G. Langergraber*, J.K. Gupta**, A. Pressl*, F. Hofstaedter**, W. Lettl**, A. Weingartner** and N. Fleischmann*** * Institute for Sanitary Engineering and Water Pollution Control, University of Natural Resources and Applied Life Sciences, Vienna, Muthgasse 18, A-1190 Vienna, Austria (E-mail:
[email protected]) ** s::can Messtechnik GmbH, Herminengasse 10, A-1020 Vienna, Austria (E-mail:
[email protected]) *** hydrophil, Lerchenfeldergürtel 43, Top 6/3, A-1160 Vienna, Austria (E-mail:
[email protected]) Abstract A submersible UV/VIS spectrometer was used to monitor a pilot-scale sequencing batch reactor (SBR). The instrument utilises the whole UV/VIS range between 200 and 750 nm. With just one single instrument nitrate, organic matter and suspended solids can be measured simultaneously. The spectrometer is installed directly in the reactor, measures in real-time, and is equipped with an auto-cleaning system using pressured air. The paper shows the calibration results for measurements in the SBR tank, time series for typical SBR cycles, and proposes possible ways for optimisation of the operation by using these measurements. Keywords Control; in-situ; on-line; real-time; SBR; UV/VIS spectroscopy
Introduction
Monitoring and control of sequencing batch reactors (SBRs) is a very active field of research nowadays. Several strategies have been applied to improve the performance of SBR systems including: soft sensors (Zyngier et al., 2001), computer simulation (Mikosz et al., 2001), feedforward aeration control (Wett and Ingerle, 2001), and on-line control of ORP, pH and DO (Andreottola et al., 2001). This paper suggests using a submersible UV/VIS spectrometer for on-line monitoring and control of a SBR. The instrument used is a submersible spectrometer that utilises the whole UV/VIS range between 200 and 750 nm. It is installed directly in the reactor and measures several parameters simultaneously in real-time. Using nitrate nitrogen measurements it is possible to monitor and control nitrification and denitrification. Suspended solids measurements can be used to optimise settling times. The paper shows calibration results and time series for typical SBR cycles.
Water Science and Technology Vol 50 No 10 pp 73–80 © IWA Publishing 2004
On-line monitoring for control of a pilot-scale sequencing batch reactor using a submersible UV/VIS spectrometer
Materials and methods The pilot-scale SBR
The pilot-scale SBR is located in the technical laboratory of the Institute for Sanitary Engineering and Water Pollution Control, Vienna University of Natural Resources and Applied Life Sciences, and has a maximum volume of 1.5 m3 with an exchange ratio of about 30%. The SBR tank has a rectangular shape. The technical equipment consists of a stirrer, an air diffuser (rubber membrane), a water level control, and a swimming decant device. The plant is automatically run by a computer and a LABVIEW®-programme. The typical operation cycle of the pilot-scale SBR consists of the following phases: settling (30 min), withdrawal (ca. 5 min), filling and mixing (ca. 15 min), mixing (45 min), and aeration (180 min). Filling and withdrawal are controlled by water level. Usually the pilot-scale SBR is operated with this fixed time schedule. Additionally the SBR can be operated manually.
73
The technical laboratory is connected to a nearby sewer. Typical concentrations of the mechanically pre-treated influent wastewater are as follows: 400 mg COD/l, 60 mg NH4N/l and 8 mg TP/l. The UV/VIS spectrometer
