基于臭气测定值的渗滤液恶臭浓度机器学习预判模型

Machine learning prediction model of leachate malodor concentration based on odor unit

为了评估生活垃圾运输过程中,发酵垃圾渗滤液中气味浓度的变化.采用控温发酵对含水量不同的渗滤液样本进行处理,并应用仪器测量与人工嗅辨相结合的方法来确定气味浓度.结果表明,仪器测量结果与人工嗅辨值之间存在显著的相关性(r=0.916),从而验证了所采用方法的有效性.进一步地,研究分析了水分含量、温度和发酵时间对气味浓度的具体影响.通过比较随机森林、XGboost和LightGBM等先进机器学习模型的性能,使用MAE、MSE、MAPE等评价指标后,确认随机森林模型在预测气味浓度方面的优越性.这些结果为理解和控制垃圾处理过程中的气味扩散提供了实用的参考,并为机器学习技术在环境科学研究中的应用奠定了基础.

This study delves into the assessment of odor concentration in fermented landfill leachate during the transportation of domestic waste.Leachate samples with varying moisture content underwent controlled temperature fermentation,and their odor concentration was evaluated using both equipment-based measurements and manual olfactory assessments.The results established a significant and robust correlation(r=0.916) between the two evaluation methods.Furthermore,the study identified the influential role of moisture content,temperature,and fermentation duration on the resulting odor concentration.To predict the odor concentration,state-of-the-art machine learning models,including random forest,XGboost,and LightGBM,were employed.Evaluation metrics such as MAE,MSE,MAPE,and R^(2) were utilized to determine the model′s fitting performance,ultimately highlighting the random forest model as the optimal choice.