Figure 7 - uploaded by Carl Kriger
Content may be subject to copyright.
4: The SCADA system graphical user interface allows the user to view realtime plant variables, trend information, enter data in manual mode, run the neural network software for training and validation, and view and enter data in the database
Citations
... Le nombre varie de 500 à 1000 données. Il y'a des résultats où les époques sont variés aussi bas que 100 et la performance du réseau est similaire à celle où le nombre d'époques est élevé (Kriger, 2007 ...
... Il est conseillé de partager l'ensemble des données en deux sous-ensembles. Kriger, 2007). ...
The performance and reliability of a wastewater treatment plant is an important consideration
especially if treated wastewater will be recovered for future reuse. Safety margins for public
health and environmental protection must be ensured.
The performances of treatment process are usually influenced by many factors such as
qualitative and quantitative changes in waste water and the inherent variability of the process
wastewater. Algerian legislation has established criteria of quality of treated water and
discarded, thus it should be possible to evaluate the performance of the process and the
reliability of facilities to ensure compliance.
The performance of municipal wastewater treatment plant in Khenchela was assessed based
on reliability. Data studied and analyzed statistically included wastewater flow rates and some
important water quality parameters such as Chemical Oxygen Demand (COD), five-day
Biochemical Oxygen Demand (BOD5), and Total Suspended Solids (TSS).
Results of goodness of fit tests (Kolmogorov-Smirnov test, Anderson-Darling test, Cramervon Mises test) indicated that among the several distribution models investigated to fit the
daily records of water quality parameters, the lognormal distribution were appropriate for the
data, and the most adequate to describe collected data.
Using the lognormal function, a probabilistic reliability model was developed and found
suitable to quantitatively the performance of the plant. A significant feature of this model is
that the model parameters are based on properties of original data.
The development of this reliability model is the first contribution of this study; a model that
can provide a quantitative performance of the studied wastewater treatment plant, and can
also be used to estimate mean values of effluent quality.
In order to develop an effective control strategy for the activated sludge process of wastewater
treatment plant, an understanding of the reliability disturbances to the wastewater treatment
plant is necessary. Biological systems are among the most difficult to control and predict. Due
to the complex biological reaction mechanisms, the highly time-varying and multivariable
aspects of the wastewater treatment plant (WWTP), the diagnosis of the WWTP are still
difficult in practice. The application of intelligent techniques, which can analyze the
multidimensional nonlinear process data using a visualization technique, can be useful for
analyzing and diagnosing the activated-sludge process in the WWTP.
This complex capability for nonlinearity representation combined with the fact that no
reliability model exists for the WWTP, makes neural networks an ideal choice for a solution.
Forecasting the behavior of complex systems has been a broad application area for neural
networks. Applications such as economic forecasting, electricity load / demand forecasting,
and forecasting natural and physical phenomena have been extensively studied, hence the
numerous papers presented at annual conferences in this focus area. The cognitive ability of
artificial neural networks to map nonlinear complex input-output relationships, which would
allow for better prediction and corrective control of processes, make them particularly
attractive.
The second contribution of this work presents the development of neural network models for
prediction of reliability and process rate of failure based on historical plant data and a
probabilistic model especially developed for this study.
Six different neural networks divided on two types; simple model and complete model, based
on Multi-Layer Perceptrons (MLP) are developed for the prediction of reliability level, and
process rate of failure based on the following quality parameters of treated wastewater,
Chemical Oxygen Demand (COD), five-day Biochemical Oxygen Demand (BOD5) and Total
Suspended Solids (TSS).
The application area is the prediction of the reliability level at a local municipal wastewater
treatment plant. The forecast result is used for the determination of the set point to control the
process, in order to optimize plant performance.
The results will hopefully provide useful information about the scope and possibilities for the
application of neural networks in the field of wastewater treatment and serve as a diagnostic
tool of process failures and especially help operator in the daily management of the plant.
Develop artificial neural networks models based on a probabilistic model for predicting the
reliability and failure rate of wastewater treatment process based on activated sludge is the
first attempt in the literature and research.