Best AI Stocks to Watch in August 2025

Data Science And Machine Learning Applied To Silicon Photovoltaic Solar ...

Presented by Roger French, Kyocera Professor, Materials Science & Engineering; Director, SDLE Research Center; Faculty Director, Applied Data Science Program 

See French's Presentation 

Abstract: Advances in computing, communication, and data collection have facilitated collection of petabyte-scale datasets from which data-driven models can be built. This digital transformation affects society, industry, and academia, since data-driven models can challenge how things are done and offer new opportunities for developing how things work.

At CWRU we have offered the university-wide Applied Data Science (ADS) program since 2015. The ADS program teaches non-computer science students, producing "T-shaped" graduates with deep knowledge in their domain plus strong data science skills. The ADS program provides both an undergraduate minor and graduate level courses for which a University Certificate is being developed. ADS students learn the foundations: coding, inferential statistics, exploratory data analysis, modeling and prediction, and they complete a semester long data science project for their ADS portfolio. The courses are taught using a practicum approach, with an open data science toolchain consisting of R, Python, Git, Markdown, Machine Learning, and TensorFlow on GPUs.

We utilize data science and big-data analytics to address critical problems in energy science. As solar power grows, we need to fully understand and predict the power output of photovoltaic (PV) modules over their entire > 30 year lifetimes. Degradation science [reference 1] combines data-driven statistical and machine learning with physical and chemical science to examine degradation mechanisms in order to improve PV materials and reduce system failures. We use distributed and high performance computing, based on Hadoop2 and the NoSQL Hbase, to ingest, analyze, and model large volumes of time-series datasets from 3.4 GW of PV power plants [reference 2]. We have developed an automated image processing and deep learning pipeline applied to electroluminescent (EL) images of PV modules to identify degradation mechanisms and predict their associated power losses [reference 3]. Unbiased, data-driven analytics, now possible using data science methodologies, represents a new front in our research studies of critically important and complex systems.

References

1. R.H. French, et al., Degradation science: Mesoscopic evolution and temporal analytics of photovoltaic energy materials, Curr. Op.Sol. State & Matls. Sci. 19 (2015) 212–226.

2. Y. Hu, et al., A Nonrelational Data Warehouse for the Analysis of Field and Laboratory Data From Multiple Heterogeneous Photovoltaic Test Sites, IEEE Journal of Photovoltaics. 7 (2017) 230–236.

3. A. M. Karimi, et al., Automated Pipeline for Photovoltaic Module Electroluminescence Image Processing and Degradation Feature Classification, IEEE Journal of Photovoltaics. (2019) 1–12.

 

Machine Learning Method Could Speed Path To Cleaner Energy Solutions

By Wick Eisenberg

/ Published Oct 9, 2024

The process of testing new solar cell technologies has traditionally been slow and costly, requiring multiple steps. Led by a fifth-year PhD student, a Johns Hopkins team has developed a machine learning method that promises to dramatically speed up this process, paving the way for more efficient and affordable renewable energy solutions.

"Our work shows that machine learning can streamline the solar cell testing process," said team leader Kevin Lee, who worked with fellow electrical and computer engineering graduate students Arlene Chiu, Yida Lin, Sreyas Chintapalli, and Serene Kamal, and undergraduate Eric Ji, on the project. "This not only saves time and resources but opens new possibilities for clean energy technology development."

The team's results appear in Advanced Intelligent Systems.

Image caption: In this new method, the scanning instrument (right) is used to measure current and optical maps (lower left) of solar cell arrays (top left).

A major hurdle in commercializing new solar materials and devices is the lengthy fabrication-testing-iteration cycle. Optimizing a new solar cell material for the market is an arduous process: After a device is made, multiple time-consuming measurements are needed to understand its material properties. This data is then used to adjust the fabrication process, repeating the cycle.

