User Perspectives on Project Feasibility Data Marcel Šúri Tomáš Cebecauer GeoModel Solar s.r.o., Bratislava, Slovakia marcel.suri@geomodel.eu http://geomodelsolar.eu http://solargis.info Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [1]
Solar resource information - user requirements Global coverage, harmonized, validated Data available at any location Long-climate record (up to 15-20 years) harmonized and without gaps High accuracy, low uncertainty High level of detail (temporal, spatial) Modern data products (long-term averages, RMY, time series) Real-time data supply: historical, monitoring, nowcasting, forecasting + Meteo and other geodata for energy modeling All this is possible with satellite-based data, supported by high-quality ground measurements! Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [2]
Project stages 1. Prospecting, prefeasibility and site assessment 2. Feasibility, design optimization, financing and due diligence 3. Performance assessment 4. Management of solar power and energy markets (not considered in this presentation) Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [3]
Project stages 1. Prospecting, prefeasibility and site assessment 2. Feasibility, design optimization, financing and due diligence 3. Performance assessment 4. Management of solar power and energy markets (not considered in this presentation) Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [4]
Prospecting, prefeasibility and site assessment Data needed: Representative and homogeneous annual long-term averages and aggregated monthly statistics of satellite-based data at high spatial resolution At least 10 years of data are needed to represent climate reliably or uncertainty has to be increased Terrain and air temperature are needed for energy modeling Other GIS data (infrastructure, population, etc.) for country analysis Uncertainty (bias) for long-term annual values to be typically expected: DNI ±4 to 15% GHI ±2 to 7% Uncertainty is higher in the complex land cover (land/sea/desert/islands), in mountains (snow/ice) and high latitudes, and in regions with extreme aerosols/humidity Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [5]
Prospecting, prefeasibility and site assessment Services needed: Fast access: on-the-click information Interactive online tools for energy simulation Automated computer access Preliminary assessment site reports Country reports - resource potential analysis Maps, GIS data Uncertainty in energy modeling contributes by few more percent; however some simple algorithms produce systematic errors Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [6]
Project stages 1. Prospecting, prefeasibility and site assessment 2. Feasibility, design optimization, financing and due diligence 3. Performance assessment 4. Management of solar power and energy markets (not considered in this presentation) Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [7]
Feasibility, design optimization, financing and due diligence Data needed: Site-specific solar and meteo time series with long-term data record Representative Meteorological Year (RMY) Ancillary meteo data (air temperature, relative humidity, wind speed and direction) High-resolution terrain data Services needed: Site adaptation of satellite-based data by correlating them to local solar measurements Generation of Representative Meteorological Year Bankable reports: Site analysis of solar resource, Energy yield study Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [8]
Feasibility, design optimization, financing and due diligence Time series This is the only bankable data, which can provide full climate statistics average, median percentiles, interannual variability, P(50), P(90) expectances, probability distributions, etc. 12 to 20+ years of high quality data are available worldwide at primary resolution of approx. 3 to 5 km Quality parameters: Minimum bias, low RMSD (if possible adapted for the site) Representative distribution statistics skewed distribution of parameters (GHI and DNI) results in errors when simulating tilted solar radiation or energy yield For due diligence and financing, the satellite-based time series and derived data products have to be validated using independent sites from the similar climate zone, or site adapted (correlated) with local measurements Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [9]
Site adaptation of satellite-based time series Ground measurements available for a short time period (few months, 1-2 years) They are correlated with time series of satellite-derived irradiance to: Correct systematic errors (reduce bias) Match data frequency distribution Conditions to be fulfilled for successful adaptation: Systematic deviation in satellite data should exist Magnitude of deviation is invariant in time High quality ground measurements are available Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [10]
Site adaptation of satellite-based time series Example: Tamanrasset (Algeria) Simple bias correction Original DNI ground satellite data scatterplot: Bias: -4.2% Correction of frequency distribution Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [11]
Site adaptation of satellite-based time series Example: Tamanrasset (Algeria) Ground measurements must be of high quality and must be properly quality-checked. Result of the procedure: Eliminated (reduced) bias - when compared to local measurements reduced RMSD improved statistical distribution of values Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [12]
