Monday, 2024 May 27

China’s AI solar energy breakthrough calls for cautious optimism

In February, scientists from Tsinghua University and the National Tibetan Plateau Data Centre unveiled findings from a research study, showcasing a new artificial intelligence model that can potentially be used to enhance the efficiency of double-sided solar panels.

According to the study published in the Journal for Remote Sensing, the scientists used machine learning models and data augmentation techniques to analyze sunshine duration data collected from over 2,453 weather stations throughout China. The application of machine learning algorithms effectively overcame the limitations posed by sparse and unevenly distributed ground-based observations, enabling the team to predict solar radiation patterns with unprecedented accuracy.

Photovoltaic (PV) panels, also referred to as solar panels, are designed to convert sunlight directly into electricity using semiconductor materials. When sunlight strikes a PV panel, it energizes the electrons within its material, causing them to move and generate an electric current.

Unlike traditional panels, double-sided panels have a transparent back sheet, allowing sunlight to pass through. Diffused rays may subsequently be reflected off surfaces to reach the back side of the panels, generating electricity. In addition to being difficult to transport and maintain, the power-generating potential of a double-sided PV panel depends greatly on how much diffuse solar radiation reaches its rear. Thus, finding locations with optimal solar radiation is crucial to maximize solar output.

The new methodology is particularly groundbreaking because it does not rely on local data for calibration, making it a universally replicable solution. This, combined with the model’s notable precision, signals great potential for the technology’s application on a global scale.

Furthermore, according to Yang Kun, a senior author of the study, the AI system revealed great solar potential within remote areas of China currently lacking in power line infrastructure, such as the eastern Tibetan Plateau and the Taklamakan Desert in Xinjiang. The study’s findings could not only pave the way for future research and policy planning in China, but also provide data that could be applied in other fields such as agriculture, as plants have been found to carry out photosynthesis more efficiently under diffuse light conditions.

China already enjoys a dominant position in the field of solar energy. According to the New York Times, it installed more solar panels in 2023 alone than the US has in its history, exports of fully assembled solar panels rose by 38%, and key component exports nearly doubled. Such success has arisen largely due to China’s formidable advantage in both land and labor, enabling it to produce polysilicon PV modules at a much lower cost compared to the US, India and especially Europe. Moving forward, the country plans to accelerate the expansion of its solar projects as part of a trio of emerging technologies alongside electric vehicles and lithium batteries.

However, the AI model detailed in the study is not without its drawbacks, as it may require a considerable amount of energy and water to run. For context, according to the University of California, training OpenAI’s GPT-3 alone consumed 700,000 liters of clean freshwater, while Schneider Electric estimated that AI consumes nearly as much power annually as the nation of Cyprus.

Tackling these issues will require finding better methods of training AI systems, and ensuring that the systems run on renewable energy. In this vein, US lawmakers introduced a bill in February requiring the federal government to assess AI’s current environmental footprint and develop a standardized system for reporting future impacts. Similarly, the European Union’s AI Act will require “high-risk AI systems” to report energy and resource use from 2025 onwards, while the International Organization for Standardization (ISO) said it will issue criteria for “sustainable AI” later this year.

MORE FROM AUTHOR

Related Read