Modular neural networks prediction model based A2/O process control system

The Anaerobic/Anoxic/Oxic (A2/O) process is a traditionally well-established biological wastewater treatment process (WWTP). In order to control a biological process in a laboratory environment, engineers typically adopt a methodology that relies mostly on their prior knowledge of transient and steady-state behaviors of micro-organisms. Based on this prior knowledge, our A2/O process is designed to keep proper reaction time in check as well as the state defining conditions of micro-organisms. However, converse to our expectation, unforeseen experimental changes in our biological samples could cause the entire experimental process to deviate from its original course of progress. Practically, to mitigate these unexpected changes, modular neural networks (MNNs) prediction model for dissolved oxygen (DO) concentration is proposed. The MNNs consists of numerous neural network models. Each neural network model is only concerned with a single cluster. Therefore, when suitable neural network models can be selected for each condition, stable and advanced treatment performance is expected. DO concentration is an observable and controllable factor in a reactor. On the whole it affects the biological reactions and water quality. DO set-point is decided by prediction model and its control remains as an essential feature of WWTP. The predicted DO concentration is used to control the air blower, thus enabling scheduling for a stable NH4-N concentration in the effluent wastewater. The proposed model has time-series static neural networks with multiple inputs and one output. The inputs — the air blower speed and DO concentration — are time-dependent variables. The output is the DO concentration in the immediate following time step. Our prediction results are compared with those of other prediction methodologies; proposed prediction model shows that it can achieve better accuracy for the DO concentration estimate than by other models. Therefore, the proposed system can be applied to a time-delayed compensation system to meet the target DO set-point tracking.

[1]  Guoqiang Peter Zhang,et al.  Time series forecasting using a hybrid ARIMA and neural network model , 2003, Neurocomputing.

[2]  Rafael Gadea Gironés,et al.  A hardware design of a massive-parallel, modular NN-based vector quantizer for real-time video coding , 2008, Microprocess. Microsystems.

[3]  Yi-Ti Shih,et al.  Advanced parameter identification for a linear-motor-driven motion system using disturbance observer , 2009 .

[4]  Zhong-Qiu Zhao,et al.  A novel modular neural network for imbalanced classification problems , 2009, Pattern Recognit. Lett..

[5]  Oscar Castillo,et al.  Optimization of Artificial Neural Network Architectures for Time Series Prediction Using Parallel Genetic Algorithms , 2008, Soft Computing for Hybrid Intelligent Systems.

[6]  Jung-Hwan Ahn,et al.  Improvement of positioning accuracy of magnetostrictive actuator by means of built-in air cooling and temperature control , 2011 .

[7]  K. Chau,et al.  Prediction of rainfall time series using modular artificial neural networks coupled with data-preprocessing techniques , 2010 .

[8]  Michael Y. Hu,et al.  Forecasting with artificial neural networks: The state of the art , 1997 .

[9]  Young-Koo Lee,et al.  A modular classification model for received signal strength based location systems , 2008, Neurocomputing.

[10]  Bijaya Ketan Panigrahi,et al.  Detection and classification of power quality disturbances using S-transform and modular neural network , 2008 .

[11]  Olaf Wolkenhauer.  Data engineering - fuzzy mathematics in systems theory and data analysis , 2001 .

[12]  M. Xia,et al.  Adaptive neural network model for time-series forecasting , 2010, Eur. J. Oper. Res..

[13]  Stefano Alvisi,et al.  Fuzzy neural networks for water level and discharge forecasting with uncertainty , 2010, Environ. Model. Softw..

[14]  David West,et al.  Predictive modeling for wastewater applications: Linear and nonlinear approaches , 2009, Environ. Model. Softw..

[15]  Amit Kumar,et al.  An evaluation of artificial neural network technique for the determination of infiltration model parameters , 2006, Appl. Soft Comput..

[16]  J. Heijnen,et al.  Microbiology and biochemistry of the enhanced biological phosphate removal process , 1998 .

[17]  R. Hancock,et al.  Biochemical model for enhanced biological phosphorus removal , 1986 .

[18]  Farouq S Mjalli,et al.  Use of artificial neural network black-box modeling for the prediction of wastewater treatment plants performance. , 2007, Journal of environmental management.

[19]  Yongzhen Peng,et al.  Enhanced Biological Nutrients Removal in Modified Step-feed Anaerobic/Anoxic/Oxic Process , 2009 .

[20]  H. Chris Tseng,et al.  Modular neural networks with applications to pattern profiling problems , 2009, Neurocomputing.

[21]  Manuel Carcenac,et al.  A modular neural network for super-resolution of human faces , 2009, Applied Intelligence.

[22]  Tugba Taskaya-Temizel,et al.  2005 Special Issue: A comparative study of autoregressive neural network hybrids , 2005 .

[23]  Durdu Ömer Faruk.  A hybrid neural network and ARIMA model for water quality time series prediction , 2010, Eng. Appl. Artif. Intell..

[24]  António Completo,et al.  Accuracy control of complex surfaces in reverse engineering , 2011 .

[25]  George Kourakos,et al.  Pumping optimization of coastal aquifers based on evolutionary algorithms and surrogate modular neural network models , 2009 .

[26]  H. D. Stensel,et al.  Wastewater Engineering: Treatment and Reuse , 2002 .

[27]  Patricia Melin,et al.  A hybrid approach for image recognition combining type-2 fuzzy logic, modular neural networks and the Sugeno integral , 2009, Inf. Sci..

[28]  Christopher MacLeod,et al.  Incremental growth in modular neural networks , 2009, Eng. Appl. Artif. Intell..

[29]  Pan Wang,et al.  A novel Bayesian learning method for information aggregation in modular neural networks , 2010, Expert Syst. Appl..