You are on page 1of 10

Bioresource Technology 193 (2015) 513–522

Contents lists available at ScienceDirect

Bioresource Technology
journal homepage: www.elsevier.com/locate/biortech

Review

Automatic process control in anaerobic digestion technology: A critical


review
Duc Nguyen a, Venkataramana Gadhamshetty c, Saoharit Nitayavardhana b, Samir Kumar Khanal a,⇑
a
Department of Molecular Biosciences and Bioengineering, University of Hawai’i at Mānoa, 1955 East-West Road, Honolulu, HI 96822, USA
b
Deparment of Environmental Engineering, Faculty of Engineering, Chiang Mai University, Chiang Mai 50200, Thailand
c
Civil and Environmental Engineering, South Dakota State University, 501 E. St Joseph Street, Rapid City, SD 57701, USA

h i g h l i g h t s

 Critical review on automatic controls for diverse scales of AD processes.


 Monitoring in AD system was focused in chromatography and spectrometry techniques.
 Advanced controls are usually equipped with simple sensors and vice versa.
 Centralized AD plants could promote applications of automatic controls.

a r t i c l e i n f o a b s t r a c t

Article history: Anaerobic digestion (AD) is a mature technology that relies upon a synergistic effort of a diverse group of
Received 1 May 2015 microbial communities for metabolizing diverse organic substrates. However, AD is highly sensitive to
Received in revised form 16 June 2015 process disturbances, and thus it is advantageous to use online monitoring and process control tech-
Accepted 17 June 2015
niques to efficiently operate AD process. A range of electrochemical, chromatographic and spectroscopic
Available online 24 June 2015
devices can be deployed for on-line monitoring and control of the AD process. While complexity of the
control strategy ranges from a feedback control to advanced control systems, there are some debates
Keywords:
on implementation of advanced instrumentations or advanced control strategies. Centralized AD plants
Anaerobic digestion
Bioenergy
could be the answer for the applications of progressive automatic control field. This article provides a
Monitoring critical overview of the available automatic control technologies that can be implemented in AD
Automatic control processes at different scales.
Process control Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction complex polymers into simple soluble products; (ii) fermentation


of simple soluble products (e.g., amino acids, fatty acids and sug-
Anaerobic digestion (AD) processes have successfully been ars) into short-chain fatty acids (C > 2), ethanol, lactic acids etc.,
employed to remediate waste ranging from high strength indus- also known as acidogenesis; (iii) anaerobic oxidation of
trial wastewater, sewage sludge, municipal and agricultural solid short-chain fatty acids or volatile fatty acids (VFAs) into acetic acid
wastes to animal manures while generating renewable energy and hydrogen also known as acetogenesis and (iv) methane pro-
resources. The major advantages of AD process includes low energy duction from acetate and hydrogen via methanogenesis. The AD
consumption, less sludge yield, and above all, its ability to stabilize process could be inhibited by its own intermediates such as
diverse organic wastes with concomitant production of renewable VFAs, ammonia and hydrogen produced during digestion (Chen
energy, biogas (Metcalf and Eddy, 2004; Khanal, 2008). et al., 2008), and it is critical to design and operate an AD bioreac-
The fundamental science underlying the AD process has been tor that enables precise control over the operational parameters
well covered in many reviews (Speece, 1996; Khanal, 2008). AD and promote the syntrophic growth of diverse microorganisms. A
process is complex due to the involvement of diverse microbial lack of meticulous control of operational parameters results in a
communities supporting a series of interdependent biochemical catastrophic failure of AD processes. Lusk and Wiselogel (1998)
reactions. The major steps in AD process include (i) hydrolysis of reported that among 57 operating farm-based anaerobic digesters
in the U.S, the failure rate was over 50%. Besides poor design and
⇑ Corresponding author. Tel.: +1 808 956 3812; fax: +1 808 956 3542. high maintenance cost, major reasons for cessation of AD operation
E-mail address: khanal@hawaii.edu (S.K. Khanal). were due to generation of corrosive hydrogen sulfide gas, inability

http://dx.doi.org/10.1016/j.biortech.2015.06.080
0960-8524/Ó 2015 Elsevier Ltd. All rights reserved.
514 D. Nguyen et al. / Bioresource Technology 193 (2015) 513–522

to maintain optimum pH, and low biogas yield. Lack of profit asso- respectively, and ultimately promote the stable operation of AD
ciated with low biogas yield was stated to be a typical reason for process. In this section, key monitoring parameters in the AD pro-
closing down of AD plants (Beddoes et al., 2007). As a result, many cess and techniques for on-line monitoring of those parameters are
industries/clients are hesitant to implement AD technology despite discussed.
its inherent advantages.
Automatic AD process control system enables quick process sta-
bilization with less operation and maintenance inconveniences. 2.1. Monitoring parameters
The ultimate objective of automatic control is to allow AD pro-
cesses stably operate at their maximum capacity. The system con- In the AD process, the degradation of complex organic matter
sists of two constituents, monitoring and control. First, the process occurs through a series of biochemical reactions catalyzed by a
parameters that are sensitive to process disturbances, such as pH, diverse group of microbial communities as indicated earlier. Each
biogas/methane production rate, ammonia, VFAs composition, and microbial community requires different physiological conditions
total VFAs to alkalinity (VFA/ALK) ratio are closely monitored. for their growth, and further these diverse communities work syn-
Next, the monitoring data is sent to control system to regulate/con- trophically to metabolize the organic matter and the degradation
trol these parameters around desired ranges. Diverse process con- products. As a result, maintaining an optimal condition for the
trol systems have been studied and tested in AD systems. The whole microbial community in the AD system is crucial. Besides
control strategy could be as simple as a feedback on/off control, the fundamental operating parameters (e.g., substrate composi-
or as complicated as adaptive, fuzzy logic, neural network control, tion, biogas production and composition, pH, and temperature) of
or their combinations (Pind et al., 2003b; Batstone et al., 2004; the AD process, the parameters indicating process disturbances
Drosg, 2013). Advanced control system could be equipped with (e.g. individual VFAs, total VFAs to alkalinity ratio, ammonia and
basic monitoring techniques; on the other hand, simple control hydrogen) are of greater interest (Björnsson et al., 2000; Drosg,
strategies could be compensated by advanced monitoring equip- 2013). The process disturbances associated with an increasing
ment (Fig. 1). organic loading rate are due to accumulation of intermediates dur-
The goal of this paper is to critically review the existing litera- ing digestion (Batstone et al., 2004). Studies have shown that the
ture on recent advances on emerging process monitoring and con- inhibition effects occur when the key operational parameters are
trol techniques as applied to AD process. Further, this review not controlled in their optimum ranges (Chen et al., 2008;
provides details on the automatic process control techniques as Yenigün and Demirel, 2013). The optimum ranges of these param-
applied to industrial-scale AD plants. Finally, a broader perspective eters are shown in Table 1.
on the further research needs in the domains of automatic AD pro-
cess control is briefly discussed.
2.2. Monitoring techniques

