Water Quality Prediction Based on BP Neural Network at Dahuofang Reservoir, China

To ensure the safety of drinking water, understanding the trends of water quality in water resource and to provide a scientific basis for water quality management, a three-layer BP neural network is selected to simulate and predict six water quality indicators of the outbound of Dahuofang Reservoir. The six water quality indicators are dissolved oxygen, five days' biochemical oxygen demand, permanganate index, ammonia nitrogen, total nitrogen and total phosphorus. Training the model with water quality data from 2005 to 2011, Levenberg-Marguardt optimization algorithm is adopted to train samples. After reaching the error requirement, simulate the model with the water quality monitoring data in 2012 and test the model accuracy. Simulation results show that the accuracy of the model prediction is higher in 2012. It is proved that this model can be used to predict water quality of the outbound mouth in Fushun section, and the model provides a theoretical basis for improving the water quality of the reservoir area and can be used to guide the actual water quality management.