The new method drastically reduces this time by extracting all the materials' important characteristics from a single measurement. Unlike other methods trained on computer-simulated data—which often produce inaccurate results—the Hopkins team's approach uses real-world data. Their neural network collects thousands of data points from one solar cell, capturing complex properties and variations caused by defects, such as spin-casting streaks, cracks, and contaminants, and eliminating the need to fabricate thousands of solar cells.

"Kevin's method has the potential to speed up photovoltaic development times," said Lee's adviser and study co-author Susanna Thon, an associate professor of electrical and computer engineering at JHU's Whiting School of Engineering and associate director of the university's Ralph O'Connor Sustainable Energy Institute. "Instead of laboriously making multiple measurements on many devices to learn what you need to know about device behavior, Kevin, thanks to his [machine learning] algorithm, can now tell you everything you'd want to know about a device and its properties from a single measurement that takes about 30 seconds."

The other novel feature of Lee's system is that it takes spatial maps of data from solar cells and converts them into images.

"Normally one of the most common measurements you get after creating a new solar cell is called a JV curve, and what it does is measure the cell's response to light," Lee said. "We had the idea of converting these JV curve maps into images so we could take advantage of advanced machine learning models developed for applications not in materials science, but in computer vision, to learn patterns in solar cell behavior."

Another benefit of the new method is its applicability to various materials and devices beyond solar cells, potentially accelerating the timeline from material discovery to market adoption.

"In theory, the system we developed could be used to measure other devices, such as transistors and light sensors," Lee said. "The time saved, and the accuracy of this system could lead to a wide array of new technologies being created much more quickly, which I am excited to see happen."

Machine Learning Shows Promise For Predicting Building Energy Use

A new machine learning approach developed through an international collaboration between Polytechnic University of Milan and Drexel University could help architects and urban planners better predict neighborhood energy consumption during early design stages.

The study, published in the journal Buildings, tested an artificial intelligence model that forecasts building energy use with 88% accuracy using just four key data points, a major improvement over current methods that require extensive inputs. The research team used the CatBoost machine learning model, which outperformed several other artificial intelligence approaches in their testing.

"This framework gives designers quick insights about energy impacts when decisions matter most," said co-author Simi Hoque, PhD, PE, professor of civil, architectural and environmental engineering at Drexel University. "We can now make energy-smart choices from the start of neighborhood projects."

The research team, led by Milan's Andrea Giuseppe di Stefano, developed and validated their approach using data from over 22,865 buildings. The model analyzes building size, primary use, number of floors, and climate zone to generate energy predictions. These four factors were identified as the most influential through detailed statistical analysis of the dataset.

When tested on mixed-use buildings in New York City, the model's predictions differed from traditional energy modeling calculations by no more than 8.69% lower to 11.04% higher. This level of accuracy is particularly impressive given the minimal input data required.

The simplified approach addresses a key challenge in sustainable urban development - the need to consider energy efficiency before designs are finalized. Current modeling tools require detailed information about building materials, mechanical systems, and operational schedules that isn't typically available in early planning stages. By contrast, this new method needs only basic information that architects and planners typically have at the start of a project.

The model's effectiveness stems from its training on a comprehensive dataset combining information from both residential and commercial buildings. The researchers merged data from the Commercial Buildings Energy Consumption Survey and the Residential Energy Consumption Survey, creating a robust foundation for predictions across different building types.

This research marks the first phase of a larger framework to optimize neighborhood energy use. Future phases will analyze building shapes and evaluate district-level energy systems. The team plans to incorporate additional features such as solar exposure analysis and district heating potential in subsequent stages.

The work is particularly timely as cities face pressure to reduce carbon emissions from buildings, which account for about 40% of energy use in the European Union. With urban populations expected to double by 2050, the need for efficient energy planning tools becomes increasingly critical.

"We're giving designers practical tools to create more sustainable neighborhoods," Hoque said. "Making good energy choices early leads to better performing buildings."

---