Deep learning-based transmembrane pulsed electro-chemisorption for improved energy efficacy of recovering ammonia from wastewater

作者：Zhiqiang Zhang, Beiqi Deng, Zhiyong Cao, Shuchen Zhang, Jiao Zhang, Jiao Zhang, Junwen Zhang, Junwen Zhang, Dequan Wei, Pengyu Xiang, Xinchao Liu, Siqing Xia · 发表于：Energy & Environmental Sustainability · 年份：2025 · DOI：10.1016/j.eesus.2025.100029 · 被引用次数：7 · 研究领域：Ammonia Synthesis and Nitrogen Reduction、Membrane-based Ion Separation Techniques、Wastewater Treatment and Nitrogen Removal

The transmembrane electro-chemisorption technology for recovering ammonia from wastewater has gained growing attention for its chemicals saving and easy scalability. However, high energy consumption remains a challenge for its application. Herein, a deep learning (DL)-based transmembrane pulsed electro-chemisorption method was first developed to efficiently improve the energy efficacy of a stacked transmembrane electro-chemisorption (sTMECS) system for recovering ammonia. When the sTMECS system was adjusted from direct current power supply (DCPS) mode to default pulse power supply (PPS) mode with duty cycle of 50 %, the specific energy consumption ( SEC ) for recovering ammonia from synthetic wastewater decreased from 144.5 kWh/kg N to 36.1 kWh/kg N. Furthermore, a backpropagation neural network (BPNN) model was constructed and trained using experimental datasets of the sTMECS system, and an intelligent dynamic strategy for adjusting the duty cycle was proposed. The SEC for recovering ammonia from real urine under intelligent PPS mode was 18.9 kWh/kg N, showing reductions of 80.7 % and 23.1 % compared to DCPS mode and default PPS mode, respectively. Therefore, this study provides a new way for recovering ammonia from wastewater with relatively low energy consumption via combining the artificial intelligence technologies with the present ammonia-recovering methods, which is also conducive to reducing the carbon footprint.