Development of novel process designs for simultaneous oxidation and denitrification of wastewaters

Development of novel process designs for simultaneous oxidation and denitrification of wastewaters

European Symposiumon ComputerAidedProcessEngineering- 11 R. Ganiand S.B. Jorgensen(Editors) 9 2001 ElsevierScienceB.V. All rightsreserved. 493 Devel...

332KB Sizes 0 Downloads 24 Views

European Symposiumon ComputerAidedProcessEngineering- 11 R. Ganiand S.B. Jorgensen(Editors) 9 2001 ElsevierScienceB.V. All rightsreserved.

493

Development of Novel Process Designs for Simultaneous Oxidation and Denitrification of Wastewaters 9

a*

S.Rlgopoulos , P. Linke b and A. Kokossis bw Department of Chemical Engineering, UMIST, PO Box 88, Manchester, M60 1QD, UK b Department of Process Integration, UMIST, PO Box 88, Manchester, M60 1QD, UK a

A systematic synthesis approach is adopted to the problem of activated sludge process design. The conventional designs as well as all novel schemes for combined oxidation/denitrification of wastewater are explored. The process is optimised using a novel methodology for optimal reaction/separation network synthesis, supplied with a comprehensive and general-purpose kinetic model (IAWPRC Activated Sludge Model No.l). The optimisation results suggest a counter-intuitive optimal design policy: Simultaneous oxidation and denitrification are carried out simultaneously within the same reactor at very low oxygen concentrations. The new designs feature significantly improved nitrogen removal as compared to conventional processes. 1. INTRODUCTION The activated sludge process is the most commonly encountered form of biological wastewater treatment. Despite having received the attention of many researchers, most of the work so far has addressed the optimisation of individual operating parameters whilst no study to date has attempted the overall process optimisation including the basic structure of the process. The primary aims of the activated sludge process are to oxidise the organic content of wastewater into harmless inorganic compounds, and to convert organic and ammonia nitrogen to gaseous components. These two objectives are contradictory: the former process takes place in the presence of oxygen while the later is anoxic and each process imposes additional processing requirements onto the other. Current systems attempting to accomplish both aims employ separate aerobic and anoxic reactors, such as in the empirically designed Ludzack-Effinger and Wuhrmann systems (Van Haandel et al., 1981). This study combines a detailed activated sludge model with a robust synthesis methodology for multiphase reaction-

Current address: Dept. of chemical engineering, UniversityCollege London, London WC1E 7JE Corresponding author. Current address: Departmentof Chemical & Process Engineering, Universityof Surrey, Guildford, GU2 5XH, UK. Tel. ++44(0)1483 876573, Fax: ++44(0)1483 876581. Email: [email protected]

494 separation networks to gain insights into the complex process trade-offs and to seek pathways towards improved designs. 2. MODELLING OF WASTEWATER TREATMENT Modelling activated sludge systems is still an active research area. To date, the IAWPRC Group Activated Sludge Model No. 1 (Henze et el., 1987) is the most complete and is widely accepted by the wastewater treatment industry. It has been developed by a task group of researchers appointed by the International Association on Water Pollution Research and Control (IAWPRC) and subsequently used by many practitioners. Dold and Marais (1986) verified the model under a variety of process conditions. The model comprises of 13 reactants, participating in 7 biochemical reactions (Table 1). This complexity, as compared to the simple Monod model, accounts for the model's flexibility and applicability to different wastewater compositions. The highly non-linear kinetic equations are shown in Table 2. For a detailed model description it is referred to the IAWPRC report (1987) and Henze et al. (1987). Table 1. Reaction paths in wastewater treatment according to the IAWPRC model

Carbonaceous Organics Reaction Path 02 ,NO,X BH

X S

SS .~_02

Ss +NO3-

Organic Nitrogen Reaction Path

>S S

SNH + 02

XBlf ~ X B H

xn.

02 ,NO,XsH

XND

>XB H ..1_N2

Ss +NO3,

>aND,

SND

Xsg

~ SNH

X.. >XBA + NO 3xn. >X8 n + N2 ,

,,,

Table 2. Kinetics of the IAWPRC model

Process rate (ML3T 1)

Process

Aerobic growth of heterotrophs Anoxic growth of heterotrophs

PH Ks + Ss ~u~ Ks +Ss

Aerobic growth of autotrophs Hydrolysis of entrapped organics, Rh

I

Ko,H+So

'uA K ~ + SNH

kh

XBH Kx + - ~

KO,H+ So KN~ +SN~ .ngXB~

II

Ko,A +So

.itKo,H+So) L "! Ks +Ss)lKNo +Suo

Process

Process rate (Mt'sr ')