2. Process monitoring Laboratory analysis or off-line monitoring of control parameters


is often associated with potential errors due to improper sampling
Process monitoring is the initial step and crucial component of and analysis techniques, as well as human bias (Drosg, 2013).
any AD automatic control systems. Advances in instrumentation Further, it takes a significant amount of time to receive the output
enable the on-line (real-time) monitoring of critical parameters data, thereby making it difficult to implement corrective measures
in the AD system for early detection of process disturbances. on time in order to prevent reactor disturbance and eventual fail-
Such instruments will send early warning signal to operators or ure. Therefore, developments of on-line monitoring techniques
control algorithm in case of manual and automatic control, were focused recently.

Fig. 1. Monitoring techniques and control strategies in the AD system.


D. Nguyen et al. / Bioresource Technology 193 (2015) 513–522 515

Table 1
Optimum operational range of the AD process. Khanal (2008) and Deublein and Steinhauser (2011)

Parameters Mixed-culture Hydrolysis/acidogenesis Methanogenesis


pH 6.8–7.4 5.2–6.3 6.7–7.5
Temperature (°C) Mesophilic: 35 25–35 Mesophilic: 32–42
Thermophilic: 55 Thermophilic: 50–58
Solids retention time (SRT) (days) High-rate: 15–30
Low-rate: 30–60
Total VFAs (mg/L as acetic acid) 50–250
Acetic acid (mg/L) <1000
Propionic acid (mg/L) <250
Oxidation reduction potential (ORP) (mV) 200 to 350 +400 to 300 < 250
Alkalinity (ALK) (mg/L as CaCO3) 1500–3000
VFA/ALK ratio 0.1–0.2
Carbon to nitrogen (C/N) ratio 10–45 20–30
Carbon to nitrogen to phosphorus (C:N:P) ratio 350:7:1 100:5:1 120:5:1
NH+4-N (mg/L) 50–1,000 61,500
H2 (mg/L) <100

While basic sensors provide robust method for operating the AD followed by ultrafiltration and sample preparation unit. The disad-
systems (e.g., gas/liquid flow meter, temperature, pH, and oxida- vantage of this system is that the filtration units require occasional
tion reduction potential (ORP) sensors), intricate sensors/analytical replacement, and the system is applicable only to the samples with
instruments (e.g., chromatography, spectroscopy) are valuable for total solids (TS) content less than 5%. To overcome these limita-
controlling the AD process (Vanrolleghem and Lee, 2003; tions, Boe et al. (2007) established the headspace gas chromatogra-
Spanjers and Lier, 2006). Among parameters indicating process dis- phy (HS/GC) system for VFA analysis. This ex-situ system consists
turbance, sensors for ammonia, hydrogen and auto-titrator for of VFA stripping unit connected to GC-FID. The headspace
VFA/ALK ratio analysis are already commercially available. This solid-phase micro-extraction gas chromatography-mass spectrom-
led to the focus on development of on-line analysis of individual etry (HS-SPME/GC–MS) is another technique for VFAs analysis
VFAs in recent years. Table 2 presents methods available for mea- (Pickl et al., 2011). The HS-SPME uses specific fibers to capture
suring various parameters in the AD process, which can be used for volatile compounds in headspace of sample vial, and injects coated
AD process control. fibers to GC–MS. The HS-SPME technique offers higher sensitivity
and selectivity (Pickl et al., 2011). However, the complexity of
the system could be the reason for its limited application in AD
2.2.1. Chromatographic devices
process.
In recent years, chromatographic methods have been developed
to monitor critical parameters in the AD process. Chromatographic
technique can be used to identify and quantify compounds of 2.2.2. Spectroscopic devices
interest in both liquid and gas phases. On-line monitoring by gas Spectroscopic devices use light at specific wavelength to quan-
chromatography (GC) is commonly applied to analyze biogas com- tify the concentration of chemicals in a given sample by detecting
position by thermal conductivity detector (GC-TCD), and individual the proportions of light that was absorbed, transmitted and scat-
VFAs by flame ionization detector (GC-FID) (Ward et al., 2011). tered. Unlike chromatographic devices, the spectroscopic device
Individual VFAs (e.g. acetate, propionate, butyrate, valerate and offers non-invasive analytical technique, as the sample is not
capriorate) have been found to be an effective early indication destroyed during analysis. Therefore, on-line monitoring with
for AD upset (Nielsen et al., 2007; Chen et al., 2008). However, recirculation loop could be applied (Holm-Nielsen et al., 2008).
the analysis of VFAs requires an additional step to convert VFAs The major advantage of spectroscopic over chromatographic tech-
from liquid phase to gas phase for GC detection. Therefore, tech- nique is that it could measure a wide range of parameters.
niques for on-line analysis of individual VFAs have been developed Infrared spectroscopy (IRS) has been applied for on-line moni-
recently. For instance, Pind et al. (2003a) developed an on-line GC toring of dissolved CO2, total VFAs, acetate, COD, TOC, alkalinity
system that continuously analyze VFAs with 15 min interval. The in the liquid phase of AD process by coupling ultrafiltration with
upstream of GC apparatus consists of in-situ microfiltration IRS working in the mid infrared range (i.e., 2.5 to 25 lm) (Steyer

Table 2
Measuring techniques of parameters in anaerobic digestion.

