 15 Summary of Differences .......................................................................................................... 15 Appendix 1: Power System Issues .......................................................................................... 17 Hydro System Features ................................................................................................ 17 Defining a Drought ...................................................................................................... 18 System Operation During Droughts............................................................................. 18 Ministry of Energy Planning Criteria .......................................................................... 19 Appendix 2: Power System Models ........................................................................................ 20 The OCCAM Model .................................................................................................... 20 SPECTRA .................................................................................................................... 20 SDDP ........................................................................................................................... 21 References ................................................................................................................................ 22 About the Author ..................................................................................................................... 22 314055-1 5 Objective The objective of this report is to compare some aspects of two reports prepared to assess security of supply of electricity in New Zealand. Sinclair Knight Merz in conjunction with the Centre for Advanced Engineering is producing the sixth edition of “Electricity Supply and Demand to 2015”. A draft copy of this was made available for review, by its principal author, Bryan Leyland. Transpower has recently published its “System Security Forecast 2001/02” which has a transmission focus, but analyses the demand and supply balance on the basis of existing plant. The objective of this study is to: review the projections and highlight the main sources of difference in the assessments of security of supply risks quantify the relative effects on security of supply of these main sources comment on the future security of supply risk, in particular the next two to three years Factors to be considered include: demand growth dry inflows thermal generation plant factors potential loss of water rights Modeling Methodologies and Conclusions of Reports Both reports perform simple adequacy tests on the electricity system using stack models these compare the maximum output over a year of the various generation resources with the load. Both models use a one year time step. They do not attempt to simulate operation of the system in performing this analysis. For a dry year, some measure of low inflows is applied to reduce the output of hydro plants, while an assumed maximum utilization factor for thermal stations takes account of the need for maintenance, breakdowns and the effect of demand patterns on the ability of the system to absorb that station’s generation. Neither analysis is a capable of considering the effect of: annual load pattern - greater energy consumption in winter seasonal inflow variability the decision making process - real decisions are made with imperfect foresight initial storage levels in lakes random plant failures transmission constraints - (Leyland considers HVDC constraints and losses) After presenting its overall view of generation adequacy using as above using a stack model, Transpower perform further studies using the OCCAM model. It is described in Appendix C of the System Security Forecast and some model details are given in Appendix 2 of this report. OCCAM takes some account of inflow variability (using high, mean and low inflow scenarios), reservoir operation and transmission constraints. OCCAM uses weekly time steps. In conjunction with a detailed transmission system model, SCNPD, it is used to generate plots of average prices and unserved energy for various regions. These are 314055-1 6 presented in Appendix D of the System Security Forecast. Some examples of more detailed power system models are given in Appendix 2 of this report. The following table shows when generation shortfalls first occur in the Leyland analysis: Year in Which Demand Exceeds Supply Leyland Model Scenario Mean Year Dry Year Water rights New Plant loss? Yes Committed 835 GWh, 2009-10 1609 GWh, 2003-04 No Committed 1218 GWh, 2011-12 805 GWh, 2003-04 Yes All options (1) 821 GWh, 2019-20 291 GWh, 2003-04 (1) No All options 