ARTICLE pubs.acs.org/est
Novel Online Monitoring and Alert System for Anaerobic Digestion Reactors Fang Dong,† Quan-Bao Zhao,† Wen-Wei Li,*,† Guo-Ping Sheng,† Jin-Bao Zhao,† Yong Tang,† Han-Qing Yu,*,† Kengo Kubota,‡ Yu-You Li,‡ and Hideki Harada‡ † ‡
Department of Chemistry, University of Science and Technology of China, Hefei, 230026 China Department of Civil Engineering, Tohoku University, Sendai 980-8579, Japazn
bS Supporting Information ABSTRACT: Effective monitoring and diagnosis of anaerobic digestion processes is a great challenge for anaerobic digestion reactors, which limits their stable operation. In this work, an online monitoring and alert system for upflow anaerobic sludge blanket (UASB) reactors is developed on the basis of a set of novel evaluating indexes. The two indexes, i.e., stability index S and auxiliary index a, which incorporate both gas- and liquid-phase parameters for UASB, enable a quantitative and comprehensive evaluation of reactor status. A series of shock tests is conducted to evaluate the response of the monitoring and alert system to organic overloading, hydraulic, temperature, and toxicant shocks. The results show that this system enables an accurate and rapid monitoring and diagnosis of the reactor status, and offers reliable early warnings on the potential risks. As the core of this system, the evaluating indexes are demonstrated to be of high accuracy and sensitivity in process evaluation and good adaptability to the artificial intelligence and automated control apparatus. This online monitoring and alert system presents a valuable effort to promote the automated monitoring and control of anaerobic digestion process, and holds a high promise for application.
’ INTRODUCTION Anaerobic digestion (AD) processes are sensitive to environmental fluctuations and frequently suffer from operation instability. Thus, real-time monitoring and robust control of AD processes are essential to avoid possible instability due to disturbance. For this purpose, intensive studies have focused on fast, reliable, and online monitoring of AD processes as well as the development of efficient feedback alert and control strategies.1 A number of indexes have been frequently adopted in the process monitoring, including volatile fatty acids (VFAs), pH, redox potential, biogas production rate, and composition. Compared with the easy-to-measure parameters like pH and redox potential, VFA and biogas production are widely considered as the two most crucial and direct indicators of the system status.2,3 In a normally operating AD system, VFAs produced by acidogens and acetogens can be immediately utilized by the methanogens for methane production, leaving a low VFA concentration.4 Thus, the VFA accumulation is usually interpreted as the result of methanogenesis inhibition or organic overloading, and implies a risk of process upset.5 However, VFA alone is insufficient to reveal the overall reactor status. In this context, the biogas composition and production rate can also provide useful information.6,7 Therefore, a combined detection of both liquidand gas-phase parameters is considered as an efficient strategy to provide comprehensive insight into an AD process.810 r 2011 American Chemical Society
A variety of techniques for online monitoring of VFAs are available, such as fluorescence spectroscopy,11,12 near-infrared (NIR) spectroscopy,13 titration,14 and gas chromatography.15,16 Among them, titration is the most practical option, attributed to its high simplicity, rapidness, and cost-effectiveness.14,1719 On the other hand, NIR spectroscopy technique is gaining a great popularity for online monitoring of gas-phase parameters.20 Thus, incorporating titration and NIR spectroscopy techniques might offer a relatively simple way to achieve a comprehensive and accurate monitoring of AD process. In addition to parameter detection, effective analysis of the obtained data is of equal importance for reactor status evaluation and diagnosis. Because of the high complexity of AD systems and the unclear interrelations of many involved parameters, it is difficult to extract useful “hidden” information from the large data sets of individual parameters and to find out the true situation as well as potential risks. For AD process evaluation or diagnosis, a common practice is to set threshold values for some individual indicators like pH and VFA, and to judge the reactor status on the basis of the detected values.8,2123 However, such an evaluation is usually inaccurate, because information Received: June 30, 2011 Accepted: September 14, 2011 Revised: September 11, 2011 Published: September 14, 2011 9093
dx.doi.org/10.1021/es202245f | Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology
ARTICLE
Table 1. Operation Conditions for Shock Tests test
duration
6-fold organic overload shock
24 h
operation condition influent COD of 60 000 mg/L influent NaHCO3 of 3000 mg/L HRT of 24 h
5 °C temperature shock
96 h
influent COD of 10 000 mg/L influent NaHCO3 of 3000 mg/L HRT of 24 h temperature of 5 °C
6-fold hydraulic shock
24 h
influent COD of 1667 mg/L influent NaHCO3 of 500 mg/L HRT of 4 h
toxicant shock
slug dose influent COD of 10 000 mg/L influent NaHCO3 of 3000 mg/L HRT of 24 h chloroform dosage of 80 mg/L
Figure 1. Schematic of online monitoring and alert system for a UASB reactor (dotted lines: data; solid lines: liquid or gas pipes).
