The performance of wind farms Evidence from the UK, Denmark and the US Professor Gordon Hughes University of Edinburgh 28 th January 2013
Background � Wind power is a capital intensive technology � Projections of load factors are critical for any assessment of costs and future investment � Difficulty of interpreting raw data � Systematic variations in wind conditions � Variations across sites & operational regimes � Changes in composition of turbine types, ages, etc � Lack of published evidence on performance of wind farms over time and by location 2
Texas load factors - sparkline graphs 55578 55579 55581 55747 62 0 55795 55796 55968 55992 Load factor (%) 62 0 56111 56211 56212 56225 62 0 0 50 100 150 0 50 100 150 0 50 100 150 0 50 100 150 Months since Jan 2000 Graphs by plant_id
Error components specification
Period fixed effects vs wind speeds
Statistical methods � Standard panel fixed effects estimation used with robust standard errors � Consistent estimates of coefficients under quite weak assumptions � Standard errors consistent if errors are heteroskedastic or serially correlated � Cross-check using bootstrap standard errors � No assumptions required about the distribution of the errors
Analysis of UK onshore wind farms � Analysis of ROC monthly output for 2002-12 � Strong incentive to report reliable data � Standard panel fixed effects models allowing for age, period (wind availability), location � Multiplicative (log-linear) specification for load factor � Unit fixed effects capture site & location effects � Total of 296 separate reporting units analysed with total installed capacity of 4200 MW � Extensive testing to identify additional factors which influence performance 7
United Kingdom onshore : quadratic vs age effects in additive model 8
United Kingdom onshore : quadratic vs age effects in multiplicative model 9
United Kingdom onshore : equal vs capacity weights in additive model 10
United Kingdom onshore : equal vs capacity weights in multiplicative model 11
Analysis of Danish wind farms � Data reported to Danish Energy Agency � Nominally for each turbine but aggregated for sites linked to a single meter � Data summed to wind farms defined by location, turbine type, installation date, etc � Onshore wind farms in Denmark are smaller and older than in the UK � 823 wind farms commissioned in 1992 or later with total capacity of 2570 MW � Limited sample of offshore wind farms � all in shallow water & most constructed in 2002 or later � 30 wind farms with total capacity of 860 MW 12
Denmark: onshore wind farms 13
Denmark: offshore wind farms 14
Analysis of US wind farms � Data from Energy Information Agency (EIA) � Annual figures for plant characteristics � Monthly data for output – reported either monthly (largest plants) or annually � Data cleaning � Inconsistencies over time in reported capacity, etc � Sample of 673 plants over US � Initial analysis focuses on Texas [92 plants, 11.2 GW] and Minnesota [121 plants, 2.6 GW] � Extended to WSC (TX & OK – 13.6 GW), WNC (MN, IA, KS, ND, SD & NE – 12.3 GW) census regions covering about 50% of wind farms in US 15
Texas: equal vs capacity weights in additive model 16
Texas: equal vs capacity weights in multiplicative model 17
Minnesota: equal vs capacity weights in additive model 18
Minnesota: equal vs capacity weights in multiplicative model 19
UK wind farms – age distribution 20 15 Wind farm age 10 5 0 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 20
UK wind farms – capacity & turbine size 21
UK – additive model by turbine size 22
UK – multiplicative model by turbine size 23
UK – additive model by country 24
