ODREĐIVANJE VETROENERGETSKOG POTENCIJALA VOJVODINE LINEARNOM INTERPOLACIJOM PODATAKA SA METEOROLOŠKIH STANICA ASSESSMENT OF WIND POTENTIAL IN VOJVODINA BY LINEAR INTERPOLATION OF METSTATIONS DATA N. Vasiljević * and M. Zlatanović ** PDRB Kolubara D.O.O. Lazarevac * Elektrotehnički fakultet, 11120 Beograd, Bulevar Kralja Aleksandra 73 ** Abstract: By linear interpolation of data from meteorological stations the wind potential map of Vojvodina, north region of Serbia, was generated. As a reference source of long term wind data eight met stations were selected. From this data and local terrain roughness, orography and obstacles data in surrounding area, the regional wind climate tables were calculated using WAsP methodology and licensed software. Digital topography maps were used to generate local wind climate atlases. Validity of each local wind atlas in 20km*20km area around the met station was assumed based on mostly plain and homogeneous orography of Vojvodina. For the regions outside 20km*20km areas covered by met station data, linear interpolation of wind data was used between minimum two, and maximum three nearest wind atlases to complete Vojvodina wind atlas. As the result, 66 single wind atlases were derived, 58 of them by linear interpolation method and 8 from original met station data. By assembling regional wind atlases the wind atlas of Vojvodina was generated. Key words: Observed Wind Climate, Renewable Energy, WAsP, Wind Atlases, Wind Energy 1. INTRODUCTION In April 2008 the total world installed capacity of wind turbines reached the figure of 100GW. Today wind energy supplies 5% of the European Union electricity while 12% is planned for 2020, which assumes installation of 180GW wind turbines power. The benefits of using wind energy include economy, employment, security of supply, electricity market and environment benefits. As the world leader in wind energy sector EU has today 150 000 jobs and 380 000 are expected in 2030. General opinion based on existing data is that all European countries have good wind energy potential. European wind atlas prepared by Risø Institute in 1989 was completed by atlases of new EU and candidate countries [1]. The wind atlas of Serbia prepared by the same methodology still is not published. Some reports contain preliminary data based on meteorological stations observation with indicated areas of good wind potential [2,3]. Two projects based on specialized wind measurement stations data financed by Serbian Ministry of Science and Spanish government have objective to produce wind and solar energy potential atlas of Serbia, but the final results are still not presented. Wind potential map of South East Banat prepared by the team of professor Zlatanovic confirms good wind energy potential of selected region according to measurements performed on 50m height wind mast and on two stations at 33m [4]. Questions related to sensors properties and validation of meteorological data collected from different internet web sites were discussed in references [5,6]. Wind map of former Yugoslavia region using model of the most probable atmospheric state carried out by the ARMINES Institute does not predict very good wind potential of Serbia. 1
Most existing data predict significant wind potential of Vojvodina, the region in which the wind energy was used back in history. The terrain of Vojvodina is very open and flat or gently undulating and corresponds to roughness class 1 (z 0 =0.03 m) according European wind atlas, except Vršac s mountain and Fruška Gora that together make no more than 10% of whole area. We present wind atlas of Vojvodina generated by licensed WAsP software tool and linear interpolation method of metstations data [7,8]. The application of WAsP is fully justified since the terrain is mostly homogeneous and accuracy of simulation of wind over this region is accurate. 2. METHOD OF WIND ATLAS GENERATION Publicly available data were used in wind map of Vojvodina preparation and processed according to European Wind Atlas methodology including verification procedure. Several databases were created such as wind characteristics, orography, roughness and obstacles. The internet is selected as the only source of data including digital orography maps, so the reported results may be reproduced by using WAsP software tools and linear interpolation method. The step by step procedure of map generation and verification is described. 2.1 Meteorological data As the member of world network of meteorological stations the Republic Hydrometeorological Service of Serbia sends minimum 8 hours values of meteorological data for international exchange. This data are publicly available for example on the internet site http://meteo.infospace.ru as integer values. The validity of such data is discussed in Ref. [5]. In our procedure we used data from 8 meteorological stations located in the region of Vojvodina. These stations are: Palić, Sombor, Kikinda, Novi Sad, Sremska Mitrovica, Banatski Karlovac, Vršac and Surčin. Data were collected from 3.3.2000 to 3.3.2008 and, in the best case, consist of data recorded every 3 hours. Using the software tool OWC Wizard, which is the part of WAsP Climate Analysis, OWC (Observed Wind Climate) files based on filtered wind speed and direction data were calculated. 2
