State Electricity Commission of Victoria

Victorian Solar Energy Council





Wind Monitoring Study

Final Report




CR 249/42
July 1987
Natural Systems Research Pty. Ltd.
Environmental Consultants


CONTENTS

1. INTRODUCTION
2. INSTALLATION
2.1 Tower Foundations
2.2 Towers
2.3 Anemometers
2.4 Data Loggers
3. DATA RETRIEVAL
4. ANEMOMETER CALIBRATION
5. MONITORING RESULTS
5.1 Data Recovery
5.2 Wind and Energy Roses
5.3 Speed Distributions
5.4 Average Speeds
5.5 Maximum Speeds
5.6 Wind Energy Flux
5.7 Ranking of Sites
6. WEATHER DURING MONITORING PERIOD
6.1 Climatic Summary
6.2 Wind Speed at Tullamarine Airport
6.3 Wind Speed Comparison and Prediction
7. CONCLUSIONS
8. REFERENCE
9. ACKNOWLEDGEMENTS
FIGURES
APPENDICES

1. INTRODUCTION

This is the final report for a two-year monitoring study to assess the potential for wind-powered electricity generation in Victoria. The study was commissioned jointly by The State Electricity Commission of Victoria and The Victorian Solar Energy Council. In 1983 and 1984, these organizations conducted a survey of Victoria to identify locations expected to have a high potential for wind-powered generation of electricity. Based upon criteria such as anticipated wind regime, landowner cooperation and ease of access, ten sites were selected along the coastline, with five each to the east and the west of Melbourne, as shown in the frontispiece map.

The contract, awarded in October 1984, required:

(a) Erection of a tower and provision and installation of wind monitoring equipment at each site.

(b) Monthly collection of data from all sites.

(c) Processing and analysis of collected data.

(d) Reporting of results every month, quarter and year during the two-year study period.


The contract further specified that:

(e) The anemometers were to be calibrated in a wind tunnel by an accredited testing organisation, before installation and at intervals of six months throughout the two year monitoring period.

(f) Data recovery at each site was to be a minimum of 90% for each of the two years of monitoring.

2. INSTALLATION

2.1 Tower Foundations

Tower foundations were excavated and poured in December 1984. Two alternative forms were used:

(a) Pier type, where the legs on the tower base were each set in a concrete pier l.5m deep and 0.3m in diameter, with the foot bellowed to a diameter of 0.5m. This method required 0.3m3 of concrete.

(b) Slab type, where the tower base was welded or bolted together to make a frame which was set in a slab of concrete 1.5m square and 0.6m deep. This foundation required 1.4m3 of concrete.


Since less concrete was needed, pier footings were constructed at the first three sites where foundations were installed (Queenscliff, Apollo Bay and Port Campbell). The next site, Port Fairy, is situated on a rocky hill where it was too difficult to dig holes l.5m deep, so a partially-elevated slab foundation was constructed. Although requiring more concrete, this alternative proved to be easier to construct. Slabs were therefore installed at the remaining six sites. These were sunk 0.3m to 0.5m below the surface so they could be left in the ground when the towers were dismantled at the completion of the project.

Plate 1 shows the pier foundation at Port Campbell, while Plates 2 to 6 illustrate progressive stages in the construction of slab foundations at various sites.


2.2 Towers

Nine-metre, triangular-section, self-supporting towers manufactured by Hills Industries Ltd., were assembled and erected at the ten sites during December 1984 and January 1985. The towers were assembled on the ground and raised using ropes and a vehicle as depicted in Plate 7.


2.3 Anemometers

Because of their durability and reputation for reliability, Synchrotac heavy-duty anemometers, type 706, were chosen to withstand the corrosive environment and strong winds expected at the coastal sites. Following pre-deployment calibration in December 1984 (see Section 4), Synchrotac anemometers were installed at a height of ten metres on top of the tower at each site. The anemometer and tower at Cape Liptrap are shown in Plate 8. To preclude attack by birds or animals, the cables connecting the anemometer to the data logger were encased in galvanized piping, as can be seen in Plates 9 and 12.

2.4 Data Loggers

In order to meet the tight schedule for commencing monitoring within three months of commissioning, as well as facilitating prompt servicing of faults, data loggers were selected which were designed and assembled by a Melbourne manufacturer, Tain Electronics Pty. Ltd.

The data loggers were designed to log wind speed and direction at intervals dependent upon the speed (every 140 metres of wind run). Every hour the logged data was processed and the following values were stored in the logger memory and output to a connected printer:
average wind speed, standard deviation of wind speed, maximum wind speed, minimum wind speed, predominant average wind direction, standard deviation of wind direction.

The methods used to calculate the predominant average wind direction and the standard deviation of wind directions are outlined in Appendix I.

Specified data accuracies were:
+/- 2% for wind speed,
+/- 5° for wind direction,
+/- 2 minutes/month for time.


