Use of artificial neuronal networks for prediction of the control parameters in the process of anaerobic digestion with thermal pretreatment. Issue 10 (24th August 2018)

• Record Type: Journal Article
• Authors: Flores-Asis, Rita; Méndez-Contreras, Juan M.; Juárez-Martínez, Ulises; Alvarado-Lassman, Alejandro; Villanueva-Vásquez, Daniel; Aguilar-Lasserre, Alberto A.
• Is Part Of: Journal of environmental science and health. Volume 53:Issue 10(2018)
• Page Start: 883
• Page End: 890
• Publication Date: 2018-08-24
• Subjects: Anaerobic co-digestion -- artificial intelligence -- artificial neuronal network -- bioenergy; Environmental engineering -- Periodicals; Environmental sciences -- Periodicals; Ecology -- periodicals; Hazardous Substances -- periodicals; 628
• DOI: 10.1080/10934529.2018.1459070
• Languages: English
• ISSNs: 1093-4529

Abstract

ABSTRACT: This article focuses on the analysis of the behavior patterns of the variables involved in the anaerobic digestion process. The objective is to predict the impact factor and the behavior pattern of the variables, i.e., temperature, pH, volatile solids (VS), total solids, volumetric load, and hydraulic residence time, considering that these are the control variables for the conservation of the different groups of anaerobic microorganisms. To conduct the research, samples of physicochemical sludge were taken from a water treatment plant in a poultry processing factory, and, then, the substrate was characterized, and a thermal pretreatment was used to accelerate the hydrolysis process. The anaerobic digestion process was analyzed in order to obtain experimental data of the control variables and observe their impact on the production of biogas. The results showed that the thermal pre-hydrolysis applied at 90°C for 90 min accelerated the hydrolysis phase, allowing a significant 52% increase in the volume of methane produced. An artificial neural network was developed, and it was trained with the database obtained by monitoring the anaerobic digestion process. The results obtained from the artificial neural network showed that there is an adjustment between the real values and the prediction of validation based on 60 samples with a 96.4% coefficient of determination, and it was observed that the variables with the major impact on the process were the loading rate and VS, with impact factors of 36% and 23%, respectively.