- Email: [email protected]
Annual and seasonal variations in mean wind speed and wind turbine energy production
Solar EnergT,"Vol. 45, No, 5. lap. 285-289. 1990 Printed in the U.S.A.
0038--092X/90 $3.00 + .00 Copyright ~ 1990 Pergamon Press pie
ANNUAL AND SEASONAL VARIATIONS IN MEAN WIND SPEED AND WIND TURBINE ENERGY PRODUCTION R. W. BAKER. S. N. WALKER, and J. E. WADE Department of Mechanical Engineering, Energy Resources Research Laboratory', Oregon State University, Corvallis, OR 97331, U.S.A. Abstract--A significant uncertainty in the estimate of energy from a wind plant is the interannual and interseasonal variation in energy output, These estimates are typically based on wind data that have been collected on-site for one or more years. Rarely have the wind data records exceeded three years. Variations in the mean annual wind speeds have been assumed to be _ 10% of the long-term mean wind speed with a 90% confidence interval. This assumption, however, is based on previous work using long-term airport wind data and it has not been verified at high wind locations. In addition, most of the strong wind sites that have been developed for wind energy display distinct seasonal variations in the wind speed. In the AItamont and San Gorgonio Passes about 80% of the energy is produced during the spring and summer months. If the winds are abnormally weak during any of these months (particularly May-August). the annual energy production can be significantly reduced. Similarly, stronger than normal winds during this time period can greatly enhance the annual production. Over l0 years of hourly wind data have been collected at several moderate-to-high wind sites in the Pacific Northwest for the Bonneville Power Administration. This data base is used to determine the yearto-year and the interseasonal variations in the mean speeds and in wind turbine energy production. Wind turbine power curves are used with the speed frequency' distributions to produce seasonal and annual energy estimates.
1. I N T R O D U C T I O N
2. Y E A R - T O - Y E A R V A R I A T I O N S IN M E A N
Several years of hourly wind data have been collected at many windy sites in the Pacific Northwest. This wind monitoring work to assess the wind power potential of the region has been done for the Bonneville Power Administration (BPA). The wind data collected from three of the BPA wind monitoring locations will be discussed in this paper. These sites are Sevenmile Hill located on a 600-m ( M S L ) ridge about 100 km east of Portland, Oregon, Kennewick which is situated on a 700-m ridgeline about 300 km east of Portland, and Cape Blanco located on a 60-m coastal bluff about 100 km north of the California/Oregon border. The data recovery, rates over the 10- to 12-year data period have varied from 90% at Sevenmile Hill to 83% at Cape Blanco. Data have been analyzed through the year 1988 and all three sites are still in operation. The long and mostly continuous data base at these three sites provides an opportunity to inspect the interannual as well as the interseasonal variations in the mean speed and speed frequency distributions. The latter is best measured by transposing wind speed to power output using a turbine power curve and then by analyzing the yearly or seasonal variations in the estimated turbine energy production. A power curve o f a 250-kW machine with a cut-in speed of 4 m/sec, a rated speed of 13 m/sec, and a cut-out speed of 24 m / s e c was used to estimate energy production. Along with a discussion of the year-to-year variations in the mean annual speed (energy) and seasonal speeds (energy), the paper also addresses the relationship of changes in the mean annual wind speed to the changes in the estimated turbine energy production. 285
ANNUAL WIND SPEED AND ENERGY
At most wind energy sites or potential developments, a long period of continuous wind data on the order of 10 to 20 years duration or greater are not available. Information on the interannual and interseasonal variations in the strength of the wind is needed, however, to predict likely variations (i.e., revenue variations) in turbine production. Work done previously by Corotis[l] and J ustus et al. [2] on long periods of data from locations of the National Weather Service ( N W S ) has indicated relatively small variations in the mean annual wind speed. Their research indicates that there is 90% confidence that the mean annual wind speed from a single year will be within +_10% of the true long-term mean speed. This assumption has been used by the wind industry, Unfortunately, the correlation between the wind speeds at NWS locations and high wind sites is poor and, therefore, there is still a high degree of uncertainty in estimating interannual variations without a long period of site data to inspect. Although the l0 to 12 years of wind data collected at the three BPA sites does not represent a climatological long record (preferably 20 to 30 years), it does give an opportunity to investigate at the high wind sites interannual variations in wind speed and turbine energy production and to discuss some of the reasons for the variations. The yearly mean speeds and estimated turbine energy production at the three sites show different patterns with Cape Blanco having the largest variations and Kennewick the smallest. The annual statistics are listed in Table I. For the 12 years of record, the mean speed at Kennewick was 7.0 m / s e c and the m a x i m u m and