Parameters Techniques
Physical sensors Electrochemical sensors Gas chromatography Liquid chromatography Spectroscopy Titration
pH + +
Alkalinity + + + +
Total VFAs + + + +
Individual VFAs + + +
Biogas volume +
Biogas composition + + + +
NH+4–N + + + + +
Chemical oxygen demand (COD) + +
Total organic content (TOC) + + +
H2S + + + + +
H2 + +
516 D. Nguyen et al. / Bioresource Technology 193 (2015) 513–522

et al., 2002a). This system was evaluated for nearly 15 years for range from simple on/off and proportional integral derivative
monitoring of multiple parameters in the pilot-scale (Steyer (PID) control to intricate model-adaptive control, fuzzy logic and
et al., 2002b, 2006) and industrial scale AD systems (Spanjers artificial neural network schemes, and their combinations. In addi-
et al., 2006). Results from the IRS were in agreement with data tion, higher level of control algorithm could compensate for infor-
from the industrial TOC analyzer and titrimetric sensor (Fig. 3). mation from monitoring the AD process. In this section, the
Despite the ability to analyze multiple parameters, the disadvan- contemporary control strategies for AD system are critically
tage of spectroscopy technique is that it requires intense calibra- reviewed. The review focuses on the outcome of the control strate-
tion for the parameters of interest. Further, frequent clogging of gies rather than gaining deeper insight into control algorithm and
filtration membrane in sample pretreatment is another drawback development.
of this technique.
Near infrared spectroscopy (NIRS) using 0.8–2.5 lm wavelength
3.1. Basic control elements and structures
is a state-of-the-art technique to monitor multiple parameters in
AD process. Hansson et al. (2003) monitored lab-scale AD system
There are three basic elements in any loop-control system,
by on-line NIRS for 2 years. The viability of NIRS was tested in
namely sensor, controller and actuator. As discussed earlier, sen-
the digester that was subjected to various disturbances including
sors are implemented to measure the output of the AD process.
overloading, C:N ratio shifting, mixing and foaming problems.
The data from the sensors can be transmitted to the display, stored
The authors also demonstrated that the NIRS can be used to mon-
in disk, and send to controller for subsequent calculation. The con-
itor propionic acids in response to disturbances within short span
troller (e.g., microprocessor controller or computer) receives data
of time (i.e., as short as 2.5 min). Compared to other real-time gas-
from sensors for tracking error/deviation by comparing with
eous monitoring techniques (e.g., GC or membrane-inlet mass
desired set point, and then the tracking error is plugged in the con-
spectrometry), NIRS has been reported to yield better result with
trol algorithm to calculate the variable output. The actuator or final
less maintenance (Ward et al., 2011). NIRS technique has also been
control element receives variable output in the form of electrical
adapted to analyze composition of feedstock (Jacobi et al., 2011).
signal and converts it into physical actions, like open/close valves,
Data from this on-line substrate monitor was used to control feed-
activate/deactivate pumps or regulate flow rate.
ing strategies of the biogas plant. However, authors indicated that
Feedback and feed forward controls are two fundamental struc-
inconsistent size and dryness of samples could result in high offset
tures that constitute the basis for developing advanced control
of the achieved data.
schemes. For example, cascade control is a multiple feedback loop
The state-of-art literature on the diverse instrumentation tech-
control, and ratio or adaptive control uses the theory of feed for-
niques used in the anaerobic technology can be found elsewhere
ward control. The schematic diagram of the two basic
(Vanrolleghem and Lee, 2003; Batstone et al., 2004; Spanjers and
loop-control system is illustrated in Fig. 2; the controller and actu-
Lier, 2006; Madsen et al., 2011).
ator are combined to prevent tangle.
The advantage of feedback control is its simplicity; however,
the correction happens only after the error has occurred.
3. Control systems
Accordingly, feed forward system could be implemented to make
correction before the disturbance affects the process. Yet, accurate
The ultimate goal of controlling the AD process is to meet two
measurement should be established in order to precisely estimate
conflicting objectives, i.e., maximizing the efficiency of AD process
the incoming disturbance. As a result, estimation could be made
while maintaining its stability. In other words, it is desirable to
using adaptive (Rincón et al., 2012) or predictive models (Gaida
operate AD system that can steadily handle high loading rate with-
et al., 2011). These advanced control systems, however, require
out affecting the performance. However, such case requires a
complex algorithm and mathematical experience.
proper control strategy since increasing organic loading rate often
introduces disturbances to the system. Operators prefer operating
digester at low loading rates to avoid the process disturbances, 3.2. Controlled inputs
which often results significant economic loss. For example, a
reduction in 10% biogas yield of a 500-kW electricity equivalent Controlled inputs or manipulated variables in automatic control
biogas plant was reported to cause corresponding loss of nearly of AD process are corrected actions triggered from the control algo-
11% in annual revenue from selling electricity (Wiese and Haeck, rithm. Therefore, these variables should have a quick and signifi-
2006). cant impact on the process. In AD processes, pH, alkalinity, and
The degree of complexity of the control algorithms is directly feeding rate are the typical controlled inputs. pH is one of the most
proportional to the number of required objectives. As a result, a fundamental parameters in bioprocess control. In practice, how-
variety of process control strategies have been developed that ever, it is difficult to regulate the pH inside the reactor at a stable

Fig. 2. Basic control concepts: (A) Feedback; (B) Feedforward.


