Abstract: Appropriate choice of a model to improve the accuracy and reliability of the forecast to provide a scientific basis for regional water quality management, is the key to solve the problem of water quality prediction. In order to solve this problem, the change values of CODMn of the Dongling Bridge monitoring section were forecasted in Liaohe Basin, according to the Liaohe Basin reality with the auto-regressive moving average (ARMA) model. The results showed that: the ARMA (1,1) model was the most appropriate model for the water quality prediction of the Dongling Bridge section by comprehensive consideration of the autocorrelation function, partial autocorrelation function and the BIC principle. The fitting results showed that the relative error was between 2.60% and 25.98%, the average relative error was 13.69%, indicating that the model could accurately predict future water quality trends by taking full advantage of recent water quality data. The forecast of the values of CODMn of Dongling Bridge monitoring sections showed that future CODMn was growing, water environment management tasks were still hard in Liaoning Province. Finally, the problems and development directions of ARMA model application in the prediction of water quality were discussed.