about the relationship between the individual parameters and the reactor status is still lacking and thus such an evaluation relies heavily on the professional knowledge and experiences of operators. Moreover, these parameters can only reveal the current reactor status, but it is actually, in most cases, too late for an effective process control once the threshold values are reached. Indeed, no effective method for AD reactor diagnosis and risk prediction is currently available. As such, it is imperative to propose and apply objective indexes for AD reactor diagnosis. Such indexes should be ideally accurate and sensitive to environmental fluctuations, reveal the change dynamics of reactor status, and be adaptive to online monitoring, autoalert, and control systems. In this work, an effective online monitoring and alert system is developed for an upflow anaerobic sludge blanket (UASB) reactor. Particularly, two new evaluating indexes, stability index S and auxiliary index a, are proposed for a quantitative assessment of reactor status, which for the first time take into account the changing dynamics of both liquid- and gas-phase parameters of UASB reactors. On the basis of a comprehensive investigation into the relationship between the indexes and reactor performance, a database of S and a under various disturbances is established to offer a foundation for online diagnosis and early warning of AD process.
’ MATERIALS AND METHODS UASB Reactor and Online Monitoring and Alert System. In this work a bench-scale Plexiglas-made UASB reactor was used, which consisted of a reaction zone of 2.0 L and a gassolids separator zone. The online monitoring system consisted of a gasphase detection unit and a liquid-phase detection unit that connected to a computer through a data acquisition card (PCI1602, ICP DAS Co., China), as illustrated in Figure 1. The liquidphase detection unit was comprised of an automatic sample loading and ejecting device and an online titration device, controlled by computer through 485 serial ports. A precise meter pump (BT100-1F, Longer Co., China) was used as the sample loading and ejecting device. The online titration device consisted of an inject pump (TJ-3A, Longer Co., China), a pH glass
electrode (LE409, Mettler Toledo Co., USA), and a titration cell. A magnetic stirrer was used for the liquid mixing in the titration cell. The gas-phase detection unit consisted of methane and carbon dioxide sensors (AGM10 and AGM32, Onward Co., China) and a gas flow measurement device for monitoring of biogas production rates. The software of alert system was designed on a LabView virtual instrument platform (National Instruments Co.,USA), which incorporates multifunctions of data acquisition, analysis, recording, and display. Specifically, the indexes S and a profiles were calculated, recorded, and displayed real-time based on the LabView software. Experimental Operations. The seed sludge for the UASB reactor was taken from an anaerobic digester in a local citrateprocessing wastewater treatment plant. The pH and volatile suspended solids (VSS) of seed sludge were 7.2 and 42.2 g/L, respectively. The reactor temperature was maintained at 35 ( 1 °C using a heating jacket, except during the temperature shock testing period. The reactor was operated at a fixed loading rate of 10 g/L chemical oxygen demand (COD) and 24-h hydraulic retention time (HRT) except as otherwise specified. Sucrose-rich synthetic wastewater was used as the feedstock, which contained 3000 mg/L NaHCO3 as the buffer and sufficient amount of inorganic nutrients as follows (in mg/L): NH 4 HCO3 , 405; K2HPO4 3 3H2O, 155; CaCl2, 50; MgCl2 3 6H2O, 100; FeCl2, 25; NaCl, 10; CoCl2 3 6H2O, 5; MnCl2 3 4H2O, 5; AlCl3, 2.5; (NH4)6Mo7O24, 15; H3BO3, 5; NiCl2 3 6H2O, 5; CuCl2 3 5H2O, 5; and ZnCl2, 5. After the reactor reached a steady state in terms of COD removal and methane production, four consecutive shock tests were conducted to investigate the response of the UASB reactor to various operating shocks, including organic overloading, hydraulic loading, temperature, and toxicant shocks (Table 1). The organic overloading shock was imposed by switching the feed COD concentration to six fold while keeping the flow rate unchanged. The hydraulic loading shock was provided by elevating the flow rate while accordingly reducing the feed COD and bicarbonate concentrations to maintain a constant organic loading rate (OLR). The temperature shock was achieved by directly shifting the reactor temperature from 37 to 5 °C. For the toxicant shock, toxicants that need a relatively high concentration to induce significant toxic effect, e.g., formaldehyde, 9094
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology
ARTICLE
Figure 2. Index variations under 6-fold organic overloading shock: (A) S and a; and (B) total VFA concentration and MPR.