R. W. BAKER,S. N. WALKER.and J. E. WADE
286
Table 1. Annual statistics for the three sites (the subscript "a'" is for annual)
Wind speed (m/sec) Mean (~'0) Standard deviation (aa)
odvo High Low Years of data Estimated turbine energy (MWH) Mean Standard deviation High Low
Sevenmile Hill
Kennewick
Cape Blanco
7.5 0.6 .08 8.2 (+ 10%) 6.4 (-14%) 10
7.0 0.3 .04 7.5 (+7%) 6.3 (-9%) 12
8.6 0.9 . I0 10.0 (+ 16%) 6.7 (-22%) 12
691.4 82.3 811.6 (+17%) 532.0 (-23%)
minimum values were within ___10% of this mean speed. The highest turbine energy year, however, was 19% above the 12-year mean value and the lowest energy year was 26% below the mean value. For Sevenmile Hill, the highest mean speed year was only 10% above the long-term mean, but the lowest speed year was 14% lower than the mean. Only one of the 10 years of record fell outside ___10%of the 10-year mean speed. With respect to turbine energy, the highest year was 17% above the mean and the lowest year was 23% below the mean. At Cape Blanco, the largest variations occurred. The highest wind year at 10.0 mtsec exceeded the 12-year mean speed of 8.6 m/sec by 16% and the lowest wind year at 6.7 m/see was 22% lower than the mean. The energy differences were +23% and -31%. The mean annual wind speeds for 9 o f th e 12 years were outside the bounds of---10% of the 12-year mean.
592.5 74.8 707.2 (+19%) 438.6 (-26%)
778.7 103.2 960.2 (+23%) 538.3 (-31%)
J ustus et aL [2] found for regions comprised of several National Weather Service (NWS) sites the annual values of cra/~, varied from 0.05-0.07 lbr sites in the continental U.S. ( 10 to 18 years of data). The values ofaa/~a shown in Table 1 vary from 0.04 at Kennewick to 0.10 at Cape Blanco. The latter is significantly higher than those measured at the NWS sites. Using the 10 to 12 years of data, an analysis was done to determine the long-term mean speed using the Student's T distribution. For a 90% confidence limit the true mean speed lies between 6.8 and 7.1 m/see at Kennewick. 7.1 and 7.8 m/sec at Sevenmile Hill, and 8.1 and 9.0 m/sec at Cape Blanco. The annual turbine energy production at the sites is directly related to the shape of the site wind speed frequency distribution. The annual wind speed distributions for the three sites indicate that each distribution has a different shape as shown in Fig. 1. The Kennewick
300
rn
SEVENMILE
)¢
KENNEWICK C. BLANCO
250
200 ~3 V
',4
z LIJ
n," IJ.J
Or2 LIJ rw LI.
0 I:L
150
100
50
10
20 WIND
SPEED
30
O, 4O
(M/S)
Fig. 1. Annual speed frequency distributions and the 250 kW turbine power curve.
Variations in mean wind speed and wind turbine energy production distribution displays a definite single frequency peak at about 4.5 m/see and the frequency drops off steadily for winds above 6.7 m/see. In contrast, the distributions at Sevenmile Hill and Cape Blanco have a much broader shape and there are two relative frequency maximums at Sevenmile Hill (3.5 m/see and 9 m / see). At Cape Bianco there is a high frequency of winds above 18 m/see. Year-to-year variations in the wind speed frequency distributions at Sevenmile Hill are shown in Fig. 2 for 1987, the lowest wind year, and 1985, one of the highest wind years recorded at the site. In 1987 the estimated energy production was 532 MWH (~ = 6.4 m/see) and in 1985 it was 769 MWH (~ = 8.0 m/see), a 45% difference in energy and a 25% difference in mean speed. Both distributions have a relatively similar shape (bi-modal maximum), but there was a much higher frequency of wind speeds above I 1 m/sec in 1985 than in 1987. Further analysis of wind speed versus energy production reveals that each site has a different relationship. Plots of energy production vs. mean annual wind speed are shown in Fig. 3. A linear fit approximation was made to the data points. The most scatter in the data was at Kennewick and the least at Sevenmile Hill. The slope of the Kennewick curve indicates about a 25% change in the energy production with a 10% change in mean annual wind speed. At Sevenmile Hill there is about a 20% change in energy with a 10% change in mean speed, and at Cape Blanco there is only about a 13% change in energy for a 10% change in the mean speed. For a given turbine power curve these variations are caused by the shape of the speed frequency distributions. The three sites display different distributions, as shown in Fig. 1. With a turbine power curve like that shown in Fig. 1, which is typical of most machines, the total energy is dependent on the amount
