Prediction of Effluent Parameters of Wastewater Treatment Plant Based on Improved Least Square Support Vector Machine with PSO

Most of wastewater effluent parameters are difficult to measure online. Many soft-sensing methods have been used to predict them, such as the mechanistic approaches and artificial neural networks. In this paper, improved least squares support vector machines for regression (LS-SVR) is proposed. Benchmark Simulation Model No.1 (BSM1) is used to generate input-output data, then effluent parameters, COD (chemical oxygen demand), BOD (biochemical oxygen demand), TN (total nitrogen), SNH (ammonium nitrogen) and Tss (total suspended solids) forecast model is built. The parameters of LS-SVR are optimized by particle swarm optimization (PSO) in order to obtain a more accurate model. The study shows that the improved LS-SVR modeling approach is capable of predicting the wastewater treatment plant effluent parameters with a good degree of accuracy and is adapted to the changes of the weather.