Microbial Diversity and Key Metabolic Pathways in Lignite-Promoted Anaerobic Fermentation with Residual Sludge

To enhance methane production efficiency in lignite anaerobic digestion and explore new ways for residual sludge utilization, this study employed the co-fermentation of lignite and residual sludge for biomethane conversion. The bacterial colony structure, metabolic pathways, and interactions between residual sludge and lignite in anaerobic methanogenic fermentation with different mass ratios were analyzed using macrogenomics sequencing. This study aimed to explore the mechanisms involved in the co-anaerobic fermentation of lignite and residual sludge. The results indicated that the addition of sludge enhanced the metabolic pathways in hydrolysis acidification, hydrogen-acetic acid production, and methanation phases. Notably, the enhancement of acetate- and carbon dioxide-nutrient metabolic pathways was more pronounced, with increased activity observed in related enzymes such as acetic acid kinase (k00925) and acetyl coenzyme synthetase (K01895). This increased enzymatic activity facilitated the microbial conversion of biomethane. The results of the study indicated that the sludge exhibited a promotional effect on the methane produced through the anaerobic fermentation of lignite, providing valuable insights for lignite and residual sludge resource utilization.

基于优化决策树的时延敏感流智能感知调度

目前流量调度策略无法做到智能按需化,尤其对于网络突发故障造成的拥塞以及高价值业务的护航场景,无法按需保障时延敏感的业务体验。通过分析研究不同网络业务流量时延敏感性属性需求,探索挖掘不同网络业务流量的行为特征与其时延敏感性需求之间的内在关联关系。然后利用AI技术对这种内在的关联关系进行学习,构建其映射关系,实现了时延敏感流智能感知调度。同时,考虑AI模型的可解释性及可部署性实际问题,采用强化学习剪枝优化可解释性决策树模型,提高模型的鲁棒性同时使模型更轻量化,易于设备部署实现。通过真实网络流量实验,强化学习优化后的决策树模型在单次推理情形下感知正确率提高1.75%,推理速度提升约30%;同时,实验也证明了使用局部微观统计特征多次推理有助于提高模型感知正确率。在所有实验中,强化学习优化的决策树模型规模缩小了60.0%~87.2%,并且Saras比Q-learning具有更好的优化表现。