TEAM: EMMA COOPER, KEVIN ANDERSON, DAN RILEY, MADHURI KUMARI | CONTRIBUTING WRITER: RORY MCCLANNAHAN, MONICA BIGNEY
Sun tracking algorithms have made utility-scale solar installations more profitable. What happens when solar is installed on uneven terrain?
The most popular design for single-axis tracking systems consists of a torque tube oriented in the north-south direction. Graphic adapted from Sinovoltaics, a solar developer.
Over the last 10 years, technology allowing rows of panels to rotate in place has become ubiquitous at large solar installations. These rotating installations, known as single-axis tracking (SAT) systems, are designed to allow panels mounted on a torque tube (the “single axis”) to track the position of the sun over the course of the day. By allowing the panels to directly face the sun for the whole day – known in the industry as minimizing the angle of incidence – SAT systems boast a higher power output per row than a traditional “fixed tilt” system.
Local weather patterns have an impact on how much more power per row is realized. In Albuquerque, where conditions are usually clear, SAT systems will show large gains over fixed-tilt systems. But under the “socked in” cloudy conditions that are frequent in the northwest, solar panels can often collect more irradiance overall by lying flat than by tracking the position of the sun.
A single-axis tracking system deployed in the field. Image from Valsa, a solar supplier.
In this project, the team quantified how weather influences SAT system output across the United States by using local weather data provided by the National Solar Radiation Database (NSRDB) to simulate the power produced by four different sun-tracking algorithms in over 800 locations across the country. To parallelize these simulations, the team leveraged Kahuna, a high-performance data analytics (HPDA) research cluster hosted by Sandia California. Kahuna’s implementation of a queue management system allowed the team to run batch simulations for hundreds of locations and monitor the outputs. By allocating calculations to multiple different nodes on Kahuna, the team was able to complete the calculations within a couple of days—as opposed to the several months it likely would have taken on a regular computer.
Diffuse fraction is one of the metrics used in the solar community to quantify the proportion of days that are overcast. A higher diffuse fraction indicates that the region experiences overcast weather more frequently. This graphic was created using HPDA resources and data from the NSRDB.
The study focused specifically on four sun-tracking algorithms that adapt their controls based on the surrounding terrain. Some of the algorithms work by finding a single optimal angle of rotation for all the rows in a system, while more computing-intensive algorithms optimize rotations row-by-row. For these latter algorithms, Kahuna’s computing speed, enabled by its data processing unit (DPU) chips, came in handy. For our 400+ row SAT system, the DPU chips that make up Kahuna’s hardware allowed for quick calculations of tracker rotation angles for the entire system in mere minutes.
Kahuna enabled the team to compare energy outputs from the four different sun-tracking algorithms and to identify regions of the country that experienced the highest and lowest gains. This demonstrated that an algorithm using a global optimization approach results in the largest energy gains, with increases in annual yield between 40 to 90 kWh/kW relative to the baseline algorithm the results were compared against. The Southwest region experienced the highest annual yield from using the global optimization approach, while the Midwest saw the lowest gains from using the algorithm.
Tracker rotation angles calculated using HPDA resources with a global optimization algorithm. The example tracking system used in this study comprised 445 rows running north-south on synthetic terrain, which varies in the east-west direction. Here, the optimal rotations identified by the algorithm alternate between rows facing towards the rising sun (negative rotation angles) and rows facing away from the sun to avoid shading their neighbors (positive rotation angles).
Annual yield improvement was strongly correlated with annual diffuse fraction. In other words, using the better sun-tracking algorithm led to greater increases in production in sunnier areas than in cloudy regions. However, even the cloudy regions saw an annual energy increase of 2% relative to the baseline when using the global optimization algorithm.
The research results make a promising case for the use of adaptive tracking algorithms across the United States, but the energy gains found in this study are highly dependent on choice of terrain. Utility-scale solar sites are almost always graded, and there’s not a lot of data available on the potential cost-savings or additional construction costs of building a system on an ungraded surface. Mechanical trackers that can follow adaptive algorithms also are more expensive, and it’s unclear whether the additional expenses can be negated by grading-related cost savings.
This study is a first step towards a comprehensive value proposition for the use of uneven terrain over traditional graded terrain, but more research on cost is needed to make a compelling case to solar developers.