312 GWh, 20019-20 1141 GWh, 2014-15 (1) “All options” implies the commissioning of committed, probable, and possible stations. (2) Shortfalls do not exceed 622 GWh until 20014-15, and are zero in several years before then. Figure 2.12 in Transpower’s report gives a less pessimistic view from its stack model of overall system adequacy than the Leyland Report. The table below presents results for Transpower’s medium and high growth scenarios. The medium growth scenario is used throughout this report unless otherwise stated. Year in Which Demand Exceeds Supply Transpower Stack Model Inflows Load Growth Scenario Mean Year Dry year Medium 2011-12 2006-07 High 2009-09 2004-05 In Appendix D of Transpower’s report, OCCAM model results are given. This model is used to generate offer stacks for generators which are fed into the SCNPD model which simulates transmission system operation. OCCAM shows small amounts for expected shortfalls in OCCAM Model Results Year Shortfalls Region Begin North Isthmus 2001-02 Auckland 2008-09 Waikato 2001-02 Bay of Plenty 2001-02 Central North Island 2009-10 Taranaki 2008-09 Hawke’s Bay 2006-07 Wellington Nelson2006-07 Marlborough West Coast 2007-08 Canterbury 2001-02 South Canterbury 2008-09 Southland - Otago - 314055-1 7 Energy supply, beginning in 2001-02 in four of the thirteen regions with most others following at later dates. It is not known whether these shortfalls are due only to transmission system effects, or whether generation shortages contribute. Data for the Leyland model were adjusted until results similar to the Transpower stack model were obtained. Assumption of water rights losses was removed, and Huntly’s maximum plant factor increased to 80%, with committed plant only - mean year shortfall is delayed for 3 years to 2012-13, as for Transpower Assumption of water rights losses was removed, dry year definition changed from 30% loss of inflows to 15% loss, 80% Huntly plant factor - dry year shortfall is delayed from 2003-04 until 2009-10, later than in Transpower’s model. Assumption of water rights losses was removed, dry year definition remaining at 30% loss of inflows, with committed plants, 85% Huntly plant factor, 95% New Plymouth plant factor, and operation of both these stations unconstrained by fuel supplies - dry year shortfall delayed one year to 2004-05. The additional energy due to the Manapouri second tunnel received special treatment by Leyland to account for dry year effects. He considers that the second tunnel has no benefit in a dry year when the station would be operating at low outputs. For this comparative analysis, Leyland’s special treatment of Manapouri has been ignored when apportioning differences in the models to various causes. Leyland’s model effectively shows dry year problems are unavoidable, if his definition of a dry year is accepted. He defines a dry year to have total New Zealand inflows 15% below mean, but models the effect that this would have if the low flows occurred over a six month period, with mean flows over the remainder of the year. Appendix 1 shows hydro that total hydro storage is less than 20% of total annual inflow. Hence a dry period of even less than six months can have a serious impact. Appendix 1 also shows how North and South Island inflows are not strongly correlated. A low flow period in one island alone might cause supply problems if the HVDC link becomes constrained Demand Forecasts Preparation of Transpower’s demand forecasts involved separately considering three electricity use segments: domestic, commercial and direct supply industrial. The first two are forecast using a time series model, while the third is treated as a constant, with some likely growth increments added. Growth is allocated to regions in proportion to forecast population changes. The forecast represents generation, and so includes transmission system losses. No specific allowance is made for HVDC transmission losses. As this is a function of generation, it can not be predicted at the forecasting stage. Transpower bases its forecasts on the historical metered injections data, so it does not include embedded generation that does not enter the grid. These small embedded generators simply reduce off-takes from the grid. To produce the table below, embedded generation estimates from the Leyland report have been added to Transpower’s data. The addition comprises Leyland’s total cogeneration output minus those cogeneration resources specifically itemized in the Transpower report. 314055-1 8 This effectively adds to the forecast those cogeneration resources treated by Transpower as embedded. The Leyland report specifically forecasts load in each island, but excludes demand met by cogeneration resources, adds 5% for AC transmission system losses, determines where generation will be provided, and then calculates HVDC transfer. 