Integral of the CRAC in the doctorate. candidate dissertation on solar forecasting: UNM Newsroom

Guillermo Terrén-Serrano, a Ph.D. candidate in the Department of Electrical and Computer Engineering (ECE) at the University of New Mexico (UNM), has devised a new way for solar forecasters to predict future cloud cover.

According to his advisor, ECE Professor and holder of the King Felipe VI Chair in Information Sciences and Related Technologies Manel Martínez-Ramón, Terrén-Serrano has taken an innovative and multidisciplinary approach to predicting solar cloud cover by building a state state of the art forecasting system entirely from scratch.

“I think this work could have an impact on energy forecasting, which is very important right now and will continue to be very important in the future,” Martínez-Ramón said.

According to Terrén-Serrano, there is a paradoxical problem with renewable energies. Renewable energy is generated by wind and sunlight, which means it also depends on the availability of wind and sunlight. Therefore, weather conditions, including cloud cover, can affect the collection of these renewables. Adverse weather conditions can have a serious impact on renewable energy providers, who are prevented from connecting to power grids if they have a decrease in their energy production.

“Network operators limit ramp rates, [or] the energy drops you can send through the transmission lines. So if there’s a drop in your ramp rate, they’re going to cancel you and not allow you to connect to networks,” Terrén-Serrano explained.

To solve this problem, Terrén-Serrano designed an end-to-end system to predict future cloud cover known as the Clear Sky Index. By knowing when clouds will obscure the sun, providers can use batteries to compensate for the drop in solar power, ensuring a stable energy load. This will prevent drops in their ramp rates and eliminate the possibility of them being kicked off the power grid for violating ramp rate guidelines.

“What we are trying to do is increase renewable resources [by making it] possible for PV (photovoltaic) operators to connect their plants to power grids without violating their ramp tariffs and without risking being disconnected,” noted Terrén-Serrano.

To do this, Terrén-Serrano and Martínez-Ramón determined that they would approach the problem of solar forecasting from a different perspective than what they actually see on the ground today. Their point of view is “different from other published articles that we know of”, asserted Martínez-Ramón. In the past, solar forecasts were based on algorithms published in a computer journal or found in a repository of software assistants that have been adapted to perform solar forecasts.

Professor Manel Martinez-Ramon

“It creates a lot of problems, and we talk about it in the papers,” Terrén-Serrano said.

“We want to get rid of the black box concept in solar forecasting. By black box, we mean the use of an already published technique, applied to the problem posed, without modifying or interpreting the behavior of the machine in terms of solar forecasting processes,” Martínez-Ramón clarified.

Terrén-Serrano and Martínez-Ramón’s multidisciplinary approach borrows concepts from fluid dynamics, atmospheric dynamics, physics, and standard and non-standard image processing techniques, and combines these concepts with learning machine learning, kernel learning, and deep learning to create their solar forecasting system.

This use of machine learning instead of end-to-end learning is very important, according to Terrén-Serrano. Their system, which he calls “Image Understanding,” is based on a sequence of images taken by a camera that tracks the sun throughout the day. The machine learning algorithms he created then analyze the images to plot what is happening with the clouds. The result is a solar forecast of future cloud cover based on this analysis.

For example, on a cloudy afternoon sky, clouds in the distance appear to be barely moving, while near clouds appear to be moving much faster. But this appearance is misleading because it is based on our point of view. In reality, the clouds in the distance are moving quite quickly. One of the key elements of Terrén-Serrano’s new system is that it analyzes each image taken of cloud formations pixel by pixel, obtains an accurate dimension of each pixel and determines the true velocity of the cloud formation. Martínez-Ramón explained what Terrén-Serrano created “is a methodology to undo perspective so that all the clouds [in the transformed image] move at real speed.

Martínez-Ramón continued, “It is of notable significance that Terrén-Serrano has developed an integral end-to-end system that brings together camera and sensor data and provides an accurate solar energy forecast. . All elements of the system are innovative, including data processing, physical models, information retrieval, and machine learning applied to that information. Everything was developed from two cameras mounted on a servo system to track the sun, electronics developed by us, and a small computer. This is not a standard thesis: it has over 250 pages of publishable work. I hope he will get recognition for his thesis.

When Terrén-Serrano started designing his system for his thesis, he was told he would never get the wind speed vector right. He accepted this challenge and succeeded. He was able to establish an accurate wind speed vector. His paper on geospatial perspective reprojection is currently under review for publication and explains how he arrived at his results. And this is not the only article resulting from Terrén-Serrano’s research. He has written 12 other papers on his forecasting system that have been published, presented at conferences, or are being reviewed for publication.

Terrén-Serrano credits UNM’s Center for Advanced Research Computing (CARC) with the ability to build its new solar forecasting system. Terrén-Serrano and Martínez-Ramón acknowledge that his work might have been impossible, or at least much more difficult, without CARC. Both expressed their gratitude to CRAC not only for the computer systems, but also for the help they provide to the students.

Martínez-Ramón emphasized: “We are really, really grateful to CARC because [they] do a great job helping students with their work when they are having difficulty. I only have good words for you.

Terrén-Serrano recalls with a laugh: “Matthew Fricke always helps me and tells me how to do things right because sometimes I’m wrong. I remember the first email I received from CRAC. I had submitted 1000 jobs and was only supposed to submit 20, so I crashed the scheduler that orders the jobs!

Terrén-Serrano continued: “We had ideas on how to do this work and we talked about this program, [but it was] CARC which gave me the traceability. I could get all the data and manage it efficiently. Using CARC’s central processing units (CPUs) and graphics processing units (GPUs), he was able to search for specific areas from the sky imagers and link those areas to other processes such as tracking clouds, cloud detection, image processing and solar forecasting. Terrén-Serrano was then able to combine it with machine learning and artificial neural networks.

“With CARC systems, I can use kernel learning or deep learning, which is computed in GPUs, so I have all these technological possibilities for improvement in these research areas,” said Terrén. -Serrano.

Martínez-Ramón concluded: “It’s the work of Guillermo; I’m just the adviser. I had a lot of fun working with him – many. Because there is a lot of theory and a lot of novelties. Terrén-Serrano is currently finalizing his thesis in anticipation of a graduation in the spring of 2022.