G. Langergraber et al.
The submersible UV/VIS spectrometer (Figure 1, spectro::lyser by s::can Messtechnik, Vienna) has a length of about 0.6 m and a diameter of 44 mm. The instrument records light attenuation in the wavelength region between 200 and 750 nm (256 pixel with a Xenon flash lamp as a light source). The length of the measurement path available is from 1 to 100 mm opening a wide range of applications from ultra-pure waters (DOC > 10 µg/l) up to concentrated wastewaters with a COD of several 1,000 mg/l. A path length of 1 mm was used for the measurements in the SBR tank. Spectrometric measurement methods are defined for single substances (e.g. nitrate, nitrite, benzene, phenol) as well as for surrogate parameters (e.g. SAC254, turbidity, TSSeq, CODeq, TOCeq, DOCeq). A single measurement typically takes about 15 s. The results are displayed and/or communicated in real-time. The measurement takes place directly in-situ without sampling or sample treatment. Thus, measurement errors due to sampling, transport, storage, dilution etc. are not relevant. The instrument is equipped with an auto-cleaning system using pressurised air that prevents film growth on the optical parts of the instrument. An ex-proof version is available for sewer and industrial applications in explosive atmosphere. Up to now the instrument has been successfully applied for measurements in sewers (e.g. Hofstaedter et al., 2003; Lorenz et al., 2002), wastewater treatment plants for municipal and industrial wastewater (e.g. Winkler et al., 2002; Langergraber et al., 2004, 2003b), river water (e.g. Fleischmann et al., 2002), and groundwater (e.g. Langergraber et al., 2002). Sensor calibration
lamp electronics with Xenon flash lamp
measuring beam
cleaning nozzles
reference beam
74
Figure 1 The submersible UV/VIS spectrometer
collecting optics
beam alternator
inserts for adaptors of optical path
enlarging optics
For typical waters (e.g. municipal wastewater – raw and treated, river water, drinking water, . . .) a so-called “global calibration” is provided as default configuration of the instrument. Usually high precision can be achieved using this standard parameter set. For many purposes, such as plant control, precision is more important than trueness and the global calibration often delivers sufficient results. Using a second calibration step (local calibration) improvements concerning trueness, precision and long-term stability of the results can be achieved. The global calibration is based on a partial-least-square (PLS) regression for the parameters of concern. The broad range of available wavelengths allows high flexibility for the choice of the best correlating wavelengths for the calibration function and the avoidance of cross-sensitivities. This is an advantage compared to systems, which provide absorption
256 pix array detector with µprocessor electronics
Results and discussion Calibration results
G. Langergraber et al.
measurements of a single or two wavelengths only. A detailed description of the calibration procedure implemented is given in Langergraber et al. (2003a). Due to the different composition of wastewaters, e.g. due to industrial inlets, a local calibration can be required. The local calibration is based on grab samples analysed for the parameters of interest and can be performed without demounting of the probe. The local calibration can be performed either by running a complete calibration procedure based on PLS regression (advanced local calibration) or by a simple change of the slope-intercept of the regression function that can be easily done by the user and most of the time this is enough to improve the correlation significantly.
The sampling for the local calibration was done at the end of the mixing period where low nitrate nitrogen concentrations occur. The settled sludge samples were mixed with a standard nitrate nitrogen solution of 50 mg/l. The UV/VIS absorption spectra of these mixtures were used for calibration. The results were obtained with advanced local calibration (PLS regression). Figure 2 shows the calibration results for total suspended solids (TSS). The correlation coefficient R2 was 0.995. The 95% confidence interval was 0.13 g/l at the mean lab value of 4.3 g/l. The confidence interval contains all possible calibration curves with a probability of 95%. The 95% prediction interval was 0.63 mg/l at the mean lab value. A future value should be in this interval with a probability of 95%. The lower and upper limits of detection were 0 and 15 g/l, respectively. The calibration results for nitrate nitrogen (NO3-N) are shown in Figure 3. For nitrate nitrogen the correlation coefficient R2 was 0.98. The 95% confidence interval of the calibration function and the 95% prediction interval were 1.6 mg/l and 7.4 mg/l, respectively at the mean lab value of 12.8 mg/l. The lower and upper limits of detection were 0 and 30 mg/l, respectively. The detection limit of nitrate nitrogen is decreasing with increasing TSS concentration. Figure 4 shows the nitrate nitrogen detection limit based on analysis of the fingerprints. For the used path length of 1 mm the detection limit is e.g. 20 mg/l of nitrate nitrogen in the presence of 8.5 g/l TSS. 16
Sensor TSSeq (g/l)
14 95 % confidence interval (calibration)
12 10 8 6
95 % prediction interval
4 2 0 0
2
4
6
8
10
12
14
16
Laboratory TSS (g/l) Figure 2 Calibration results for TSS
75
30
G. Langergraber et al.
Sensor NO3-Neq (mg/l)
25
95 % confidence interval (calibration)
20 15 10 5 95 % prediction interval
0 -5 -5
0
5
10
15
20
25
30
Laboratory NO3-N (mg/l) Figure 3 Calibration results for NO3-N (o . . . outlier, TSS concentration too high) 30
25
NO3-N (mg/l)
Limit of detection
20 Fingerprints OK
15
10
5
0 0
5
10
TSS (g/l) Figure 4 Detection limit NO3-N
Additionally a calibration was performed for filtered COD (CODf, Figure 5). The obtained correlation coefficient R2 for CODf was 0.90, the 95% confidence interval of the calibration function 24 mg/l at the mean lab value of 185 mg/l, and the 95% prediction interval 93 mg/l. The lower and upper limits of detection were 20 and 540 mg/l, respectively. Although the focus of the experiments was on nitrate and suspended solids measurement and less data were available compared to the calibrations for NO3-N and TSS the results obtained for CODf were quite good. Time series
76
Figure 6 shows a typical result obtained for a single SBR cycle. Five characteristic points can be observed: 1. The TSS concentration drops rapidly after switching off the aeration. In this case the instrument was installed 10 cm below the minimum water level (level after withdrawal) of the SBR. Once the low TSS concentration is reached settling is finished and withdrawal can start. During settling and withdrawal the NO3-N concentration stays constant.