D. Nguyen et al. / Bioresource Technology 193 (2015) 513–522 517

level due to the dynamics of AD process. Therefore, adjusting pH by control is based on rate of change of the error (time derivative of
dosing concentrated acid or base solution is a challenge to control error). As a result, output from integral and derivative control helps
the AD systems. The misinterpretation of pH data has been the system quickly reach desired set point without large oscilla-
reported as another drawback of using pH as process variables tion. Details of PID algorithm and related issues were presented
(Björnsson et al., 2000; Pind et al., 2003b). The reason for misinter- elsewhere (Åström and Hägglund, 2001; Ang et al., 2005).
preting pH data comes from the fact that pH value is affected by The PID control scheme has been successfully applied in AD sys-
factors like bicarbonate alkalinity, ammonia, or organic acids. tems, especially for non-biological manipulated parameters, such
Similarly, pH value has also been found to be less sensitive to as pH (Heredia-Molinero et al., 2014), temperature control
changes in the system (Boe et al., 2010). Thus, pH has not been rec- (Aguilar et al., 2002), and feeding rate (García-Diéguez et al., 2011).
ommended as a main process control input, instead it is monitored Overall, PID control is a simple, robust and effective control
and interpreted along with other parameters in the AD systems strategy for anaerobic process. However, due to the fact that PID
(Björnsson et al., 2000; Boe et al., 2010) algorithm is highly depended on proportional, integral, and deriva-
Supplementation of alkalinity can be used to maintain the sta- tive gain, identifying optimal values of these constants (i.e., tuning
bility of the system under high organic loading rate by increasing constants) could be a challenge. The details on the automatic tun-
the buffering capacity. Addition of alkalinity has been successfully ing of PID controller are described in other studies (Liu and Daley,
applied to prevent abrupt decrease in pH due to excessive accumu- 2001; Robles et al., 2014). However, the tuning of PID is compli-
lation of VFAs during overloaded condition. This has been con- cated in non-linear systems, where optimal set-points and
trolled by simple on–off system (Guwy et al., 1997), PID control responses are highly dynamic (Liu and Daley, 2001). As a result,
(Marsili-Libelli and Beni, 1996) or advanced artificial neural net- other advanced control schemes must be developed for a better
work (Wilcox et al., 1995). To maintain an adequate buffering control of dynamic processes.
capacity, alkalinity level in AD should be maintained in the range
of 1000 to 3000 mg/L as CaCO3 (Wilcox et al., 1995; Speece, 3.3.2. Adaptive control
1996) or the VFA/ALK ratio of below 0.3 (Drosg, 2013). Although Since anaerobic process is a dynamic and non-linear system, the
fine control schemes for dosing alkalinity were studied, the high optimal condition of the system varies with time. As a result, a
chemical cost is still a problem, especially on industrial-scale bio- fixed-gain control scheme like PID control could be ineffective
gas plants. As a result, there are very limited studies on alkalinity when the system changes its state. In such a situation, adaptive
dosing control in recent years. Some of the alternative methods controllers are better substitutions since the parameters could be
for adjusting the bicarbonate alkalinity in AD systems are: (i) add- automatically adjusted to compensate with variations in the pro-
ing nitrogenous organic matter (e.g., cattle manure) to induce the cess. Based on the objective function, the adaptive control is able
generation of natural alkalinity (Speece, 1996), (ii) recirculating to self-adapt to the changes in the process and yield the optimal
dissolved CO2 in effluent (Turkdogan-Aydinol and Yetilmezsoy, manipulated input that fits the current state of the system. The
2010), or (iii) recycle biogas (Robles et al., 2013). optimal input values are decided based on model, which is nor-
As a result, feeding rate is the commonly used variable input for mally constructed from mass balance of parameters in the system
controlling disturbances in AD system. A wide variety of control (Bernard et al., 2001). The performance of adaptive control is vali-
strategies were studied to manipulate feeding rate, which included dated based on the capability to predict the future response of out-
proportional integral (PI) control (Steyer et al., 1999; put. Details of algorithms and applications of adaptive controllers
Alvarez-Ramirez et al., 2002), PID control (García-Diéguez et al., can be found in Åström and Wittenmark (2013).
2011), adaptive control (Hilgert et al., 2000; Luo et al., 2014), Adaptive controller was developed and effectively applied in
robust adaptive (Petre et al., 2013), fuzzy logic (Murnleitner various AD process to control feeding rate (Rincón et al., 2012),
et al., 2002; Djatkov et al., 2014), neural network (Holubar et al., recycling flow rate (Petre et al., 2013), feed-discharge pattern
2002, 2003), and neural fuzzy (Waewsak et al., 2010). The feeding (Luo et al., 2014) and intermediate alkalinity to total alkalinity
rate can be used to simultaneously regulate the retention time and ratio (Bernard et al., 2001). The advantage of this control scheme
organic loading rate, allowing microbial communities in the sys- is that the system is well-regulated even when large uncertainty
tem to adapt to the disturbances. Unlike previous methods, this input and disturbances are introduced to the process. Further,
control approach avoids the sudden interruption to the system. since the adaptive model is based on mass balance and kinetic of
However, the ability to control the feeding rate could be restricted the AD process, only basic monitoring data is required (e.g., feed-
due to the continuous incoming feedstock that need to be digested. stock composition, biogas volume) (Bernard et al., 2001). The adap-
To overcome this limitation, regulation of recycling flow rate tive control scheme could be an effective control strategy,
within the reactor or to the acidification reactor in a two-staged especially considering that the microbial kinetics exert a key influ-
reactor configuration was applied as variables with successful ence on the functionality of anaerobic process. Besides providing a
result (Murnleitner et al., 2002). better control system in non-linear and dynamic processes, adap-
tive algorithm could also be applied as software sensor (i.e., esti-
3.3. Control strategies mate values of parameters using models instead of
instrumentation). Adaptive model could successfully compute
3.3.1. Proportional integral derivative (PID) control value of COD, VFAs and alkalinity in the pilot- and
The PID control is the widely used feedback control algorithm industrial-scale anaerobic digesters (Bernard et al., 2001).
for automatic process control (Åström et al., 2001). The PID control Therefore, the applications of mathematical models to replace
consists of proportional, integral, and derivative algorithm, which advanced instrumentation in the AD process are promising. The
is based on present, past, and future error, respectively. The pro- drawback of adaptive control could be in the complexity of con-
portional control is given by multiplying the error with a constant structing model that requires advanced algorithms and knowledge
(i.e., proportional gain). However, the proportional control creates of process kinetics (Steyer et al., 2006).
an off-set between the set-point and output, even when the output
reaches steady state (i.e., steady state error). To eliminate this 3.3.3. Fuzzy logic control
off-set, the integral term is added to the algorithm. The integral Besides control strategies based on algorithms and mathemati-
control is given by multiplying the integral of error after time with cal models, the knowledge-based control schemes have also been
a constant (i.e., integral gain). On the other hand, the derivative applied in anaerobic process control. In contrast to previous
518
Table 3
Advanced control system in anaerobic digestion process.

