Water Quality Prediction in Wastewater Treatment Process Using Lasso Echo State Network

Aiming at the effluent water quality prediction in the wastewater treatment process (WWTP), a lasso echo state network (LESN) is proposed. LESN uses the lasso regularization method instead of linear regression to learn the output weights, the coordinate descent method is used to train LESN. The simulations confirm that LESN can well handle the collinearity problem compared with other ESN models. Effluent water quality prediction is well done, LESN has higher prediction accuracy and generalization ability.