500
300
G. Langergraber et al.
Sensor CODfeq (mg/l)
95 % confidence interval (calibration)
400
200
95 % prediction interval
100
0 0
100
200
300
400
500
Laboratory CODf (mg/l) Figure 5 Calibration results for filtered COD
2. A rapid drop of the NO3-N concentration occurs after filling and mixing starts. Readily biodegradable organic matter in the influent water is utilised as the carbon source for denitrification. The TSS concentration increases due to mixing. 3. After filling of the SBR (the maximum water level is reached) mixing without aeration is continued for further denitrification. A change of the slope of the NO3-N curve can be observed because during this period less readily biodegradable organic matter is available for denitrification. 4. The depletion of nitrate nitrogen shows the end of the possible denitrification and the aeration can be turned on. 5. During the aeration phase nitrification occurs and the nitrate nitrogen concentration is increasing. The change of the slope indicates that the free available ammonia nitrogen is consumed. The aeration can be turned off at this point. The slower increase afterwards is caused by nitrification of ammonia released due to ammonification and decay processes. 40
4 2
NO3-Neq
35 3
NO3-Neq (mg/l)
3.5
TSSeq
5
30
3
25
2.5
20
2
15
1.5 4
10
1
5
0.5
0 09:00
TSSeq (g/l)
1
0 11:00
13:00
15:00
17:00
19:00
Time (hh:mm) Figure 6 Typical curve during a SBR cycle
77
G. Langergraber et al.
The very high fluctuation of the TSS concentration during the aeration phase is most probably caused by air bubbles passing the measurement pathway and therefore influencing the spectral measurement. The fluctuation is lower in the denitrification phase. Figure 7 shows the same period as shown in Figure 6 but CODf instead of TSS. As for TSS high fluctuations occur for CODf. In contrast to TSS the fluctuations are of the same amplitude for both the denitrification and nitrification phases. The results for TSS and CODf indicate that further investigations regarding the installation of the sensor in the reactor have to be carried out. Figure 8 shows a comparison of spectrometric data and reference grab samples analysed for nitrate nitrogen. The grab samples were taken close to the position where the instrument was installed to guarantee the identity of the samples. The reference analyses match the online data very well. Possibilities for optimising the system
It is evident that by using the on-line measurements of NO3-N and TSS the time schedule of a SBR cycle can be optimised compared to a fixed time schedule. The time needed for 40
240
NO3-Neq
35
3
NO3-Neq (mg/l)
210
CODfeq
5
30
180
25
150
20
120
15
90
4
10
60
5
30
0
0
09:00
11:00
13:00
15:00
17:00
19:00
Time (hh:mm) Figure 7 Time series for NO3-N and filtered COD during a typical SBR cycle 25 Sensor NO3-Neq 20 Reference grab samples NO3-N
(mg/l)
15
10
5
0 11:00
12:00
13:00
14:00
Time (hh:mm) 78
Figure 8 Comparison of spectrometric with reference method
15:00
16:00
17:00
CODfeq (mg/l)
2
Summary and conclusions
G. Langergraber et al.
settling can be measured by the TSS concentration whereas the measured NO3-N concentration determines the mixing and aeration time needed to guarantee denitrification and nitrification, respectively (points 4 and 5 in Figure 6). By minimising the times needed for single phases more cycles can be run and therefore higher amounts of water can be treated. The amount of water withdrawn could be optimised by installing the sensor on the flexible part of the decanter. During withdrawal the sensor is moving downwards and withdrawal can continue until the TSS concentration increases indicating that the sensor is approaching the sludge blanket level. Higher volumes of water can be withdrawn and therefore treated by using such a system. No experiments regarding this optimisation strategy have been carried out up to now.