AD systems Control objectives Control strategies Monitoring parameters Controlled inputs Remarks References
Reactor configuration Reactor Substrate
scale
Two-staged reactor Lab Food Identify the biological state of Fuzzy logic – Temperature – Temperature – Successfully in state identification Murnleitner
(fixed bed methane processing reactors and regulate the system – pH – pH – Effectively regulate the system et al. (2002)
reactor) wastewater under shock loading rates – ORP – Feeding rate under shock loading and start-up
– Conductivity – Recirculation state
– Biogas volume and rate

D. Nguyen et al. / Bioresource Technology 193 (2015) 513–522


composition – Flow rate
- Liquid level between two
reactors
Up-flow anaerobic Pilot Molasses Predicting the biogas and Fuzzy logic – pH N/A – Successfully predicted the biogas Turkdogan-
sludge blanket (UASB) wastewater methane production rate – Alkalinity and methane production rate using Aydinol and
– TOC multiple inputs multiple outputs Yetilmezsoy
– Biogas volume and (MIMO) fuzzy logic model (2010)
composition
Hybrid reactor Lab Synthetic Using neural network model to Neural-fuzzy – pH Feeding rate – The system was well regulate during Waewsak
wastewater predict pH, total VFAs, alkalinity – Total VFAs start-up, overloading, and recovery et al. (2010)
as inputs for fuzzy logic to control – Alkalinity phase
feeding rate. – COD
– Biogas volume and
composition
Submerged anaerobic Full Municipal Regulate filtration cycle to PID-fuzzy logic – pH Filtration and backflush – The system used PID control at lower Robles et al.
membrane wastewater minimize membrane fouling rate – Temperature rate level and MIMO fuzzy logic controller (2013)
bioreactors – Pressure, water at upper level in order to regulate the
(SAnMBRs) level, flow rate filtration and biogas recycling rate
– Biogas composition for membrane scouring
– Solids, nutrient, – Membrane fouling rate was success-
sulphide, COD, fully control closed to minimum level
VFAs, alkalinity with less membrane scouring time
D. Nguyen et al. / Bioresource Technology 193 (2015) 513–522 519

controllers, fuzzy logic controllers are a decision-making system (Wilcox et al., 1995). The system was trained over 80 h with vari-
that is based on knowledge of operators rather than precise data ous monitoring data of BA and showed good performance in
from on-line monitoring. Exploiting merits of computational tech- adjusting alkalinity level at different loading rates. Holubar et al.
nology, knowledge from experts could be converted into functions (2002) developed the hierarchical structure of neural network to
to be applied in automatic control. In fuzzy logic control, the mon- model and control methane production in four anaerobic reactors.
itored parameters are divided into subsets, which have values The system has been reported to model pH, VFA, biogas production
ranging from 0 to 1. From the interrelationship between subsets, with high precision. By successfully predicting the gas production
the manipulated input is estimated based on empirical rules. rate and composition in advance, the system was well regulated
Fuzzy logic controllers have effectively been used to control even under shock loading induced by sludge-surplus disturbances.
anaerobic processes by using simple monitoring parameters. The same neural network was then applied to control an AD sys-
These control systems perform well in process with large number tem in start-up and recovery state (Holubar et al., 2003).
of input uncertainties. Details of fuzzy logic controllers applied in
anaerobic process are presented in Table 3.
Besides the application of fuzzy logic in controlling AD systems, 3.3.5. Hybrid control system
it has also been applied in the dynamic modeling of AD process Recently, hybrid control strategies were developed by exploit-
(Polit et al., 2002). In the research, biogas flow rates at various ing advantages of individual control schemes. Heredia-Molinero
pH and temperature conditions were modeled that showed good et al. (2014) successfully associated fuzzy logic with PID control
correlation with real measurement. The fuzzy modeling to create an effective algorithm to maintain pH level. Tay and
method has been found advantageous compared to theoretical Zhang (2000) combined the artificial neural networks with the
mass-balance modeling due to the high flexibility and adaptation fuzzy logic control for predicting system disturbances in advance.
features. Djatkov et al. (2014) used knowledge-based fuzzy logic The system was applied in various anaerobic reactor configurations
law to estimate performance of five industrial-scale agricultural (i.e., anaerobic fluidized bed reactor, anaerobic filter, and UASB)
biogas plants in Germany. The system has been found effective in and could accurately predict VFA, TOC, methane production rate
assisting plant operators to operate and optimize the biogas plants in advance time of one hour. The similar control system was then
with combined heat and power systems. The limitation of fuzzy adapted in automatic control of anaerobic hybrid reactor
logic control is that it solely depends on knowledge level of oper- (Waewsak et al., 2010) (Table 3). By using the artificial neural net-
ators. As a result, the performance of this control system greatly work component to predict the disturbance, and fuzzy logic com-
varies with AD plants. ponent to control the influent flow rate, the reactor was
regulated under organic and hydraulic overloading.

