| Calendar-Based Seasonal versus Fuzzy C-Means Clustering for ANN Prediction of Wastewater Treatment Plant Effluent Quality |
| کد مقاله : 1411-IWWA |
| نویسندگان |
|
الهام رجبی *1، محمدجواد غلامی2 1هیات علمی 2فارغ التحصیل دانشگاه پیام نور |
| چکیده مقاله |
| Artificial neural networks (ANNs) can learn the nonlinear relationships of the treatment process directly from operational data and thereby support effluent-quality prediction and plant management. The main question of this study is whether partitioning the training data before network training, either by calendar season or by a data-driven method, improves prediction accuracy compared with a single model. For this purpose, 1,159 days of daily data from the Sari wastewater treatment plant (Iran), with eight input and four output variables, were used. A feedforward network with one hidden layer, trained with the Levenberg–Marquardt algorithm, was run in three configurations: without clustering (control), with calendar-based seasonal clustering following the method of Zhao et al. (K = 4), and with fuzzy C-means (FCM) clustering (K = 3). The data were randomly divided into training, validation and test subsets in a 70/15/15 ratio, and the normalized root mean square error (NRMSE) was the main comparison criterion. FCM achieved the lowest error for turbidity (0.861 versus 1.009 for calendar-based clustering), was slightly better for total suspended solids, and showed less scatter across different random splits. Differences for chemical oxygen demand were negligible, and the prediction of dissolved oxygen, which is almost independent of the other outputs, remained poor in all three cases. A complementary analysis showed that the high-load FCM cluster coincides with a multi-year increasing trend in influent load, indicating that part of the heterogeneity of the data arises from the non-stationarity of the system rather than from seasonal variation. |
| کلیدواژه ها |
| Artificial neural network, Fuzzy C-means, Seasonal clustering, Wastewater treatment plant, Effluent quality, Non stationarity |
| وضعیت: پذیرفته شده برای ارسال فایل های ارائه پوستر |
