The predicting of reservoir algae viscosity based on independent component analysis and Back Propagation Artificial Neural Networks
With the rapid development of industry and agriculture, an increasing of nitrogen phosphorus and other nutrient emission has accelerated the eutrophication process and stimulated the abnormal reproduction of algae. Frequent outbreaks of algal bloom in some large reservoirs pose a serious threat to the use of water resources and the safety of drinking water. How to predict an abnormal reproduction of algae, control the eutrophication process of reservoirs and protect the safety of drinking water has become a top priority for water sources protection. To achieve the prediction and an early warning of raw water algae bloom, the independent component analysis and BP neural network algorithm was adopted. In this paper a typical 3layer neural network model is used to predict the algae concentration of the next day with the latest 20 days related data as a network input. Algae related factors such as nitrogen phosphorus and other nutrient have different sampling frequency. In order to ensure data validity and reliability, all monitoring data is used for modeling and prediction. The best data is selected by the independent component correlation algorithm. In this paper, we use 100 days monitoring data of 9 factors of Songshan Lake reservoir in Dongguan as the training set, and the next 20 days of monitoring data as the test set. Experimental results show that the prediction error is less than 30％, which has proved the validity of the prediction model.
algal bloom algae concentration prediction independent component analysis BP neural network Songshan Lake reservoir
Chang Xu Hongliang Zhou Hongjian Zhang
State Key Laboratory of Industrial Control Technology, Hangzhou, China Department of Control Science and Engineering, Zhejiang University, Hangzhou, China
国际会议
三峡
英文
442-445
2012-05-18（万方平台首次上网日期，不代表论文的发表时间）