287
of time the winds are between cut-in and rated speed and between rated and cut-out. The power output increases significantly with wind speed between cut-in and rated speeds but there is no power increase at rated to cut-out speed. As the speed frequency distribution becomes more skewed to higher speeds more rated power is produced. Powell ( 1981 ) [ 3 ] as well as others have done similar analysis of energy production and mean wind speed assuming a Rayleigh speed frequency distribution. As the rated speed to mean speed ratio decreases the energy production increases significantly. Given the mean annual wind speeds for the three locations, annual turbine energy output was calculated assuming a Rayleigh distribution. This curve is included in Fig. 3. As shown, for Sevenmile Hill and Kennewick the Rayleigh distribution underpredicts the annual energy output. There is about a 15% underestimate for Sevenmile Hill and about an 8% underestimate for Kennewick. In contrast, for the higher speed site, Cape Blanco, there was about a 10% overestimate assuming a Rayleigh distribution. In addition, the speed frequency distribution shapes varied at all three sites on an annual basis, so year-to-year differences between the actual and Rayleigh estimates varied considerably. These differences ranged from - 13% to -22% at Sevenmile Hill, -1% to -15% at Kennewick, and -1% to -,26% at Cape Blanco. It is not uncommon for high wind areas to display distinct seasonal patterns in wind speed and direction. Two of the three California wind farm development areas, Altamont Pass and San Gorgonio Pass. have the strongest winds during the spring and summer months. These periods provide an ideal match to load demand and on-peak/off-peak buyback rates. Due to its higher elevation in the "l'ehachapi Mountains and its exposure to storm winds, the "l-ehachapi area displays good spring
1985 × 4.'
~'3-
V
z LIJ :::::)2Or
LI.I
E
0 ,5
10
15
20
25
WIND SPEED ( M / S ) Fig. 2. lnterannual variations in the wind speed frequency distributions at Sevenmile Hill.
1987
R.W. BAKER,S. N. WALKER,and J. E. WADE
288
rn
A: SEVENMILE HILL
x
B: KENNEWICK
v
C: CAPE BLANCO
=1,
D: RAYLEIGH
V
Z 0
gO0-
S v
0 7OO L~ Lid Z
2
A//a,/.? 4OO 5
×
;
I
8
6
!
lo
SPEED (M/S) Fig. 3. Annual mean wind speed vs. turbine energy production.
and summer winds as well as frequent strong flow during the winter. Of the three sites only Sevenmile Hill displays a definite seasonal wind cycle as 70% of the estimated annual turbine energy is generated during the spring and summer months. The seasonal statistics are shown in Table 2. The mean wind speed during the summer is 10.2 m/sec compared to 5.2 m/sec in the winter. At the other two sites the energy production is distributed about evenly over the four seasons, although the seasonal mean speeds are not as uniform. Interannual variations in the seasonal mean speeds and energy values are quite substantial at all of the sites. At Sevenmile Hill and Kennewick, the highest and lowest seasons on record in general varied by _+15% to 30% from the mean seasonal speed and _+20% to 50% from the mean seasonal energy. At Cape Blanco the extreme seasonal mean speeds varied from the longterm mean by _+20% to _+35% and _+25% to _+50% from the mean energy. At all three sites the largest seasonal variations in the speed occurred during the winter. Winter was also the season for the largest turbine energy production variations at Sevenmile Hill and Kennewick. However, during the summer variations in energy production were much larger at Cape Blanco. Analysis of the significance of any one season of energy production on the annual production revealed that at Kennewick significantly higher or lower production years (i.e., > +10% of the long-term mean) were most affected by variations in the winter energy production. For the four years that exceeded the > _+10% value, high (low) winter production related to high (low) annual production. For Sevenmile Hill, 5 out of the 10 years had annual energy values > _+10% of the long-term mean and all of these years were as-
