Neural Networks–Driven Soft-Sensors with Applications in Wastewater Treatment

To achieve optimal operations of wastewater treatment, proper monitoring of the difficult-to-measured variables, such as chemical oxygen demand (COD), biochemical oxygen demand (BOD), and so on, is indispensable. Soft-sensors are usually designed for hardware sensor replacement with implications of cost-saving. With the advent of artificial intelligence and the availability of big data, data-driven techniques have reformulated wastewater plants into a smart infrastructure. In this chapter, to deal with the highly dynamic and nonlinear issues in wastewater data, a novel particle swarm optimization (PSO) algorithm and a novel square-root unscented Kalman filter (SR-UKF) are proposed for Elman Neural Networks–based soft-sensor modeling enhancement, aiming to facilitate Elman Neural Networks to adapt to the dynamic data behaviors. The proposed methods are verified by a dataset from the University of California database (UCI database). The results showed that the proposed model can achieve better prediction performance in terms of root mean square error (RMSE) and coefficient relationship (R) than the traditional models.