Feedforward artificial neural networks for nutrient removal simulation in a multiple stage enhanced biological nutrient removal process
| Autor: | Chiao-Fuei Ouyang, Yu-Jan Chou, Wei-Liang Kuo |
|---|---|
| Rok vydání: | 2003 |
| Předmět: | |
| Zdroj: | Journal of the Chinese Institute of Engineers. 26:211-219 |
| ISSN: | 2158-7299 0253-3839 |
| DOI: | 10.1080/02533839.2003.9670771 |
| Popis: | Serial artificial neural networks (ANNs) describing different metabolic behaviors of autotrophic and heterotrophic microorganisms were employed to simulate dynamic characteristics of nutrient removal in a multiple stage enhanced biological nutrient removal (EBNR) process. In the sludge cultivation phase, the applied ranges of total chemical oxygen demand (COD t ), total phosphorus (T-P) and total nitrogen (T-N) concentration in synthetic wastewater were 100.0∼300.0, 1.7∼5.0 and 13.3∼ 40.0 mg/L, respectively. Empirical results indicated effluent nutrient concentrations of T-N and T-P were less than 8.1mg/L and 0.5mg/L when a three step-feeding mode was applied, and water quality data was collected to train and pre-test the ANNs. Further simulation was finished when the step-feeding ratio was Q 1 :Q 2 :Q 3 = 0.8:0.2:0.0 (recycling sludge ratio= 0.25). Statistical coefficients of correlation between simulated and experimental values were 0.58, 0.95, 0.89, and 0.90 for soluble COD (COD s ), phosphate (PO 4 -P), ammonia-nitrogen (NH 3 -N) and nitrate/nitrite (NO x -N), respectively. Due to low root mean squared error and high correlation coefficient between simulated and experimental data, artificial intelligence technology offered an alternative approach to simulating nutrient removal in an EBNR system. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|