Printed recordings for 1400 hours, 1500 hours, 1600 hours and 1700 hours on 17 April 1986 are visible in Plate 11. The printers were intended to serve the dual functions of:

(i) a visual indicator to landowners, who were asked to check periodically, that wind data was being recorded;

(ii) a back-up data record should the logger malfunction and the data stored in memory be lost

The loggers were powered by four, six-volt, rechargeable batteries. These and the connected printer were located inside a lockable steel enclosure bolted to a concrete housing which was covered by a galvanized iron lid. Plate 10 shows the logger, printer and batteries inside the steel enclosure and concrete housing, while the galvanized iron lid can be seen upside down in Plate 12.

The hourly data loggers have a memory capacity of 32 kilobytes, sufficient to store hourly recordings for 44 days.

The hourly loggers were installed and connected at all ten sites on 30 and 31 January 1985.

Another logger, which records data every two minutes, was connected in parallel with the hourly logger at Kilcunda North on 19 April 1985. This two-minute logger has a memory capacity of 160 kilobytes, again sufficient to store data for approximately six weeks.

3. DATA RETRIEVAL

At monthly intervals, all sites were visited by an experienced meteorologist to retrieve stored data from the loggers. This was effected by means of a portable Cromemco C-10 computer transported inside a vehicle and powered by a twelve-volt battery and an inverter. The computer was connected to the logger and data was transferred to 5.25-inch diskette. The data was read twice and these files were compared before being checked by a program which flags recordings outside a series of predetermined limits. The flagged recordings were screened for veracity and unacceptable values were deleted on site by the meteorologist. The validated data was then copied to a back-up diskette before the stored data was erased from the logger memory.

During the monthly site visits, several other functions were performed:

(a) Four charged, six-volt batteries replaced the four that had been powering the logger and printer for the preceding month. This was accomplished without interrupting the power supply by a two-way switch in the battery lead, which enabled one pair of batteries to be substituted while power was provided by the other pair. The four batteries and the two-way switch can be seen in Plate 10.

(b) A full chart roll was fed into the printer after the roll with recordings for the previous month was removed from the take-up spool.

(c) A canister of recharged silica gel was exchanged for that which had absorbed moisture inside the logger during the previous month (visible inside logger case in Plate 10).

(d) The anemometer, logger and printer were checked for correct operation and any which had malfunctioned were replaced by spares.

Plate 12 shows data retrieval in progress at Apollo Bay on 17 April 1986, while Plate 13 is a close-up of the Cromemco computer with data retrieved on this date displayed on the screen.

During winter and spring, or immediately following rain, tracks to some sites, particularly the steep hills at Apollo Bay, Kilcunda North and Toora, and the access road to the dairy at Port Campbell, became impassable for conventional vehicles. If conditions were not too slippery or boggy, the track could be traversed by fitting chains (as seen in Plate 12). If the ground conditions were too muddy, the sites were reached on foot and a spare logger substituted for that which had recorded for the previous month. This logger was then returned to the vehicle for data extraction before proceeding to the next site.

4. ANEMOMETER CALIBRATION

The Synchrotac anemometers were calibrated in the wind tunnel at the Department of Civil and Aeronautical Engineering, Royal Melbourne Institute of Technology, prior to installation and at roughly six-monthly intervals thereafter. Fifteen cup and speed generator units were used during the monitoring study, with ten being deployed in the field at any one time. After the pre-deployment calibration of all fifteen cup and speed generator sets in December 1984, subsequent calibrations were in groups of five, with five spare sets replacing five of those in the field. Recalibrations were carried out in August and October 1985, March and April 1986, August 1986 and February 1987. Table 1 summarises results from all calibrations, and shows that after two years all recalibrated sensitivities were within +/-1.7 % of the pre-deployment calibration, with the percentage change averaging + 0.6% for the 15 anemometers.

Apparent decreases in sensitivity of 2.4% and 3% respectively occurred in one speed generator (No.10) after roughly six months and in another (No.13) after some eighteen months. As the sensitivity changed between calibrations, it was not possible to determine exactly when the change occurred. It has therefore been assumed that anemometer calibrations remained valid until recalibration showed a change. The revised calibration coefficients for these speed generators were applied to data recorded during the succeeding six months at the sites where these units were deployed (Queenscliff from September 1985 to March 1986, and Swan Reach from August 1986 to February 1987 respectively), by altering the calibration coefficients in the firmware of the connected loggers. The two cup and generator sets which appeared to exhibit variations in sensitivity of more than 2% were intentionally redeployed at the two sites which recorded the lowest wind speeds. The sensitivities of both these cups and speed generator units reverted to within 0.7% of their pre-deployment values at the subsequent recalibration.


5. MONITORING RESULTS

5.1 Data Recovery

Monitoring of hourly data commenced on 1 February 1985, with the two year period ending on 31 January 1987. Table 2 details data recovery percentages from all sites for each month, the eight quarters and the two years of monitoring. As faults in the hourly loggers were progressively rectified, the combined data recovery improved from 89.7% during the first quarter to 98.5% in the second quarter. The combined data recoveries were 96.4% for the first year, 98.3% for the second year, and 97.3% for both years of the monitoring study. For the two-year period, data recovery at each of the ten sites ranged from a maximum of 100% at Port Fairy to a minimum of 94% at Port Campbell. With such high data recoveries, analysis results for the hourly data are unlikely to be influenced significantly by sampling variations.