and surfactant, which shows chronic toxicity,24 were not chosen here in order to avoid OLR interference.25 In this study, chloroform was selected as the toxicant, because of its significant toxic effect on anaerobic microorganisms even at a very low level.26 The toxicant shock was imposed through slug dose of chloroform at 80 mg/L. An interval of over two months was adopted between every two shock tests to ensure a stable operation before another shocking test. The variations of the gas- and liquid-phase parameters during the entire test period were measured online and the indexes S and a were calculated. Monitoring and Analysis. The VFA concentration and pH were measured online using the liquid-phase detection unit with five-point titration technique.19 The volume and composition of methane and carbon dioxide were detected online by the gasphase detection unit. All detections were conducted once per hour except during the temperature shock test, in which a 4-h measurement frequency was adopted. The pH electrode and titration cell were cleaned or washed regularly to prevent possible fouling. The monitoring data were collected by the data acquisition card and real-time recorded by computer. VSS and COD concentrations were measured according to the Standard Methods.27 Evaluating Indexes. Two evaluating indexes are proposed here for reactor status evaluation. The stability index S is used to describe the VFA accumulation in the liquid phase, while the auxiliary index a is used to evaluate the variation trend of
methane production in the gas phase, as follows: S¼
1 dVFA 100 QCH4 dt
ð1Þ
a¼
dQCH4 dt
ð2Þ
where (dVFA)/(dt) (mmol/L/h) is the change rate of total VFA concentration at time t and QCH4 (mmolCH4/L/h) is methane production rate (MPR). The signs of indexes also serve as useful signals for reactor state evaluation. To benefit an automatic process diagnosis, a diagnosis software based on the statistic analysis and logic judgment programs is designed for data analysis. In this study, if the majority (i.e., 70%) of the index data show the same sign in a time span of at least 5 h, this sign would be automatically assigned to the index. Specifically, if a change of sign occurs, the system will not give an immediate judgment. Instead, it would evaluate whether such a sign change is caused by a real change of reactor status or just a temporary fluctuation according to the statistics of this point and the subsequent points in at least 5 h. On the basis of the signs and values of both indexes, the current state and possible risks of reactor can be determined. In this study, the dangerous state is defined as sustaining VFAs accumulation and corresponding inhibition in methane production. Under this state, the majority of S would be positive and a would be negative according to the eqs 1 and 2, and a higher value 9095
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology
ARTICLE
Table 2. Operation Conditions and Shock Types of Tests in Literature shock type
duration
shock condition
references 25
4-fold organic overload shock
hours 03
influent COD increased from 5000 to 20 000 mg/L
4-fold hydraulic shock
hours 03
HRT decreased from 12 to 3 h
25
toxicant shock
slug addition
10 mg chloroform/L was dosed
25
temperature shock
days 070
temperature increased from 37 to 55 °C
30
temperature shock
days 3038
temperature increased from 55 to 65 °C
31
of S implies a higher degree of risk. To better reflect the degree of danger and stability, several threshold values are set here. Under the premise of the positive S coupled with negative a, the alert system will give alarm if the absolute value of S exceeds 50.0. Notably, this threshold for alarm is only a type of emergency alarm that indicates a serious situation. Actually, even at the beginning of shock, the diagnosis system has already detected the occurrence of shock according to signs of S and a. In a special case, if both indexes are close to zero, i.e., |S| is less than 4.0, and |a| is less than 1.5, this refers to a typical steady state. Or, if both indexes have frequently changed signs (positive and negative signs occur alternately and both occurrences are below 70%), this can be regarded as an unsteady state. In this way, the system status can be diagnosed and a warning signal will be given by the alert system in advance if a dangerous state or trend is identified.
