Adaptive neural-fuzzy inference system vs. anaerobic digestion model No.1 for performance prediction of thermophilic anaerobic digestion of palm oil mill effluent

摘要

Palm oil industry generates high volume of palm oil mill effluent (POME) albeit contributing significantly to the economy of several ASEAN countries. This necessitates effective waste management methods Thermophilic high-rate anaerobic reactor accompanied by an accurate model to define and to predict the process performance can be a promising solution for POME treatment. Various mechanistic and meta-heuristic models had been developed, but not specifically designed for thermophilic anaerobic digestion of POME. This study explores the possibility of using ADM1 for estimating the performance of a thermophilic anaerobic reactor for POME treatment and compares it to Adaptive Neural-Fuzzy Inference System (ANFIS) model. A total of six prediction models were developed using ADM1 and ANFIS to estimate effluent pH, COD (Chemical Oxygen Demand), Total Suspended Solids (TSS) and methane composition. Results indicated that all ANFIS models were better than ADM1 models, with difference in the average error of up to 6.81%. However, ADM1 is more suited for better understanding of overall reaction of the system particularly via sensitivity analysis performed on the models.