Modeling and optimization of activated sludge bulking for a real wastewater treatment plant using hybrid artificial neural networks-genetic algorithm approach |
| |
| Affiliation: | 1. System Engineering Department, University of Malaga, Campus de Teatinos s/n, 29071, Malaga, Spain;2. System Engineering Department, University of Malaga, Campus de Teatinos s/n, 29071, Malaga, Spain;3. Dept. of Mechanical, Chemical and Materials Engineering, University of Cagliari, Via Marengo, 2, I-09123 Cagliari, Italy;4. Dept. of Civil and Environmental Engineering, Aalto University, P.O. Box 12100, FI-00076 Aalto, Finland;5. Department of Chemical Engineering, Federal University of Campina Grande, 58429-140 Campina Grande, Brazil;6. Centre for Intelligent Systems, Faculty of Science and Technology, University of Algarve, Campus de Gambelas, 8005-139 Faro, Algarve, Portugal;7. System Engineering Department, University of Malaga, Campus de Teatinos s/n, 29071, Malaga, Spain |
| |
| Abstract: | Prediction of sludge bulking is a matter of growing importance around the world. Sludge volume index (SVI) should be monitored to predict sludge bulking for a wastewater treatment plant. This study was an effort to develop hybrid artificial neural network-genetic algorithm models (MLPANN-GA and RBFANN-GA) to accurately predict SVI. Operating parameters, including MLVSS, pH, DO, temperature, TSS, COD and total nitrogen were the inputs of neural networks. Genetic algorithm was utilized in order to optimize weights and thresholds of the MLPANN and RFBANN models. Training procedures for SVI estimation were successful for both the MLPANN-GA and RBFANN-GA models. The training and validation models showed an almost perfect match between experimental and predicted values of SVI. The results indicated that with low experimental values of input data to train ANNs, the MLPANN-GA compared with the RBFANN-GA is more accurate due to higher coefficient of determination (R2) and lower root mean squared error (RMSE) values. The values of RMSE and R2 for the optimal models approached 0 and 1, respectively. The mean average error for the ANN models did not exceed 3% of the input values of the measured SVI. The GA increased the accuracy of all the MLPANN and RBFANN models. |
| |
| Keywords: | Activated sludge bulking Sludge volume index Artificial neural networks Multi-layer perceptron Radial basis function Genetic algorithm |
| 本文献已被 ScienceDirect 等数据库收录！ |
|