’ RESULTS Reactor Performance under Steady-State Conditions. The gas production rate and compositions as well as the effluent pH, COD, and VFAs during the reactor steady-state operation period were monitored to evaluate the reactor performance. The average gas production rate was in a range of 407.7431.5 mL/h. The methane and carbon dioxide contents were 65.568.0% and 20.127.0%, respectively. The effluent pH was 7.147.21. The effluent COD and VFA concentrations were 41.596.0 and 39.095.5 mg/L, respectively. Organic Overloading Shock. The typical response of the AD process to an organic overloading shock is as follows:28 increase in VFA concentration, sludge volume index, and MPR; decrease in COD removal efficiency and methane content; change of pH depending on the buffer capacity of the feed. According to the variations of operating conditions and S and a values, the organic overloading shock test could be divided into 4 phases (Figure 2). When the organic overloading shock was applied in phase 1 (P1, hours 014), initially no obvious inhibition was observed at this stage. The S value was positive because of a rising VFA concentration. Meanwhile, a also showed positive values as the MPR kept increasing. In phase 2 (P2, hours 1524), the shock condition continued and an obvious inhibition occurred. S remained positive values, but a turned negative as the MPR started to decline. The shock was stopped since phase 3 (P3, hours 2537), and the reactor rapidly restored, with S and a values changed to negative and positive, respectively. In phase 4 (P4, after hour 38), the reactor became restabilized, and both S and a values were close to zero. A lasting warning signal was given by the alert system throughout P2 according to the identification result of diagnosis software, because most S showed positive values while a was mostly negative. The absolute value of S exceeded 1000.0 at hour 24, indicating a high risk of reactor collapse due to severe VFA accumulation. From hour 25, the organic overloading shock was terminated, and the S and a in P3 changed to negative and
positive values, respectively. Such changes suggest that the shock strength was weakened or stopped, resulting in a reduced VFA concentration and recovery of the methane production capacity. Accordingly, the alert stopped with the recovery of reactor status. In P4, both the S and a values approached zero. The MPR and VFA concentration became stabilized, and the reactor was tending toward stable operation. Hydraulic Loading Shock. The variations of S and a during the 6-fold hydraulic loading shock are shown in Figure S1 (Supporting Information). Most of the S showed negative values, and the absolute values were all below 3.0 in P1 (hours 124). Alternative positive and negative values of a were observed almost in half-and-half, with all the absolute values below 1.4. These results suggest that the MPR remained stable in P1. After hour 24 (P2), the hydraulic shock stopped. The value of S mostly became positive and its absolute values were below 2.4 during hours 2536. The total VFA concentration increased slightly. The a values also showed positive and negative fluctuation, and the absolute values fell to below 0.8. The MPR remained stable in this stage. After hour 37, both the absolute values of S and a were close to zero, and the reactor restored stable performance. No significant change in both indexes was observed during the hydraulic loading shock, indicating that the reactor performance was less affected by such a hydraulic loading shock. This was consistent with previous studies, in which the COD removal varied only slightly under the condition of decreased HRT while unchanged OLR.29 Temperature Shock. The changes of S and a during the temperature shock are shown in Figure S2. During the temperature shock (P1, hours 196), most of S exhibited positive values, indicating an increase in VFA concentration. The absolute values of S mounted from 3.0 up to a maximum of 60.0 at hour 28, suggesting an apparently increasing trend of VFA accumulation. Meanwhile, negative values were found for a in P1, indicating a considerable suppression of methane production. Accordingly, warning signals were given by the alert system throughout P1. Noticeably, the signs of a here were apparently different from those in the hydraulic and organic overloading shocks, which mostly showed positive values. With temperature shock, the methane production was almost completely inhibited during hours 6096 and the absolute values of a were below 0.1. With the ending of temperature shock after hour 96 (P2), the S value changed drastically from positive to negative. A rapid drop in the total VFA concentration was observed. At the same time, the a value changed from negative to positive and a gradual restoration of methane production was observed. As a consequence, the alert signal automatically terminated. Although both S and a values fluctuated between positive and negative values within P3 (hours 188264), they became stable and approached zero in P4 (after hour 265), indicating that the UASB reactor restored stable performance. Toxicant Shock. The response of S and a to the toxicant shock is illustrated in Figure S3. Both S and a exhibited violent 9096
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology
ARTICLE
exhibited negative values during this period. This indicates a rapid VFA accumulation and a severe inhibition in methane production. A warning signal was given by the system in P1. After hour 5 (P2), a remained negative and the biogas production dropped to near zero, while S value changed from positive to negative in P2. Because the VFA could not be consumed by the methanogens, as evidenced by the completely ceased gas production, this means that the VFA production process was also inhibited. The slight decreased VFA concentration in the reactor might be attributed to the effluent discharge. The sharp increase in the absolute value of S and the ceased methane production at hour 12 indicate a severe system collapse. After hour 12, a was near zero and no positive value of a was observed, analogue to that under other shock-ceasing conditions. This further confirms the full stop of methane production and collapse of the UASB system. Verification. To further evaluate the practicability of the two indexes S and a, the experimental data of shock tests reported in previous studies were used for verification. The operating conditions and shock types in these tests are summarized in Table 2, whereas the calculated results of S and a are shown in Figure 3. As shown in Figure 3C, the S and a calculated from the literature data show changing profiles highly similar to those in this study, except for the toxicant shock test. For the toxicant shock test by Tay and Zhang,25 the calculated values of a increased for a short period before dropping to zero again, indicating the recovery of methane production and stable performance after the ending of shock. In contrast, in our study an irreversible collapse of reactor occurred due to the much higher dosage of chloroform, which had caused unrecoverable toxic damage. Generally, these two evaluating indexes demonstrate good applicability to various AD systems.