3.3.4. Neural network control


Neural network is technique of controlling process by mimick- 3.3.6. Summary of control strategies
ing the working principle of human brain. Neural networks consist To sum up, there are diverse control strategies available for AD
of nodes (same as neurons on human brain) categorized in series of processes occurring at different level of complexities. Each control
layers that are interconnected with each other in various ways. strategy has its own inherent merit and drawback, which made the
Measurement data from sensory devices is fed as inputs to the implementation greatly depend on distinct control objectives.
nodes in first layer (input layer) and interconnected with other Advantages and disadvantages of each control strategy are listed
nodes in intermediate layers (hidden layers) to generate suitable in Table 4. For application in real scale anaerobic digesters, simple
output signals in nodes of output layer. After setting the multilayer feedback closed loop control like PID control is preferable and
structures, the neural network has to be trained by weighting mostly adopted to control valve and heating system. Advanced
nodes in the system, and adjusting with various training algorithm. monitoring techniques could be used to reinforce these robust con-
Such process is called training period, where the neural network is trol schemes in order to produce a better control result. On the
taught a wide variety of operating conditions of the system. other hand, advanced control strategies were demonstrated to be
Eventually, a well-trained neural network controller could be powerful in modeling and controlling anaerobic systems with sim-
effectively applied in regulating the process. ple monitoring inputs (Fig. 1). It was assumed that high level of
The artificial neural network was applied to monitor and con- experience from expertise is a limiting factor for applying complex
trol the bicarbonate alkalinity (BA) in anaerobic filter system control strategies in real scale biogas plants.

Table 4
Merits and demerits of various control strategies.

Control strategy Merits Demerits


On–off  Simple control  Inadequate for fine control
 Suitable for valve, pump control  No significant impact on important process control
PI/PID  Simple, robust control  Applicable only for linear system
 Fine control  Limited for single input and single output system
 No model required
Adaptive control  Non-linear/dynamic system  Model required
 Capable of on-line estimating unknown parameters  Detailed information of input required
 Automatic tuning  Large uncertainties from model kinetics
 Able to anticipate future disturbance  Complex mathematic calculation required
Fuzzy logic  Multiple input multiple output system  Highly depend on expertise of operator
 Non-linear/dynamic system
 Easy communication with operator
Artificial neural network  No model or expertise required  Required training time
 Suitable when plant behavior is unknown  Large information required
520 D. Nguyen et al. / Bioresource Technology 193 (2015) 513–522

4. Automatic control in industrial-scale AD plants prompts that the centralized biogas plants may be a better solution
to automate the AD process in a given geographical location. The
Different from process control experiments carried out in labo- investment cost for adequately instrumented and automated bio-
ratory or pilot-scale anaerobic reactor, automated control system gas plant with electricity generation capacity larger than 300 kW
in industrial scale plant has to deal with many uncertainties. For has been estimated to be just 5–10% of total capital cost (Wiese
industrial scale AD plants, simple parameters, which include pH, and Haeck, 2006). This investment can be profitable in long term
temperature, mixed liquor level, gas pressure, mixed liquor and operations, since a 10% drop in efficiency of a biogas plant can
biogas flow rate, are monitored on-line (Vanrolleghem and Lee, result in 11% decrease in the annual revenue (Wiese and Haeck,
2003; Spanjers and Lier, 2006; Wiese and Haeck, 2006). Only 10% 2006). Even though the operational cost was not included in the
among 400 industrial scale anaerobic reactors worldwide are report, this seems to be a worthy investment for large-scale indus-
equipped with on-line analysis of COD, TOC, VFAs, alkalinity, trial AD plants in long term.
and biogas composition (Spanjers and Lier, 2006). This situation Steyer et al. (2006) mentioned that despite the availability of
could be explained from the complexity in operation and advanced modeling and control strategies, all the disturbances in
maintenance of these advanced analyzers. Additionally, high capi- industrial-scale AD plants could not be eliminated. Suitable
tal and operation costs of these state-of-the art devices make it on-line diagnosis system is also required to optimize an
economically unattractive for biogas operators to embrace the industrial-scale AD plant (Lardon et al., 2004). Based on several
technology. years of operating a fully instrumented AD reactor (Fig. 3), Steyer
The survey on manure-based biogas plants in Europe also et al. (2002b) summarized the following factors as obstructions
revealed poor management of data monitoring and process control for automation of AD process: pipe clogging/leaking, foaming,
(Wiese and Haeck, 2006). In fact, continuous on-line monitoring water level dropping below sensors, and dosing system malfunc-
was not performed in many plants. For plants with on-line moni- tion. As a result, besides instrumentation and control strategy of
toring systems, real-time control was rare and even the periodical AD processes, an effective diagnosis system plays an equally
data analysis was skipped. For the plants with real-time con- important role in controlling the whole system.
trollers, the control system was simple, time-based, equipped with Development of monitoring and control technologies for AD
on–off controller. The main reason for this poor state of automa- system is a never-ending process. In fact, over the last decade there
tion is that the number of these biogas plants are small-scale have been significant improvements in automatic control and
plants with electricity production capacity under 125 kW. Similar monitoring system. Techno-economic and management of auto-
scenarios were reported in the United States, where most of the matic control and instrumentation in AD plants should be deeply
farm-based AD plants were shut down due to high maintenance investigated to evaluate the pros and cons of investment.
cost and low biogas production (Beddoes et al., 2007). As a result, Recently, the automatic control systems have become robust and
these small-scale AD plants cannot afford the expenses to acquire feasible for anaerobic processes due to key developments in hard-
the instruments and automate the plant operation. This situation ware instrumentation and software control algorithms.

Fig. 3. AD process with instrumentation, control and automation system (Steyer et al., 2002b).
D. Nguyen et al. / Bioresource Technology 193 (2015) 513–522 521

