Monitoring of MSW Incinerator Leachate Using Electronic Nose Combined with Manifold Learning and Ensemble Methods

Abstract

Waste incineration is regarded as an ideal method for municipal solid waste disposal (MSW), with the advantages of waste-to-energy, lower secondary pollution, and greenhouse gas emission mitigation. For incineration leachate, the information from the headspace gas that varies at different processing processes and might be useful for chemical analysis, is ignored. The study applied a novel electronic nose (EN) to mine the information from leachate headspace gas. By combining manifold learnings (principal component analysis (PCA) and isometric feature mapping (ISOMAP), and uniform manifold approximation and projection (UMAP) and ensemble techniques (light gradient boosting machine (lightGBM) and extreme gradient boosting (XGBT)), EN based on the UMAP-XGBT model had the best classification performance with a 99.95% accuracy rate in the training set and a 95.83% accuracy rate in the testing set. The UMAP-XGBT model showed the best prediction ability for leachate chemical parameters (pH, chemical oxygen demand, biochemical oxygen demand, ammonia, and total phosphorus), with R2 higher than 0.99 both in the training and testing sets. This is the first study of the EN application for leachate monitoring, offering an easier and quicker detection method than traditional instrumental measurements for the enforcement and implementation of effective monitoring programs.

Keywords

• electronic nose
• incinerator leachate
• data mining
• prediction
• classification