9% loss is assigned to the HVDC transfers, and finally North Island cogeneration plants outputs are added to get the total generation requirement. As can be seen from the tables below, the two reports use generally similar forecasts for total New Zealand demand, when Transpower’s mean growth scenario is considered. Leyland forecasts significantly more growth in the South Island. This higher growth must be questioned because a significant part of the South Island load is accounted for by the Tiwai aluminium smelter, so for the remaining smaller base the growth is effectively occurring at an even higher rate than Leyland’s 2%. In a mean hydrological year, the higher South Island growth results in lower HVDC losses, but in a dry year, it could cause problems through constrained southward transfer on the HVDC link. Forecast Demand, GWh - Including Transmission Losses Leyland (1) Transpower (Medium Scenario) 2001-02 2002-03 2003-04 2004-05 2005-06 2006-07 2007-08 2008-09 2009-10 2010-11 South Island North Island New Zealand South Island North Island New Zealand 14322 14608 14900 15198 15502 15812 16129 16451 16780 17116 24225 24604 24990 25370 25769 26175 26588 27007 27434 27867 38843 39484 40119 40771 41432 42122 42853 43711 44501 45313 13923 14120 14323 14542 14758 14948 15132 15305 15474 15638 25477 26142 26826 27556 28263 28932 29579 30217 30849 31473 39400 40262 41149 42098 43021 43880 44710 45523 46322 47112 % Transpower exceeds Leyland 1.41% 1.93% 2.50% 3.15% 3.69% 4.01% 4.15% 3.98% 3.93% 3.82% (1) Leyland’s New Zealand total includes HVDC link loss. Individual island forecasts do not. Forecast Demand Growth, % - Including Transmission Losses Leyland Transpower (Medium Scenario) South North New South North New Island Island Zealand Island Island Zealand 2002-03 2003-04 2004-05 2005-06 2006-07 2007-08 2008-09 2009-10 2010-11 2.00% 2.00% 2.00% 2.00% 2.00% 2.00% 2.00% 2.00% 2.00% 1.57% 1.57% 1.52% 1.57% 1.57% 1.58% 1.58% 1.58% 1.58% 1.65% 1.61% 1.63% 1.62% 1.66% 1.74% 2.00% 1.81% 1.83% 1.42% 1.44% 1.53% 1.48% 1.29% 1.23% 1.15% 1.10% 1.06% 2.61% 2.62% 2.72% 2.57% 2.37% 2.23% 2.16% 2.09% 2.03% 2.19% 2.20% 2.31% 2.19% 2.00% 1.89% 1.82% 1.76% 1.70% Ministry data for years ending 31 March 2001 and 2002 show total generation requirements of 38646 GWh and 38148 GWh respectively. The drop for the 2002 year was due to the high prices experienced during the winter of 2001. Transpower’s 2001-02 forecast represents a 1.95% increase on the Ministry’s actual 2000-01 data. Therefore it is a reasonable estimate of the generation requirement that would have occurred in the absence of the unusually high prices produced by low inflows in the winter of 2001. 314055-1 9 Leyland and Transpower demand forecasts for 2001-02 are very similar, with Transpower 1.41% higher. Transpower’s growth forecasts for its medium scenario are higher, giving 3.82% higher demand than Leyland by 2010-11. Demand forecasts are not a significant factor in the different conclusions of the two reports. 314055-1 10 Generation Transpower’s analysis assumes no new generation, to illustrate the need for new plant. The analysis of Leyland’s report presented here uses the option giving the largest amount of new plant - committed, probable, and possible options are all included. Large Hydro Schemes Leyland obtains mean hydro data by taking the average of actual outputs over the period 1984 - 2001. Transpower data is from an unknown source - data used for a previous study was simply re-used. It is thought to represent a seventy two year average. Transpower staff believe the Waikato figure to be too low, and Waitaki to be too high, but the New Zealand total to be reasonable. Comparison of Mean Hydro GWh Aniwhenua Mangahao Matahina Patea Tongariro Scheme Waikaremoana Waikato River Wheao North Island total 2000-01 1931 - 1989 Leyland Transpower CAE 1993 (1) 126.5 146 123.9 130 230.9 303 311 117.5 138 1288.2 1437 2624 (2) 439.6 540 504 4608.6 3173 3173 130.4 123 7065.6 5990 Clutha Cobb Coleridge Highbank Manapouri (3) Waipori Waitaki South Island total 2673.6 190.6 243.9 82.6 4967.0 181.7 8194.2 16533.7 3905 195 295 101 5168 220 8554 18438 New Zealand total 23599 24428 4064 243 216 5062 (4) 8378 (1) This column’s data is from ref [1]. The data refers to inflows, not specifically to generation, ie this suggests that no allowance has been made for high flow periods when spill is unavoidable. (2) Tongariro resource consents may have changed this number. Improvements in modeling of diversions prior to their commencement have been made since this data was prepared. (3) Leyland and Transpower data for Manapouri includes the effect of the second tailrace tunnel throughout this report, for ease of comparisons (4) CAE data does not include the second tailrace tunnel at Manapouri. It is likely that the CAE number includes the potential energy contained in water that must be spilled from Manapouri and at the Mararoa diversion. 