Effluent COD of SBR Process Prediction Model Based on Fuzzy-Neural Network

Abstract

The measurements of many key parameters and effluent qualities in WWTP (wastewater treatment plant) are impossible due to the lack of precise online sensors and strong time-delay of WWTP process. The fuzzy neural network (FNN) based effluent COD (chemical oxygen demand) of activated sludge SBR (sequential batch reactor) prediction model is built in this paper, before which preprocessing of SBR simulation data is done using PCA (principal component analysis) to extract the valid information of vast multi-dimension data. The gaining principal components are treated as the inputs of the FNN model to predict effluent COD with an adaptive genetic algorithm (AGA) method to rectify the prediction model. The result indicates that hybrid FNN can extract valid information from dataset and describe complex non-linear properties of WWTP to predict effluent qualities accurately