Fig. 1. Orography map of Vojvodina with the locations of meteorological stations. The color scale represents relative terrain altitude These files represent wind climate data base for certain location, i.e. contain 16 sectors wind rose with mean speed data for each sector and overall mean speed for all sectors. Also they consist of statistical Weibull distribution parameters, like shape parameter k (non-dimensional quantity) and the scale parameter A (dimensions of speed m/s), for each sector as well as overall non directional data. 2.2 Orography For preparation of Vojvodina s orography map, data were downloaded from NASA s Shuttle Radar Topography Mission web site, actually its ftp server ftp://e0srp01u.ecs.nasa.gov. After importing these data in to the application Global Mapper, Vojvodina s orography map was derived as shown in Fig. 1. Obtained orography map were converted in the form that can be used in WAsP software tool by generating contours in required form. Procedure of generating contours was done also in Global Mapper so that contour lines were generated with 10m spacing to make suitable vector terrain map. 2.3 Generating wind atlases For terrain roughness of Vojvodina region generalization was applied. Since this region is a plain which surface is mostly farmed for cereal cultures, roughness class 1 (z 0 =0.03 m) was used in general. Description of obstacle groups surrounding meteorological stations was made on the basis of satellite pictures using Google Earth. After all of these steps databases were generated and prepared to be imported in WAsP for the purpose of wind atlas generation. Each wind atlas project file in WAsP contains the following layers: OWC file, contour vector map, rose of terrain roughness and obstacle group surrounding the meteorological station. The last step prior to generation of wind atlas in WAsP was locating meteorological station on contour vector map from which data were derived in the form of OWC file. After this last step WAsP was fully prepared for selected region wind atlas calculation. Wind atlases contain the site-independent or regional wind climate, derived from the wind measurements at a meteorological station through the wind atlas analysis. The observed wind climate (OWC) has thereby been reduced to certain standard conditions, i.e. wind roses and wind speed distributions for five standard heights and four roughness classes in a number of sectors (usually 12). Since we had wind data from 8 meteorological stations together with terrain description we derived 8 so called primary or reference wind atlases. 2.4 Interpolation Area of Vojvodina amounts approximately 24 500 km 2. In the best case if we have terrain of the lowest roughness, which is equivalent to the calm sea surface, and using the assumption that the terrain is absolutely flat, single wind atlas can be applied on to the area no more than 50km*50km=2500 km 2. In our case if we presume that meteorological stations are located ideally, we need at least data from 10 meteorological stations. Even in this idealized case we still have lack of data. Because of this reason, linear interpolation independent of direction was applied between existing wind atlases for the purpose of filling the lack of data for specific areas of Vojvodina. The whole region of Vojvodina was divided into 66 sectors of 20km*20km area covering complete Vojvodina (Fig. 2). In Figure 2 the primary atlas regions are indicated by pink squares. They represent areas in centre of which the meteorological stations are located and wind atlases for these areas were derived from local meteorological stations data. For each sector map beyond the extent of primary atlases, wind atlases were derived using linear interpolation between minimum two, and 3
