Download - Presentation musfique ahmed
INTERNATIONAL CONFERENCE ON ARCHITECTURE AND
ENGINEERING IN URBAN DEVELOPMENT 2013
The Development of Modeling Techniques for
Biological Wastewater Treatment : A review
Musfique Ahmed
Lecturer
Department of Environmental Science
Independent University, Bangladesh (IUB)
WASTEWATER TREATMENT
Removal of Physical, Biological and Chemical
constituents
Complex process
Quality fluctuation
Composition fluctuation
Adaptive behaviour of microorganisms
MODELING IN WASTEWATER TREATMENT
Models
Describe
Predict
Control
This Review focuses on the models using for
biological treatment process.
MODEL DEVELOPMENT
`Model Goals Select model purpose, required
model accuracy, model boundaries
Data Collection
Data Analysis
Model Setup &
Calibration
Model Verification
Model Simulation
tank dimensions, piping and inter
connections, flow dynamics and
influent characterization.
data screening, mass balances,
and design hand calculations
Modifying input parameters
MODELS
Models are classified into three groups
Aerobic Process – ASM models
Anaerobic Process – ANN and UASB
Hybrid models
ACTIVATED SLUDGE MODEL NO. 1(ASM1)
First model of ASM family
Developed by International Water Association (1987)
Developed to describe organic carbon removal, nitrification and de-
nitrification with instantaneous use of oxygen and nitrate as electron
acceptors
Useful as a predictor of oxygen demand and sludge production in an
activated sludge system
ACTIVATED SLUDGE MODEL NO. 2(ASM2)
Henze et al. first introduced the ASM2 model in 1995 by including biological
phosphorus (bio-P) removal in ASM1
Increase the capability of ASM1 model
Introduced a new group of organisms to the biomass – PAOs
Phosphorus Accumulating Organisms
Capable of gathering phosphorus and stocking them in the form of cell
internal polyphosphates (XPP) and poly-hydroxyalkanoates (XPHA).
ACTIVATED SLUDGE MODEL NO. 3(ASM3)
Developed with the same objectives as ASM1 for biological N removal
Insertion of internal cell storage compounds in heterotrophs
Developed by considering the importance of storage polymers in the
heterotropic activated sludge alteration.
All readily biodegradable substrate (SS) first taken up and stored into an
internal cell component (XSTO) prior to growth.
PARAMETERS
ANAEROBIC PROCESS MODELLNG
Very complex and complicated to model
high sensitivity to the influent characteristics
operational conditions
different environmental conditions
Most powerful methods for modelling the complex and non liner
anaerobic system is using artificial neural networks (ANN).
ARTIFICIAL NEURAL NETWORK
Predict the performance of the process
Develop a precise nonlinear mapping from input-output
couples of data without recognizing their functional
relationship
Models Reason Inputs & Output
Parameters
Hanbay,
Turkoglu &
Demir (2007)
Prediction and analysis of the COD
removal in effluent
Temperature,pH, COD,
TN, TSS
Hamed,
Khalafallah &
Hassanien
(2004)
Performance prediction of a WWTP
in Cairo, Egypt
BOD
SS concentrations
Hong et al.
(2007)
For the real time estimation of
nutrient concentrations to
overcome the problem of delayed
measurements
NO3-
NH4+
PO43+ concentrations
UPFLOW ANAEROBIC SLUDGE BLANKET (UASB)
To remove carbonaceous BOD
To stabilize the waste and
Conduct denitrification
Models for describing the aspects
• fluid flow
• rheological behavior of the sludge
• extremely long start-up period
• transport phenomena
UPFLOW ANAEROBIC SLUDGE BLANKET (UASB)
Bolle et. al. (1985) developed a hydrodynamic
model of the fluid flow based on previous scale
model experience and some physical intuition.
Assumption: both the sludge bed and sludge
blanket were behaving like completely stirred tank
reactors and the liquid flow settler volume was
explained as a plugflow reactor.