The submersible UV/VIS spectrometer can be used for measurements in a SBR tank. The instrument measures in-situ directly in the reactor and communicates the results in realtime. The auto-cleaning system which is operated with pressured air prevents trends in the time series due to film growth on the optical devices. Compared to a conventional nitrate nitrogen analyser the instrument is able to measure in-situ. Therefore no sampling and sampling preparation (e.g. ultra-filtration) is needed. Additionally no chemicals are required. In addition to nitrate nitrogen also TSS and CODf can be measured simultaneously with just a single instrument. High correlation coefficients of R2 = 0.98 and 0.995 were obtained for NO3-N and TSS, respectively, by running an advanced local calibration. The measurement of NO3-N was possible also for high TSS concentrations. However, for increasing TSS concentrations the detectable maximum nitrate nitrogen concentration is decreasing. The obtained correlation coefficient for CODf was lower (R2 = 0.90). Several characteristic points of a SBR cycle could be monitored using on-line UV/VIS spectrometry. Compared to a fixed time schedule the times for settling, denitrification and nitrification can be minimised and therefore the volumes of water that can be treated can be maximised. By mounting the instrument on the flexible part of the decanter additionally the amount of withdrawn water could be increased. Further tests have to be carried out regarding the optimal position for installation of the instrument in the reactor and the implementation of an optimised control strategy. The results of the first investigations showed that the submersible UV/VIS spectrometer is a promising instrument for optimising the operation of a SBR. References Andreottola, G., Foladori, P. and Ragazzi, M. (2001). On-line control of a SBR system for nitrogen removal from industrial wastewater. Wat. Sci. Tech., 43(3), 93–100. Fleischmann, N., Staubmann, K. and Langergraber, G. (2002). Management of sensible water uses with real-time measurements. Wat. Sci. Tech., 46(3), 33–40. Hofstaedter, F., Ertl, T., Langergraber, G., Lettl, W. and Weingartner, A. (2003). On-line nitrate monitoring in sewers using UV/VIS spectroscopy. In: Wanner, J., Sykora, V. (eds): Proceedings of the 5th International Conference of ACˇE CˇR ‘Odpadni vody – Wastewater 2003’, 13–15 May 2003, Olomouc, Czech Republic, pp. 341–344. Langergraber, G., Fleischmann, N., Van der Linden, F., Wester, E., Weingartner, A. and Hofstaedter, F. (2002). In-situ measurement of aromatic contaminants in bore holes by UV/VIS-spectrometry; In: Breh, W. et al. (eds): Field Screening Europe 2001, Kluwer, Dordrecht, The Netherlands, pp. 317–320. Langergraber, G., Fleischmann, N. and Hofstaedter, F. (2003a). A multivariate calibration procedure for UV/VIS spectrometric quantification of organic matter and nitrate in wastewater. Wat. Sci. Tech., 47(2), 63–71. Langergraber, G., Fleischmann, N., Hofstaedter, F., Weingartner, A. and Lettl, W. (2003b). Detection of (unusual) changes in wastewater composition using UV/VIS spectroscopy. In: Ruzickova, I., Wanner, J.
79
G. Langergraber et al. 80
(eds): Preprint of the 9th IWA conference on ‘Design, Operation and Costs of Large Wastewater Treatment Plants’ – Poster papers, 1–4 September 2003, Prague, Czech Republic, pp. 135–138. (Water Intelligence Online (2004) 2004 08010). Langergraber, G., Fleischmann, N., Hofstaedter, F. and Weingartner, A. (2004). Monitoring of a paper mill wastewater treatment plant using UV/VIS spectroscopy. Wat. Sci. Tech., 49(1), 9–14. Lorenz, U., Dettmar, J. and Fleischmann, N. (2002). Adaptation of a new online probe for qualitative measurements to combined sewer systems. In: Stricker, E.W., Huber, W.C. (eds): Proceedings of the 9th International Conference on ‘Urban Drainage’, 8–13 September 2002, Portland, Oregon, USA, pp. 427–428. Mikosz, J., Plaza, E. and Kurbiel, J. (2001). Use of computer simulation for cycle length adjustment in sequencing batch reactor. Wat. Sci. Tech., 43(3), 61–68. Wett, B. and Ingerle, K. (2001). Feedforward aeration control of a Biocos wastewater treatment plant. Wat. Sci. Tech., 43(3), 85–92. Winkler, S., Rieger, L., Thomann, M., Siegrist, H., Bornemann, C. and Fleischmann, N. (2002). In-line monitoring of COD and COD-fractionation: Improving dynamic simulation data quality. In: IWA (ed.): Preprints of the 3rd IWA International World Water Congress, April 7–12, 2002, Melbourne, Australia. Zyngier, D., Araújo, O.Q.F., Coelho, M.A.Z. and Lima, E.L. (2001). Robust soft sensors for SBR monitoring. Wat. Sci. Tech., 43(3), 101–106.