maximum three nearest wind atlases, which were derived from meteorological stations data and terrain description. The squares of different colors are indicated in Fig. 2 which center represents virtual locations of wind atlases derived using linear interpolation. Fig. 2. Map of Vojvodina divided into 66 sectors of 20km*20km areas. Reference sectors are pink colored (see the text) For the illustration on how interpolated atlases were generated the next example is given. Wind atlas derived from Sombor meteorological station data (UTM coordinates: 356149.9, 5069788) fully describes regional wind climate for area of 20km*20km, which center is in fact location of this meteorological station. However, wind climate of areas surrounding Sombor wind atlas cannot be describe only by this atlas data due to the influence of the other wind atlases. Location SNP1 with UTM coordinates 377835.398, 5070068.243 is also center of an 20km*20km area positioned east of meteorological station Sombor and has the regional wind climate that can be described taking into consideration two more wind atlases beside Sombor atlas (Fig. 3). These two atlases are wind atlases of Palić (UTM coordinates: 404672.6, 5105898) and Novi Sad (UTM coordinates: 409889.5, 5020624). At this point we have three wind atlases, and the simplest way of interpolation between them is linear interpolation independent of direction. Each of these three atlases takes a part in interpolated atlas with certain weight. The distance of the location representing the center of considered sector area to certain of surrounding atlases determines the weighting factor for primary sector atlas data to be used in interpolation procedure. In this particular case of wind atlas for the location SNP1, wind parameters can be calculated from the following formula (SNP1)= 57%(Sombor) + 25%(Palić) + 18%(Novi Sad) (1) where weighting factors are given in percents and terms in pharantheses represents value of 4
considered parameter at indicated location. The weighted factors are calculated as the relative distance from SNP1 location to the meteorological stations of selected reference atlases. Fig. 3. Illustration of weighted factor calculation for SNP1wind atlas interpolation procedure This example describes the method used in generation of interpolated wind atlases for 58 locations that do not contain meteorological station. In this way, from 8 reference wind atlases made using local meteorological stations data and terrain description 58 atlases were obtained by the interpolation procedure. LIB Interpolator LT software tool made by Risø Institute was used for interpolation. 2.5 Resource grids After generation of necessary atlases we obtained wind climate database for whole region of Vojvodina in the form of 66 sector tables. To get wind atlas of Vojvodina in Geographic Information System (GIS) form we have to insert these atlases into topography map areas for which 5
they describe wind climate. Previously made contour vector map of Vojvodina was also divided in 66 sector maps of 20km*20km area as shown in Fig. 2. Calculated 66 single wind atlases directly corresponds to 66 vector maps that make whole contour vector map of Vojvodina. Importing corresponding wind atlases with wind turbines in center into contour vector map, we can derive and export wind resource grids from WAsP program. Optionally roughness description can be added to get more site specific data. Over resource grid points WAsP calculates the following data: the elevation, the mean wind speed, the mean power density, the Weibull-A value, the Weibull-k value, ruggedness index RIX (optional) and the annual energy production if a wind turbine generator is associated. Particularly in the case of wind potential map we only calculated all sectors mean wind speed distribution for standard heights 10m and 50m and we added 33m wind potential map to be compared with specific wind data measured at this height. Putting in one map all 66 local wind atlases the Wind atlas of Vojvodina was generated. 3. WIND ATLAS OF VOJVODINA As the result of successive applying method of creation of resource grids for all 66 wind atlases and appropriate 66 contour vector maps, we derived 66 resource grids of the mean wind speed (all sectors) at standard heights of 10m and 50m and additionally at 33m. Connecting these resource grids in Global Mapper high resolution map of wind potential was made. Actually three maps were made for heights of 10m, 33m and 50m above the ground level. Finally applying similar method as used for contour vector map generation the regions with mean wind speed in steps of 0.5 m/s were shown. Wind atlases of Vojvodina at three heights above ground obtained by described method are shown in figures 4, 5 and 6. Fig. 4. Mean wind speed for all sectors at 10 m height 6
Fig. 5. Mean wind speed for all sectors at 33 m height Fig. 6. Mean wind speed for all sectors at height 50 m 7