Outcome: the short-circuiting flow over the sludge
bed increases with the increasing superficial gas
velocity
UPFLOW ANAEROBIC SLUDGE BLANKET (UASB)
Skiadas and Ahring (2002) proposed a model for
UASB reactors by using Cellular Automata (CA)
concept.
A cellular automation is a simulation, which is
discrete in time, space and state
The CA theory is used to predict the granules’
structure which appears different in outer and inner
granule layers
HYBRID MODELS
Integration of two different models
Improved in predicting process dynamics
variability of bacteria growth rates variable retention times for phosphorus and nitrogen removal
A group from Taiwan National University gave the solution
By incorporating a biofilm model into the general dynamic model
To predict the effluent quality of a combined activated sludge and biofilm process.
HYBRID MODELS
Neural Fuzzy System – Fuzzy system + Neural Netwroks
Adaptive neuro-fuzzy interference system (ANFIS) –
functional neural fuzzy models
Tay & Zhang developed a fast predicting neural fuzzy model
for high rate wastewater anaerobic system to simulate and
predict the response of a system to different system
disturbances
HYBRID MODELS
Input and Output Parameters
Liquid Phase- include pH,
volatile fatty acids (VFA), alkalinity,
COD or TOC,
COD reduction and
redox potential (ORP)
Gas Phase - Gas production rates
CH4
CO2
H2
CO
DISCUSSION
Aerobic Process Modelling – Deterministic in nature
- derive a direct link between the inputs, outputs,
state variables and parameters
the state variables are represented by the
parameters and previous states of the model
ANN modelling - Stochastic model - use random
data generation for non linear mapping
Calibration is easier than the conventional
deterministic models.
DISCUSSION
Models applied in UASB reactors-
Deterministic - the model developed by Skiadas
and Ahring (2002) by using CA theory
Used real data and mathematical equations
Stochastic – Using artificial neural networks in
UASB reactors for the prediction of COD removal
efficiency
LIMITATIONS
Experimental basis of activated sludge modeling is very significant
The experimental backup lagged behind because of the fast pace of progressing in the modeling of activated sludge.
Over parameterized - a given parameter is treated with minor significance that can cause major propagation towards all estimated parameters
ANN training data - The problem of overfitting occurs in case of noisy and uncertain training data
Models for UASB reactors usually do not consider non-ideal conditions in full-scale reactors.
REFERENCES
Bolle, WL, Breugel, Jv, Eybergen, GCv, Kossen, NWF & Zoetemeyer, RJ 1985, 'Modeling the Liquid Flow in Up-Flow Anaerobic Sludge Blanket Reactors', Biotechnology and Bioengineering, vol. 28, pp. 1615-20.
Hamed, MM, Khalafallah, MG & Hassanien, EA 2004, 'Prediction of Wastewater Treatment Plant Performance Using Artificial Neural Networks', Environmental Modeling & Software, vol. 19, no. 10, pp. 919-28.
Hanbay, D, Turkoglu, I & Demir, Y 2007, 'Prediction of Chemical Oxygen Demand (COD) Based on Wavelet Decomposition and Neural Networks', Clean – Soil Air Water, vol. 35, no. 3, pp. 250 – 4.
Ng, ANL & Kim, AS 2006, 'A mini-review of modeling studies on membrane bioreactor (MBR) treatment for municipal wastewaters', Desalination, vol. 212, no. 1-3, pp. 261-81.
Pena-Tijerina, AJ & Chiang, W 2007, 'WHAT DOES IT TAKE TO MODEL A WASTEWATER TREATMENT PLANT?', paper presented to TEXAS WATER 2007, Texas.
Tay, J-H & Zhang, X 2000, 'A FAST PREDICTING NEURAL FUZZY MODEL FOR HIGH-RATE ANAEROBIC WASTEWATER TREATMENT SYSTEMS', Water Research, vol. 34, no. 11, pp. 2849-60.