314055-1 11 South Island 2000-01 2001-02 2002-03 2003-04 2004-05 2005-06 2006-07 2007-08 2008-09 2009-10 2010-11 16534 16021 16038 15844 15861 15698 15714 16057 17673 18374 19190 Mean Year Hydro Generation, GWh Leyland Transpower New South New North Island North Island Zealand Island Zealand 7066 6572 6590 6602 6608 6615 6622 6628 6635 6642 6648 23599 22593 22628 22446 22469 22313 22336 22685 24309 25016 25839 18438 18438 18438 18438 18438 18438 18438 18438 18438 18438 18438 5990 5990 5990 5990 5990 5990 5990 5990 5990 5990 5990 24428 24428 24428 24428 24428 24428 24428 24428 24428 24428 24428 % Transpower exceeds Leyland 3.39% 7.51% 7.37% 8.11% 8.02% 8.66% 8.56% 7.13% 0.49% -2.41% -5.78% Leyland assumes a loss of energy when resource consents are renewed. 500 GWh per year is assumed to be lost in the North Island from 2001-02. Assumed total losses in the South Island are 530, 740 and 920 GWh beginning in years 2001-02, 2003-04 and 2005-06 respectively. This estimate is based on that given by RJ Aspden in a paper written in November 1991, revised in 1994, and with the comment made in 1996 by Aspden that he still considered this to be reasonable - now six years ago. This estimate is too old to be reliable. Outstanding consents to be renewed in the near future include Clyde, Roxburgh, the Waikato River stations, and an appeal has been lodged regarding the Tongariro scheme. In the case of the Waikato and Clutha systems, it is unlikely that there will be significant drops in the total energy output of the hydro systems as no diversions are involved. Transpower estimates of total mean hydro potential are 829 GWh (3.39%) higher than Leyland’s in 2000-01, although differences for each island are significantly larger. Transpower’s estimate becomes increasingly higher until 2006-07 due to the increasing losses in water rights assumed by Leyland. This is a contributor to the greater shortfalls reported by Leyland. Large Thermal Stations One of the most critical aspects of this type of study is the selection of maximum plant factors for thermal stations. The plant factor determines the maximum total annual output of the station, and represents the proportion of the year that the station could operate at full load. Plant factor must generally be reduced from 100% to allow for forced outages (outages caused by breakdowns), routine maintenance outages, planned overhauls, and the reduction in output occurring during startups and shutdowns. Constraints on plant operation due to fuel supplies, transmission system capacity etc would further reduce practical output capability. Due to load variation over the year, and within each day, there is not always sufficient demand to absorb the output of all thermal stations. At low load times of the year, in the early hours of the morning in particular, it may be necessary to shut down the higher cost thermal plants due to lack of load requirements, even during a dry period. 314055-1 12 Transpower and Leyland make different assumptions for Huntly and new Plymouth plant factors: Comparison of Plant Factors (%) Huntly mean year Huntly dry year New Plymouth Stratford Power CC Otahuhu A Otahuhu B Southdown Leyland 66.3 82 (1) 67.0 90.0 0.0 90.8 95.0 Transpower 85.6 85.6 95.0 93.5 0.0 93.9 100.1 (1) Applies a 90% plant factor for 8 months but during a dry year only, giving an effective plant factor of 82% for the year. Transpower’s figure of 95% plant factor for a sustained period at New Plymouth is unrealistic for a station that is now approximately 27 years old and is one of the most inefficient of the large thermal stations. Furthermore, Transpower does not assume that this station decommissions, whereas Leyland assumes it closes at the end of 2008-09. Leyland assumes New Plymouth capacity drops from 400 to 300 MW in 2002-03, whereas Transpower assume it drops from 440 MW to 330 MW in 2003-04. Huntly is now approximately 18 years old, which may affect its ability to run for prolonged periods at high output. A further constraint on Huntly is imposed by restrictions on cooling water supplies from the Waikato River - at periods of low flows and higher water temperatures, the station output is limited. The possibility of prolonged operation at high loads was demonstrated over the four months of the 1992 