Scientists have environmental recycling processes that convert farm and forestry waste and polluted soil into useful energy and other useful materials. However, it has been very challenging to precisely control these processes. A research team is using machine learning to regulate the environmental processes, so they are more effective and predictable. Their work provides a technical path for achieving precise, intelligent, and sustainable remediation of polluted soil.
Their research was published in the journal Environmental Chemistry and Safety on July 1, 2026.
Co-pyrolysis technology is an environmental recycling process that heats waste products from farms and forests, along with polluted soil, in an oxygen-limited environment. This heating reaction converts the waste products and polluted soil into fuel and other useful materials.
The challenge that scientists have faced with the co-pyrolysis technology lies with the differences in the raw materials. Because each batch of raw materials behaves a little differently, it is difficult for scientists to accurately control the co-pyrolysis process.
The research team from Panzhihua University, Tsinghua University, and Harbin Institute of Technology has developed a theoretical framework that combines machine learning with the co-pyrolysis technology to provide a precise and sustainable solution to turn polluted soil to clean soil.
The world produces over 2 billion tons of farm and forest waste every year. Most of this waste is not used in any way. At the same time, more than one-third of the world’s farmlands cannot be used for farming because the soil is polluted with toxic chemicals. Much of the plant waste is dumped in landfills or burned, which causes air pollution. Cleaning the polluted soil requires methods that harm the soil’s health and create new pollution. These two environmental problems are connected to each other.
The co-pyrolysis technology offers a solution for farm and forest waste and polluted soil. This process “bakes” the plant waste and polluted soil under oxygen limited conditions. The plant waste and polluted soil react better together than either of them would work on their own. The reaction that occurs cracks and destroys the organic pollutants and traps the heavy metals so they can no longer harm the environment. Finally, the process turns the waste and soil into a high-performance charcoal-like substance called biochar.
While the co-pyrolysis technology offers a promising solution to these environmental problems, the process has been hard to control. Each batch of plant waste and polluted soil is different, making it hard to predict and control the exact repair effects they can achieve.
The advances in machine learning technology in recent years give scientists the tools they need to better manage the co-pyrolysis process. Algorithms, such as deep learning, neural networks, and reinforcement learning, help scientists better understand the complex data related to the co-pyrolysis process. Machine learning tools let them to fix the polluted soil at different sizes, ranging from the molecular level all the way to the larger ecosystem.
Using machine learning with the co-pyrolysis methods allows scientists to predict the connections across entire process. “Compared to traditional experience-oriented process development models, the core advantage of machine learning driven co-pyrolysis closed-loop design lies in its ability to model and predict high-dimensional nonlinear relationships across the entire chain of raw materials, processes, products, and ecological responses,” said Yuanchuan Ren, Panzhihua University, China.
Looking to the future, the research team suggests that the co-pyrolysis closed-loop system that recycles waste could have potential for use beyond Earth. “The technical framework and engineering methods of the co-pyrolysis closed-loop system are expanding beyond the scope of Earth's environmental remediation, gradually extending to the utilization of in-situ resources in extraterrestrial celestial bodies and the construction of interstellar human settlements,” said Ren.
The research team includes Yuanchuan Ren, Yuhang Lin, Xuejun Zhu, Hongbo Han, Shiyong Zhao, Renjie Huang, Tingfeng Su, Yan Guo, Fenghui Wu, Qiang Niu, Dandan Chen from Panzhihua University; Cheng Wang from Tsinghua University; and Nanqi Ren from Harbin Institute of Technology.
This research is funded by Sichuan Science and Technology Program, the Panzhihua Key Laboratory of Chemical Resource Utilization Open Science Project, Panzhihua Association for Science and Technology Youth Science and Technology Talent Support Project, the Key Laboratory of Dry-hot Valley Characteristic Bio-Resources Development at University of Sichuan Province, and The College Students' Innovation and Entrepreneurship Training Program.
D OI Link:
https://doi.org/10.26599/ECS.2026.9600049
Environmental chemistry and safety
Machine learning driven closed-loop system for co-pyrolysis of polluted soil and biomass: Design principles and multi-scale regulation mechanism for soil health
1-Jul-2026