Fuzzy-decision tree modeling for H2S production management in an industrial-scale anaerobic digestion process

Abstract

H2S is a byproduct in anaerobic digesters (ADs), causing significant challenges due to its negative impact on biogas quality and equipment corrosion. As a solution, previous researchers developed complex mechanistic models that are computationally expensive due to solving equations with many variables. These models were only tested within normal H2S ranges. This research introduces a novel hybrid fuzzy-decision tree (F-DT) model to predict H2S production from primary/secondary sludge ratio (P/S), dry solids (DS), and volatile solid (VS) reduction parameters in the Qom wastewater treatment plant in Iran. Notably, the H2S concentrations in this facility consistently surpass the acceptable thresholds reaching levels as high as 6500 ppm, in contrast to the designed accepted limit of 3000 ppm. The process involved collecting and pre-processing data, creating a decision tree classifier for rule generation, developing a fuzzy model suited for complex systems like AD, and optimizing it with a genetic algorithm. The analysis of the final model revealed that the P/S ratio significantly influences elevated H2S levels, indicating that maintaining a P/S ratio > 1.2 holds promise as an effective strategy. The model is both simple and interpretable, with decent accuracy in predicting H2S compared to previous models. This study not only introduces a useful method for optimizing industrial-scale ADs, but also aims to uncover the reasons behind high H2S levels.