Dissolved Oxygen (DO), one of the water quality indicators, was selected as estimate factor to investigate the application of three learning algorithms of artificial neural network, i.e.,  BP Method, bpx Method and L-M Method, in the water quality prediction. Results showed that the prediction effect of the three learning algorithms were all ideal, among which the L-M method which is an optimization algorithm of BP was the most accurate. BP neural network has good application prospect and promotion value in water quality prediction. 