4. VERIFICATION OF RESULTS A general method for verification whether regional statistics based on particular metstation data is adequate to represent wind potential around the station is to compare the estimates and measurements at specific location. We selected metstation at Vrsac airport to compare one year measurements at 33m height with prediction from metstation. The distance between the measuring mast and station was less than 8km. For the mean wind speed calculation we assumed logarithmic wind profile with height. Measured mean speed for one year was 4.1m/s, while based on metstation data 4m/s speed was predicted which confirms that both measurement locations are in the same wind climate region. From wind atlas wind speed between 4m/s and 4.5m/s was expected, but more slose to 4.5m/s value. Intercomparison between two metstations data situated in basically the same climatological condition is also suggested as a verification method. Since we used linear interpolation this method gives good results in flat terrain. The highest pick of Vojvodina is located at Vršačka brda. Metstation Banatski Karlovac is located 24km west from Vršačka brda in flat region while Vršac metstation is situated 2km north from Vršačka brda where the wind is significantely influenced by orography. The results of metsations data intercomparison are given in Table 1. Measured mean wind speed at Vršac was 3.2m/s and based on Weibull statistics (WAsP) 3.3m/s. For Banatski Karlovac both measured and Weibull mean wind speed are the same. The estimated wind speed at Vršac based on Banatski Karlovac data is 3.64m/s and measured was 3.2m/s. The estimated wind speed at Banatski Karlovac based on Vršac data is 3.52m/s while measured speed was 3.87m/s. Wind speed at Vršac was overestimated nearly 14% and wind speed at Banatski Karlovac was underestimated 10%. This indicates the regional wind climate difference at metstations locations. Table 1. Cross prediction of annual mean wind speed at metstations Vršac Banatski Karlovac Vršac WAsP measured 3,3 3,2 3,52 Banatski Karlovac WAsP measured 3,64 3,84 3,87 5. CONCLUSIONS The method of generating wind atlas applied to Vojvodina region was presented. It was shown that south-east region of Vojvodina has the best wind potential. Wind measurement campaign performed during two and half years at the heights up to 50m above the ground level at three locations confirmed this conclusion. The specialized wind potential measurements and verification procedure support wind statistics application for wind atlas generation and show that direction independent mean wind speed prediction of presented map is accurate with the error inside 10%. As an exception, it was found that orography near Vršačka brda hills significantly influences wind speed leading to intercomparison error of nearly 14%. Wind potential map does not overestimate the wind potential since at some regions annual mean wind speed over 6m/s was measured at 50m above ground level. Generated wind potential atlas of Vojvodina may serve as a useful tool for selection of the regions for siting procedure intended for wind energy exploitation. ACKNOWLEGMENT The authors gratefully thanks to the Dr. Morten Nielsen, Risø National Laboratory for Sustainable Energy, for giving permition to use the LIB Interpolator LT software. 8
REFERENCES [1] Ib. Troen and Erik Lundtang Petersen, "European Wind Atlas," Risø National Laboratory, Roskilde, Denmark, 1989. [2] R. Putnik, S. Grkovica, R. Milojkovic et.al. Possibililty for wind conversion to electricity, EPS, Belgrade, 2002 (In Serbian) [3] Gburčik Petar et. al., Study of potential of Serbia for exploitation of wind and solar energy, MNZŽS, National energy efficiency program, EE704-1052A, 2005 (in Serbian) [4] M. Zlatanović, Assessment of wind potential at the site Rošijana for eco house energy supply, EWEC 2008, Bruxelles, http://www.ewec2008proceedings.info/ [5] M. Zlatanović, Comparison of wind mast measurements and meteorological wind data at the Deliblatska Peščara site, EWEC 2006, Athens, http://www.ewec2006proceedings.info/ [6] M. Zlatanović, I. Popović and Z. Gršić, Field test comparison of different wind sensors, EWEC 2008, Bruxelles, http://www.ewec2008proceedings.info/ [7] Niels G. Mortensen, Duncan N. Heathfield, Lisbeth Myllerup,Lars Landberg and Ole Rathmann, " Getting Started with WAsP 9," Risø National Laboratory, Technical University of Denmark Roskilde, Denmark, June 2007. [8] A. J. Bowen and Niels G. Mortensen, "WAsP prediction errors due to site orography," Risø National Laboratory, Roskilde, Denmark, December 2004. 9