The market for solar cells and panels is advancing rapidly. According to publicly available estimates, the market value of solar panels worldwide stood at around US$82 billion in 2023 and is forecast to reach US$260 billion in 2031, more than threefold growth in eight years.
The progress of solar panels relies on improvements in solar cells, the fundamental components of solar panels. The panels are made of monocrystalline or polycrystalline silicon solar cells soldered together and sealed under an anti-reflective glass cover. Undoubtedly, any progress in solar technology depends on improvements in solar cells. We will discuss one such breakthrough in the next segment.
Deploying Machine Learning to Accelerate the Search for New Semiconductor Molecules for Perovskite Solar Cells
Researchers at the Karlsruhe Institute of Technology, working as part of an international team, could find new organic molecules that increased the efficiency of perovskite solar cells1. In achieving their mission, the team of researchers used an intelligent combination of AI and automated high-throughput synthesis. As the report highlights, the research is compatible with applications in other areas of materials research, including the search for new battery materials.
The scientific paper detailing the research states that the inverse design of tailored organic molecules for specific optoelectronic devices of high complexity holds enormous potential that remains unrealized. The models that existed before the research results came out relied on large data sets that generally did not exist for specialized research fields.
The researchers demonstrated a closed-loop workflow that combined high-throughput synthesis of organic semiconductors to create large datasets and Bayesian optimization to discover new hole-transporting materials with tailored properties for solar cell applications. The predictive models were based on molecular descriptors that allowed the scientists to link the structure of these materials to their performance.
Altogether, the research could identify a series of high-performance molecules from minimal suggestions to achieve up to 26.2% (certified 25.9%) power conversion efficiency in perovskite solar cells.
While the scientific explanation of what the research team had achieved might sound too technical to grasp, there are easier ways to understand the pain points and how the solutions came to the fore.
The job of finding out which of a million different molecules would conduct positive charges and make perovskite solar cells particularly efficient would require synthesizing and testing all of them. The deployment of AI helped reduce this cumbersome load, as the team of researchers, headed by Tenure-track Professor Pascal Friederich, who specializes in the applications of AI in materials science at KIT’s Institute of Nanotechnology, and Professor Christoph Brabec from the Helmholtz Institute Erlangen-Nürnberg (HI ERN), deployed an AI model to serve their purpose with the least possible attempts.
According to Brabec:
“With only 150 targeted experiments, we were able to achieve a breakthrough that would otherwise have required hundreds of thousands of tests.”
The researchers believed the workflow they developed could open up new ways to quickly and economically discover high-performance materials for a wide range of applications.
But how did the researchers achieve what they did? Let us delve a bit deeper!
Training AI With Molecular Data
It all started at the Helmholtz Institute Erlangen-Nürnberg (HI ERN), with a database full of structural formulae for approximately one million virtual molecules that could be synthesized from commercially available substances. The researchers selected 13,000 at random from these virtual molecules. To determine their energy levels, polarity, geometry, and other properties, the KIT researchers used established quantum mechanical methods.
The selection of 101 molecules out of the 13,000 was based on the diversity of their molecular properties. In other words, the scientists chose these 101 with the greatest differences in their properties.
Subsequently, the scientists synthesized these molecules with robotic systems to produce solar cells that were otherwise identical. Measuring the efficiency of these solar cells was one of the most crucial aspects of this research.
Speaking about this efficiency assessment, Brabec had the following to say:
“Being able to produce truly comparable samples thanks to our highly automated synthesis platform, and thus being able to determine reliable efficiency values, was crucial to our strategy’s success.”
The AI model the researchers trained and developed suggested 48 other molecules to synthesize. These recommendations were based on two criteria:
- High expected efficiency
- Unforeseeable properties
The recommendation of AI yielded positive results. The AI-suggested molecules made it possible to build solar cells with above-average efficiency. Some of these could even surpass the capabilities of today’s most advanced materials.
The result reassured researchers of their decision to deploy AI. While characterizing the level of achievement, Friedrich had the following to say:
“We can’t be sure we’ve really found the best one of a million molecules, but we’re certainly close to the optimum.”
The deployment of AI was really encouraging, as it helped researchers gain insight into the molecules it suggested. The AI’s suggestions were partly based on the presence of certain chemical groups, such as amines, that chemists had previously overlooked. Highlighting these overlooked chemical groups was undoubtedly a significant positive outcome of deploying AI.
Broadly, the deployment of AI to advance the cause of solar tech is not new. The International Renewable Energy Agency carried out an extensive study on how AI applications could improve renewable energy integration processes, especially those involving wind and solar power.
In the next segment, we look into some of the report’s insights to get a comprehensive overview of how AI might prove effective in solar tech.
Diverse Application of AI in Advancing Solar Tech
AI can help solar tech become more efficient by helping it generate improved renewable energy generation forecasts. Big data and AI could produce accurate power generation forecasts that would make it feasible to integrate much more renewable energy into the grid. The report cites the example of IBM, which, in 2015, was able to show an improvement of 30% in solar forecasting while working with the US Department of Energy’s SunShot Initiative.
The self-learning weather model and renewable generation forecasting technology leveraged large datasets of historical data and real-time measurement integrated into it from local weather stations, sensor networks, satellites, and sky image cameras.
System operators could also benefit from it with accurate short-term forecasting, improving unit commitment, increasing dispatch efficiency, reducing reliability issues, and reducing the operating reserves needed in the system.
The report highlighted two projects. One was EWeLiNE, a German research project that used machine learning software and ended in 2017. The other was Gridcast, its follow-up. Both used AI to predict power generation. They pulled data from solar sensors, wind turbines, and weather forecasts. This helped cut down wasted power.
