Prediction of total organic acids concentration based on FOS/TAC titration in continuous anaerobic digester fed with food waste using a deep neural network model

Abstract

In this work, the complexities of anaerobic digestion fed with highly degradable feedstock are investigated, focusing on accumulation of organic acids (OA) as a critical monitoring parameter, and the significance of prediction models for total OA concentration. The anaerobic digestion of food waste (FW) was conducted under the organic loading rate (OLR) range of 2.55-8.80 g COD/L/d and hydraulic retention time (HRT) of 30-15 days. The feasibility of fl & uuml;chtige organische sa<spacing diaeresis>uren (FOS), totales anorganisches carbonat (TAC), and the FOS/TAC was investigated by predicting the total OA using a deep neural network (DNN) model. Two digesters, Digester 1 and 2, were fed with FW from four distinct sites. When the OA concentration exceeded 2 g/L as CH3COOH, the feeding was paused to recover the methanogens activity. The total OA concentration was successfully predicted with FOS, TAC, and FOS/TAC using the DNN regression model even though applying on the datasets from two distinct digesters, indicating a R-value of 0.9557, R2 of 0.9133, and mean square error of 0.0329. The predictive capability of DNN regression model shows the feasibility of total OA prediction based on the titrimetric method for monitoring and optimizing continuous anaerobic digestion of highly degradable feedstock.