A wavelet neural network hybrid model for monthly ammonia forecasting in river water

Abstract

Forecasting water quality is always an effective approach for water environmental management. This study presents acombined Wavelet transform (WA) and Artificial Neural Network (ANN) model for monthly ammonia nitrogen series predictionin river water. The WA decomposed original time series into different subseries, in which the most significant one waschosen as the training data instead of the original series. Compared to the traditional ANN, the WA-ANN models were found more accurate and reliable. The results of the study indicate that WA could remove the noise of the original datasets and the WA-ANN could help environment decision-maker manage water quality more effective. © Maxwell Scientific Organization, 2013.