Data-driven techniques for fault detection in anaerobic digestion process

Anaerobic digestion (AD) is an appropriate process for bio-energy (biogas) production from waste and wastewater receiving a high level of attention at both academic and industrial scale due to increasing public awareness regarding environmental protection and energy security. Monitoring such processes is an imperative task to ensure optimized operation and prevent failures and serious consequences during the operation of the plant. To fulfill this task, a practical data-driven framework for fault detection in AD is proposed and validated on a simulated data set obtained using the benchmark simulation model No.2 (BSM2) from the International Water Association (IWA). The proposed framework is based on data-driven soft-sensors predicting total volatile fatty acids (VFA), mainly consisting of acetate, propionate, valerate and butyrate concentrations inside the digester. The VFA concentration is considered because it does not only reflect the current process health, but it is also sensitive to the incoming feeding imbalances. VFA soft-sensors using different advanced techniques such as support vector machine (SVM), extreme learning machine (ELM) and ensemble of neural network (ENN) are tested and compared in terms of accuracy and fault detection (FD) robustness. A principal component analysis (PCA) model was also developed to compare the proposed approaches with the traditional FD method. By applying soft-sensors, the residual signal, i.e., the difference between estimated and measured VFA values can be generated. This residual signal can then be used in combination with univariate statistical control charts to detect the faults. A comparison of the proposed FD framework with PCA method clearly demonstrates the over performance and feasibility of the proposed monitoring framework. (C) 2020 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.