Causes of data losses were:

* logger failure 2.1%
* logger theft 0.2%
* printer switch accidentally left on by inspecting landowner, resulting in flattened batteries 0.2%
* lightning strike 0.1%
* anemometer failure 0.1%

Data coverage from the two-minute logger at Kilcunda North was 30.4% for the first year, 16.4% for the second year, and 23.4% for both years. Unlike the hourly loggers, no printer back-up of two-minute recordings was feasible because of the large volume of data generated. Any malfunction during the month therefore generally resulted in the loss of all data for that month. Both the hourly logger and the original two-minute logger were stolen from Kilcunda North on 31 March 1986, just when the latter was recording reasonably reliably (data recovery exceeded 90% in December 1985 and January 1986). Almost three months elapsed before a replacement was manufactured, tested, and installed on 20 June 1986. The replacement two-minute logger, like the original, suffered a series of failures for several months before its performance improved significantly in December 1986, when 100% data recovery was achieved.

When the two-minute logger was deployed at Kilcunda North and connected in parallel with the one-hour logger, the slave hourly logger did not record directions although it did record speeds, even if the two-minute logger was not working. This is why data recoveries for the roses at site 6 (Figures 6, 6a, 6b) are less than for analyses of speed only (Tables 2 and 8 and Figure 16, 16a, 16b).

Similar small discrepancies in data recoveries at Port Campbell (Figures 3 and 13), Apollo Bay (Figures 4 and 14) and Swan Reach (Figures 10 and 20) were the result of logger failures at these sites necessitating recourse to chart recordings, on which overprinting masked more values of wind direction than of wind speed.

Following the theft of loggers from Kilcunda North, it was decided to improve security at this site, which is the only one of the ten that is adjacent to, and clearly visible from, a road. The enclosure was modified by drilling holes in the existing concrete walls and attaching locking bars with inserted dyna bolts, as depicted in Plate 14.

An additional layer of concrete was then poured over the outside of the locking bars and bolts. Finally, a cover, fabricated from 6 mm steel plate, was locked to the protruding locking bars. Welded guards surround the locks, thereby precluding access by hacksaw and/or hammer. The steel security cover and lock guards are shown in Plate 15.

Monitoring ceased at the five sites west of Melbourne in February 1987. To the east of Melbourne, monitoring continued until 10 March 1987 at Swan Reach and until 2 July 1987 at Toora, Yanakie, Cape Liptrap and Kilcunda North. Data recovery was 100% for the hourly loggers at all sites, and 81% over the additional five months for the two-minute logger at Kilcunda North. By the end of July 1987, equipment was dismantled and removed from all sites except Kilcunda North and Cape Liptrap. Further recordings are proposed at these two sites for the Latrobe Valley Airshed Study.


5.2 Wind and Energy Roses

Wind and energy flux roses for the two-year monitoring period are plotted in Figures 1 to 10 for sites 1 to 10 respectively. Figure 21 presents roses for the two-minute logger at site 6.

Figures la to l0a are annual roses for the first year of monitoring (1 February 1985 to 31 January 1986), and Figures lb to l0b are annual roses for the second year of monitoring (1 February 1986 to 31 January 1987). Figures 21a and 21b are the corresponding annual roses for the two-minute logger at Kilcunda North.

For the two-year monitoring period at each site, average monthly wind roses are plotted in Figures lc to l0c, while average three-hourly wind roses are plotted in Figures 1d to l0d.

For the two-year monitoring period, the wind regimes can be divided into four broad groups:

(i) Reasonably uniform distribution of directions

Of all sites, Bridgewater (the most westerly) had the most uniform distribution of directions, with a small peak in the occurrence of west-northwesterlies and a compensating minimum occurrence of winds from the easterly octant.

(ii) Prevalence or wind from the north and south

This regime was exhibited by four of the six sites in Western and Central Victoria - Port Fairy, Port Campbell, Queenscliff and Kilcunda North. At Port Fairy, Port Campbell and Kilcunda North, winds were most frequently from the northern octant with winds from the southwestern quadrant also prevalent. The least frequent winds were easterlies at Port Fairy and Port Campbell and westerlies at Kilcunda North.

Queenscliff also had a high incidence of winds from the south and north although, unlike Kilcunda North, westerlies occurred most frequently.

(iii) Prevalence or wind from the east and west

This regime was displayed by the four most southern sites - Cape Liptrap, Yanakie, Toora and Apollo Bay - the latter two of which have the highest elevations. There was a predominance of winds from both the east-northeastern and the west-southwestern octants, with the former being most frequent at Yanakie and the latter most frequent at C, ape Liptrap. At the most elevated sites the predominant wind increased m strength and veered one semi-cardinal point to between west and northwest. In contrast to Kilcunda North, the occurrence of southerly and northerly winds was generally very low at these four sites. This is probably caused by the blocking effect of the Otway and Strzelecki Ranges immediately north of these sites. The Great Dividing Range, higher and further away, may also affect the three sites in South Gippsland.