electricity shortage when Huntly was run at plant factors exceeding 80% with an average of 96.2% from March to June. Plant Factors (%) During 1992 Shortage from ref [2] New Month Huntly Plymouth Nov 91 47.8 3.4 Dec 91 48.4 9.4 Jan 92 58.8 17 Feb 92 63.2 17.2 Mar 92 86.5 16.6 Apr 92 96.5 33.6 May 92 99.3 94.6 June 92 102.5 103.5 Comparison of Mean Maximum Annual GWh, 2001-02 Station Huntly - mean year Huntly - dry year New Plymouth Stratford Power CC Otahuhu A Otahuhu B Southdown Total - mean year Total - dry year 314055-1 Leyland Transpower % Transpower exceeds Leyland 5808 7183 2348 2791 0 2903 982 14832 16207 7500 7500 3662 2950 0 3250 1000 18362 18362 22.6 4.2 35.9 5.4 0.0 10.7 1.8 19.2 11.7 13 2001-02 2002-03 2003-04 2004-05 2005-06 2006-07 2007-08 2008-09 2009-10 2010-11 Mean Year Thermal Generation Capability, GWh % Transpower Leyland Transpower exceeds Fossil Geothermal Fossil Geothermal Leyland 14832 2442 18362 2436 16.9% 14245 2482 18362 2436 19.6% 14245 2482 17446 2436 15.9% 14245 2947 17446 2436 13.5% 17554 (1) 3342 17446 2436 -5.1% 17554 3342 17446 2436 -5.1% 17554 4235 17446 2436 -9.6% 17554 4651 17446 2436 -11.7% 15793 (2) 4851 17446 2436 -3.8% 15793 4851 17446 2436 -3.8% (1) A new gas fired thermal plant is assumed to be commissioned for 2005-06 (e.g. Huntly combined cycle). (2) Leyland assumes New Plymouth decommissions at the end of 2008-09. Transpower does not. While operation of Huntly and New Plymouth at high plant factors for periods of a few months might be possible, a full year at 85% and 95% output respectively, as assumed by Transpower, is not credible. Leyland allows Huntly to operate at a plant factor of 66% for the full year, which may be little pessimistic. However, when calculating dry year shortfalls, Leyland allows Huntly to operate for four months of the six month dry period at 90% output. The intent is to capture actual likely performance - the existence of a dry period is not likely to be clear for some time, so Huntly would not begin maximum generation immediately. The adjustment is a means of representing the lack of perfect foresight that exists in reality. This is a reasonable assumption, given the limitations of the modeling methodology. Plant factor assumptions are a major contributor to the different conclusions of the Leyland and Transpower reports. Fuel Supplies Transpower does not model any fuel constraints in its stack model, and the OCCAM model does not include these. Leyland goes to considerable lengths to model fuel supplies - this is a major concern of his. Huntly can burn gas or coal. New Plymouth can run on gas or oil. There is the possibility of optimizing where gas is burnt to maximize the total output from these two stations but the trade-off is not made explicitly in the model. Instead, Huntly is constrained until 2003 to a maximum of 1000 GWh of coal burn, and is unrestricted on coal after that. New Plymouth is able to generate up to 1000 GWh on oil at any time. The size of gas reserves data used by Leyland are similar to the Ministry’s information: 314055-1 14 Maui reserves: Leyland models reserves of 767 PJ on 1 October 2001, compared with the Ministry’s estimate of 740 PJ on 1 January 2002. Pohokura reserves: Leyland also has the same value for the Ministry (600 PJ) and uses the same production start date (2006). Leyland ramps production up in successive years, allowing offtakes of 20 PJ, 40 PJ and then 60 PJ for remaining years. The Ministry offtake estimate is 55 PJ per year. Kupe reserves: Leyland uses a total of 285 PJ, while the Ministry estimates 300 PJ. Dry Year Shortfall Sensitivity in Leyland Report A study was carried out using the Leyland model to determine the impact of various features. The base case discussed throughout this report indicates a dry year shortfall of 1609 GWh in 2003-04. If the assumption of water rights loss is removed, i.e. all water rights are retained, the 2003-04 shortfall drops to 805 GWh. If it is further assumed that New Plymouth oil firing capability is unrestricted, the shortfall drops a further 174 GWh to 631 GWh in 200304. Unlimited Maui gas and the assumption of a very high plant factor for New Plymouth drop shortfall to 158 GWh. Leyland Model - Sensitivity Study Dry Year Shortfall (GWh) 2003-04 1609 2004-05 2350 2005-06 1300 2006-07 3538 2007-08 4234 2008-09 4521 No water rights loss 805 1546 0 1789 2883 3403 Add: No New Plymouth oil restriction 631 1372 0 1615 2709 3230 Add: Unlimited Maui 158 1372 0 0 80 777 Add: 85% New Plymouth plant factor 158 899 0 0 0 304 Base case - committed plant Changes: This sensitivity study shows that to significantly reduce dry year shortfalls in the Leyland model it is necessary to make