sociated with significantly higher or lower production in the spring and summer. At Cape Blanco, four out of the 12 years were abnormally high or low with respect to energy output. For three of these years summer energy production was the single biggest contributor. However, in 1981 all seasonal energy values were significantly below normal (22%-40%) which resulted in 1981 being 31% below normal in energ.v production. The continental U.S. NWS site data analyzed by Justus[ 2 ] showed that the mean seasonal as~vs values varied from 0.09-0.14 (winter), 0.08-0.12 (spring), 0.10-0.13 (summer), and 0.09-0.15 (fall). Except for the winter months when the northwest site data indicated much higher as/v~ values (0.13-0.19 ) there was reasonably good agreement between the spring, summer, and fall values with those reported by Justus. 3. CONCLUSIONS The long record of hourly wind data at the three BPA wind monitoring sites provided an opportunity to investigate interannual and interseasonal variations in mean wind speed, the wind speed distribution, and, most importantly, the potential turbine energy production. Significant findings include 1. At two of the three sites the mean annual wind speeds were within 10% of the long period mean 90% of the time. This is in agreement with findings of Corotis and Justus. At Cape Blanco only three of the 12 years fell within the +10% bound. However, the annual maximum and minimum turbine energy values were typically 15%-25% above or below the long period mean at all of the sites. 2. Two of the three sites did not show definite seasonal cycles in energy production. However, at Sevenmile Hill 70% of the energy is produced during the spring
289
Variations in mean wind speed and wind turbine energy production Table 2. Seasonal statistics for Sevenmile Hill (the Subscript "'s'" is for seasonal) Winter Sevenmile Hill Wind speed (m/sec) Mean (G) Standard deviation (as)
adG High Low Estimated turbine energy (MWH) Mean Standard deviation High Low Kennewick Wind speed (m/sec) Mean (6,) Standard deviation (a,)
adff, High Low Estimated turbine energy (MWH) Mean Standard deviation High Low Cape Blanco Wind speed (m/sec) Mean (G) Standard deviation (as) a,/ff~ High Low Estimated turbine energy (MWH) Mean Standard deviation High Low
Spring
Summer
Fall
5.2 0.7 0.13 6.5 (+25%) 4.0 (-23%)
8.2 0.8 0.10 9.7 (+19%) 7.0 (-15%)
10.2 0.8 0.08 11.3 (+10%) 8.6 (-15%)
6.3 0.7 0.11 7.1 (+13%) 5.1 (-18%)
80.0 23.9 124.2 (+55%) 35.0 (-56%)
200.1 30.4 260.9 (+30%) 155.5 (-22%)
283.5 32.4 329.9 (+16%) 220.7 (-22%)
127.8 21.9 147.0 (+15%) 95.1 (-26%)
0.16 9.2 (+36%) 4.6 (-33%)
7.5 0.9 0.12 9.1 (+21%) 6.0 (-20~c)
6.7 0.8 0.12 8.6 (+28%) 5.7 (-15%)
6.8 0.9 0.13 8.0 ( + 1 7 ~ ) 5.4 (-21%)
147.0 44.8 230.1 (+57%) 76.0 (-48%)
167.2 35. I 218.5 (+31%) 107.6 (-36%)
139.0 24.5 186.2 (+34%) 110.5 (-20%)
149.6 27.2 188.5 (+26%) 105.8 (-29%)
9.6 1.8 0.19 13.1 (+37%) 7.1 (-26%)
8,3 1.0 0,12 9.6 (+ 16%) 6,2 (-25~)
8.2 1.0 0.12 10.4 (+27%) 6.5 (-21%)
8.2 1.2 0.15 10.2 (+24%) 6.7 (-19%)
194.2 33.4 249.1 (+28%) 146.8 (-24%)
192.5 32.9 233.6 (+21~) 125.3 (-35%)
215.8 47.3 319.8 (+48%) 129.2 (-40%)
176.3 29.0 237.8 (+35%) 137.0 (-22%)
6.8 1.1
and summer. The m a x i m u m and m i n i m u m seasonal energy values typically varied by 25%-50~ from the mean seasonal energy value. 3. A particular season's energy production significantly contributed to the annual production for abnormally high or low energy production years. For the Kennewick site winter was the most significant season. At Sevenmile Hill spring and summer production was the most crucial and summer production was the most important season to affect annual production at Cape Blanco. 4. The annual and seasonal a/F~ values for Kennewick and Sevenmile Hill were in general agreement with those reported by Justus. However, the Cape Blanco values were significantly greater. 5. The relationship o f annual turbine energy produc-" tion and site mean annual wind speed was different at each of the three locations because the speed frequency distribution shapes were different. A comparison to Rayleigh distribution estimates revealed a significant underestimate of the energy by 8% to 15% for Kennewick and Sevenmile Hill, respec-
tively, and about a 10% overestimate for Cape Blanco. Thus, it is important to have actual site speed frequency distribution information to make accurate estimates of turbine energy production. Using theoretical speed frequency distributions can produce large errors in estimated turbine energy production.
.4cknowledgments--This research was performed at Oregon State University for the Bonneville Power Administration under Contract DE-A 179-86BP63406.
REFERENCES
1. R. B. Corotis. Stochastic modeling of site wind characteristic~ RLO/2342-77/2 NTIS (RLO/2432-77/2/MF401) (1977). 2. C. G. Justus, K. Mani, A. S. Mikhail, lnterannual and month-to-month variations ofwind speed, J. Appl. Met.. Ig, 913-920 (1979). 3. W. R. Powell. An analytical expression for the average output power of a wind machine, Solar Energy. 26, 7780 ( 1981 ).
