A soft-sensing technique for wastewater treatment based on BP and RBF neural networks

With the modern industry, the water resources which are essential for human survival have been greatly destroyed. With the goal of managing wastewater effectively, economically, and ecologically, scientists have been working years for the simplicity and effectiveness of a wastewater treatment system. However, the quality parameters of wastewater treatment usually cannot be detected on line or otherwise the measurement meters are very expensive. In this paper, the soft-sensing method based on the combined back-propagation (BP) and radial basis function (RBF) neural networks is proposed to solve this problem. Wastewater treatment technique is analyzed systematically. BOD, COD, N and P which cannot be detected on-line are taken as the primary variables. ORP, DO, PH and MLSS which can be detected on-line are taken as the secondary variables. Neutral network for soft-sensing is proposed and trained using the testing data of practical treatment processes. The simulation results show that the soft-sensing system of wastewater treatment based on BP and RBF neural networks can correctly estimate the quality parameters in real time.