COD_Mn forecast based on BP neural network at Yuqiao Reservoir in Tianjin

To insure the safety of drinking water, find out the change current of water quality, and offer scientific methods for managing water quality, the forecast of COD_Mn is studied based on back propagation (BP) neural network technique at Yuqiao reservoir in Tianjin of PRC. To achieve a higher convergent speed, the forecasting model adopts Levenberg-Marquardt (LM) algorithm, and the model adopts early stop method to improve extended capacity of the model. Sample data of artifical neural network (ANN) are from daily measured values of Yuqiao reservoir from 2003 to 2005. To evaluate veracity of the model roundly, the forecast periods of Yuqiao reservoir are divided into abundant rain period, freezing period and other periods. The effect of the model is reviewed in different periods. The results show the forecast efficiency is the worst in abundant rain period, best in freezing period and the middle in other periods. The whole forecasting effect is good, and the model can be used for guidance in water quality management at Yuqiao reservoir.