Assessment of Solar Resource. Upington Solar Park, South Africa. Reference No. 58-01/2011 Representative Meteorological Year (RMY) Two data sets are derived from site-adapted time series one for P(50) and one for P(90). In assembling RMY, the values of DNI, GHI and Air Temperature are only considered, where the weights are set as follows: 0.6 is given to DNI, 0.4 to GHI, and 0.1 to Air Temperature (divided by the total of 1.1). Includes The P(50) hourly RMY data data set represents, of one for each representative/typical month, the average climate conditions year and the most derived representative from the time cumulative distribution function, therefore extreme situations (e.g. extremely cloudy weather) are not represented this dataset. Therefore, to capture all possible weather situations it is recommended in power series production representing simulations to use long full (17 period years) time series of the data. The P(90) RMY data set represents for each month the climate conditions which after summarization of DNI for the whole year result in the value close to P(90) derived by statistical analysis of uncertainties and interannual variability (the conservative DNI value 2729 kwh/m It is constructed on a monthly basis, comparing months of individual years with 2, see Section 10, Tab. 16). Thus RMY for P(90) represents the year with the lowest annual value of DNI over the period of 17 years. Both RMY data sets include the following parameters: two long-term monthly characteristics: cumulative distribution function and the Direct Normal Irradiance, DNI [W/m 2 ] Global Horizontal Irradiance, GHI [W/m mean. The representative months 2 ] are concatenated into RMY Diffuse Horizontal Irradiance, Diff [W/m 2 ] Azimuth and solar angle [ ] RMY is comparable Air temperature at 2 to metres, the Temper TMY [ C] file, only the weighting is tuned to meet the Wind speed at 10 metres, Wspeed [m/s] modeling needs of either PV (focus on GHI) or CSP/CPV (focus on DNI). Wind direction, Wdir [ ] Relative air humidity, Rh [%] Fig. 11: Snapshot of the P(50) Representative Meteorological Year, RMY Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [13]
Representative Meteorological Year (RMY) Representative Meteorological Year: P(50) RMY data set represents, for each month, the average climate conditions and the most representative cumulative distribution function, therefore extreme situations (e.g. extremely cloudy weather) are not represented in this dataset. To characterize year with very conservative values of solar resource P(90) RMY data set can be derived. It represents for each month the climate conditions which after summarization of DNI (GHI) for the whole year result in the value close to P(90). The P(90) annual value is derived from the uncertainty and interannual variability, thus RMY for P(90) represents closely a year with the lowest annual value of DNI (GHI) over the longer period. There may be some other types of RMY constructed, depending on the criteria Solar resource data in RMY can be supplemented by air temperature, relative humidity, wind speed, and wind direction Meteo data can be supplied from meteorological models Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [14]
Calculation of tilted irradiance and energy simulations These type of calculations have non-linear nature and therefore the results are affected if proper distribution of values (direct and diffuse irradiance and temperature) is not maintained This is a case of average daily profiles and also synthetic time series. Therefore, it is not advised to apply these older data products, if RMY or full time series can be used. Satellite-derived time series and RMY are available today for almost any location worldwide, except polar regions Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [15]
Project stages 1. Prospecting, prefeasibility and site assessment 2. Feasibility, design optimization, financing and due diligence 3. Performance assessment 4. Management of solar power and energy markets (not considered in this presentation) Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [16]
Performance assessment Satellite-derived time series have numerous advantages (compared to ground sensors): Good quality, stable radiometry Available for any location Time availability 99.5%, just few gaps have to be filled by intelligent algorithms Known quality and uncertainty over large areas No problems with pollution, misalignment, data cleaning, calculation of timeintegrated statistics Therefore they can be used for: Performance assessment of power plants Validation of on-site measured irradiance Easy and cheap service for production evaluation of PV systems Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [17]
Conclusions Traditional approaches based on interpolation of point measurements and application of synthetic generators are substituted by satellite-derived time series which have a number of advantages: They are available for (almost) any site globally Have often better overall quality and reliability High-quality data products can be derived: Representative Meteorological Year for planning and design Aggregated statistics for reporting Customized time series for monitoring and system performance assessment Complementary data to ground measurements In the absence of high quality ground measurements satellite-based data offer the only alternative for system monitoring and performance assessment. Solar Resources and Forecasting Workshop, NREL, Boulder CO, 20-22 June 2011 [18]