Prediction of wastewater treatment plants performance based on NW multilayer feedforward small-world artificial neural networks

Abstract

In order to provide a tool for predicting wastewater treatment performance and form a basis for controlling the operation of the process, a reliable model is essential for any wastewater treatment plant. This would minimize the operation costs and assess the stability of environmental balance. For the multi-variable, uncertainty, non-linear characteristics of the wastewater treatment system, a NW multilayer feedforward small-world artificial neural network prediction model is established standing on the actual operation data in the wastewater treatment system. The model overcomes several disadvantages of the conventional BP neural network. Namely: slow convergence, low accuracy and difficulty in finding the global optimum. The results of model calculation show that the predicted value can better match measured value, played an effect of simulating and predicting and be able to optimize the operation status. The establishment of the predicting model provides a simple and practical way for the operation and management in wastewater treatment plant, and has good research and engineering practical value. © 2011 Springer-Verlag.