5. Conclusions Hilgert, N., Harmand, J., Steyer, J.-P., Vila, J.P., 2000. Nonparametric identification
and adaptive control of an anaerobic fluidized bed digester. Control Eng. Pract.
8, 367–376.
Two major strategies to promote the stable performance of the Holm-Nielsen, J.B., Lomborg, C.J., Oleskowicz-Popiel, P., Esbensen, K.H., 2008. On-
AD process include monitoring and control over the process. While line near infrared monitoring of glycerol-boosted anaerobic digestion
processes: evaluation of process analytical technologies. Biotechnol. Bioeng.
there is a constant debate on the importance of instrumentation
99, 302–313.
versus control strategy, the experience on both aspects is limited Holubar, P., Zani, L., Hager, M., Fröschl, W., Radak, Z., Braun, R., 2002. Advanced
in industrial scale AD operations. With the rapid development of controlling of anaerobic digestion by means of hierarchical neural networks.
Water Res. 36, 2582–2588.
instrumentation and automatic control, implementation and oper-
Holubar, P., Zani, L., Hager, M., Fröschl, W., Radak, Z., Braun, R., 2003. Start-up and
ation cost of this advanced system is expected to decrease. The recovery of a biogas-reactor using a hierarchical neural network-based control
optimistic outlook for the next decades is that small scale AD tool. J. Chem. Technol. Biot. 78, 847–854.
plants would be equipped with automatic control system for a bet- Jacobi, H.F., Moschner, C.R., Hartung, E., 2011. Use of near infrared spectroscopy in
online-monitoring of feeding substrate quality in anaerobic digestion.
ter performance, meanwhile, centralized AD plant is still a feasible Bioresour. Technol. 102, 4688–4696.
option. Khanal, S., 2008. Anaerobic Biotechnology for Bioenergy Production: Principles and
Applications. John Wiley & Sons Inc., Ames, Iowa.
Lardon, L., Punal, A., Steyer, J.P., 2004. On-line diagnosis and uncertainty
Acknowledgements management using evidence theory––experimental illustration to anaerobic
digestion processes. J. Process Control 14, 747–763.
Liu, G., Daley, S., 2001. Optimal-tuning PID control for industrial systems. Control
This work is supported by the National Institute of Food and Eng. Pract. 9, 1185–1194.
Agriculture, U.S. Department of Agriculture, under Agreement No. Luo, T., Yang, M., Han, J., Sun, P., 2014. A novel model-based adaptive control
2013-67022-21177. Gadhamshetty acknowledges the partial fund- strategy for step-feed SBRs dealing with influent fluctuation. Bioresour.
Technol. 167, 476–483.
ing support from NASA Award No. NNX13AB25A. Lusk, P., Wiselogel, A.E., 1998. Methane recovery from animal manures: the current
opportunities casebook. National Renewable Energy Laboratory. Available from
<http://agrienvarchive.ca/bioenergy/download/methane.pdf> (accessed: May
References 2015).
Madsen, M., Holm-Nielsen, J.B., Esbensen, K.H., 2011. Monitoring of anaerobic
Aguilar, R., Poznyak, A., Martı´nez-Guerra, R., Maya-Yescas, R., 2002. Temperature digestion processes: a review perspective. Renew. Sust. Energy Rev. 15, 3141–
control in catalytic cracking reactors via a robust PID controller. J. Process 3155.
Control 12, 695–705. Marsili-Libelli, S., Beni, S., 1996. Shock load modelling in the anaerobic digestion
Alvarez-Ramirez, J., Meraz, M., Monroy, O., Velasco, A., 2002. Feedback control process. Ecol. Model. 84, 215–232.
design for an anaerobic digestion process. J. Chem. Technol. Biot. 77, 725–734. Metcalf, E., Eddy, H., 2004. Wastewater Engineering: Treatment, Disposal, and
Ang, K.H., Chong, G., Li, Y., 2005. PID control system analysis, design, and Reuse. McGraw-Hill, New York.
technology. IEEE Trans. Control Syst. Technol. 13, 559–576. Murnleitner, E., Becker, T.M., Delgado, A., 2002. State detection and control of
Åström, K.J., Hägglund, T., 2001. The future of PID control. Control Eng. Pract. 9, overloads in the anaerobic wastewater treatment using fuzzy logic. Water Res.
1163–1175. 36, 201–211.
Åström, K.J., Wittenmark, B., 2013. Adaptive control, second ed. Courier Dover Nielsen, H., Uellendahl, H., Ahring, B., 2007. Regulation and optimization of the
Publications, New York. biogas process: propionate as a key parameter. Biomass Bioenergy 31, 820–830.
Åström, K.J., Albertos, P., Quevedo, J., 2001. PID control. Control Eng. Pract. 9, 1159– Petre, E., Selisßteanu, D., S
ß endrescu, D., 2013. Adaptive and robust-adaptive control
1161. strategies for anaerobic wastewater treatment bioprocesses. Chem. Eng. J. 217,
Batstone, D.J., Gernaey, K.V., Steyer, J.P., 2004. Instrumentation and control in 363–378.
anaerobic digestion. In: 2nd IWA Leading-Edge Conference on Water and Pickl, K.E., Adamek, V., Gorges, R., Sinner, F.M., 2011. Headspace-SPME-GC/MS as a
Wastewater Treatment Technologies, p. 173. simple cleanup tool for sensitive 2, 6-diisopropylphenol analysis from lipid
Beddoes, J.C., Bracmort, K.S., Burn, R., Lazarus, W.F., 2007. An analysis of energy emulsions and adaptable to other matrices. J. Pharmaceut. Biomed. 55, 1231–
production costs from anaerobic digestion systems on US livestock production 1236.
facilities. US Department of Agriculture, Washington, DC (2007 Oct.). Pind, P.F., Angelidaki, I., Ahring, B.K., 2003a. A new VFA sensor technique for
Bernard, O., Polit, M., Hadj-Sadok, Z., Pengov, M., Dochain, D., Estaben, M., Labat, P., anaerobic reactor systems. Biotechnol. Bioeng. 82, 54–61.
2001. Advanced monitoring and control of anaerobic wastewater treatment Pind, P.F., Angelidaki, I., Ahring, B.K., Stamatelatou, K., Lyberatos, G., 2003b.
plants: software sensors and controllers for an anaerobic digester. Water Sci. Monitoring and control of anaerobic reactors. Biomethanation, vol. II.
Technol. 43, 175–182. Springer, pp. 135–182.
Björnsson, L., Murto, M., Mattiasson, B., 2000. Evaluation of parameters for Polit, M., Estaben, M., Labat, P., 2002. A fuzzy model for an anaerobic digester,
monitoring an anaerobic co-digestion process. Appl. Microbiol. Biot. 54, 844– comparison with experimental results. Eng. Appl. Artif. Intell. 15, 385–390.
849. Rincón, A., Erazo, C., Angulo, F., 2012. A robust adaptive controller for an anaerobic
Boe, K., Batstone, D.J., Angelidaki, I., 2007. An innovative online VFA monitoring digester with saturated input: guarantees for the boundedness and
system for the anerobic process, based on headspace gas chromatography. convergence properties. J. Process Control 22, 1785–1792.
Biotechnol. Bioeng. 96, 712–721. Robles, A., Ruano, M.V., Ribes, J., Ferrer, J., 2013. Advanced control system for
Boe, K., Batstone, D.J., Steyer, J.-P., Angelidaki, I., 2010. State indicators for optimal filtration in submerged anaerobic MBRs (SAnMBRs). J. Membrane Sci.
monitoring the anaerobic digestion process. Water Res. 44, 5973–5980. 430, 330–341.
Chen, Y., Cheng, J.J., Creamer, K.S., 2008. Inhibition of anaerobic digestion process: a Robles, A., Ruano, M.V., Ribes, J., Seco, A., Ferrer, J., 2014. Model-based automatic
review. Bioresour. Technol. 99, 4044–4064. tuning of a filtration control system for submerged anaerobic membrane
Deublein, D., Steinhauser, A., 2011. Biogas from Waste and Renewable Resources: bioreactors (AnMBR). J. Membrane Sci. 465, 14–26.
An Introduction. John Wiley & Sons, Weinheim. Spanjers, H., Lier, J.B.V., 2006. Instrumentation in anaerobic treatment – research
Djatkov, D., Effenberger, M., Martinov, M., 2014. Method for assessing and and practice. Water Sci. Technol. 53, 63–76.
improving the efficiency of agricultural biogas plants based on fuzzy logic Spanjers, H., Bouvier, J., Steenweg, P., Bisschops, I., Gils, W.V., Versprille, B., 2006.
and expert systems. Appl. Energy 134, 163–175. Implementation of in-line infrared monitor in full-scale anaerobic digestion
Drosg, B., 2013. Process monitoring in biogas plants, IEA Bioenergy. Available from process. Water Sci. Technol. 53, 55–61.
<http://www.rohkraft.net/wp-content/uploads/2014/03/20140319_DROSG_ Speece, R.E., 1996. Anaerobic Biotechnology for Industrial Wastewater. Archae
process_montoring_RZ_web1.pdf> (accessed: May 2015). Press, Nashville, Tennessee.
Gaida, D., Luis, S., Wolf, C., Back, T., 2011. Optimal control of biogas plants using Steyer, J.-P., Buffière, P., Rolland, D., Moletta, R., 1999. Advanced control of anaerobic
nonlinear model predictive control. ISSC, June 23–24, 2011, Trinity College digestion processes through disturbances monitoring. Water Res. 33, 2059–
Dublin. Available from <http://eprints.maynoothuniversity.ie/3648/1/SMcL_ 2068.
optimal_control.pdf> (accessed: May 2015). Steyer, J., Bouvier, J., Conte, T., Gras, P., Harmand, J., Delgenes, J., 2002a. On-line
García-Diéguez, C., Molina, F., Roca, E., 2011. Multi-objective cascade controller for measurements of COD, TOC, VFA, total and partial alkalinity in anaerobic
an anaerobic digester. Process Biochem. 46, 900–909. digestion processes using infra-red spectrometry. Water Sci. Technol. 45, 133–
Guwy, A.J., Hawkes, F.R., Wilcox, S.J., Hawkes, D.L., 1997. Neural network and on-off 138.
control of bicarbonate alkalinity in a fluidised-bed anaerobic digester. Water Steyer, J., Bouvier, J., Conte, T., Gras, P., Sousbie, P., 2002b. Evaluation of a four year
Res. 31, 2019–2025. experience with a fully instrumented anaerobic digestion process. Water Sci.
Hansson, M., Nordberg, A., Mathisen, B., 2003. On-line NIR monitoring during Technol. 45, 495–502.
anaerobic treatment of municipal solid waste. Water Sci. Technol. 48, 9–13. Steyer, J., Bernard, O., Batstone, D.J., Angelidaki, I., 2006. Lessons learnt from 15
Heredia-Molinero, M.C., Sánchez-Prieto, J., Briongos, J.V., Palancar, M.C., 2014. years of ICA in anaerobic digesters. Water Sci. Technol. 53, 25–33.
Feedback PID-like fuzzy controller for pH regulatory control near the Tay, J.-H., Zhang, X., 2000. A fast predicting neural fuzzy model for high-rate
equivalence point. J. Process Control 24, 1023–1037. anaerobic wastewater treatment systems. Water Res. 34, 2849–2860.
522 D. Nguyen et al. / Bioresource Technology 193 (2015) 513–522