’ DISCUSSION
Figure 3. Profiles of S and a based on data in literature: (A) 4-fold organic overloading shock;25 (B) 4-fold hydraulic loading shock;25 (C) toxicant shock;25 (D) 55 °C temperature shock;30 and (E) 65 °C temperature shock.31
responses to the toxicant shock in the initial 4 h (P1). The S index mostly showed positive values above 50.0, while a mostly
Merits of the Monitoring and Alert System. The constructed monitoring and alert system shows several distinct advantages over the conventional systems. First, comprehensive but reasonably simple online monitoring of AD process is realized, which offers the basis for a rapid and accurate evaluation of AD reactor status. Second, the alert system demonstrates high capacity in process diagnosis and early warning, attributed to the introduction of the two new indexes, S and a. Third, the system, which is characterized by automated execution, modularized design, and easy use, holds great promise for future practical application. All monitoring processes are automatically executed with software program. Simple and commercially available components are adopted for the hardware of this system to facilitate modularization. Most of all, no complex operation and maintenance are required for this system, and customized services for reactor monitoring and online data processing can be achieved by using the virtual instrument software program. The two indexes, S and a, are proposed for diagnosis of AD reactor status. They can reflect reactor status and changing trend accurately, and have a rapid response to the operating condition disturbance. On the basis of the diagnosis results, an early alert can be given by the system to prevent the upcoming risks. Accuracy of Indexes. The S and a indexes, which incorporate the key operating parameters in both liquid- and gas-phase, offer meaningful information on anaerobic reactor status. Specifically, a combined evaluation of values and signs of both indexes reveals the hidden pattern of dynamic change in VFA accumulation and 9097
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology
ARTICLE
Table 3. Correlation between the Sign of Indexes and Reactor Status S
a
positive
positive
organic overload; methanogens are
positive
negative
methanogens are
P2 of Figure 2A, P1 of Figure S2(A),
evidently suppressed
P1 of Figure S3(A), P2 of Figure 3A, P1 of Figure 3C,
reactor status
example P1 of Figure 2A, P1 of Figure 3A
not evidently suppressed
P1 of Figure 3D, P1 of Figure 3E negative
positive
activity of methanogens has restored.
P3 of Figure 2A, P2 of Figure S2(A), P2 of Figure 3C, P2 of Figure 3E
negative
negative
acidogenesis and methanogenesis are both inhibited
P2 of Figure S3(A), P3 of Figure 3A
frequent
frequent
reactor state is unstable; at the
P3 of Figure S2(A), P2 of Figure 3D
change of sign
change of sign
transitional stage
close
close
reactor is operated at
to zero
to zero
steady state or all the biochemical
P4 of Figure 3A, P1P4 of Figure 3B, P3 of Figure 3C,
processes in the reactor cease
P3 of Figure 3D, P3 of Figure 3E
methane production. Based on the experimental results of the shock tests and literature data verification, a database about the correlation between the two indexes and anaerobic reactor status was established (Table 3). Such a combination of different signs and magnitudes of the two indexes corresponds well to different reactor status and shock types. Here, the absolute values of S and a reflect the intensity of changes or disturbance in anaerobic reactors, whereas their signs indicate a specific reactor status and shock type. As shown in Table 3, positive S is found under many different operating conditions and reactor status, implying that the value of S alone is insufficient to clearly identify reactor status. Noticeably, the signs of a also vary significantly and offer valuable supplementary information. On the basis of a combined analysis of S and a at the initial stage of shock, it is possible to preliminarily distinguish the shock type and find out whether anaerobes are significantly inhibited. Generally, a positive S and positive a correspond to an organic overloading shock, while a positive S and negative a indicate a temperature or toxicant shock. At the late stage of an organic overloading shock (Figures 2A and 3A), a positive S and negative a are obtained regardless of the shock intensity, which suggests a methanogenesis-inhibited state. In contrast, a negative S and positive a at the late stage of shock means consumption of VFA by methanogens and a gradual adaptation and reactor recovery after shock. In addition, the changing trends of S and a are also informative. If S value changes from positive to negative, while a value remains negative for a long time, the inhibition of both VFA and methane production is suggested. Apparently, there exist some relationships between the indexes of S and a and the reactor status. Compared with the conventional individual parameters, such as pH, VFA, and MPR, which rely on their preset threshold values for judgment and usually lead to subjective and incorrect conclusions, these two integrated evaluating indexes proposed in this study offer more objective, accurate, and comprehensive assessment on the status and shock types of anaerobic reactors. Sensitivity of Indexes. Both S and a demonstrate high sensitivity, mainly attributed to an integrated analysis of both the absolute value and the sign as well as the use of MPR as a regulatory factor. With this regulatory factor, the changes of S can be properly and significantly magnified or buffered, so that the reactor changing
P4 of Figure 2A, P1, P2 of Figure S1(A), P4 of Figure S2(A),