(iv) Prevalence of wind from the north and west

The site which is furthest east and north - Swan Reach - experienced a wind regime with the most frequent winds prevailing from the western and the north-northeastern octants. The winds here also differed from the other nine sites in being much slower and having a much higher incidence of calms (almost 1% compared with 0.1% or less at the others).

Data coverage for two-minute recordings (23%) at Kilcunda North is too low to expect the resultant roses to be similar to those from hourly recordings.

Comparison of the annual roses for the first and second years of monitoring revealed that the wind regimes were fairly similar during both years at seven of the ten sites. More noticeable differences between years were evident in the roses for Bridgewater (Figures 1a and lb), Port Fairy (2a and 2b) and Queenscliff (5a and 5b).

Although the wind regimes were essentially similar, the most frequently recorded direction at eight of the ten sites (the exceptions being Toora and Swan Reach) was different during both years, shifting more westerly in the second year from the predominant southerly and easterly-component winds of the first year. This can be clearly seen in Figure 23 and Table 3 (below).

The frequency occurrence of wind directions at Tullamarine Airport is also plotted in Figure 23. Like one of the most proximate sites in the wind study network, Kilcunda North, Tullamarine experiences a prevalence of wind from the north and south.

The monthly wind roses, presented in Figures 1c to l0c, show two broad patterns:

(i) Winter northerlies and summer southerlies
Variations of this regime were exhibited by five of the six sites in Western and Central Victoria (the exception being Apollo Bay), as well as Swan Reach.

At Bridgewater, Port Fairy, Port Campbell and Kilcunda North, southerly-component winds generally prevailed in the warmer months between November and March, while. northerly-component winds occurred most frequently in the colder months from April or May to August.

At Queenscliff and Swan Reach, winds in the colder months were predominantly in the quadrant between north and west.



TABLE 3


VARIATION IN PREDOMINANT WIND DIRECTION BETWEEN YEARS

____________________________________________________________________________________________
            Predominant Wind Direction and Occurrence Frequency

Site   Location 	First Year 		Second Year 		Two Years
No.   
                    Direction Occurrence      Direction Occurrence      Direction Occurrence
                                  %                         %                          %
____________________________________________________________________________________________

 1    Bridgewater     S          9.3            WNW       12.6            WNW        10.0
 2    Port Fairy      S         11.0            NNW       11.9            NNW         9.7
 3    Port Campbell  NNE        10.7             SW        9.5            NNE         9.6
 4    Apollo Bay      E         12.2            WNW       16.5            WNW        14.2
 5    Queenscliff     S          9.7             W        14.6             W         11.8
 6    Kilcunda North NNE        10.7             N        12.0             N         11.0
 7    Cape Liptrap   ENE        10.8            WSW       13.6            WSW        11.9
 8    Yanakie        ENE        15.6            WSW       13.2            ENE & E    13.4
 9    Toora           W         13.3             W        16.1             W         14.7
10    Swan Reach      W         10.0             W        10.3             W         10.2
____________________________________________________________________________________________


(ii) Winter northwesterlies and summer easterlies and westerlies

This general regime was displayed by the four most southern sites, Apollo Bay, Cape Liptrap, Yanakie and Toora.

At Apollo Bay, northwesterlies predominated in the colder semester from April to September, while in the warmer semester from October to March, the incidence of easterlies and westerlies increased.
At the three sites in South Gippsland, the northwesterlies predominated for a shorter period (May to August), and easterlies and westerlies again prevailed between October and March.

Subdivision into three-hourly wind roses in Figures 1d to l0d revealed that at Bridgewater, Port Fairy, Port Campbell, Queenscliff, Kilcunda North and Swan Reach, the southerly winds in the summer months were generally perpendicular to the coast and they occurred predominantly in the afternoon, indicating advent of a sea-breeze. Lighter northerly-component winds occurred most frequently at night in May, June, July and August, symptomatic of development of a land-breeze. At Apollo Bay, Cape Liptrap, Yanacie and Toora, the easterlies and westerlies tended to occur most frequently in the afternoon in the summer months, perhaps suggesting deflection of the sea-breeze by the ranges inland.

Appendices II, III and IV table joint frequency distributions of wind direction against each of wind speed, wind energy flux at 10 m, wind energy at 60 m, and standard deviation of wind direction, for the first year, the second year, and both years combined, at all ten sites. Also given are mean values for each parameter for each wind direction.


Figure 24 compares the variation in wind speed and standard deviation of wind direction with wind direction frequency occurrence, for the two-year monitoring period.
This figure shows:
(i) There was a close correspondence between the mean speed and the occurrence frequency of winds from different directions. That is, the mean wind speed was generally fastest for the most frequent directions and slowest for the least frequent directions. Actually, at the majority of sites there was a tendency for the wind to back one quarter-cardinal point between the direction with the highest mean speed and the direction with the highest frequency occurrence.

