Analysis of municipal wastewater treatment plant performance using artificial neural network approach
Artificial neural network (ANN) was used in this research as a statistical modeling tool for predicting the performance of wastewater treatment plant. A two years data of the waste water treatment plants’ effluent and influent parameters was collected and applied in developing and training the ANN u...
| Main Authors: | , , , |
|---|---|
| Format: | Conference or Workshop Item |
| Language: | English |
| Published: |
2011
|
| Subjects: | |
| Online Access: | http://irep.iium.edu.my/3161/ http://irep.iium.edu.my/3161/1/icbioe_ANN_modified.pdf |
| Summary: | Artificial neural network (ANN) was used in this research as a statistical modeling tool for predicting the performance of wastewater treatment plant. A two years data of the waste water treatment plants’ effluent and influent parameters was collected and applied in developing and training the ANN using the ANN toolbox in MATLAB. The data were obtained from Bandar Tun Razak Sewage Treatment Plant (BTR STP), that is managed by Indah Water Konsurtium (IWK), Malaysia's national sewerage company. The input and output parameters for the ANN were BOD, SS, and COD. It was found that the use of data screening is essential to come up with better ANNs model. Moreover, using multiple input-single output models was even a better model than single input-single output. The optimum number of hidden layer and neurons were determined which gave excellent results in predicting both the BOD and COD of the effluent which are required by the DOE. From the regression analysis, networks with one hidden layer and 20 nodes and BOD as input and COD as output were found to be the best one. The optimum number of hidden layers is 10 and the R value is improved by 30 %. The Mean Squared Error (MSE) is the lowest for the network. From the regression analysis, it is obvious that networks using screened data give better results in term of R values and MSE, and were selected for the subsequent modeling analysis in this study, that is prediction. |
|---|