S

olar energy is one of the most sought-after renewables, but it comes with its challenges. One of the main ones? Its inconsistency due to ever-changing cloud cover.

Cloud types matter when it comes to sunlight reaching the ground. For instance, cumulus clouds can significantly reduce the radiation, while clouds close to the sun might let more light through. These variations can lead to fluctuations in output from solar systems, leading to power grid instabilities and energy wastage. Large solar plants especially are affected, with many exceeding the fluctuation limits set by grid operators.

Leveraging Machine Learning and IR Imaging

To combat these challenges, researchers from the University of California Santa Barbara and the University of New Mexico rolled out a new method that can better forecast solar energy on an hourly basis, stabilizing energy supply and smoothing out variations. This solution is a fusion of data capture equipment and advanced software processing.

Predicting solar power output is a complex task, varying based on technology, installation type, and even geographical location. Current methods include:

  • Numerical Weather Prediction (NWP): While detailed, these models are costly and resource-intensive.
  • Satellite Imagery: Offers a broad view but can have transmission delays, which make real-time predictions less accurate.
  • Visible Light Cameras: These are great for capturing high-resolution images of the sky, but the sun's brightness can sometimes distort the results. Even with filters, the issue isn't wholly addressed.
  • Infrared (IR) Cameras: Although they handle sunlight well and can potentially be a low-cost substitute for expensive ceilometers, they come with their own set of challenges, especially in capturing a wide view and maintaining image clarity.

To mitigate the limitations of IR cameras, the research team introduced a solution that leverages machine learning. This process helps identify and correct any data discrepancies, ensuring that the most accurate information is sent from the camera to the software for analysis.

The study offers advanced methods to process IR images of the sky, helping to differentiate between atmospheric backgrounds and clouds more effectively. A unique algorithm categorizes sky conditions in these images into four groups: clear sky, cumulus clouds, stratus clouds, and nimbus clouds. Based on this sorting, the algorithm then works with the overall sunlight data (Global Solar Irradiance or GSI) to figure the amount of irradiance on the image.

Solar for home is a sustainable energy solution, harnessing sunlight to power residential properties and reduce electricity bills while minimizing environmental impact.

What's more, the algorithm is equipped to filter out dirt effects on the camera, eliminating the need for daily camera cleaning. A special image-processing technique minimizes the effects of debris, ensuring cleaner IR images even in less-than-ideal conditions.

Implications and Future Prospects

So, where do we go from here? The team suggests that fine-tuning our short-term solar predictions can lead to significant savings on smart grid operations since there's a need to boost energy storage to balance voltage fluctuations. The scientists underline, “The IR sky imager and the hybrid data processing methods introduced in this investigation are low-cost and scalable to PV systems of any size, easing the development of distributed generation systems supported with solar energy.”

Sources:

https://www.sciencedirect.com/science/article/pii/S0038092X23006023

Posted 
Nov 27, 2023
 in 
Blog
 category

More from 

Blog

 category

View All

Get Accurate & Competitive Quotes in Minutes

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.