Diagnostic concept for dynamically operated biogas production plants
Introduction
Producing biogas through anaerobic digestion of waste and energy crops is an important option for generating electric energy from renewable resources. Unlike solar and wind power, electricity from biogas can be produced on demand and so can compensate for fluctuations in the electricity grid. Three possible approaches are generally identified [1], [2]. The first and easiest option is to store the biogas locally and convert it on demand into electricity in a combined heat and power plant. The second option is to upgrade biogas to biomethane and distribute it via the natural gas grid (e.g. Luo and Angelidaki [3], Bensmann et al. [4], Díaz et al. [5]). This option makes use of the storage capacity of the natural gas grid. The third option is to operate the biogas plant dynamically, producing biogas on demand. This can be done by changing the substrate composition or the retention time, and therefore the organic loading rate. In this paper we focus on the third option.
Changing the organic loading rate dynamically is risky: Large disturbances can lead to washout of the methanogenic microorganisms and consequently to undesired operating states with no methane generation. To prevent this occurring, advanced controller systems such as adaptive online optimization or fuzzy control are possible (see, for example, Mailleret et al. [6] and Murnleitner et al. [7]). However, using this approach in large-scale industrial biogas plants is difficult for a number of reasons, including the absence of suitable instrumentation (e.g. sensors, process computers), staff are often not trained in such methods, and the need for parameterized process models. To overcome these problems, we propose in this paper an intuitive, deterministic approach. In particular, we address the following key questions: Is it possible to detect an upcoming system failure with the help of the dynamic system response early enough to prevent it? And which variables would enable a reliable diagnosis?
Various diagnosis variables are discussed in the literature from recent decades (e.g. Switzenbaum et al. [8], Ahring et al. [9], Kleybocker et al. [10]). These variables can be grouped with respect to their measurement medium. Thus, possible diagnosis variables in the gas phase are the composition of the gas (methane, carbon dioxide, hydrogen) and the gas production rate. Diagnosis variables in the liquid phase include the pH value and the concentration of dissolved hydrogen, inorganic carbon or volatile fatty acids (VFA). Thanks to recent improvements in measuring technology, specific VFAs are a promising indicator [11]. Diagnosis variables in the solid phase also exist, such as the concentration of methanogenic microbes [12] and the composition of the microbial community [13]. However, both of these are hard to determine online.
Previous studies have analyzed similar questions using an experimental approach. For example, investigators have looked at different measurements following disturbances in the hydraulic retention time, and changes in the composition of the substrate in step or pulse form (e.g. Mechichi and Sayadi [14]). In one of the most comprehensive studies, Angelidaki et al. [15] analyze a laboratory-scale continuous stirred tank reactor (CSTR) fed with cattle manure exposed to a variety of disturbances. The researcher measure gas production, gas composition, pH value and individual VFAs demonstrating the sensitivity of these variables to disturbances. They conclude that the concentration of VFA (e.g. propionic acid) is the most important parameter for characterizing the process state; however, the biogas production rate should be also evaluated in order to monitor process performance. Molina et al. [16] use similar experiments and statistical methods to determine a minimal set of measured variables for discriminating the current process state. They conclude that alkalinity should also be taken into account.
These earlier studies are valid only for the specific operating points and disturbances that they analyzed. By contrast, the present paper systematically analyzes process behavior at different steady-state organic loading rates and following step changes in the organic loading rate over a wide range. This was only possible in a simulation study, as experiments of this type are both time-consuming and expensive. Clearly, the capability of a model to predict the exact state of a biological system is limited due to uncertainties in the model and side effects that are not modeled. Nevertheless, this approach is justified for performing qualitative analyses and predicting the principle trends. Our results therefore complement earlier experimental studies and provide some guidance for the operation of biogas plants.
In this paper we analyzed two different experimental systems in order to identify relevant diagnosis variables. We then suggest a practical operating scheme for detecting and preventing process failure. Within this framework, we evaluate different diagnosis variables with respect to their ability to detect disturbances early on and to discriminate between disturbances that are acceptable and disturbances that lead to process failure. We focus on a qualitative comparison of various easily measurable diagnosis variables so that the proposed operating scheme can be directly implemented in a full-scale biogas plant. We present the results in the form of an operating diagram, enabling easy comparison.
This paper is structured as follows. Following this introduction, Section 2 describes the experimental systems and mathematical model used. Next, Section 3.1 discusses the dynamic response of the experimental systems. Section 3.2 is a preliminary analysis of the dynamic systems response. Section 3.3 presents the operating scheme for detecting unacceptable disturbances and preventing system failure and then evaluates it systematically with the help of a suitable dynamic process model for different diagnosis variables. Finally, Section 4 summarizes the results and suggests conclusions.
