Solar Ecology

Solar Genome Project History


The Solar Genome Project started with the investigation of the possibility of using machine learning (ML) algorithms to estimate the different components of the solar irradiance (i.e. direct normal and diffuse irradiance) from the measurement of the solar irradiance performed by an array of multiple pyranometers (aka Mutipyranometer Arrays - MPA). The investigation proved to be successful and the concept was expanded to include the development of complete radiometric systems based on MPA and ML. The system would be later named All Seeing-Eye (ASE) radiometric system. It is important to mention that the concept of using Multi Pyranometer Arrays (MPA) to characterize solar irradiation components dates from the middle of the ’80s [1], [2]. However, due to reasons such as the limited computational power available at that period or the mathematical approach used to calculate the components of the solar irradiance, the concept did not advance considerably until after 2010. In 2015, Dr. Vivek Srikrishnan in collaboration with Dr. George Young and Dr. Frey Brownson proved that the combination of multiple pyranometer arrays with artificial neural network algorithms could perform well [3]. Srikrishnan used data from three to five pyranometers installed at different angles at the National Renewable Laboratory (NREL) in Golden, CO to calibrate an Artificial Neural Network (ANN) algorithm. Vivek showed that a combination of five pyranometers performed slightly better with a mean absolute error lesser than 20 W/m2. The success of such an approach motivated the deployment of the ASE systems at the State College area, close to Penn State University.


The All Seeing-Eye and SID Systems

The All Seeing-Eye (ASE) systems were developed to answer the following questions:

  1. How well would the ML algorithms perform in different locations?
  2. Is it possible to develop a low-cost radiometric system capable of accurately provide information about the different components of solar irradiance?

These questions are being answered for the State College area in Pennsylvania, where multiple ASE systems are installed, as shown in Figure 1 below. The development of the ASE systems started back in 2016 with the first 2 systems installed (Figure 1). After that, the approach for the development of low-cost radiometric systems was expanded to include more simple devices composed of a single pyranometer (later called SID system) to multiple uses such as education, meteorology, etc. The following sections will briefly describe the development history of the ASE and SID radiometric systems. 

Figure 1 - Map showing the location of the first two ASE radiometric systems installed near the town of State College, PA


First ASE System

The first ASE system was deployed in 2016 at the PSU Russell E. Larson Agricultural Research Center at Rock Springs, located approximately 10 miles away from PSU University Park campus, in State College PA (Figure 2). The system can measure solar irradiance from five pyranometers: one installed horizontally facing upward, and four others tilted 90 degrees facing the cardinal directions. Such configuration allows for picturing the directionality of the sunlight and to test the ML algorithms with local data. This particular system is also more affordable when compared to other radiometric devices that are capable of informing the different components of the solar irradiance. However, this particular system is still far from the final goal of having a system that is "a couple hundred dollars" in cost, a goal set by Dr. Frey Brownson.


Figure 2 -  Corryn Klien (left) and Worlasie Djameh (right) posing for a picture after having worked in the installation of the first ASE system. The system remains operational and is installed at Rock Springs, PA, right beside an NOAA SURFRAD station. Credits: Big Ten Network

Having the ASE installed at the Rock Springs was a strategic choice since the site houses one of the eight National Oceanic and Atmospheric Administration - Surface Radiation Budget Monitoring stations (NOAA-SURFRAD) in the US. The station provides precise multi-component solar irradiance information, including downwelling, albedo, direct normal irradiance, and more. The data provided by this station is used to test and calibrate the ASE systems.


Second ASE System

To study the effect of the locale in the amount of solar irradiance impinging on a surface and the response of the ML algorithms to this response, a second ASE system was installed at a site called Scotia, near State College, PA. This second system is mounted on a tripod, therefore it is mobile and can be relocated. In particular, this system also counts with a temperature sensor and a Rotating Shadowband Radiometer (RSR). The RSR measures the global horizontal irradiance (GHI) and the diffuse horizontal irradiance (DHI). By having these measured values, it is also possible to estimate the value of the direct normal irradiance (DNI) at the site. Therefore the outputs from the RSR can be used as a reference for comparison against the ASE system. The temperature sensor also allows for performing temperature correction of the data points. Figure 3 shows the second ASE system installed at Scotia site.