trend can be more clearly revealed. This ensures a high sensitivity of the indexes and a capability of the system for early warning. For example, in the organic overloading shock test (Figure 2), the positive S and a in P1 corresponded to the VFA accumulation and MPR increase. Such an MPR increase was attributed to the uninhibited methanogenesis at the beginning. As a consequence, the increase in S was weakened by the methane production, which serves as a buffer against the factor dVFA/dt. Thus, S showed no drastic increase. The moderate change of S was consistent with the reactor status. In P2, however, the MPR severed as amplifier of S, because methanogenic inhibition occurred. At this phase, the S remained positive, while a shifted to negative. The increase in dVFA/dt was magnified by the MPR, leading to a maximum S of over 1000.0 at the end of shock. This was consistent with the significantly deteriorated reactor performance. Accordingly, alert was given by the system throughout P2. It is worthwhile to note that, however, the shock had already been detected by the system even at the beginning of shock. For both S and a, their signs are dependent on the transient changing trend of the monitoring parameters. The sign will change and respond immediately once the reactor becomes instable or undergoes changing state, even when the conventional evaluating parameters are still below their threshold values. For example, in the organic overloading shock test, the VFA concentration, methane production, and pH all showed little changes at the initial shock stage (within 1 h): the VFA concentration increased slightly from 2.0 to 2.2 mmol/L, the maximum MPR increased from 6.4 to 7.5 mmol/L/h, and pH and methane component kept at around 7.3 and 66.0%, respectively. However, both S and a values changed substantially. S and a jumped rapidly from near zero to 5.6 and 3.2, respectively, with a positive sign (the sign can change instantaneously as a result of shock). These changes suggest a changing reactor status and offer hints to the type and level of the encountered shocks. The positive sign of S and a clearly suggested an organic overloading shock. As a consequence of diagnosis results, alert was given by the system. The time of earliest alert given was at hour 15, which was 5 h earlier than the occurrence of severe acidification (pH < 6.2) and 8 h before the full stop of methane production. In contrast, at hour 15 the VFA concentration was 20.1 mmol/L; MPR was 36.9 mmol/L/h, which was even much higher than that at hour 0 (6.4 mmol/L/h); pH value was 6.93 and acidification 9098
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology did not occur. Apparently, these conventional parameters revealed no sign of risk at that moment. Therefore, indexes S and a are not only more effective and accurate in assessing the reactor status and operation condition changes, but also more effective in risk prediction, because risks can be detected far in advance. Similar high prediction accuracy of risks was repeatedly demonstrated in the other shock tests. Therefore, compared to the conventional parameters that can reveal the reactor status only after occurrence of deterioration, the two indexes proposed here can give synchronous and real-time information on the reactor status and rapidly reflect any slight change early before the dangers occur. Adaptability of Indexes to Diagnosis Program. With S and a, a simple and reliable diagnosis program can be developed for a fast and automatic evaluation of anaerobic reactors, without the need for human intuition or intervention. Moreover, these two indexes are suitable for a logic working pattern of artificial intelligence programming, because the positive and negative signs resemble 0 and 1 of the computer binary system. Therefore, an integration of the two indexes with the online monitoring and alert system can provide a practical tool for online artificial intelligence diagnosis of anaerobic reactor status. By further employing an additional feedback automatic control system, intelligent operating and automatic control of anaerobic reactors could be realized. In summary, an efficient online monitoring and alert system is developed for monitoring and diagnosing the status of a UASB reactor. The core of this system is a set of integrated indexes, a stability index S and an auxiliary index a, which offer the basis for a quantitative and comprehensive evaluation of anaerobic digestion reactors. The great accuracy and sensitivity of the two indexes enable a reliable early warning of potential risks. Their good adaptability and relatively simplicity also ensure a high possibility for being incorporated into artificial intelligence and automated control systems. Therefore, this online monitoring and alert system holds great promise for practical application in online monitoring, diagnosis, and control of anaerobic digestion reactors for waste treatment. Moreover, with a further development of advanced monitoring and data interpretation techniques as well as a deeper understanding of microbial metabolic dynamics, it may also be expanded to other fields of anaerobic fermentation processes. Nevertheless, despite the good performance of our system, it may face challenges for practical application in large-scale reactors, such as liquid residual in pipeline, pipe jam, electrode fouling, and fluid mixing. In addition, the setting of threshold values for S and a indexes is important and may vary significantly with reactor type and diagnosis objective, because there is inevitably a balance between prediction accuracy and degree of foreseeability. Thus, selection of appropriate threshold setting values is necessary. These all warrant further investigations.