AI could also play a crucial role in maintaining grid stability and reliability. Four companies (Adaptricity, AEK, Alpiq, and Landis+Gyr), together with the Canton of Solothurn, are testing how AI solutions can ensure future grid stability and minimize investments in costly grid expansion in a pilot project called SoloGrid.
It was seen that AI could increase the capacity of power grids and reduce the need for new lines by making better use of existing lines based on weather conditions. AI-based systems, using large amounts of weather data, could ensure optimal use of existing power grids by adapting operations to weather conditions at any time, thereby reducing congestion.
AI could also improve safety, reliability, and efficiency in power systems by automatically detecting disturbances. AI models, trained with examples of typical system outages, could gradually learn to distinguish—and precisely categorize—normal operating data from defined system malfunctions. As a result, the algorithm could make split-second decisions on the presence of an anomaly or fault, as well as its type and location. More specifically, the algorithm could reach a decision within 20–50 ms, providing sufficient time to implement the appropriate fully automated countermeasures.
AI could also boost solar technology by improving demand forecasting. The report gave an example: BeeBryte, a French startup. It used AI to predict a building’s thermal energy demand. This lets it produce heating and cooling at the right times. Customers could set an operating temperature range, and AI kept it steady. This led to savings of up to 40% on utility bills.
AI also helped extend the lifespan of storage units. It used predictive logic to optimize charging and discharging. The report mentioned Athena, an AI system developed by the California-based company Stem. It mapped out energy usage and tracked fluctuations in energy rates. This lets customers use storage more efficiently.
Apart from innovative and ingenious startups, several large companies have used AI to advance solar tech applications. In the coming segments, we discuss two such companies and how they applied AI to move solar tech forward.
1. Google (GOOGL +1.82%)
Google is known for using its DeepMind AI systems to predict the output from its solar farms, providing accurate forecasts that help optimize energy consumption and reduce costs. One report suggests that it reduced energy used for cooling at a Google data center by 40% (a 15% overall reduction in power usage) using only historical sensor data from within the data center (e.g., temperatures, power, pump speeds, setpoints) to improve energy efficiency. The AI system predicts the future temperature and pressure of the data center over the next hour and gives recommendations on whether to turn consumption on or off.
Inspired by its success, the company took these innovations to the next level, where, instead of having its recommendations implemented by people, the AI system directly controlled data center cooling while remaining under the expert supervision of data center operators.
In 2015, Google launched Project Sunroof to help homeowners estimate their solar savings potential. Building on expertise gained from that project, the company released a new Solar API tool in 2023. It used mapping and computing resources to provide detailed rooftop solar potential data for more than 320 million buildings across 40 countries, including the U.S., France, and Japan.
Google trained its AI model to extract 3D information about the roof geometry directly from aerial imagery, along with details about trees and shade, while the Solar API accounted for factors like historical weather patterns in the area, energy costs, and more. The goal is to allow solar installers to understand solar potential and costs before ever visiting an area.
Alphabet Inc. (GOOGL +1.82%)
On October 29, 2024, Alphabet Inc. announced its financial results for the quarter ended September 30, 2024. Consolidated Alphabet revenues in Q3 2024 increased 15%, or 16% in constant currency, year over year to $88.3 billion, reflecting strong momentum across the business.
2. GE (GE -0.02%)
In 2023, GE Vernova launched Fleet Orchestration, a new software that leveraged AI and machine learning to help power utilities maximize the use of renewable energy.
While speaking about the utility and benefits the solution had to offer, Linda Rae, general manager of power generation and oil and gas for GE Vernova’s digital business, said the following:
“Renewables like wind and solar play an important role in the energy transition. They’re key to reducing the world’s (dependency) on fossil fuels. The challenge is that these energy sources are inherently variable — the wind doesn’t always blow, and the sun doesn’t always shine. But people always expect their lights to turn on when they flip the switch. This is where software solutions, such as Fleet Orchestration, can help reduce this uncertainty for power companies, giving them the confidence needed to ensure grid stability while simultaneously reducing emissions.”
According to a study released by ISO New England, reserve margins may need to increase from 15% to 300% by 2040 to reach 56% renewables use. GE believes its software’s ability to aid in setting appropriate margins – the amount of extra resources to keep power systems reliable in times of stress – comes at the right time.
General Electric Company (GE -0.02%)
A week ago, GE Vernova released its financial results for the fourth quarter and full year ending December 31, 2024. The company reported full-year revenue of $34.9B, +5%, +7%, organically driven by Electrification and Power.
AI and Solar Tech to Walk Hand in Hand in the Future
To become more useful for the solar tech regime, AI would have to take certain steps in the future. There would have to be a robust mechanism for collecting useful data. The next step would involve efficient operationalization of that data by properly structuring it to make it actionable. Thirdly, it has to be ensured that different datasets get to communicate effectively and provide value. Finally, the AI models would have to be optimized and fine-tuned to deliver the maximum they can.
Click here for a list of top solar stocks to invest in.
Study Reference:
1. Wu, J., Torresi, L., Hu, M., Reiser, P., Zhang, J., Rocha-Ortiz, J. S., Wang, L., Xie, Z., Zhang, K., Park, B.-W., Barabash, A., Zhao, Y., Luo, J., Wang, Y., Lüer, L., Deng, L.-L., Hauch, J. A., Guldi, D. M., Pérez-Ojeda, M. E., Seok, S. I., Friederich, P., & Brabec, C. J. (2024). Inverse design workflow discovers hole-transport materials tailored for perovskite solar cells. Science, 386(6727), 1256–1264. https://www.science.org/doi/10.1126/science.ads0901
