To better achieve water quality forecast of Songhua River and instruct scientific management of water quality,a water quality forecasting model is set up by ANN technology and is trained by water-quality data from Sifangtai Monitoring Station of the Songhua River.The model could be applied to forecast CODMn that is one of the main pollution indicators in Songhua River.To improve forecasting accuracy,the data is divided into 12 groups and handled by excluding abnormal data based on clustering analysis.At last a test is carried out to verify the effect of clustering analysis,and the results indicate that the clustering analysis in water-quality forecasting model can improve the forecasting effect significantly.