0038--092X/90 $3.00 + .00 Copyright ~ 1990 Pergamon Press pie
ANNUAL AND SEASONAL VARIATIONS IN MEAN WIND SPEED AND WIND TURBINE ENERGY PRODUCTION R. W. BAKER. S. N. WALKER, and J. E. WADE Department of Mechanical Engineering, Energy Resources Research Laboratory', Oregon State University, Corvallis, OR 97331, U.S.A. Abstract--A significant uncertainty in the estimate of energy from a wind plant is the interannual and interseasonal variation in energy output, These estimates are typically based on wind data that have been collected on-site for one or more years. Rarely have the wind data records exceeded three years. Variations in the mean annual wind speeds have been assumed to be _ 10% of the long-term mean wind speed with a 90% confidence interval. This assumption, however, is based on previous work using long-term airport wind data and it has not been verified at high wind locations. In addition, most of the strong wind sites that have been developed for wind energy display distinct seasonal variations in the wind speed. In the AItamont and San Gorgonio Passes about 80% of the energy is produced during the spring and summer months. If the winds are abnormally weak during any of these months (particularly May-August). the annual energy production can be significantly reduced. Similarly, stronger than normal winds during this time period can greatly enhance the annual production. Over l0 years of hourly wind data have been collected at several moderate-to-high wind sites in the Pacific Northwest for the Bonneville Power Administration. This data base is used to determine the yearto-year and the interseasonal variations in the mean speeds and in wind turbine energy production. Wind turbine power curves are used with the speed frequency' distributions to produce seasonal and annual energy estimates.
1. I N T R O D U C T I O N
2. Y E A R - T O - Y E A R V A R I A T I O N S IN M E A N
Several years of hourly wind data have been collected at many windy sites in the Pacific Northwest. This wind monitoring work to assess the wind power potential of the region has been done for the Bonneville Power Administration (BPA). The wind data collected from three of the BPA wind monitoring locations will be discussed in this paper. These sites are Sevenmile Hill located on a 600-m ( M S L ) ridge about 100 km east of Portland, Oregon, Kennewick which is situated on a 700-m ridgeline about 300 km east of Portland, and Cape Blanco located on a 60-m coastal bluff about 100 km north of the California/Oregon border. The data recovery, rates over the 10- to 12-year data period have varied from 90% at Sevenmile Hill to 83% at Cape Blanco. Data have been analyzed through the year 1988 and all three sites are still in operation. The long and mostly continuous data base at these three sites provides an opportunity to inspect the interannual as well as the interseasonal variations in the mean speed and speed frequency distributions. The latter is best measured by transposing wind speed to power output using a turbine power curve and then by analyzing the yearly or seasonal variations in the estimated turbine energy production. A power curve o f a 250-kW machine with a cut-in speed of 4 m/sec, a rated speed of 13 m/sec, and a cut-out speed of 24 m / s e c was used to estimate energy production. Along with a discussion of the year-to-year variations in the mean annual speed (energy) and seasonal speeds (energy), the paper also addresses the relationship of changes in the mean annual wind speed to the changes in the estimated turbine energy production. 285
ANNUAL WIND SPEED AND ENERGY
At most wind energy sites or potential developments, a long period of continuous wind data on the order of 10 to 20 years duration or greater are not available. Information on the interannual and interseasonal variations in the strength of the wind is needed, however, to predict likely variations (i.e., revenue variations) in turbine production. Work done previously by Corotis[l] and J ustus et al. [2] on long periods of data from locations of the National Weather Service ( N W S ) has indicated relatively small variations in the mean annual wind speed. Their research indicates that there is 90% confidence that the mean annual wind speed from a single year will be within +_10% of the true long-term mean speed. This assumption has been used by the wind industry, Unfortunately, the correlation between the wind speeds at NWS locations and high wind sites is poor and, therefore, there is still a high degree of uncertainty in estimating interannual variations without a long period of site data to inspect. Although the l0 to 12 years of wind data collected at the three BPA sites does not represent a climatological long record (preferably 20 to 30 years), it does give an opportunity to investigate at the high wind sites interannual variations in wind speed and turbine energy production and to discuss some of the reasons for the variations. The yearly mean speeds and estimated turbine energy production at the three sites show different patterns with Cape Blanco having the largest variations and Kennewick the smallest. The annual statistics are listed in Table I. For the 12 years of record, the mean speed at Kennewick was 7.0 m / s e c and the m a x i m u m and
