A solar radiation database for Chile

Abstract

Chile hosts some of the sunniest places on earth, which has led to a growing solar energy industry in recent years. However, the lack of high resolution measurements of solar irradiance becomes a critical obstacle for both financing and design of solar installations. Besides the Atacama Desert, Chile displays a large array of "solar climates" due to large latitude and altitude variations, and so provides a useful testbed for the development of solar irradiance maps. Here a new public database for surface solar irradiance over Chile is presented. This database includes hourly irradiance from 2004 to 2016 at 90 m horizontal resolution over continental Chile. Our results are based on global reanalysis data to force a radiative transfer model for clear sky solar irradiance and an empirical model based on geostationary satellite data for cloudy conditions. The results have been validated using 140 surface solar irradiance stations throughout the country. Model mean percentage error in hourly time series of global horizontal irradiance is only 0.73%, considering both clear and cloudy days. The simplicity and accuracy of the model over a wide range of solar conditions provides confidence that the model can be easily generalized to other regions of the world.

Figure 1

This scheme shows the flow of the methodology used to estimate GHI and DNI. Solid thick rectangles represents the models used. The doted rectangles show the input variables for each model. The solid thin rectangles show the output from the models.

Figure 2

Example of the methodology of the cloud detection algorithm for a particular location. Each dot represents the satellite reflectivity with the corresponding clear sky irradiance. The orange solid line is the threshold set by the algorithm, the blue dots represent the clear sky and the orange dots the cloudy cases.

Figure 3

Fitted curve for the empirical model of GHI attenuation by clouds. The yellow dots are the irradiance under cloudy conditions calculated with the first approximation model (equation 8) versus measured irradiance for the same time and location under cloudy conditions. The red line correspond to the fitted quadratic function (equation 9).

Figure 4

For each one kilometer grid point, the daily average insolation of GHI (panel A) and DNI (panel B), for the period 2004 to 2016, were calculated using the global horizontal and direct normal irradiance hourly time-series respectively (which were calculated with the methodology presented in this work). Black dots in panel A show the position of the ground stations used to validate the irradiance database. The calculation and plotting of the daily average insolation for each grid point were made using Matlab software 8.5.0.197613, Academic License number 1086178 (https://www.mathworks.com/).

Figure 5

The red dots represent the percentage of time in which the algorithm correctly classified between clear or cloudy skies at each station. The blue dots represents the percentage of time with cloudiness according to the irradiance measurements. The stations are arranged from north to south (left to right).

Figure 6

The dots represent the daily GHI average for the full period of measurements at each station (x-axis) versus the estimated GHI average for the same period. The segmented and dotted lines mark the difference ranges of 10% and 5% respectively.

Figure 7

The blue and green dots shows the MPE (Mean Percentage Error) and RMSE respectively for the GHI daily average time series at each station.