https://specials.auma.com/en/profox?utm_source=engineerlive&utm_medium=onlinebanner&utm_campaign=profox_2024

GenAI: IP understanding is key to unlocking the potential of solar

Engineer Live News Desk

From the development of advanced materials to enhance individual photovoltaic cell efficiency and longevity, to optimising the placement and operation of modules, GenAI and the application of machine learning (ML) are having a transformative effect across the full breadth of the solar industry.

Nowhere is AI having more of an impact than in the development of perovskite solar cells. Widely seen as the next disruptor in PV technology, perovskite-based solar cells promise very high solar conversion efficiencies in combination with being processable from solution, opening the door to cheaper and less energy-intensive fabrication techniques than traditional silicon PV. ML-based techniques are being successfully used to inform the design and development of perovskite solar cells, accelerating their route to commercialisation.

Historically, ensuring the long-term stability of perovskite active layers when cells are exposed to atmospheric conditions has been a challenge, and the process of integrating perovskite active layers with other PV components is not a straightforward task. In the past, researchers and engineers have largely been forced to use trial-and-error procedures when looking for potential solutions that could enhance the performance, or extend the life of perovskite devices. Now, GenAI models trained on vast quantities of scientific data about the characteristics and reactivity of chemical compounds, can predict solutions that are most likely to deliver the required improvements.

GenAI models have predicted new interface materials and surface passivation techniques for perovskite solar cells, which have been tested and shown to improve the open-circuit voltage and power conversion efficiency characteristics. New AI-derived encapsulant materials have also been scientifically tested, demonstrating durability benefits. Innovators are also using ML analysis to optimise the large-scale fabrication of perovskite semiconductor thin-films.

The use of GenAI is not limited to the production of new types of solar cell, it is also being used to improve the control of existing silicon PV installations. For example, the energy-generating potential of a PV installation can be limited by nearby structures that create shade at certain times of day or intermittent cloud cover. AI models trained on historic monitoring data can be used to better identify when low output from a module is due to normal variations in incident light or a fault condition.

Amazon is backing a project at a solar farm in Southern California, which is using machine learning to strengthen carbon-free energy production at the same time as helping to stabilise the grid. In other situations, AI can be used to detect faults and spot when components, such as microinverters, are not working properly. On a larger scale, energy network engineers are using AI models to support grid optimisation by predicting energy output spanning multiple installations.

Earlier this year, a team of researchers in China announced that they were using data from satellites and weather stations to train ML models, to pinpoint the best place to install dual-sided PV modules for maximum energy output. Several papers have also been published exploring the use of advanced AI algorithms to support the near-optimal sizing and placement of grid-connected PV installations.

Innovators operating in the solar industry, for example those looking to bring perovskite PV to market or identify enhanced control and operation techniques, should focus on building a robust intellectual property (IP) portfolio. Securing patents for innovative technologies now could pave the way to significant commercial rewards in the future.

While many innovative solar companies will be familiar with the process of patenting devices, materials and fabrication techniques, they may be less familiar with the additional factors that must be accounted for when patenting software inventions. When it comes to an AI-based invention, all the usual requirements will apply – such as, the invention must be new and inventive compared to what has gone before. However, additional factors need to be considered during the preparation of the patent application. For example, in Europe, the novel part of an AI-based invention must give rise to a tangible technical benefit. Additionally, when describing a novel technology in a patent application, the patentee must disclose sufficient detail for the invention to be repeatable by a third party. In Europe, information about the types of neural networks used and the datasets used to train their model is generally required to meet sufficiency criteria.

The solar industry has reached an exciting stage of development when advances made possible by GenAI and ML analysis are accelerating the way to significant efficiency and performance improvements. For innovative companies, IP understanding, particularly around AI inventions and methodologies, is vital to optimising commercial opportunities.

 

Andrew Hey is a senior associate, and Cleantech sector specialist at European IP firm, Withers & Rogers.

 

Recent Issues