R. W. BAKER,S. N. WALKER.and J. E. WADE
286
Table 1. Annual statistics for the three sites (the subscript "a'" is for annual)
Wind speed (m/sec) Mean (~'0) Standard deviation (aa)
odvo High Low Years of data Estimated turbine energy (MWH) Mean Standard deviation High Low
Sevenmile Hill
Kennewick
Cape Blanco
7.5 0.6 .08 8.2 (+ 10%) 6.4 (-14%) 10
7.0 0.3 .04 7.5 (+7%) 6.3 (-9%) 12
8.6 0.9 . I0 10.0 (+ 16%) 6.7 (-22%) 12
691.4 82.3 811.6 (+17%) 532.0 (-23%)
minimum values were within ___10% of this mean speed. The highest turbine energy year, however, was 19% above the 12-year mean value and the lowest energy year was 26% below the mean value. For Sevenmile Hill, the highest mean speed year was only 10% above the long-term mean, but the lowest speed year was 14% lower than the mean. Only one of the 10 years of record fell outside ___10%of the 10-year mean speed. With respect to turbine energy, the highest year was 17% above the mean and the lowest year was 23% below the mean. At Cape Blanco, the largest variations occurred. The highest wind year at 10.0 mtsec exceeded the 12-year mean speed of 8.6 m/sec by 16% and the lowest wind year at 6.7 m/see was 22% lower than the mean. The energy differences were +23% and -31%. The mean annual wind speeds for 9 o f th e 12 years were outside the bounds of---10% of the 12-year mean.
592.5 74.8 707.2 (+19%) 438.6 (-26%)
778.7 103.2 960.2 (+23%) 538.3 (-31%)
J ustus et aL [2] found for regions comprised of several National Weather Service (NWS) sites the annual values of cra/~, varied from 0.05-0.07 lbr sites in the continental U.S. ( 10 to 18 years of data). The values ofaa/~a shown in Table 1 vary from 0.04 at Kennewick to 0.10 at Cape Blanco. The latter is significantly higher than those measured at the NWS sites. Using the 10 to 12 years of data, an analysis was done to determine the long-term mean speed using the Student's T distribution. For a 90% confidence limit the true mean speed lies between 6.8 and 7.1 m/see at Kennewick. 7.1 and 7.8 m/sec at Sevenmile Hill, and 8.1 and 9.0 m/sec at Cape Blanco. The annual turbine energy production at the sites is directly related to the shape of the site wind speed frequency distribution. The annual wind speed distributions for the three sites indicate that each distribution has a different shape as shown in Fig. 1. The Kennewick
300
rn
SEVENMILE
)¢
KENNEWICK C. BLANCO
250
200 ~3 V
',4
z LIJ
n," IJ.J
Or2 LIJ rw LI.
0 I:L
150
100
50
10
20 WIND
SPEED
30
O, 4O
(M/S)
Fig. 1. Annual speed frequency distributions and the 250 kW turbine power curve.
Variations in mean wind speed and wind turbine energy production distribution displays a definite single frequency peak at about 4.5 m/see and the frequency drops off steadily for winds above 6.7 m/see. In contrast, the distributions at Sevenmile Hill and Cape Blanco have a much broader shape and there are two relative frequency maximums at Sevenmile Hill (3.5 m/see and 9 m / see). At Cape Bianco there is a high frequency of winds above 18 m/see. Year-to-year variations in the wind speed frequency distributions at Sevenmile Hill are shown in Fig. 2 for 1987, the lowest wind year, and 1985, one of the highest wind years recorded at the site. In 1987 the estimated energy production was 532 MWH (~ = 6.4 m/see) and in 1985 it was 769 MWH (~ = 8.0 m/see), a 45% difference in energy and a 25% difference in mean speed. Both distributions have a relatively similar shape (bi-modal maximum), but there was a much higher frequency of wind speeds above I 1 m/sec in 1985 than in 1987. Further analysis of wind speed versus energy production reveals that each site has a different relationship. Plots of energy production vs. mean annual wind speed are shown in Fig. 3. A linear fit approximation was made to the data points. The most scatter in the data was at Kennewick and the least at Sevenmile Hill. The slope of the Kennewick curve indicates about a 25% change in the energy production with a 10% change in mean annual wind speed. At Sevenmile Hill there is about a 20% change in energy with a 10% change in mean speed, and at Cape Blanco there is only about a 13% change in energy for a 10% change in the mean speed. For a given turbine power curve these variations are caused by the shape of the speed frequency distributions. The three sites display different distributions, as shown in Fig. 1. With a turbine power curve like that shown in Fig. 1, which is typical of most machines, the total energy is dependent on the amount
