Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Computational Imaging for Long-Term Prediction of Solar Irradiance

Leron Julian, Haejoon Lee, Soummya Kar, Aswin C. Sankaranarayanan

TL;DR

This work tackles the uncertainty in solar irradiance forecasting caused by cloud occlusion by introducing a catadioptric sky-imaging system that uses a hyperboloidal mirror to achieve near-uniform spatial resolution across an ~$180°$ field of view. It couples optical design with a wind-informed space-time-slice forecasting framework to predict sun occlusion and ground irradiance tens of minutes into the future, substantially extending the horizon compared with prior methods. The authors validate their approach with both simulated data and outdoor real-world measurements, demonstrating that occlusion predictions can reach about $18$ minutes in simulations and $10$–$20$ minutes in real data, and they release the dataset and code to enable further research. While acknowledging limitations such as linear cloud motion assumptions and the potential benefits of integrating humidity and additional data sources, the work highlights a practical and scalable pathway to improved solar energy dispatch and grid reliability through computational imaging.

Abstract

The occlusion of the sun by clouds is one of the primary sources of uncertainties in solar power generation, and is a factor that affects the wide-spread use of solar power as a primary energy source. Real-time forecasting of cloud movement and, as a result, solar irradiance is necessary to schedule and allocate energy across grid-connected photovoltaic systems. Previous works monitored cloud movement using wide-angle field of view imagery of the sky. However, such images have poor resolution for clouds that appear near the horizon, which reduces their effectiveness for long term prediction of solar occlusion. Specifically, to be able to predict occlusion of the sun over long time periods, clouds that are near the horizon need to be detected, and their velocities estimated precisely. To enable such a system, we design and deploy a catadioptric system that delivers wide-angle imagery with uniform spatial resolution of the sky over its field of view. To enable prediction over a longer time horizon, we design an algorithm that uses carefully selected spatio-temporal slices of the imagery using estimated wind direction and velocity as inputs. Using ray-tracing simulations as well as a real testbed deployed outdoors, we show that the system is capable of predicting solar occlusion as well as irradiance for tens of minutes in the future, which is an order of magnitude improvement over prior work.

Computational Imaging for Long-Term Prediction of Solar Irradiance

TL;DR

This work tackles the uncertainty in solar irradiance forecasting caused by cloud occlusion by introducing a catadioptric sky-imaging system that uses a hyperboloidal mirror to achieve near-uniform spatial resolution across an ~ field of view. It couples optical design with a wind-informed space-time-slice forecasting framework to predict sun occlusion and ground irradiance tens of minutes into the future, substantially extending the horizon compared with prior methods. The authors validate their approach with both simulated data and outdoor real-world measurements, demonstrating that occlusion predictions can reach about minutes in simulations and – minutes in real data, and they release the dataset and code to enable further research. While acknowledging limitations such as linear cloud motion assumptions and the potential benefits of integrating humidity and additional data sources, the work highlights a practical and scalable pathway to improved solar energy dispatch and grid reliability through computational imaging.

Abstract

The occlusion of the sun by clouds is one of the primary sources of uncertainties in solar power generation, and is a factor that affects the wide-spread use of solar power as a primary energy source. Real-time forecasting of cloud movement and, as a result, solar irradiance is necessary to schedule and allocate energy across grid-connected photovoltaic systems. Previous works monitored cloud movement using wide-angle field of view imagery of the sky. However, such images have poor resolution for clouds that appear near the horizon, which reduces their effectiveness for long term prediction of solar occlusion. Specifically, to be able to predict occlusion of the sun over long time periods, clouds that are near the horizon need to be detected, and their velocities estimated precisely. To enable such a system, we design and deploy a catadioptric system that delivers wide-angle imagery with uniform spatial resolution of the sky over its field of view. To enable prediction over a longer time horizon, we design an algorithm that uses carefully selected spatio-temporal slices of the imagery using estimated wind direction and velocity as inputs. Using ray-tracing simulations as well as a real testbed deployed outdoors, we show that the system is capable of predicting solar occlusion as well as irradiance for tens of minutes in the future, which is an order of magnitude improvement over prior work.
Paper Structure (37 sections, 3 equations, 16 figures, 1 table)

This paper contains 37 sections, 3 equations, 16 figures, 1 table.

Table of Contents

  1. Introduction
  2. Prior Work
  3. Large Field-of-View Sky Imagers.
  4. Predicting Cloud Movement.
  5. Photovoltaic Power Output Prediction.
  6. Computational Imaging for Atmospheric Tomography.
  7. Catadioptric Imaging Systems.
  8. Problem Definition
  9. Design Specifications
  10. Gaps in Current Sky Imagers
  11. Solution Outline
  12. Mirror Design
  13. Testbed
  14. Simulation Setup
  15. Hardware Prototype
  16. ...and 22 more sections

Figures (16)

  • Figure 1: Our work presents a computational imaging system based on a catadioptric combination of mirrors and cameras. Seen above are images of the sky when using (left) a traditional hemispherical mirror and (right) the proposed hyperboloidal mirror. Overlaid on top are circles corresponding to different angular extents of the sky. The use of a hyperboloidal mirror enables increased. angular resolution near the horizon. We also propose a learning-based framework for predicting future solar irradiance. Together, these contributions enable a more accurate prediction over a longer time horizon than the hemispherical imagers.
  • Figure 2: A rendering obtained using Blender to visualize the non-uniform resolution of a hemispherical mirror. (Top) We create a scene consisting of a checkerboard with a length and width of 50 km, placed 2 km high above the ground. Each square on the checkerboard has physical extent of 1 km. (Bottom-left) Our imaging system consists of a pinhole camera observing the sky or the checkerboard indirectly through a hemispherical mirror. (Bottom-right) The image observed on the camera has high resolution at the zenith of the image and significantly lower resolution at the periphery.
  • Figure 3: (Left) We design the mirror for a setup where a camera is placed 1 meter above the mirror. The shape is optimized so that the overall system scales the field of view of the pinhole camera from $3.58^\circ$ to $170^\circ$ uniformly. (Right) The resulting hyperboloidal mirror shape that we use in our setup.
  • Figure 4: Simulated case visualizing the uniform resolution of our hyperboloidal mirror. (Left) The same parameters as Figure \ref{['fig:checkerboard_sphere']} with the proposed mirror replacing the hemispherical mirror. (Right) Observe how the checkerboard resolution is spatially uniform---a consequence of the system acting as overall scaling operation.
  • Figure 5: Various images captures from the synthetic dataset. (Top) Captures from hemispherical setup. (Bottom) Captures from the hyperboloidal setup. Each column is captured at the same time instant
  • ...and 11 more figures