some very optimistic assumptions about the performance of plant and the availability of fuel. Summary of Differences Amounts by which Transpower Mean Year Data Exceed Leyland Data (GWh) 2001-02 2002-03 2003-04 2004-05 2005-06 2006-07 2007-08 2008-09 2009-10 2010-11 314055-1 Load Hydro Inflow 261 506 802 1124 1429 1625 1721 1558 1535 1469 1835 1800 1982 1958 2115 2092 1743 119 -588 -1411 Thermal Maximum Geothermal Production 3556 4143 3227 3227 -82 -82 -82 -82 1679 1679 -6 -46 -46 -511 -906 -906 -1799 -2215 -2415 -2415 Cogeneration 287 287 287 47 47 47 47 47 47 47 Amount Transpower Alternative Surplus Energy Exceeds Generation Leyland (GWh) 115 115 185 185 227 377 377 377 403 403 5526 5793 4833 3782 -27 -96 -1435 -3312 -2409 -3166 15 Amounts by which Transpower Dry Year Data Exceed Leyland Data (GWh) 2001-02 2002-03 2003-04 2004-05 2005-06 2006-07 2007-08 2008-09 2009-10 2010-11 Load Hydro Inflow (Approx.) 261 506 802 1124 1429 1625 1721 1558 1535 1469 3861 4566 4542 4669 4652 4762 4746 4501 3365 2870 Thermal Maximum Geothermal Production 2155 2742 1826 1826 -1483 -1483 -1483 -1483 277 277 -6 -46 -46 -511 -906 -906 -1799 -2215 -2415 -2415 Cogeneration 287 287 287 47 47 47 47 47 47 47 Amount Transpower Alternative Surplus Energy Exceeds Generation Leyland (GWh) 115 115 185 185 227 377 377 377 403 403 6151 7157 5991 5091 1109 1172 166 -331 142 -286 Note: Above tables include commissioning of Leyland’s committed, probable and possible plants. The first table above shows how the amount of extra energy assumed by Transpower declines over time as Leyland’s model commissions additional plant. The dry year table above differs in two respects from the mean year table. Leyland assumes additional output from Huntly during a dry year, and he represents the dry year differently. The dry year inflow data above compares Transpower’s 15% annual reduction with Leyland’s 30%. This is a device used by Leyland to capture the impact of a six month mean period followed by a six month dry spell in a single year, and is necessary because the modeling technique allows only one year time steps to be used. Key differences are hydro inflows, the definition of dry year inflows, thermal station plant factors and the commissioning of new stations. This last factor can be discounted as Transpower’s study is intentionally for existing plant only. Leyland deducts 1030 GWh from hydro potential from 2001-02 due to loss of water rights which is pessimistic. Transpower’s study is optimistic in using a plant factor of 95% for New Plymouth, and its 85% plant factor for Huntly is realistic only for a few months of a shortage period, not for a full year. The sensitivity study carried out using the Leyland model shows that the occurrence of shortfall in 2003-04 in the event of low hydro system inflows is not sensitive to a wide range of parameters. The quantity of shortfall varies, but it was not completely eliminated by making a number of more optimistic assumptions for plant performance, fuel availability and water rights renewals. The critical aspect that cannot be resolved by using either of these stack models is the true nature of a dry period. A model able to capture the details of hydrology and reservoir management is essential to advance understanding beyond that possible from either of these two models. 314055-1 16 Appendix 1: Power System Issues Hydro System Features Hydro storage capacity in the New Zealand system is small, apart from Lake Pukaki. The diagram below from [1] illustrates hydro storage as a percentage of inflows, in energy terms: Storage as Percentage of Total Inflows Mean inflows = 100% 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% North Island South Island New Zealand Bottom bar represents amount of storage The small proportion of inflows that can be stored means that the system is relatively more vulnerable to short duration droughts than other systems with larger storage. Ni vs SI Inflow Variances 60.00% 50.00% SI High / NI High SI High / NI Low 40.00% South Island Inflow Variance 30.00% 20.00% 10.00% 0.00% -10.00% -20.00% -30.00% SI -30.00% S I Low / NI High Low / NI Low -20.00% -10.00% 0.00% 10.00% 20.00% 30.00% 40.00% North Island Inflow Variance On the other hand the lack of any strong correlation between North and South Island inflows is an advantage, as shown in another diagram taken from [1]. When low flows occur in the 314055-1 17 South Island, there is not an especially high probability of low inflows in the North. Hence the probability of the extreme situation of inflows being 30% below mean in both islands over a six month period may be lower than assumed in Leyland’s report. Defining a Drought Leyland defines a low