287
of time the winds are between cut-in and rated speed and between rated and cut-out. The power output increases significantly with wind speed between cut-in and rated speeds but there is no power increase at rated to cut-out speed. As the speed frequency distribution becomes more skewed to higher speeds more rated power is produced. Powell ( 1981 ) [ 3 ] as well as others have done similar analysis of energy production and mean wind speed assuming a Rayleigh speed frequency distribution. As the rated speed to mean speed ratio decreases the energy production increases significantly. Given the mean annual wind speeds for the three locations, annual turbine energy output was calculated assuming a Rayleigh distribution. This curve is included in Fig. 3. As shown, for Sevenmile Hill and Kennewick the Rayleigh distribution underpredicts the annual energy output. There is about a 15% underestimate for Sevenmile Hill and about an 8% underestimate for Kennewick. In contrast, for the higher speed site, Cape Blanco, there was about a 10% overestimate assuming a Rayleigh distribution. In addition, the speed frequency distribution shapes varied at all three sites on an annual basis, so year-to-year differences between the actual and Rayleigh estimates varied considerably. These differences ranged from - 13% to -22% at Sevenmile Hill, -1% to -15% at Kennewick, and -1% to -,26% at Cape Blanco. It is not uncommon for high wind areas to display distinct seasonal patterns in wind speed and direction. Two of the three California wind farm development areas, Altamont Pass and San Gorgonio Pass. have the strongest winds during the spring and summer months. These periods provide an ideal match to load demand and on-peak/off-peak buyback rates. Due to its higher elevation in the "l'ehachapi Mountains and its exposure to storm winds, the "l-ehachapi area displays good spring
1985 × 4.'
~'3-
V
z LIJ :::::)2Or
LI.I
E
0 ,5
10
15
20
25
WIND SPEED ( M / S ) Fig. 2. lnterannual variations in the wind speed frequency distributions at Sevenmile Hill.
1987
R.W. BAKER,S. N. WALKER,and J. E. WADE
288
rn
A: SEVENMILE HILL
x
B: KENNEWICK
v
C: CAPE BLANCO
=1,
D: RAYLEIGH
V
Z 0
gO0-
S v
0 7OO L~ Lid Z
2
A//a,/.? 4OO 5
×
;
I
8
6
!
lo
SPEED (M/S) Fig. 3. Annual mean wind speed vs. turbine energy production.
and summer winds as well as frequent strong flow during the winter. Of the three sites only Sevenmile Hill displays a definite seasonal wind cycle as 70% of the estimated annual turbine energy is generated during the spring and summer months. The seasonal statistics are shown in Table 2. The mean wind speed during the summer is 10.2 m/sec compared to 5.2 m/sec in the winter. At the other two sites the energy production is distributed about evenly over the four seasons, although the seasonal mean speeds are not as uniform. Interannual variations in the seasonal mean speeds and energy values are quite substantial at all of the sites. At Sevenmile Hill and Kennewick, the highest and lowest seasons on record in general varied by _+15% to 30% from the mean seasonal speed and _+20% to 50% from the mean seasonal energy. At Cape Blanco the extreme seasonal mean speeds varied from the longterm mean by _+20% to _+35% and _+25% to _+50% from the mean energy. At all three sites the largest seasonal variations in the speed occurred during the winter. Winter was also the season for the largest turbine energy production variations at Sevenmile Hill and Kennewick. However, during the summer variations in energy production were much larger at Cape Blanco. Analysis of the significance of any one season of energy production on the annual production revealed that at Kennewick significantly higher or lower production years (i.e., > +10% of the long-term mean) were most affected by variations in the winter energy production. For the four years that exceeded the > _+10% value, high (low) winter production related to high (low) annual production. For Sevenmile Hill, 5 out of the 10 years had annual energy values > _+10% of the long-term mean and all of these years were as-
