Earthworks Estimation for Utility Scale Solar Sites Using Dynamic Programming and Simulated Annealing
Tuesday, September 9, 2025 4:00 PM to 5:00 PM · 1 hr. (America/Los_Angeles)
RE+ Posters, Level 1, Venetian Expo Hall
Information
According to the Solar Energy Industries Association (SEIA), solar accounted for 75% of new U.S. electricity-generating capacity in Q1 2024. As deployment scales, reducing costs and design time through automation becomes essential. One key challenge is earthwork estimation, a critical and resource-intensive step in solar site development. Accurately predicting cut and fill volumes requires civil engineering expertise, yet many current photovoltaic (PV) earthwork estimation tools vary widely in approach and accuracy. This often results in over- or underestimations during detailed design, with costly construction impacts discovered too late to inform project feasibility.
To address this, we investigated algorithmic methods to automate tracker placement and reduce grading requirements. Weighted regression was applied to optimize placement of linear torque tube trackers, and dynamic programming was used for terrain-following tracker arrays. These methods are constrained by PV-relevant design parameters such as pile reveal ranges (“pile stickup”), enabling realistic engineering outputs. On a case study site, these approaches achieved roughly 10x faster site design compared to some conventional tools, while balancing cut/fill minimization and tracker layout optimization in both East-West and North-South directions.
To further reduce earthwork, we explored Simulated Annealing to optimize tracker offsets within allowable reveal windows. This yielded an additional 10–44% reduction in cut/fill volume beyond weighted regression, depending on the weighting scheme, and an average 44% reduction beyond dynamic programming in terrain-following designs.
These results suggest that incorporating automated optimization methods in early PV site design can significantly reduce grading effort, lower project costs, and improve constructability—all while accelerating the design process.
To address this, we investigated algorithmic methods to automate tracker placement and reduce grading requirements. Weighted regression was applied to optimize placement of linear torque tube trackers, and dynamic programming was used for terrain-following tracker arrays. These methods are constrained by PV-relevant design parameters such as pile reveal ranges (“pile stickup”), enabling realistic engineering outputs. On a case study site, these approaches achieved roughly 10x faster site design compared to some conventional tools, while balancing cut/fill minimization and tracker layout optimization in both East-West and North-South directions.
To further reduce earthwork, we explored Simulated Annealing to optimize tracker offsets within allowable reveal windows. This yielded an additional 10–44% reduction in cut/fill volume beyond weighted regression, depending on the weighting scheme, and an average 44% reduction beyond dynamic programming in terrain-following designs.
These results suggest that incorporating automated optimization methods in early PV site design can significantly reduce grading effort, lower project costs, and improve constructability—all while accelerating the design process.