Section snippets
Material and methods
This section presents the experimental methods and equations used in the mathematical model.
Results and discussion
The following section consists of three parts: Section 3.1 discusses the experimental results; Section 3.2 presents a preliminary analysis of the model; and Section 3.3 describes the operating scheme and an evaluation of the different diagnosis variables.
Conclusions and outlook
The present analysis focuses on the dynamic operation and evaluation of diagnosis variables and conditions for biogas plants. We propose an operating scheme for detecting unacceptable disturbances and making process intervention. Within this scheme, we compare different diagnosis variables and diagnosis conditions within a simulation study. The results are shown with the help of operating diagrams in which organic loading rates both before and after the disturbances (OLR− and OLR+) are varied.
Acknowledgment
The authors would like to thank Robert Flassig from the Process Systems Engineering Group and our colleagues from the Group Computational Methods in Systems and Control Theory at the Max Planck Institute for their support and for providing the Linux-Cluster otto that was used for some of the simulations to reduce computation time.
R. Heyer was supported by the German Environmental Foundation (DBU, 20011/136).
This study was supported by the national research project ”BiogasEnzyme” regarding
References (25)
- et al.
Review of concepts for a demand-driven biogas supply for flexible power generation
Renew. Sustain. Energy Rev.
(2014) - et al.
Biological methanation of hydrogen within biogas plants: a model-based feasibility study
Appl. Energy
(2014) - et al.
A feasibility study on the bioconversion of CO2 and H2 to biomethane by gas sparging through polymeric membranes
Bioresour. Technol.
(2015) - et al.
Nonlinear adaptive control for bioreactors with unknown kinetics
Automatica
(2004) - et al.
State detection and control of overloads in the anaerobic wastewater treatment using fuzzy logic
Water Res.
(2002) - et al.
Monitoring of the anaerobic methane fermentation process
Enzyme Microb. Technol.
(1990) - et al.
Early warning indicators for process failure due to organic overloading by rapeseed oil in one-stage continuously stirred tank reactor, sewage sludge and waste digesters
Bioresour. Technol.
(2012) - et al.
Regulation and optimization of the biogas process: propionate as a key parameter
Biomass Bioenergy
(2007) - et al.
Population dynamics at digester overload conditions
Bioresour. Technol.
(2009) - et al.
Metaproteome analysis of the microbial communities in agricultural biogas plants
New Biotechnol.
(2013)
Evaluating process imbalance of anaerobic digestion of olive mill wastewaters
Process Biochem.
Reactor configurations for biogas plants - a model based analysis
Chem. Eng. Sci.
Cited by (15)
Biogas standard system in China
2020, Renewable EnergyCitation Excerpt :Lavergne et al. reported an important variability in the experimental results in an anaerobic digestion lab test and argued that experimental standardization in batch conditions are urgently necessary [26]. Meanwhile, detecting unacceptable load changes to prevent process failures is vital for the dynamic operation of biogas plants [27]. Fortunately, two national standards, “Analysis Standard of Composition of Feedstock for Methane Production” and “Test Method of Biogas Production Capability for Raw Material,” are under development by SAC/TC 515.
Machine learning powered software for accurate prediction of biogas production: A case study on industrial-scale Chinese production data
2019, Journal of Cleaner ProductionCitation Excerpt :For instance, Grim et al. (2015) applied the Dynamic Biogas plant Model (DyBiM) and ADM1 to evaluate the technical requirements and economic implications of demand-oriented biogas production in a Swedish CHP biogas plant digesting cattle manure and sugar beet. In addition, Bensmann et al. (2016) used the ADM1-based operating scheme capable of detecting process disturbances in anaerobic digestion. Another popular model used for calculating the biogas potential of anaerobic digestion is Buswell's equation.
Demand-driven biogas production by flexible feeding in full-scale – Process stability and flexibility potentials
2017, AnaerobeCitation Excerpt :The pH-value corresponds inversely proportional to the acid concentrations. Hence, the pH can be suitable for an easy-to-measure indicator, as already recommended by Ref. [25]. The experiments show that the stop of substrate addition for 2 days does not harm the general responsiveness of biogas production.
Proteotyping of laboratory-scale biogas plants reveals multiple steady-states in community composition
2017, AnaerobeCitation Excerpt :First regular feeding/sampling was performed two days later along with the installation of the “GärOnA” gas analysis system, which will in the following be referred to as the starting point of the cultivation (0 d). The basic composition of the medium was described by Bensmann et al. [20] with the following exception: meat extract, yeast extract, and peptone were replaced by 11.5 g L−1 glucose and 34.4 g L−1 cellulose as carbon source to exclude any undefined compounds. In addition 2.56, 5.13 or 15.38 g L−1 urea were used as a nitrogen source, depending on the process regime (see below).