inflow event as a year with inflows 30% below mean for six months, with mean inflows for the balance of the year. The intention is to test the effect of a six month dry period, followed by six months of mean inflows, giving a 15% reduction over the full year. Because the model only works in one year time steps, Leyland has to apply the reduced inflows over the full year to test the ability of the system to handle the six month drought. Transpower makes only the 16.6% reduction over the full year. The most severe period of low inflows into the hydro system in recent history is the 1992 event. For the eight month period November 1991 to June 1992, inflows as a percentage of mean were [2]: North Island 91% South Island 71% New Zealand 76% System Operation During Droughts With the benefit of hindsight the 1991-92 drought was clearly a very extreme event, but this was not apparent until its later stages. Total flows for the New Zealand system, month by month as percentage of mean were as follows [2]: October 1991 November December January 1992 February March April May June July 107% 76% 72% 94% 95% 79% 66% 53% 61% 112% The consequences of a lack of perfect foresight may be to delay operating all available plant at maximum capacity. With hindsight this delay would clearly not have been the optimal strategy. For example, during the 1992 drought Huntly did not begin running on all four units until April 14th, and on July 15th one unit was taken out of service for repairs. With perfect foresight, all four Huntly units would have been fully loaded months earlier. To more accurately predict the consequences of low inflows, the decisions made by system managers must be modeled using only the information that those managers have available at the time. This condition is sometimes known as non-anticipative - the model must not anticipate future events. 314055-1 18 A stochastic optimization model provides a realistic means of representing decision making by risk neutral decision makers with access to all information. This type of model is aware of the statistical behaviour of the inflows, and so as lake levels drop to lower levels than normal for the time of year is likely to increase the dispatch of thermal plant. In 1992 ECNZ planners were using a model of this type. Ministry of Energy Planning Criteria In the absence of reliable system simulation models, an adequacy of supply standard was applied by the Ministry of Energy when carrying out long term generation system expansion planning. Least cost analysis was performed to select the best way of meeting this adequacy standard. See ref [1]. The standard ensured that planned generating capacity was 7 percent above the central load forecast in a dry year. A dry year was defined as having hydro inflows 85 percent of mean. The 7% margin was intended to ensure that even in a dry year supply could be maintained in the face of delays to the commissioning of new plant, unexpected equipment breakdowns, higher than expected load growth, etc. The total annual demand only was considered - the pattern within a year was not considered, so it is quite similar to the Transpower and Leyland methods. For the margins calculation the maximum possible generation at each thermal station over the year was restricted according to fuel type: diesel oil fired stations: 5% plant factor heavy fuel oil fired stations: 15% plant factor Stratford on Kapuni gas: 50% plant factor other thermal stations: 70% plant factor HVDC link: 80% plant factor where: Plant factor total annual generation per year installed MW times hours in year The margin calculation was as follows: i. South Island dry year generation = total South Island mean year potential generation times 0.85 ii. HVDC transfer = minimum of (HVDC capacity times hours in year times 0.8, or South Island dry year generation - South Island Demand). iii. North Island dry year generation = total North Island mean year hydro potential generation times 0.85 + thermal station installed MW, times plant factor appropriate to fuel type times hours in year iv. NI DryYear Generation HVDC Transfer NI Demand Margin NZ Demand 314055-1 19 Appendix 2: Power System Models Most optimal planning models have cost minimization as their objective. The prices for electricity predicted by these models are not applicable to a market situation where prices are determined by the offer strategies of the market participants. However the short run marginal prices from the optimization provide a floor for market prices, in most cases. The quantity of electricity available predicted by these models will be at least that actually available in a market consisting of rational participants. The