(ii) A strong inverse correlation existed between frequency occurrence and mean standard deviation of direction for different wind directions.

(iii) The mean speed and the mean standard deviation of direction were generally inversely related.

Table 4 lists the mean standard deviation of wind direction and the average wind speed at each site for both years of monitoring. This table indicates that for nine of the ten sites (the exception being Swan Reach), the year with the faster average wind speed (the second year for all sites except Port Fairy, as shown in Section 5.4) had the lower mean standard deviation of direction. Although the trend was evident, there was not complete conformation between sites with the inverse relationship between average speed and mean standard deviation of direction. This reflects the fact that wind speed alone does not determine the standard deviation of wind direction.


5.3 Speed Distributions

Graphs of speed distributions (histogram, cumulative frequency and Weibull curves) for the two-year monitoring period are plotted in Figures 11 to 20 for sites 1 to 10 respectively. Figure 22 gives graphs for the two-minute logger at site 6.


Figures 11a to 20a are speed distributions for the first year of monitoring (1 February 1985 to 31 January 1986), and Figures 11b to 20b are distributions for the second year of monitoring (1 February 1986 to 31 January 1987). Figures 22 and 22b are the corresponding distributions for the two-minute logger at Kilcunda North.


The ordinates (vertical axes) on the histograms in these figures have different scales, which must be allowed for when making comparisons. Bearing this in mind, speeds recorded at Port Campbell, Apollo Bay, Kilcunda North, Cape Liptrap, Yanakie and Toora, have distributions with a higher proportion of values clustered around the mean, producing histograms with a broad, relatively flat peak (platykurtic) and cumulative frequency plots with a more gradual gradient. At the other sites the speeds form histograms with a sharper peak (leptokurtic) and cumulative frequency distributions with a steeper gradient. At Swan Reach particularly, and Apollo Bay as well, the relatively low modal value produces a positively-skewed distribution.


Weibull distributions fit the two years of data with correlation coefficients of between 0.90 and 0.93. The Weibull constant, C, which is the speed that is exceeded l/e (0.368) of the time, is between 1.10 and 1.14 times the average speed for all sites. The close correlation between the Weibull constant and the average speed at each site is illustrated in Figure 25. For the two-year monitoring period, the least-squares line of best fit is given by C = 1.159V - 0.262 and the correlation coefficient is 0.9988.

Table 4 shows the variation in the Weibull parameters, C and k (the exponent or shape factor), at all sites during the two-year monitoring period. The more variable the wind speed the smaller the value of k. The smallest value of k was produced by the wind regime at Swan Reach, which displayed the greatest diurnal variation, as shown in Figure 23.

For several sites in New Zealand, Cherry (1980) found that the average wind speed (V) and the Weibull exponent (k) were empirically related by:

k = 1 + 0.48(V-2)0.51

For the two-year period at the ten wind monitoring sites, the same form of relationship between k and V is best expressed by:

k = 1 + 0.41(V-2)0.61

with a correlation coefficient of 0.86.

5.4 Average Speeds

The average monthly speeds at all sites are given in Table 5 and Figure 23. These presentations show that for the two-year monitoring period:

(a) The annual average speed was 8 m/s or more at Kilcunda North and Toora; and it exceeded 7 m/s at Bridgewater, Apollo Bay, Port Fairy and Cape Liptrap; 6 m/s at Yanakie and Port Campbell; 5 m/s at Queenscliff; and 4 m/s at Swan Reach.

(b) Throughout both years, monthly average speeds exceeded 7 m/s at Toora; 6 m/s at Bridgewater, Port Fairy, Apollo Bay, Kilcunda North and Cape Liptrap; 5 m/s at Port Campbell and Yanakie; 4 m/s at Queenscliff; and 3 m/s at Swan Reach.

(c) The monthly average speed was slowest in May at all sites except Apollo Bay, where it was slowest in March. The monthly average speed was fastest in October at all sites except Port Campbell and Apollo Bay, where the maximum was in July, and Yanakie and Swan Reach, where the maximum was in December.

(d) Variations in wind speed were not always uniform along the coast. The most pronounced examples were in May, June and July, when winds at the four most easterly sites were generally faster in 1986 than in 1985, while at the four most westerly sites as well as Kilcunda North, speeds were slower in 1986 than in 1985. The opposite trend occurred in January, when winds along the west coast tended to be faster in 1987 than in 1986, while on the central and eastern coast, speeds were generally slower in 1987 than in 1986.

(e) The annual average wind speed was faster in the second year than in the first year at all sites except Port Fairy, with the difference being most pronounced (0.4 m/s to 0.5 m/s) at the three sites in South Gippsland. Figure 23 shows wind speeds at these three sites particularly were generally faster at all times of the day and in most months throughout the second year.

