Modeling Wastewater Treatment Process: A Genetic Programming Approach

Abstract

Wastewater treatment, recycle, and reuse cannot be underestimated in the present context. Microbial units containing bacterial and/or algal biomass are often overrated as treatment techniques providing cheap and effective alternatives. Nonetheless, they come with a setback of limitedly understood systems especially with respect to mechanisms involving pollutant degradation and relationships among diverse environmental factors. Although different modeling techniques have been in use for predicting biodegradation, data mining based soft computing tools such as artificial neural networks (ANNs) and genetic programming (GP) offer substantial control over process operation in terms of specific understanding between experimental inputs and output. ANN and GP are also used in performance prediction over wider influent fluctuations with respect to variations in pH, single- or multipollutants, biomass, and/or e^-donor/acceptor concentrations. Although these tools have found applications in the domain of wastewater treatment, the current practice is mostly on treatment plant performance while limited or no emphasis on relationships between diverse environmental factors and pollutant removal. When compared to models based on kinetic equations and the following cumbersome solving of these equations, soft computing methods decipher the information concealed in the data obtained from many experimental and real-time biological experiments. This chapter provides its readers an insight into biological system aimed at pollutant degradation and use of soft computing tools to precisely identify influential parameters among several factors involved in biodegradation of a certain pollutant. A comprehensive approach of ascertaining and optimizing influential parameters in the removal of specific pollutants from aerobic, anaerobic bioreactors and constructed wetlands (CWs) has been given herewith. It is shown that biochemical oxygen demand removal from the biological systems depended on crucial parameters such as the differential temperature (Ta–Tw; the difference between apparent temperature and wastewater temperature) in CWs and detention times in bioreactors. Root mean square error (RMSE) was used to compute the model performance, for cyanide and phenol removal using GP, and the RMSEs of 1.69 and 2.71 were obtained, respectively, while an RMSE of 5.03 was obtained for modeling CW using ANN. The case studies and modeling approaches discussed here shall direct its readers in selecting and programming data-driven models to better predict and understand the roles of influencing parameters in the removal of complex contaminants treated in biological units.