Modelling, optimisation and control of anaerobic co-digestion processes

Resumo

Anaerobic digestion (AD) is a biological process that occurs spontaneously in nature. However, its performance and methane yield varies widely depending on the type of organic matter and the surrounding environmental conditions to which the substrates are exposed. The installation and operation of anaerobic digesters under controlled operating conditions can enhance the efficiency of AD process treating different sorts of organic substrates. The aims of these facilities include: to achieve energy recovery in the gas stream (biogas), and to obtain a more stabilised liquid effluent (digestate). In the last decades, the interest in anaerobic digestion has grown to treat organic wastes of different types such as sewage sludge generated in wastewater treatment plants, organic fraction of municipal solid wastes and agro-industrial residues. More recently, the interest in the simultaneous digestion of different organic wastes, known as anaerobic co-digestion (AcoD), has emerged based on the potential synergistic effects of some types of residues. The resulting blend of substrates can be enhanced taking advantage of the different compositions and physicochemical characteristics of the individual co-substrates. The determination of the adequate blend of substrates in AcoD that leads to a stable operation is not trivial. It requires knowledge on the process and expertise on the operation since AD involves a complex reaction pathway to convert complex substrates into biogas. This thesis contributes to the modelling, optimisation and control of AcoD processes aiming at improving the performance of anaerobic co-digesters. The fundamentals of anaerobic digestion and co-digestion are introduced in Chapter 1. In addition, a review of the main substrates used in anaerobic co-digestion, the typical reactor configurations and the key physicochemical parameters used for monitor and control are presented. A general background on AD modelling is also presented, mainly focussed on the description and applications of the Anaerobic Digestion Model No. 1 (ADM1). Since its publication, ADM1 has become the standard model to simulate AD processes, on top of which many extensions have been developed. Regarding control and optimisation of AcoD systems, a list of control strategies are reviewed and classified in different categories (PID controllers, fuzzy logic controllers, neural networks...). Chapter 2 gives a detailed description of the pilot plant used to carry out the experiments of this thesis. It consists of a hybrid Upflow Anaerobic Sludge Blanket (UASB) and Anaerobic Filter (AF) highly instrumented. The pilot plant is not only equipped with conventional sensors for pH, temperature or gas flow, but also with non-conventional advanced sensors and analysers to monitor on-line variables such as volatile fatty acids (VFA), alkalinity, biogas composition or total organic carbon. The experiments described in Chapters 3, 5 and 6 have been performed entirely in this plant. A generalised modelling approach to implement diverse soluble fermentable substrates into an ADM1-based model is developed and validated in Chapter 3. The fermentation reactions of substrates such as ethanol or glycerol, not included in the original ADM1, are implemented as glucose equivalent fermentation reactions. Assuming acidogenesis as the fastest step in anaerobic digestion, an accurate description of the stoichiometry in the fermentation of soluble substrates (ethanol, glycerol, lactate...) into VFA products (butyrate, propionate and acetate) is not required in detail as long as the mass and electron balances remain accurate since all these intermediate acids are quickly converted to acetate, hydrogen and carbon dioxide in methanogenic systems. The model is implemented in MATLAB/Simulink and allowed simulating the AcoD experiments carried out in Chapter 3 and Chapter 5, treating an ethanol-containing co-substrate (diluted wine) and a glycerol-containing co-substrate (biodiesel by-product), respectively. The treatment of solid wastes with anaerobic digestion or co-digestion is attractive due to their high organic content and potential energy recovery. In the case of treating particulate substrates, the disintegration-hydrolysis stage is considered the slowest step of the process. Chapter 4 presents a novel modelling approach of the disintegration and hydrolysis stages of complex particulate substrates. Solid substrates are assumed to contain a readily- and a slowly-biodegradable fraction. The modelling approach considers a decoupled disintegration of these two fractions to better describe the degradation of solid wastes. This approach is validated with the ADM1-based model by simulating batch assays of numerous fruit and vegetable wastes (FVW) and a continuous AcoD experiment treating a set of FVW. Anaerobic co-digestion can improve the performance of biogas plants in terms of higher methane productivities and more stabilised operations taking advantage of the synergisms between different co-substrates. However, not all combinations of substrates are feasible or appropriate for AcoD processes. Chapter 5 formulates an optimisation method based on linear programming to calculate the best feeding (substrate flows and hydraulic retention time, HRT) of co-digestion systems, capable of maximising the energy recovery at each organic loading rate (OLR) applied. The resulting blend of substrates is subjected to a set of physicochemical constraints, defined based on the heuristic knowledge of AD processes. This optimisation method is validated at pilot scale treating different blends of three substrates (pig manure, glycerine and gelatine) at different OLR. Finally, Chapter 6 introduces a novel control strategy for anaerobic co-digestion. An optimum blend obtained by linear programming is fed in a continuous AcoD experiment, and after a period of time a diagnosis system assesses the performance of the process. Alkalinity ratio and methane production are defined as key diagnosis indicators to estimate the stability of the process against acidification and the extent of energy recovery, respectively. Based on the diagnosis outcome, the control action modifies the boundaries of the restrictions applied in the calculation of the optimum feedings. This control action leads to the reach of a new optimum point when the optimisation method is again evaluated, obtaining a new blend of substrates and HRT for the next period of operation. As a result, the strategy works as a closed-loop controller that optimises the feeding and diagnoses its performance over time. When the system is stable (low alkalinity ratio values are obtained) new calculated feedings are inputted to increase the methane production rate (normally at higher OLR); and when the system is unstable, lower OLR are applied to favour the system to recover from acidification or to prevent from failure.