Vortex Solar

Vortex Solar http://www.vortexfdc.com/ is an on-line modelling service which provide solar radiation and meteorological data derived from mesoscale modeling and satellite information

It provides namely solar radiation required for the simulation within PVsyst, with a spatial resolution 3 km:

Global horizontal irradiance (GHI or GlobHor)
Diffuse horizontal irradiance (DHI or DiffHor)
Air temperature (Tamb)
Direct normal irradiance (DNI or BeamHor)
Wind speed (Windvel)

The service is available for pay.

The TMY (Typical Meteorological Year) provided by Vortex Solar is based upon a 20 years period time, with Hourly time-series built by concatenation of representative months. Both the P50 and the conservative meteorological year (P90) are availables.

Details of the method

(taken from vortex_solar_specifications.pdf, Vortex Solar)

The basic tool of Vortex system is the WRF – ARW mesoscale model, version 3.6 (Skamarock et al., 2008). This is an open source meteorological model used by the atmospheric community. It consists of a dynamical core which includes all basic equations for the atmospheric flow (momentum, continuity, energy, water vapor,…) which are numerically solved on a terrainfollowing structured mesh. The model allows nesting several domains with different grid size resolution; for the Vortex Solar products, three domains are nested starting at 0.5 degrees down to 3 km for the finer resolution. The model is run under the initial and boundary conditions given by a reanalysis product; specifically, in this case the NCEP CFSR (Saha et al., 2010) reanalysis is used. Topography and land use, which play an important role in meteorological phenomena, are taken from SRTM and USGS respectively.

Solar radiation transfer through the atmosphere, therefore the solar irradiance that hits the Earth’s surface, is parameterized in WRF. Among the number of parameterization schemes that are available in the current version of WRF, Vortex solar method uses the so-called New Goddard parameterization (Chou and Suarez, 1999), which was developed by NASA Goddard Space Flight Center.

This parameterization includes the absorption due to the most important atmospheric gases (water vapor, ozone, carbon dioxide and oxygen) as well as cloud particles (ice crystals and water droplets) and aerosols. Scattering processes by clouds, aerosols and air molecules are also fully represented. Solar fluxes are integrated over the entire spectrum from 175 nm to 10.000 nm. In order to consider the cloud horizontal non-homogeneities in the sub-grid scale and the treatment of the cloud overlapping, this scheme applies the Maximumrandom overlapping method.

Clouds are of primary importance when estimating solar radiation, either at climatological or at meteorological timescales, that is for resource assessment or for forecasting applied to management of installations. Again, several parameterizations for the cloud microphysics are available in the current version of WRF; a sensitivity analysis and considerations of simplicity and efficiency, along with a literature review, were considered when choosing the scheme for simulating clouds in the Vortex solar system. The chosen scheme is the WRF Single-Moment 5-class scheme, which allows for mixed-phase processes and super-cooled water (Hong et al., 1998).).

Despite recent improvements, no meteorological model can predict perfectly where and when clouds will actually appear. Since clouds are a great modifier of solar radiation reaching the ground, cloud observations are needed for a reliable estimation of solar radiation. Given the scarcity of ground observations of clouds, which do not cover by any means the whole Earth, cloud observations from satellite platforms are preferred in solar energy applications. At the current stage, Vortex solar uses solar radiation estimates by the Climate Monitoring Satellite Application Facility project (CM SAF) from EUMETSAT. Specifically, the surface incoming shortwave radiation product is used, which has daily resolution, 0.25º latitude and longitude resolution, and is derived from polar orbiting satellites (Karlsson et al., 2013; Riihelä et al., 2012). However, for the area under the Meteosat disk (70ºS-70ºN, 70ºW-70ºE) the corresponding product derived from the geostationary satellite is available at hourly resolution and 0.03º latitude and longitude; the surface incoming direct radiation is also available for this area. These higher resolution products are used by Vortex solar system when solar data are requested for this specific area.

Thus, the output of the modeling with WRF is combined in an intelligent way with the observations derived from satellites images. The methodology used (named Re-modeling) involves extending the range of application of the information provided by mesoscale modeling in favor of improving the representativeness, consistency and overall accuracy of the long-term solar radiation series through a non-linear multivariate statistical approach. The methodology aims to provide the best approximation to satellite measurements from modeled data. More specifically, the Re-modeling makes use of a long-term reference high resolution WRF series of a number of meteorological variables at different levels and several years of satellite data carrying information on cloud behavior at the location of interest.

In the Vortex Solar approach:

(1) a multivariate analysis allows to understand the relation between the model variables and discriminate between useful and redundant information;

(2) a non-linear statistical procedure is applied to the model-derived variables together with the satellite observations for the coincident period; and

(3) the physical-statistical model thus obtained is then used to adjust prediction of solar radiation (global horizontal and direct normal irradiances) over the original longterm series (of +20 years).

The primary outputs of the Vortex solar system are Global Horizontal Irradiance (GHI) and Direct Normal Irradiance (DNI) time series, at 1 hour time resolution and 3 km spatial resolution.

Typical Meteorological Year for P50

(taken from TMY_methodology.pdf, Vortex Solar)

The computation is based on the selection of full individual calendar months (January, February, March, …) across a multiannual time series that satisfies certain criteria for the target variables (at least 15 years). Different methods for the selection have been proposed. It is recommended to give higher relevance to global and direct radiation (GHI and DNI, respectively) in the selection. The approach employed by Vortex is based on a simple selection criteria applied to 20 years hourly modeled SOLAR time series.

For each month, the following metrics are derived:

D: distance to long-term daily cumulative sum distribution
A: monthly cumulative value standardized anomaly

An index from the weighted sum of A and D for all the target variables GHI, DNI, DIF and 2m Temperature is obtained for each individual month. The selection picks the month with minimum value for each calendar month to build the TMY time series. A bicubic interpolation is applied to smooth the transition between months. The weights applied are regional specific and depend on the solar technology (PV or CSP) application.

Typical Meteorological Year for P90 (TMYP90)

(taken from TMY_methodology.pdf, Vortex Solar)

Uncertainty penalty needs to be added to the TMY computation. Moreover, bankability analysis usually requires a value of exceedance at 90% confidence based on the overall uncertainty spread, which implies a lower estimated of the annual solar resource (GHI and/or DNI) used in the finance studies. Annual P90 value can be retrieved once the uncertainty is assessed based on combination of model and inter-annual uncertainty. In the Vortex Solar approach, a fixed model uncertainty based on the validation white paper is assumed, with order of 3-4%, while annual standard deviation is used to prescribe internal variability uncertainty. An iterative method is employed to construct a typical P90 year, starting from a first guess based on lowest GHI and DNI values month's selection.