Model Complexity Analysis, Identification, and Optimal Start up Control in Anaerobic Digestion

  • Wasim Ahmed

Student thesis: Doctoral Thesis

Abstract

This thesis work was aimed at addressing some of the issues in model structure development and identification for anaerobic digestion (AD) processes and subsequently, contributing to the development of a nonlinear model predictive control (NMPC) scheme for optimal AD start-up control intended to be evaluated and tuned against a virtual plant based on reliably structured AD models. As a case study for addressing AD model complexity challenges, a comparative analysis of five different model structures of sulfate reduction (SR) extended AD process was conducted to evaluate their accuracy and provide model developers and users with better information to decide on the optimum degree of complexity (Chapter 2). The evaluated models differed in terms of the number/type of sulfate reducing bacterial (SRB) activities considered A systematic calibration of the evaluated models against a large set of experimental data was also conducted using the parameter estimation and calibration method proposed in Chapter 3 of this thesis. Results indicate that a simpler model structure (incorporating both acetate and H2 utilizing SRB activities) achieves comparable, if not better, prediction performance when compared with more complex models against the experimental data set from literature. All the models evaluated provided acceptable predictions except the model including only hydrogen utilizing SRB activity. More complex model structures (incorporating additional SRB activities) are recommended only if required in specific experimental cases. For addressing parameter identification problems, a generalized method for the estimation and calibration of biokinetic parameters in AD models was proposed (Chapter 3). The method consists of a correlation-based approach to estimate specific groups of parameters mechanistically, followed by a sensitivity-based hierarchical and sequential single parameter optimization (SHSSPO) calibration method for the remaining groups of parameters. The correlation-based approach reduces the number iii of parameters to be fitted to data while the SHSSPO approach aims at eliminating the the uncertainty on optimality (and therefore parameter identification) associated to multivariable parameter calibration problems. The method was evaluated to estimate and calibrate the parameter values for SR processes when included into the IWA Anaerobic Digestion Model No. 1 (ADM1) and simulations were compared with experimental data from literature. Application of the proposed method led to the hydrogen sulfide inhibition parameters as the only ones requiring calibration against experimental data. Comparison of the proposed SHSSPO performance with that of multi-dimensional parameter optimisation methods shows a superior performance in terms of overall error and computation times. Final simulation results and comparisons with experimental data from literature led to model predictions of similar, if not better, quality than those achieved by multivariable parameter optimization methods. The results obtained suggest that the method could be generically applied and tested with other biokinetic models frequently used in wastewater treatment. Finally, an NMPC scheme was developed and virtually tested for optimal control of AD during start-up (Chapter 4). The scheme involves controlling effluent concentration of acetate, effluent concentration of aceticlastic methanogens in the reactor, and effluent methane production rate (control variables) by manipulating volumetric inflow rates of organic substrate, dilution water, and alkali addition (manipulated variables). The NMPC objective, with constraints, consists of minimization of set-point tracking errors for the control variables plus minimization of the amount of alkali added. A simple AD model structure was used for predictions during NMPC optimizations while the highly complex ADM1 was used to virtually represent AD process plant. As a starting point, the NMPC scheme was tested using a simple case scenario involving start-up of an AD system treating readily biodegradable substrate (glucose). The results with implementation of the proposed NMPC scheme are promising (stable levels of effluent acetate, aceticlastic methanogens in reactor, and CH4 production and no process destabilizations or acidification). Additional scenarios would be needed to fully evaluate the practical feasibility of the proposed control scheme.
Date of AwardJul 2018
Original languageAmerican English
SupervisorJorge Rodriguez (Supervisor)

Keywords

  • Anaerobic digestion; model complexity; parameter estimation; parameter calibration; ADM1; anaerobic digestion start-up; start-up control; optimal control; model predictive control

Cite this

'