UK – wind & time trends Additive model Multiplicative model Full Wind & Time Full Wind & Time Age -1.159*** -1.719*** -0.0666*** -0.0861*** (0.257) (0.339) (0.015) (0.018) Age * medium 0.513** 0.659*** 0.0412*** 0.0422*** (0.227) (0.238) (0.015) (0.015) Age * large 1.288*** 1.515*** 0.0689*** 0.0735*** (0.252) (0.269) (0.016) (0.016) Age^2 0.00542 0.00417 0.00033 0.000307 (0.007) (0.008) (0.001) (0.001) Age^2 * medium -0.0321 -0.0385* -0.00250** -0.00251** (0.023) (0.021) (0.001) (0.001) Age^2 * large -0.125*** -0.126*** -0.00502** -0.00512** (0.035) (0.038) (0.002) (0.002) Wind 5.191*** 0.204*** (0.075) (0.003) Month 0.113*** 0.00529*** (0.026) (0.001) Constant 29.70*** -24.27*** 3.413*** 1.140*** (1.320) (1.552) (0.065) (0.067) R-squared 0.658 0.579 0.560 0.460 Number of groups 291 291 291 291 Robust standard errors in parentheses 25 *** p<0.01, ** p<0.05, * p<0.1
UK – trend in additive period effects
UK – trend in multiplicative period effects
Denmark – trend in period effects
Texas – trend in additive period effects
Minnesota – trend in additive period effects
UK : Median values of site effects by year Equal weights Capacity weights Additive Multiplicative Additive Multiplicative 2000 -0.61 0.06 9.55 0.71 2001 2.07 0.18 10.40 0.68 2002 3.83 0.21 11.46 0.69 2003 3.44 0.14 10.31 0.56 2004 -5.48 -0.34 -2.11 -0.04 2005 -0.87 -0.04 2.56 0.17 2006 -3.74 -0.17 -2.87 -0.15 2007 -2.36 -0.13 -2.05 -0.18 2008 -6.47 -0.33 -8.20 -0.49 2009 -3.87 -0.20 -6.00 -0.43 2010 -4.90 -0.25 -7.51 -0.53 2011 -8.14 -0.37 -12.07 -0.71 Average -3.40 -0.18 -3.63 -0.26 31
UK : Determinants of site effects Additive model Multiplicative model (1) (2) (3) (4) Northern Ireland 4.683*** 4.520*** 0.200*** 0.195*** (0.999) (1.050) (0.047) (0.047) Scotland 5.259*** 6.554*** 0.202*** 0.241*** (0.910) (1.587) (0.038) (0.054) Wales 2.725*** 2.865*** 0.133*** 0.137*** (0.924) (0.955) (0.041) (0.044) Capacity in MW -0.0551*** -0.0537*** -0.00225*** -0.00221*** (0.017) (0.016) (0.001) (0.001) Year - 2000 -0.836*** -0.780*** -0.0479*** -0.0461*** (0.064) (0.069) (0.003) (0.003) (Year - 2000) * Scotland -0.219 -0.00664 (0.182) (0.006) Constant 0.817 0.552 0.0803*** 0.0723** (0.564) (0.588) (0.030) (0.034) Observations 291 291 291 291 R-squared 0.421 0.425 0.527 0.529 Bootstrap standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 32
Performance degradation and output 33
Age effects and project load factors UK & DK onshore experience 34
Implications for energy policy 1 � Economic life of wind farms is no more than 15 years � Many onshore wind farms re-powered after 10-12 years � After 10 years the residual value of turbines is low, but there is an option value for site redevelopment � Costs of meeting renewable targets much higher than current forecasts suggest due to � Higher capacity required due to lower average load factors � Shorter operating lives implies higher replacement costs � Implications for financing investments � Not attractive to many infrastructure investors � Higher cost of capital due to uncertainty about length of investment return and residual values 35
Implications for energy policy 2 � Impact of load factors on levelised costs � Increase from £86 to £183 per MWh for onshore wind and from £128 to £218 per MWh for R3 offshore wind � Market prices and/or subsidies required to fund renewable energy targets � DECC/CCC analysis based upon erroneous assumptions � Allowance of £8-10 per MWh for price differential � Total subsidy required in range £115-145 per MWh without any allowance for integration costs � Extra cost of offshore wind not as large as often thought 36
Levelised costs : assumed vs actual performance 37
Market prices by fuel type 38
Differences between UK/TX & DK/MN � Average turbine size and scale of wind farms � Wake effects & maintenance regimes � Community ownership: capital vs operating costs � Resistance to larger turbines � Incentives for land use � Planning restrictions: upland sites and greater density of turbines � Treating wind turbines as short rotation forests � Link to cost of capital 39
Conclusions � Improving the design and location of turbines � More detailed analysis of existing data � Longer term monitoring of installations � Implications for maintenance regimes � Lessons for lease contracts � Who should take performance risk? � Financing investment in wind energy � Structuring subsidies � Impact on financing new projects 40
Recommend
More recommend