global optimization model has all information available and so can make the truly optimal decisions, whereas market participants only have a subset of that information available. A brief summary of some stochastic power planning models follows. All these models attempt to model inflow variability, reservoir operations, seasonal load variations, make nonanticipative decisions, and make optimal hydro-thermal trade-offs. OCCAM and SDDP model the transmission system. These aspects can not be handled by stack models such as those used by Transpower or Leyland. The OCCAM Model The OCCAM model is referred to in Appendix C of Transpower’s report. It is a broad overview model of the New Zealand system, designed to capture many of the essential features without going into great detail in any of them. It models six hydro reservoirs with three possible scenarios for inflows (wet, normal and dry), thermal power stations and major components of the transmission system. The default transmission system consists of 18 nodes and 22 branches. It uses weekly time steps with typically three load duration blocks per week. No fuel supply constraints are modeled. It and is practical to solve for about a three year time horizon - solve time becomes rather long beyond this. The solution process involves repeated solves of a linear program. SPECTRA This is a two reservoir stochastic dynamic programming model designed for medium term system operational planning and for long term system expansion studies and is designed specifically for the New Zealand system. Its origins are the PRISM model developed by the Ministry of Energy in the early 1980s. SPECTRA amalgamates all the hydro stations in each island into one equivalent station and storage reservoir and links them with the HVDC. It uses all the historical data to carry out a stochastic optimization to determine water values. A detailed simulation phase models each major hydro catchment separately, using rules to derive water values for each lake from the island total calculated in the optimization phase. It has a limited ability to model fuel supply constraints. 314055-1 20 SDDP Power Systems Research Inc of Rio de Janeiro developed the SDDP model about twelve years ago for a World Bank project to model Central American power systems. Since then the model has undergone continual improvement. It is a multi reservoir, hydro-thermal optimization, with transmission system constraints and losses, with fuel constraints on thermal power stations. This is the most detailed model available for long term studies of systems such as New Zealand’s, and it has been used to analyse the power systems of many countries. The model was tested on the New Zealand system in 1995 by ECNZ with good results, but was not pursued as the effort involved in implementing a new cost based global optimization was not justified as ECNZ moved into a new market environment. 314055-1 21 References [1] “Reliability of Electricity Supply - Project Report”, Centre for Advanced Engineering, Dec 1993. [2] “The Electricity Shortage 1992”, Report of the Electricity Shortage Review Committee 1992, Dec 1992 About the Author Dr Tom Halliburton holds the degrees of BE (Hons) and PhD in Electrical Engineering from the University of Canterbury, and is a New Zealand Registered Engineer. His PhD was on the subject of scheduling hydro-thermal power systems. He has spent most of his career in the power industry, concentrating on power system optimization model development and long term power planning studies. He has also worked in hydro power station operations and maintenance, and high voltage substation design. From 1996 until late in 2001, Tom worked in the United States, gaining experience in a variety of electricity markets, and analytical and modeling techniques. Initially he worked on model development at the US Department of Energy (Bonneville Power Administration) in Vancouver, Washington. Then he moved to San Francisco where he developed a number of models for hydro generation planning, scheduling and bidding at the Pacific Gas and Electric Company. Before returning to New Zealand, Tom spent two years at Enron in Houston, Texas, where he was responsible for power system and electricity market analysis in a number of countries. As a member of Enron’s Research Group, he also developed optimization based decision aid tools for natural gas traders, gas storage valuation models and models for power marketing transactions. He now works as a consultant specializing in power system modeling and analysis. Contact details: Tom Halliburton 95 Wyndham Road Upper Hutt Email: Tel: Fax: 314055-1 tom.halliburton@attglobal.net 64 4 972 9138 64 4 972 9139 22