’ ASSOCIATED CONTENT
bS
Supporting Information. Figures S1S3. This information is available free of charge via the Internet at http://pubs.acs. org.
’ AUTHOR INFORMATION Corresponding Author
*W.-W.L. Fax: +86 551-3607453; e-mail:
[email protected]. H.-Q.Y. Fax: +86 551-3607592; e-mail:
[email protected].
ARTICLE
’ ACKNOWLEDGMENT We thank the NSFC-JST Joint Project (21021140001) and Ministry of Science and Technology of China (2008BADC4B18) for the partial support of this study. ’ NOMENCLATURE a auxiliary index COD chemical oxygen demand HRT hydraulic retention time MPR methane production rate OLR organic loading rate S stability index VFA volatile fatty acid VSS volatile suspended solids UASB upflow anaerobic sludge blanket ’ REFERENCES (1) Madsen, M.; Nielsen, J. B. H.; Esbensen, K. H. Monitoring of anaerobic digestion processes: A review perspective. Renewable Sustainable Energy. Rev. 2011, 15, 3141–3155. (2) Bjorklund, R. B.; Christiansson, A.; Anders, E. W.; Ejlertsson, J. Electrode specific information from voltammetric monitoring of biogas production. Talanta 2010, 81, 1578–1584. (3) Ward, A. J.; Hobbs, P. J.; Holliman; Jones, D. L. Optimisation of the anaerobic digestion of agricultural resources. Bioresour. Technol. 2008, 99, 7928–7940. (4) Lahav, O.; Morgan, B. E. Titration methodologies for monitoring of anaerobic digestion in developing countries - A review. J Chem. Technol. Biotechnol. 2004, 79, 1331–1341. (5) Buyukkamaci, N.; Filibeli, A. Volatile fatty acid formation in an anaerobic hybrid reactor. Proc. Biochem. 2004, 39, 1491–1494. (6) Mathiot, S.; Esoffier, Y.; Ehlinger, F.; Couderc, J. P.; Leyris, J. P.; Moletta, R. Control parameter variations in an anaerobic fluidized-bed reactor subjected to organic shockloads. Water Sci. Technol. 1992, 25, 93–101. (7) Steyer, J.-P.; Buffiere, P.; Rolland, D.; Moletta, R. Advanced control of anaerobic digestion processes through disturbances monitoring. Water Res. 1999, 33, 2059–2068. (8) Boe, K.; Batstone, D. J.; Steyer, J. P.; Angelidaki, I. State indicators for monitoring the anaerobic digestion process. Water Res. 2010, 44, 5973–5980. (9) García-Dieguez, C.; Molina, F.; Roca, E. Multi-objective cascade controller for an anaerobic digester. Proc. Biochem. 2011, 46, 900–909. (10) Nielsen, H. B.; Uellendahl, H.; Ahring, B. K. Regulation and optimization of the biogas process: Propionate as a key parameter. Biomass Bioenergy 2007, 31, 820–830. (11) Li, W. H.; Sheng, G. P.; Lu, R.; Yu, H. Q.; Li, Y. Y.; Harada, H. Fluorescence spectral characteristics of the supernatants from an anaerobic hydrogen-producing bioreactor. Appl. Microbiol. Biotechnol. 2010, 89, 217–224. (12) Palacio-Barco, E.; Robert-Peillard, F.; Boudenne, J. L.; Coulomb, B. On-line analysis of volatile fatty acids in anaerobic treatment processes. Anal. Chim. Acta 2010, 668, 74–79. (13) Zhang, M. L.; Sheng, G. P.; Mu, Y.; Li, W. H.; Yu, H. Q.; Harada, H.; Li, Y. Y. Rapid and accurate determination of VFAs and ethanol in the effluent of an anaerobic H2-producing bioreactor using near-infrared spectroscopy. Water Res. 2009, 43, 1823–1830. (14) Feitkenhauer, H.; Sachs, J.v.; Meyer, U. Online titration of volatile fatty acids for the process control of anaerobic digestion plants. Water Res. 2002, 36, 212–218. (15) Lenz, M.; Hullebusch, E. D. V.; Farges, F.; Nikitenko, S.; Corvini, P. F. X.; Lens, P. N. L. Combined speciation analysis by X-ray absorption near-edge structure spectroscopy, ion chromatography, and solid-phase microextraction gas chromatography-mass spectrometry to 9099
