Water Quality Prediction in the Waste Water Treatment Process Based on Ridge Regression Echo State Network

Abstract

Echo state network (ESN), a novel recurrent neural network, has a randomly and sparsely connected reservoir. Since the output weights are computed by Moore-Penrose inverse, the ill-posed problem may exist in the ESN. To overcome this problem, ridge regression echo state network (RESN) is proposed, in which the ridge regression algorithm is used to calculate the output weights instead of linear regression. Simulation results show that the RESN has better performance than some other existing methods, thus can deal with the ill-posed problem.