Turkdogan-Aydinol, F.I., Yetilmezsoy, K., 2010. A fuzzy-logic-based model to predict chromatography, near-infrared spectroscopy, and membrane-inlet mass
biogas and methane production rates in a pilot-scale mesophilic UASB reactor spectrometry. Bioresour. Technol. 102, 4098–4103.
treating molasses wastewater. J. Hazard. Mater. 182, 460–471. Wiese, J., Haeck, M., 2006. Instrumentation, control and automation for full-scale
Vanrolleghem, P., Lee, D.S., 2003. On-line monitoring equipment for wastewater manure-based biogas systems. Water Sci. Technol. 54, 1–8.
treatment processes: state of the art. Water Sci. Technol. 47, 1–34. Wilcox, S.J., Hawkes, D.L., Hawkes, F.R., Guwy, A.J., 1995. A neural network, based on
Waewsak, C., Nopharatana, A., Chaiprasert, P., 2010. Neural-fuzzy control system bicarbonate monitoring, to control anaerobic digestion. Water Res. 29, 1465–
application for monitoring process response and control of anaerobic hybrid 1470.
reactor in wastewater treatment and biogas production. J. Environ. Sci. 22, Yenigün, O., Demirel, B., 2013. Ammonia inhibition in anaerobic digestion: a review.
1883–1890. Process Biochem. 48, 901–911.
Ward, A.J., Bruni, E., Lykkegaard, M.K., Feilberg, A., Adamsen, A.P.S., Jensen, A.P.,
Poulsen, A.K., 2011. Real time monitoring of a biogas digester with gas

You might also like