Investigation of Uncertainty to Artificial Intelligence Models in Tabriz Wastewater Treatment Plant

Document Type : Research Paper


1 PhD. Student in Water and Hydraulic Structures, Dept. of Water and Environment Engineering, Faculty of Civil Engineering, University of Tabriz, Tabriz, Iran

2 Prof., Dept. of Water and Environment Resources Engineering, Faculty of Civil Engineering, University of Tabriz, Tabriz, Iran

3 Assoc. Prof., Dept. of Civil Engineering, Faculty of Engineering, Azarbaijan Shahid Madani University, Tabriz, Iran


In this paper, the uncertainty of artificial intelligence models for evaluting performance of the activated sludge unit of the Tabriz treatment plant is assessed. In this regard, daily data of pollution parameters, particularly Biochemical Oxygen Demand and Chemical Oxygen Demand, are utilized. All data were collected daily during the years (2015-2020) and the best parameters were selected using the correlation coefficient criterion. The TSSi, TDSi, VSSi, pHi parameters and also, BODe and CODe with a one-day delay were selected as model input and BODe and CODe were selected as model output. The calculations of uncertainties were performed in two models of Feed Forward Neural Network as point prediction and lower upper bound estimation method to provide the Prediction Interval. The LUBE method, unlike the classical methods of calculating PI, estimates PI without the need for data distribution information. In this method, the FFNN was trained with two outputs indicating the upper and lower limits of the prediction. PICP assessment and comparing it with μ values, caused γ values to equal zero that, in the continuation of the calculation process caused CWC extraction with the minimum possible amount and production of PI for computational data and observations with the possibility of controlling random changes in the activated sludge section. So, the convergence of the LUBE method has the ability to effectively control the uncertainty between the parameters of the biological section of activated sludge using PI. The time required to build PI is considerably short. Numerical results show approximately 99% success in calculations and coverage of modeling uncertainties. Providing an oscillating range of uncertainties can be a valuable aid in improving economic conditions as well as reducing activated sludge control time and better treatment plant monitoring. Despite the design criteria for BODe of 20 mg per liter, PI results show a supply of 12% of the design index. However, considering the supply of the remaining 88% in terms of quality standard for the use of effluents and returned water, according to the Deputy of Strategic Supervision, publication 535, at the rate of 31 mg per liter in the activated sludge sector, the proper performance of the treatment plant is demonstrated. The LUBE method is an efficient method, so by providing an optimized range of fluctuations for computational data, the smallest abnormal changes in the activated sludge section due to controlling the amount of food for the micro-organisms present in this section; also, the pollution indicators with the least computing time are also reported. In addition, due to the high cost of activated sludge in the wastewater treatment sector, from an economic point of view, it also helps reduce costs. According to the non-linear behavior of bacteria during the reduction of food, as well as the control of mortality caused by the reduction of food, it can be considered a very effective tool.


Baghanam, A. H., Nourani, V., Sheikhbabaei, A. & Seifi, A. J. 2020. Statistical downscaling and projection of future temperature change for Tabriz city, Iran. IOP Conference Series: Earth and Environmental Science, IOP Publishing, 012009. Kerala, India.
Chryssolouris, G., Lee, M. & Ramsey, A. 1996. Confidence interval prediction for neural network models. IEEE Transactions on Neural Networks, 7, 229-232.
Dybowski, R. & Roberts, S. J. 2001. Confidence Intervals and Prediction Intervals for Feed-forward Neural Networks. Cambridge University Press. UK.
Guo, H., Jeong, K., Lim, J., Jo, J., Kim, Y. M., Park, J. P., et al. 2015. Prediction of effluent concentration in a wastewater treatment plant using machine learning models. Journal of Environmental Sciences, 32, 90-101.
Hanbay, D., Turkoglu, I. & Demir, Y. 2008. Prediction of wastewater treatment plant performance based on wavelet packet decomposition and neural networks. Expert Systems with Applications, 34, 1038-1043.
Heskes, T. 1996. Practical confidence and prediction intervals. Advances in Neural Information Processing Systems, 9, 176-182.
Huggi, M. & Mise, S. 2019. Optimized ANN model for ultrasonication wastewater treatment process. International Journal of Advanced Research in Engineering and Technology, 10(3), 94-102.
Khan, M. S., Coulibaly, P. & Dibike, Y. 2006. Uncertainty analysis of statistical downscaling methods. Journal of Hydrology, 319, 357-382.
Khosravi, A., Nahavandi, S. & Creighton, D. 2010. A prediction interval-based approach to determine optimal structures of neural network metamodels. Expert Systems with Applications, 37, 2377-2387.
Mackay, D. J. 1992. A practical Bayesian framework for backpropagation networks. Neural Computation, 4, 448-472.
Nix, D. A. & Weigend, A. S. 1994. Estimating the mean and variance of the target probability distribution. Proceedings of 1994 IEEE International Conference on Neural Networks (ICNN'94), IEEE, 55-60. Orlando, USA.
Nourani, V., Elkiran, G. & Abba, S. 2018. Wastewater treatment plant performance analysis using artificial intelligence–an ensemble approach. Water Science and Technology, 78, 2064-2076.
Nourani, V., Paknezhad, N. J., Sharghi, E. & Khosravi, A. 2019. Estimation of prediction interval in ANN-based multi-GCMs downscaling of hydro-climatologic parameters. Journal of Hydrology, 579, 124226.
Nourani, V., Paknezhad, N. J. & Tanaka, H. 2021. Prediction interval estimation methods for artificial neural network (ANN)-based modeling of the hydro-climatic processes, a review. Sustainability, 13(4), 1633.
Nourani, V., Sayyah-Fard, M., Alami, M. T. & Sharghi, E. 2020. Data pre-processing effect on ANN-based prediction intervals construction of the evaporation process at different climate regions in Iran. Journal of Hydrology, 588, 125078.
Rastegaripour, F., Saboni, M., Shojaei, S. & Tavassoli, A. 2019. Simultaneous management of water and wastewater using ant and artificial neural network (ANN) algorithms. International Journal of Environmental Science and Technology, 16, 5835-5856.
Svetunkov, I. & Petropoulos, F. 2018. Old dog, new tricks: a modelling view of simple moving averages. International Journal of Production Research, 56, 6034-6047.
Wen, C. H. & Vassiliadis, C. 1998. Applying hybrid artificial intelligence techniques in wastewater treatment. Engineering Applications of Artificial Intelligence, 11, 685-705.
Zhou, M., Zhang, Y., Wang, J., Shi, Y. & Puig, V. 2022. Water quality indicator interval prediction in wastewater treatment process based on the improved BES-LSSVM algorithm. Sensors, 22, 422.