Predictive Urban Scale Digital Twins For Optimising Solar Panel Energy Yield
This paper presents a physics-constrained AI digital twin framework that integrates open-source geospatial data with thermodynamic laws and machine learning to accurately forecast solar energy yields in developing regions by accounting for localized challenges like dust soiling and thermal degradation, thereby optimizing infrastructure planning through an interactive 3D visualization tool.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In many parts of the world, the power grid is unreliable, forcing schools, hospitals, and businesses to rely on expensive and polluting diesel generators to keep the lights on. Solar energy offers a cleaner, cheaper alternative, but installing it effectively is a complex puzzle. To get the most power from the sun, planners need to know exactly how much sunlight a specific roof will receive every hour of the year. This requires understanding not just the weather, but also how nearby trees and other buildings cast shadows, and how extreme heat or seasonal dust storms can reduce a panel's ability to work. Traditionally, creating a detailed map of these conditions for an entire city or campus has been prohibitively expensive, requiring specialized 3D scans and proprietary data that are often unavailable in developing regions.
A team of researchers has tackled this challenge by creating a new kind of virtual model, known as a digital twin, specifically designed for the University of Ibadan in Nigeria. Instead of relying on costly, high-tech scans, they built a complete three-dimensional replica of the campus using only free, open-source maps and clever computer programming. This virtual model does more than just look like the real campus; it simulates how the sun moves across the sky and how the weather changes over an entire year. By combining this spatial map with a computer program that understands the basic laws of physics, the researchers created a tool that can predict exactly how much electricity solar panels would generate on every single building, hour by hour, for a full year.
The core of this work addresses a common problem in solar planning: most tools treat the sun and the weather as separate issues, or they use generic formulas that fail to account for local realities. In Nigeria, for instance, the air is often filled with dust during the Harmattan season, which settles on panels and blocks sunlight, while intense tropical heat can cause panels to overheat and lose efficiency. The researchers solved this by building a system that first constructs a 3D view of the campus, calculating the height of every building and the position of every tree. They then fed this data into a smart computer model trained on historical weather records. This model was taught to recognize how heat and dust specifically degrade energy production, ensuring the predictions reflect the harsh realities of the local environment rather than ideal laboratory conditions.
One of the most significant hurdles the team overcame was the lack of detailed information about the campus itself. In many developing regions, maps often show the shape of buildings but leave them unnamed or without height data, making it impossible for construction crews to know where to go. To fix this, the researchers developed a method to automatically link every unnamed building on the map to the nearest street and landmark. This process created a practical, street-level directory that tells a worker exactly which building to visit and how far it is from the road, turning abstract computer coordinates into a usable guide for real-world installation.
The results of this approach were striking. When the researchers tested their system against standard methods, their model proved far more accurate at predicting energy output. While simpler models often missed the mark by failing to account for the complex way heat and shadows interact, this new system captured those details with high precision. It showed that a significant portion of potential energy is lost to shading, heat, and dust, quantifying these losses so that planners can make better decisions. For example, the model revealed that during the peak of the dusty season, energy production drops noticeably, a factor that generic calculators often ignore.
Beyond the technical accuracy, the project produced a visual dashboard that allows planners to see the financial and environmental impact of their decisions. The tool can project how much money a campus could save over twenty-five years by switching to solar, while also tracking how much carbon dioxide would be kept out of the atmosphere. It identifies the best locations for panels—roofs that are large, unshaded, and cool—and suggests the most efficient layout for a micro-grid. By proving that high-quality, physics-based planning can be done entirely with free data, this research offers a practical blueprint for other institutions in developing regions. It demonstrates that with the right mix of open data and smart modeling, communities can overcome the lack of expensive surveys to build reliable, clean energy systems that are tailored to their specific local conditions.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.