Figure 3 - Second ASE system deployed at Scotia site. On the tripod close to the truck it is possible to see the RSR and temperature sensor. The next tripod holds the ASE system housed by a 3D printed structure. Picture by Paulo Soares

The second version of the ASE system also marks the beginning of the effort to lower the cost of the system. Figure 3 shows the ASE system housed by a plastic cube, designed and printed by Matthew Callen, a former member of the Solar Ecology research group. Figure 4 shows Matt holding the printed version of the "cube". 3D printing would become an important ally to lower the cost (and the size) of the subsequent systems developed by the Lab.

Figure 4 - Matthew Callen holding his creation: the 3D printed housing for the ASE pyranometers. Picture by Paulo Soares 


The SID Project and DIY Radiometric Systems

The first two ASE radiometric systems proved to be reliable, but their cost was still higher than US$ 1000, much higher than the desired cost. This motivated the research team to build customized versions of the system, not only the complex systems like the ASE but also more ubiquitous radiometric systems, which was especially valuable for the undergraduate students part of the Solar Ecology Collaborative team, who were free to explore different designs and parts options. For example, the students brought the idea of using 3D printing for enclosing the system electronics. Moreover, the students also tried different computer boards to tackle the goal of logging solar irradiance data. The group initially used Arduino computers, later migrating to Raspberry Pi boards, and both options added flexibility and lowered the cost of the system. The first working prototype of a SID system was assembled in 2017 by Worlasie Djameh (who is in Figure 2) and consisted of an Arduino board with an SD card reader and a pyranometer connected to a prototyping board, as shown in Figure 5.

Figure 5 - First SID system: on the left the Arduino-based data logger (data acquisition unit); on the right the DIY pyranometer with 3D printed housing. Credits: Worlasie Djameh

This first version of SID was not self-powered nor had a proper enclosure, thus being a rough prototype of the system. Nonetheless, it was capable of showing interesting results: when compared side-by-side to a conventional radiometric system (also composed of a single pyranometer), it showed similar outcomes. That was the fuel the group needed to dig deeper into this exploration.
A second version of the system, also based on the Arduino board, was ready by the beginning of 2018. This new version included a battery and 3D printed housing. This version had its first out-of-lab test in Honduras (Figure 6), where another side-by-side test was performed. However, no conclusive results could be taken this time, as the Arduino system did not work as expected and no data was recorded.

Figure 6 - Second version of the SID system (orange pyranometer) being tested in Honduras. The SID pyranometer response was compared to an Apogee pyranometer (black sensor). Picture by Kevin Brower 

Working with the Arduino board has proven to be difficult for this application since the system was not reliable. For this reason, the group decided to use a different approach replacing the Arduino by a Raspberry-Pi computer. The new configuration proved to be more stable, reliable, and easy to configure when compared to the previous versions. The success of the Raspberry-Pi based system boosted the design and deployment of the first low-cost radiometric system by the Solar Ecology research group. This version was then called SID (short for Solar Irradiance Detector) system and was composed of a data logger, a pyranometer, and a temperature sensor, as shown in Figure 7 below.

Figure 7- First SID system working prototype. Designed & assembled by Paulo Soares with the help of Mohammad Alkhazraji. Picture taken by Paulo Soares

The first working prototype of the SID system shown in Figure 6 was then calibrated against conventional pyranometers and deployed at a solar farm, where it remained operational from one year, from the summer of 2018 to the summer of 2019. The performance of this SID system was evaluated and the results can be found at the page Case Study - PSU 2MW Solar Array

Finally, the low-cost version of the ASE system was also developed and combines the power of 3D printing with the flexibility of the Linux-based Raspberry Pi OS and Python language is showing good results, that can be found on the section of this website dedicated to the All Seeing-Eye Radiometric System.

The full list of people who contributed to the Solar Genome Project can be found in the Hall of Fame page!


[1] Brownson, Jeffrey R. S. Solar energy conversion systems. Academic Press, 2013.

[2] HAMA LAINEN, M., P. Nurkkanen, and T. Slaen. "A multisensor pyranometer for determination of the direct component and angular distribution of solar radiation." Solar energy 35, no. 6 (1985): 511-525.

[3] Srikrishnan, Vivek, George S. Young, Lucas T. Witmer, and Jeffrey RS Brownson. "Using multi-pyranometer arrays and neural networks to estimate direct normal irradiance." Solar Energy 119 (2015): 531-542.