sociated with significantly higher or lower production in the spring and summer. At Cape Blanco, four out of the 12 years were abnormally high or low with respect to energy output. For three of these years summer energy production was the single biggest contributor. However, in 1981 all seasonal energy values were significantly below normal (22%-40%) which resulted in 1981 being 31% below normal in energ.v production. The continental U.S. NWS site data analyzed by Justus[ 2 ] showed that the mean seasonal as~vs values varied from 0.09-0.14 (winter), 0.08-0.12 (spring), 0.10-0.13 (summer), and 0.09-0.15 (fall). Except for the winter months when the northwest site data indicated much higher as/v~ values (0.13-0.19 ) there was reasonably good agreement between the spring, summer, and fall values with those reported by Justus. 3. CONCLUSIONS The long record of hourly wind data at the three BPA wind monitoring sites provided an opportunity to investigate interannual and interseasonal variations in mean wind speed, the wind speed distribution, and, most importantly, the potential turbine energy production. Significant findings include 1. At two of the three sites the mean annual wind speeds were within 10% of the long period mean 90% of the time. This is in agreement with findings of Corotis and Justus. At Cape Blanco only three of the 12 years fell within the +10% bound. However, the annual maximum and minimum turbine energy values were typically 15%-25% above or below the long period mean at all of the sites. 2. Two of the three sites did not show definite seasonal cycles in energy production. However, at Sevenmile Hill 70% of the energy is produced during the spring
289
Variations in mean wind speed and wind turbine energy production Table 2. Seasonal statistics for Sevenmile Hill (the Subscript "'s'" is for seasonal) Winter Sevenmile Hill Wind speed (m/sec) Mean (G) Standard deviation (as)
adG High Low Estimated turbine energy (MWH) Mean Standard deviation High Low Kennewick Wind speed (m/sec) Mean (6,) Standard deviation (a,)
adff, High Low Estimated turbine energy (MWH) Mean Standard deviation High Low Cape Blanco Wind speed (m/sec) Mean (G) Standard deviation (as) a,/ff~ High Low Estimated turbine energy (MWH) Mean Standard deviation High Low
Spring
Summer
Fall
5.2 0.7 0.13 6.5 (+25%) 4.0 (-23%)
8.2 0.8 0.10 9.7 (+19%) 7.0 (-15%)
10.2 0.8 0.08 11.3 (+10%) 8.6 (-15%)
6.3 0.7 0.11 7.1 (+13%) 5.1 (-18%)
80.0 23.9 124.2 (+55%) 35.0 (-56%)
200.1 30.4 260.9 (+30%) 155.5 (-22%)
283.5 32.4 329.9 (+16%) 220.7 (-22%)
127.8 21.9 147.0 (+15%) 95.1 (-26%)
0.16 9.2 (+36%) 4.6 (-33%)
7.5 0.9 0.12 9.1 (+21%) 6.0 (-20~c)
6.7 0.8 0.12 8.6 (+28%) 5.7 (-15%)
6.8 0.9 0.13 8.0 ( + 1 7 ~ ) 5.4 (-21%)
147.0 44.8 230.1 (+57%) 76.0 (-48%)
167.2 35. I 218.5 (+31%) 107.6 (-36%)
139.0 24.5 186.2 (+34%) 110.5 (-20%)
149.6 27.2 188.5 (+26%) 105.8 (-29%)
9.6 1.8 0.19 13.1 (+37%) 7.1 (-26%)
8,3 1.0 0,12 9.6 (+ 16%) 6,2 (-25~)
8.2 1.0 0.12 10.4 (+27%) 6.5 (-21%)
8.2 1.2 0.15 10.2 (+24%) 6.7 (-19%)
194.2 33.4 249.1 (+28%) 146.8 (-24%)
192.5 32.9 233.6 (+21~) 125.3 (-35%)
215.8 47.3 319.8 (+48%) 129.2 (-40%)
176.3 29.0 237.8 (+35%) 137.0 (-22%)
6.8 1.1
and summer. The m a x i m u m and m i n i m u m seasonal energy values typically varied by 25%-50~ from the mean seasonal energy value. 3. A particular season's energy production significantly contributed to the annual production for abnormally high or low energy production years. For the Kennewick site winter was the most significant season. At Sevenmile Hill spring and summer production was the most crucial and summer production was the most important season to affect annual production at Cape Blanco. 4. The annual and seasonal a/F~ values for Kennewick and Sevenmile Hill were in general agreement with those reported by Justus. However, the Cape Blanco values were significantly greater. 5. The relationship o f annual turbine energy produc-" tion and site mean annual wind speed was different at each of the three locations because the speed frequency distribution shapes were different. A comparison to Rayleigh distribution estimates revealed a significant underestimate of the energy by 8% to 15% for Kennewick and Sevenmile Hill, respec-
tively, and about a 10% overestimate for Cape Blanco. Thus, it is important to have actual site speed frequency distribution information to make accurate estimates of turbine energy production. Using theoretical speed frequency distributions can produce large errors in estimated turbine energy production.
.4cknowledgments--This research was performed at Oregon State University for the Bonneville Power Administration under Contract DE-A 179-86BP63406.
REFERENCES
1. R. B. Corotis. Stochastic modeling of site wind characteristic~ RLO/2342-77/2 NTIS (RLO/2432-77/2/MF401) (1977). 2. C. G. Justus, K. Mani, A. S. Mikhail, lnterannual and month-to-month variations ofwind speed, J. Appl. Met.. Ig, 913-920 (1979). 3. W. R. Powell. An analytical expression for the average output power of a wind machine, Solar Energy. 26, 7780 ( 1981 ).