(f) All sites exhibited the normal diurnal variation in wind speed, with the maximum occurring in the afternoon and the minimum generally occurring before midnight in Western Victoria and after midnight in Central and Eastern Victoria. The diurnal variation was relatively small, 1.5 m/s or less, at all sites except Yanakie and Swan Reach. At these latter sites the diurnal range in wind speed was more pronounced, exceeding 2 m/s and 3 m/s respectively. The marked diurnal variation at Swan Reach is more typical of a continental location.

5.5 Maximum Speeds

The time taken for a wind run of 140 m was monitored by the data logger and the speed over the resultant interval calculated. At the end of every hour, the average, fastest and slowest speeds recorded during the hour were stored in the memory of the hourly logger and output to the connected printer.

Table 6 details the fastest speed, and the highest average one-hour and one-day speeds recorded at each site during the two years of monitoring. The fastest speed was 39 m/s (over 3.6 seconds) between 1200 hours and 1300 hours on 29 July 1985 at Apollo Bay. The highest one-hour average speed of 26.1 rn/s was recorded at both Apollo Bay, at 0900 houm on 11 June 1985, and Kilcunda North, at 2300 hours on 1 September 1985. The highest daily average speed was 19.3 m/s on 2 July 1985 at Apollo Bay.

On 29 July 1985, a deep low pressure system with embedded cold fronts, tracking south of Tasmania, was responsible for the fastest recorded speed during the two-year monitoring period. This synoptic situation, shown in Figure 26, produced fastest speeds in one of the three categories at seven of the ten sites (the exceptions being Port Fairy, Kilcunda North and Swan Reach).

Although average speeds were generally faster during the second year, maximum speeds at most sites were faster during the first year.


5.6 Wind Energy Flux

The wind energy flux is calculated from the wind speed using the formula:

F = 1/2 p v3
where F = wind energy flux (W/m2),

v = wind speed (m/s),

p = air density (1.225 kg/m3). looks like "p" = Greek letter "rho" !

The wind energy flux is primarily dependent on wind speed, being proportional to its third power. However, the density of air can vary by some 20% over the range of temperature and pressure likely to be experienced along the coast of Victoria. The density will increase from a likely minimum of 1.1 kg/m3 at a pressure of 980 hPa and a temperature of 30°C to a likely maximum of 1.3 kg/m3 at a pressure of 1035 hPa and a temperature of 0°C.

The wind energy flux at 60 m is estimated by extrapolating the wind speed at 10 m to 60 m using the power law:
v60 = v10 (60/10)0.143
where v60 = wind speed extrapolated to 60 m,
v10 = wind speed recorded at 10 m.

The monthly wind energy fluxes at 10 m for all sites are listed in Table 7 and plotted in Figure 23. These presentations show that for the two years of monitoring:

(a) The annual average flux at 10 m exceeded 500 W/m2 at Apollo Bay, Kilcunda North, and Toora; 400 W/m2 at Bridgewater and Port Fairy; 300 W/m2 at Cape Liptrap and Yanakie; 200 W/m2 at Port Campbell; and 100 W/m2 at Queenscliff and Swan Reach.

(b) As with speeds, the annual average wind energy flux was higher for the second year than for the first year at all sites except Port Fairy. Again the difference was most pronounced (60 W/m2 to 100 W/m2) at the three sites in South Gippsland.

(c) Throughout both years, monthly wind energy fluxes at 10 m exceeded 300 W/m2 at Kilcunda North and Toora; 200 W/m2 at Bridgewater, Port Fairy and Apollo Bay; 100 W/m2 at Port Campbell, Cape Liptrap and Yanakie; while at Queenscliff and Swan Reach, one and eleven respectively of the 24 months were below 100 W/m2.

(d) The monthly wind energy flux was highest in July or October at all sites except Queenscliff, where it was highest in November, and Yanakie, where it was highest in February. The monthly wind energy flux was lowest in May at all sites except Apollo Bay, where it was lowest in March.

(e) Being proportional to the cube of wind speed, the wind energy flux exhibits much greater variations from month to month. The ratio of the maximum and minimum monthly wind energy fluxes at 10 m during the two years was lowest at Toora (2.5) and also less than three at Bridgewater (2.6), Port Fairy (2.6), Kilcunda North (2.8) and Cape Liptrap (2.8); less than four at Yanakie (3.0), Queenscliff (3.3) and Port Campbell (3.5); less than five at Apollo Bay (4.4); and highest at Swan Reach (5.4). The absolute difference between the highest and lowest monthly fluxes at Apollo Bay was 860 W/m2.

A measure of the variability in wind energy flux is the coefficient of variation (equal to the standard deviation divided by the mean, expressed as a pecentage), which is calculated for each site in Table 8. The coefficient of variation for the two-year period increased from 20% at Port Fairy to 24% at Queenscliff and Kilcunda North, 25% at Bridgewater, Port Campbell and Toora, 27% at Cape Liptrap, 29% at Yanakie, 36% at Apollo Bay and 39% at Swan Reach.