Decay of heterotrophs

bh XBH

Decay of autotrophs

bAXBA

Ammonification of nitrogen

khSNDXBH

Hydrolysis of entrapped organic nitrogen

XND Rh'~ Xs

495 3. OPTIMISATION METHODOLOGY The activated sludge process is synthesised using a systematic reaction/separation network synthesis framework (Linke et al., 2000) to determine the optimal biochemical reactor network design along with the sludge separation and recycle policies. The methodology builds upon previous efforts in reaction system synthesis (Kokossis & Floudas, 1990; Marcoulaki & Kokossis, 1999; Mehta, 1998) and employs superstructures of generic shadow compartment and separation task units that capture all possible novel and conventional process design options that exist for the multiphase system. All generic units are interconnected in a network of streams through a superstructure scheme in each of the contacting phases and in each pair of contacting phases where state change is possible. The representation provides for all possible mixing, contacting, and reaction/separation features. The superstructures are optimised using Simulated Annealing. The search produces performance targets in the form of stochastic optima with confidence levels appropriate for this problem and a multitude of designs with close-to-target performances, which offers a major advantage in understanding the process trade-offs. A detailed description of the algorithm is provided by Aarts and Van Laarhoven (1985). Two sets of representative values of municipal wastewater compositions, adopted from the IAWPRC report (1987), are studied. Two sets of values of kinetic parameters were tested, both within ranges suggested by the IAWPRC (1987). Oxygen mass transfer is modelled using the film theory. The weighted objective function is formulated so as to minimise the effluent COD and total nitrogen content: Objective = Min

COD +

N xl O0

(1)

No The stochastic nature of the optimisation means that it will yield a different design every time it is operated with different parameters. Yet these results will share some common features, and an insightful examination of them will provide the guidelines towards an improved process. For statistical confidence 16 runs were performed in each case, and about 2000-5000 different designs were evaluated during each run. 4. RESULTS AND DISCUSSION Before looking at the novel configurations, two conventional designs were evaluated according to the objective function (1) to establish a basis for comparison (see Figure 2 and Table 3). It is apparent that, although conventional designs succeed in removing COD, they do a poor job in reducing the nitrogen content. Accomplishing denitrification whilst maintaining high COD removal ratios appears to be a main challenge for process design. Using the first feed composition and parameter set, an optimisation without volume bounds on the reaction equipment yields a target (lowest value of objective function) of 267, corresponding to a 96.8% reduction in COD and 84.8% in N, but with too high residence time to be compared with the conventional processes. For this reason volume bounds are introduced to limit the mean hydraulic residence time up to 8 days, at which the conventional

496 processes were evaluated. Surprisingly the target did not deteriorate significantly, its new value being 271. COD was reduced by 97.4%, and N by 84.9%, a performance much better than those attained by any conventional process, especially as far as denitrification is concerned. An inspection of the structures revealed a striking feature: many structures did not include an anoxic reactor, and yet yielded excellent denitrification results. Moreover, the system appeared to seek ways to hinder or control the dissolution of oxygen, in striking contrast to current practice, which aims at dissolving the maximum amount possible into the oxic reactors. To find out how such excellent denitrification could be accomplished without the inclusion of an anoxic reactor, the detailed oxygen profiles within the aerated reactors (mostly PFRs) were examined (Figure 2). The concentration of oxygen is controlled within extremely low levels (0.1-0.2 ppm), an order of magnitude less than those currently used in industrial practice (2-9 ppm). This leads to the conclusion that both organic matter stabilisation and denitrification processes occur simultaneously in these designs, due to the very low oxygen concentrations - a policy that seems to achieve maximum efficiency when low volume is required. In order to keep the oxygen concentration in such low levels while retaining the necessary rate of oxygen dissolution to fulfil the requirements of the oxidation process, the new designs employed several ways: recycling the liquid phase in order to dilute the oxygen, recycling the gas phase to reduce the driving force and introducing side streams for better control of dissolution. Both liquid and gas recycling are expected to increase the operating cost of the process, though, and in practice this aim may be easier to achieve by proper selection of the aeration equipment, which provides some control over the overall mass transfer coefficient. The effect of feed composition is investigated by carrying out the optimisation with a more concentrated feed; however, the resulting structures are still based on the same design principles and achieved similar high values of N reduction (88%). Any optimisation relies on the set of parameter values used in the model, and though it is desirable to investigate novel designs, it is the conventional ones that were used to determine the kinetics. An additional study was performed with kinetic parameters that assume values up to the extreme end of their range. The trend of the resulting designs was, however, maintained: optimal solutions still relied on carrying out organic matter oxidation and denitrification simultaneously, the only difference being that dissolved oxygen concentrations were even lower. The main features of the novel designs are shown in Fig.2, while Table 3 summarises the numerical results. 5. CONCLUSIONS A superstructure-based stochastic reaction-separation network optimisation methodology has been applied to investigate and optimise the activated sludge process for combined nitrogen and COD removal. The method indicated a new concept for process design: instead of employing separate aerobic and anoxic reactors it is suggested that both processes are be carried out within the same unit. This can be accomplished by the use of very low dissolved oxygen concentrations that allow both aerobic and anoxic reactions to proceed at reasonable rates. The new design policy yielded a significant improvement in