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology
ARTICLE
evaluate biotreatment of concentrated selenium wastewaters. Environ. Sci. Technol. 2011, 45, 1067–1073. (16) Ward, A. J.; Hobbs; Bruni, E.; Lykkegaard, M. K.; Feilberg, A.; Adamsen, A. P. S.; Jensen, A. P.; Poulsen, A. K. Real time monitoring of a biogas digester with gas chromatography, near-infrared spectroscopy, and membrane-inlet mass spectrometry. Bioresour. Technol. 2011, 102, 4098–4103. (17) Moosbrugger, R. E.; Wentzel, M. C.; Ekama, G. A.; Marais, G. v. R. A 5 pH point titration method for determining the carbonate and SCFA weak acid/bases in anaerobic systems. Water Sci. Technol. 1993, 28, 237–45. (18) Anderson, G. K.; Yang, G. Determination of bicarbonate and total volatile acid concentration in anaerobic digesters using a simple titration. Water Environ. Res. 1992, 64, 53–59. (19) Lahav, O.; Morgan, B.; Loewenthal, R. Rapid, simple, and accurate method for measurement of VFA and carbonate alkalinity in anaerobic reactors. Environ. Sci. Technol. 2002, 36, 2736–2741. (20) Pejcic, B.; Eadington, P.; Ross, A. Environmental monitoring of hydrocarbons: A chemical sensor perspective. Environ. Sci. Technol. 2007, 41, 6333–6342. (21) Michaud, S.; Berner, N.; Buffiere, P.; Roustan, M.; Moletta, R. Methane yield as a monitoring parameter for the start-up of anaerobic fixed film reactors. Water Res. 2002, 36, 1385–1391. (22) Nielsen, J. B. H.; Lomborg, C. J.; Popiel, P. O.; Esbensen, K. H. Online near infrared monitoring of glycerol-boosted anaerobic digestion processes: evaluation of process analytical technologies. Biotechnol. Bioeng. 2008, 99, 302–313. (23) Punal, A.; Roca, E.; Lema, J. M. An expert system for monitoring and diagnosis of anaerobic wastewater treatment plants. Water Res. 2002, 36, 2656–2666. (24) Ginkel, S. W. V.; Kortekaas, S. J. M.; Lier, J. B. V. The chronic toxicity of alcohol alkoxylate surfactants on anaerobic granular sludge in the pulp and paper industry. Environ. Sci. Technol. 2007, 41, 4711–4714. (25) Tay, J. H.; Zhang, X. Y. Stability of high-rate anaerobic systems. I: Performance under shocks. J. Environ. Eng. 2000, 126, 713–725. (26) Speece, R. E. Anaerobic Biotechnology for Industrial Wastewaters; Archae Press: Nashville, TN, 1996. (27) APHA-AWWA-WPCF. Standard Methods for the Examination of Water and Wastewater, 20th ed.; Washington, DC, 1999. (28) Nachaiyasit, S.; Stuckey, D. C. The effect of shock loads on the performance of an anaerobic baffled reactor (ARB). 1. Step changes in feed concentration at constant retention time. Water Res. 1997, 31, 2737–2746. (29) Nachaiyasit, S.; Stuckey, D. C. The effect of shock loads on the performance of an anaerobic baffled reactor (ARB). 2. Step and transient hydraulic shocks at constant feed strength. Water Res. 1997, 31, 2747–2754. (30) Bouskova, A.; Dohanyos, M.; Schmidt, J. E.; Angelidaki, I. Strategies for changing temperature from mesophilic to thermophilic conditions in anaerobic CSTR reactors treating sewage sludge. Water Res. 2005, 39, 1481–1488. (31) Lau, I. W. C.; Herbert, H. H. P. Effect of temperature shock to thermophilic granules. Water Res. 1997, 31, 2626–2632.
9100
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