The annual average wind energy flux at 60 m, listed in the last column of Table 8, is 2.16 times ( equal to [6 0.143]3 ) the energy flux at 10 m.


5.7 Ranking of Sites

The ten sites are ranked according to wind energy flux for the two-year monitoring period in Table 8. As well as the wind energy flux at 10 m, also listed for each site in this table are elevation, data recovery, annual average wind speed and standard deviation of monthly average wind speeds, standard deviation and coefficient of variation of wind energy flux at 10 m, and annual average wind energy flux at 60 m.

For the two years of monitoring, Apollo Bay recorded the highest 10 m wind energy flux, 579 W/m2. This was only slightly more than the site ranked second, Kilcunda North; with 563 W/m2, and almost 10% more than the site ranked third, Toora, with 533 W/m2. The wind energy flux at Bridgewater, 445 W/m2, the fourth-ranked site, was approximately 25% less than Apollo Bay. Port Fairy, 402 W/m2, Cape Liptap, 361 W/m2, and Yanakie, 358 W/m2, ranked fifth, sixth and seventh, had wind energy fluxes roughly two-thirds of the value at Apollo Bay. The sites ranked eighth, ninth and tenth, Port Campbell, 279 W/m2, Queenscliff, 198 W/m2, and Swan Reach, 113 W/m2, had wind energy fluxes roughly one-half, one-third and one-fifth respectively of that at Apollo Bay.

With the exception of Apollo Bay, which recorded the highest wind energy flux and only the equal third fastest average wind speed, the ranking of sites by wind energy flux conformed with the ranking by wind speed. Apollo Bay recorded the highest energy flux of all sites because it experienced the greatest incidence of fast winds (greater than 15 m/s), as can be seen in the skewed histogram of Figure 14. Since energy flux is proportional to the cube of wind speed, these fast winds made a disproportionate contribution to the wind energy flux.

Although Apollo Bay had the highest wind energy flux, Kilcunda North, Toora and Bridgewater would probably be better sites for harvesting wind energy (depending upon turbine characteristics) because their fluxes are more consistent throughout the year (as shown in Tables 7 and 8 and Section 5.6).

For the two-year monitoring period, a reasonable correlation existed between site elevation and wind energy flux. Although not strictly in order, the three sites with the highest elevations (Apollo Bay, 287 m, Toora, 257 m and Kilcunda North, approximately 165 m) experienced the highest energy fluxes, and the lowest energy fluxes were recorded at sites with the lowest (Swan Reach, approximately 10 m) and third-lowest (Queenscliff 64 m) elevations. The relationship between site elevation and wind energy flux is plotted in Figure 27. The straight line of best fit has a correlation coefficient of 0.83, while Spearman's coefficient of rank correlation for the ranking of site elevation and wind energy flux is 0.88.


6. WEATHER DURING MONITORING PERIOD

6.1 Climatic Summary

During the first year of monitoring, Melbourne Regional Office experienced weather characterised by slightly higher than average mean temperature (15.5°C compared with 14.9°C), surface pressure (1016.8 hPa compared with 1016.2 hPa), and rainfall (684 mm compared with 656 mm), while evaporation (1076 mm compared with 1382 mm) and wind speed (2.4 m/s compared with 3.3 m/s) were considerably lower than normal.

During the second year of monitoring, Melbourne Regional Office experienced weather in which the mean temperature (15.1°C) and surface pressure (1016.1 hPa) were very close to their long-term average values (14.9°C and 1016.2 hPa respectively). Rainfall (562 mm compared with 656 mm), evaporation (1107 mm compared with 1372 mm) and wind speed (2.8 m/s compared with 3.3 m/s) were considerably (more than 15%) lower than normal. The growing number and height of multistorey buildings in the Central Business District would be increasingly sheltering the site exposure of the observatory, thereby contributing to a reduction in the wind speed and evaporation.


6.2 Wind Speed at Tullamarine Airport

The Dines Anemograph at Tullamarine Airport was used as the baseline station for the purpose of comparing wind speeds for the two-year monitoring period with the long-term average. There were several reasons for choosing Tullamarine Airport:


(i) Being near the centre of the coastline of Victoria, it would probably be the most representative location for all ten sites.

(ii) Wind speed has been measured at the site for more than 15 years, providing an extensive long-term record.

(iii) The anemometer exposure is unobstructed, unlike the Melbourne Regional Office.

(iv) The recordings are readily available.

The data used in comparing the monitoring period and the long-term average comprise three-hourly recordings. The long-term averages are compiled from recordings for 15 years from 1971 (just after the airport commenced operations) to 1985 inclusive.


Table 9 and Figure 23 compare the average wind speed for each month from February 1985 to January 1987 with the 15-year mean at Tullamarine. Table 9 shows that average speeds for the first year (4.7 m/s), the second year (4.8 m/s), and the 15-year mean were within 0.1 m/s. These results for Tullamarine confirm that the decrease from the long-term average in the wind speed during the monitoring period at Melbourne Regional Office was largely caused by a progressive increase in sheltering from the growing city skyline.