497 nitrogen removal. The ability of the optimisation methodology to deal with complex problems was demonstrated, as well as its potential to detect novel designs based on concepts radically different from those of conventional processes. The optimisation has, however, employed the reaction model within the whole domain of possible configurations and the optimum did not occur in the region where the kinetic parameters were determined. Therefore the new concept should not be regarded as an immediate design proposal but as a direction for further research, both theoretical and experimental, to explore the behaviour of the process into the new unexplored region.

0.4 E r

03

,-

02

~

01

o

I..................................

T ....................................

0

10

i

..................

20 Reactor length (m)

T ..............................................

30

40

Fig. 1. Dissolved oxygen concentration profile in combined anoxic/aerobic reactor

I

I

a) Modified Ludzack Ettinger

-

-

- ~------,---~--:-:

,

b) Wuhrmann

i

-

-

Cobine:aerobi/ano:c

-

-'-

*

Air Water Sludge

c) Optimised Scheme Fig. 2. Conventional processes (Van Haandel, Ekama and Marais, 1981) and general structure of the optimised process for single sludge denitrification

498 Table 3. Comparison of conventional processes and novel optimised designs

Processes Ludzack-Effinger (Fig.3a) Wuhrmann (Fig.3b) Optimised, no bounds Residence time 8 days Residence time 5 days 8 days, concentrated feed 8 days, different kinetics

Objective function 825 698 267 271 310 415 324

COD % removal 98.3 98.8 96.8 97.5 98.2 98 98.5

N% removal 47 55 84.8 84.9 82.8 88.1 80.8

Comments Anoxic reactor first Aerobic reactor first Huge volume Combinedanoxic/aerobic >> >> Very low oxygen conc.

LIST OF SYMBOLS

XI XAUT XHET

XND Xs

Rho bu,ba,kh,ng,nh, Kx,Ko,A,KNo

Inert insoluble organics, ppm Autotrophs, ppm Heterotrophs, ppm Insoluble org. nitrogen, ppm Insoluble COD, ppm Hydrolysis rate Coefficients of the IAWPRC model (Henze et al., 1987)

SNO

SND SNH SI Ss So COD0, No

Nitrates, ppm Soluble organic nitrogen, ppm Ammonia nitrogen, ppm Inert soluble organics, ppm Soluble COD, ppm Dissolved oxygen, ppm Initial COD and nitrogen molar flows, kgmol/hr

REFERENCES

Aarts, E.H.L., & P.G.M. van Laarhoven (1985). Statistical Cooling: A General Approach to Combinatorial Optimisation Problems. Philips J. Res., 40, 193. Dold, P.L., & G.v.R. Marais (1986). Evaluation of the General Activated Sludge Model Proposed by the IAWPRC Task Group. Wat. Sci. Tech., 18, 63. Henze, M., C.P.L. Grady Jr, W. Gujer, G.v.R. Marais, & T. Matsuo (1987). A General Model for Single-Sludge Wastewater Treatment Systems. Wat. Res., 21,505. IAWPRC (International Association on Water Pollution Research and Control) (1987). IA WPRC Scientific and Technical Reports No. 1, IAWPRC, London. Kokossis, A.C. & C.A. Floudas (1990). Optimisation of Complex Reactor Networks - I. Isothermal Operation. Chem. Engng. Sci. 45, 3,595. Linke, P., V.L. Mehta, & A.C. Kokossis. In S. Pierucci: Computer Aided Chemical Engineering, 8 (pp. 1165-1170). Amsterdam: Elsevier Science (2000). Marcoulaki, E.C., & A.C. Kokossis (1999). Screening and Scoping Complex Reaction Networks Using Stochastic Optimization. AIChE J., 45(9), 1977. Mehta, V.L. (1998). Synthesis and Representation of Multiphase Reactor Networks. Ph.D. Thesis, University of Manchester Institute of Science and Technology. Van Haandel, A.C., G.A. Ekama, & G.v.R. Marais (1981). The Activated Sludge Process3: Single Sludge Denitrification. Wat. Res., 15, 1135.