A comparison was also made of synchronous and long-term wind direction frequencies at Tullamarine Airport. The results, graphed in Figure 23, show that during both years of monitoring, directions were close to average apart from southerlies, which occurred more often, and westerlies, which occurred less frequently, than the long-term average in the first year, and south-westerlies, which were slightly more prevalent, and northerlies, which were correspondingly less frequent, in the second year.

TABLE 9

COMPARISON OF WIND SPEED (m/s) DURING MONITORING

PERIOD WITH LONG-TERM MEAN AT TULLAMARINE AIRPORT

_________________________________________________________________________
              Fifteen-Year
               Average            lFeb 1985 to               1Feb 1986 to
Month         (1971-1985)         31 Jan 1986                31 Jan 1987
_________________________________________________________________________
January          4.5                 4.8                        4.5
February         4.2                 4.8                        4.5
March            4.0                 4.6                        4.1
April            4.1                 4.3                        4.6
May              4.7                 4.0                        4.1
June             4.7                 5.2                        4.2
July             5.5                 5.2                        5.5
August           5.6                 5.5                        5.1
September        5.4                 4.4                        4.7
October          4.8                 4.8                        5.6
November         4.4                 4.2                        5.1
December         4.5                 4.1                        5.0
_________________________________________________________________________
Annual           4.7                 4.7                        4.8


6.3 Wind Speed Comparison and Prediction

Figures 28 to 32 graph the monthly variation in wind speed for each site (two per figure) and Tullamarine Airport during the two-year monitoring period. Also plotted are the 15-year averages for Tullamarine.

Crude predictions of the long-term average speed at each site have been calculated from the ratio of long-term and concurrent average speeds at Tullamarine. These predictions are also included in Figures 28 to 32. The variation between synchronous and long-term wind speeds may not be uniform across the state, as happened in May, June and July, when winds in the east were faster in 1986 than in 1985, while in the west the opposite situation occurred. This variability, together with the fact that the monitoring period comprises only two years of data for each site, means that predictions are only very approximate.

These crude predictions suggest that the long-term average speeds could be:

5% to 15% lower than the averages recorded in February, March, April and June of 1985 and January, February, April, October, November and December of 1986;

5% to 15% higher than the averages recorded in July, November and December of 1985, and May, June, August and September of 1986;

20% higher than the averages recorded in May and September of 1985.

On balance the annual average wind speeds recorded during both years of monitoring are likely to be close to the long-term annual average.

7. CONCLUSIONS

Two years of monitoring in a study to evaluate the potential for wind-powered electricity generation at ten sites along the coast of Victoria have produced the following ranking of sites (with annual average wind energy fluxes at l0m in W/m2):

                            Both Years         First Year       Second Year

Apollo Bay                     579                 569              588
Kilcunda North                 563                 544              583
Toora                          533                 485              580
Bridgewater                    445                 436              453
Port Fairy                     402                 417              388
Cape Liptrap                   361                 329              390
Yanakie                        358                 326              393
Port Campbell                  279                 274              283
Queenscliff                    198                 193              202
Swan Reach                     113                 103              123


Annual average wind speeds and energy fluxes were higher in the second year than in the first year at all sites except Port Fairy. The increases were greatest (around 20%) at the four sites in South and East Gippsland (Cape Liptrap, Yanakie, Toora and Swan Reach). The consequent interchanging of fifth and seventh positions by Port Fairy and Yanakie was the only change to the ranked order of sites between the first and second years of monitoring.

Although Apollo Bay had the highest annual wind energy flux, Kilcunda North, Toora and Bridgewater would probably be better sites for harvesting wind energy (depending upon generator characteristics) because their fluxes are more consistent throughout the year.

In the two years of monitoring, the fastest recorded gust was 39 m/s at Apollo Bay, the highest one-hour average speed was 26.1 m/s at both Apollo Bay and Kilcunda North, and the highest daily average speed was 19.3 m/s at Apollo Bay.

There was a general trend to higher wind energy flux with increasing site elevation.

Comparison of synchronous and long-term wind speed data for Tullamarine Airport has suggested that long-term energy fluxes could be higher than the recorded values in some months (particularly May and September) and lower in others (February and April), although the long-term annual averages are likely to be similar to averages for the two years of monitoring.


8. REFERENCE

Cherry, N.J. (1980). Wind Energy Resource Methodology. Journal of Industrial Aerodynamics, 5: 247-280.


9. ACKNOWLEDGEMENTS

Thanks are due to all the landowners, one property manager and one neighbour, whose regular checking of loggers has made an important contribution to the high data recovery.

The assistance readily given by Steve Howard, of Tain Electronics Pty. Ltd., in responding to requests to service loggers at all times of the day and night, warrants special mention.

Also gratefully acknowledged is the help provided by Gavin Sinnott and Fred Bolwell during site installation, when the full aeolian potential of some of the sites was realized. Tower dismantling and site restoration was undertaken with assistance from Anthony Piggin and David Croft.

Important notes concerning this electronic version.

End Ende Fin