Ammonia nitrogen level forecasting based on PCA-NARX neural network
YUAN Hong-chun1, ZHAO Yan-tao1, LIU Jin-sheng2
1.College of Information Technology, Shanghai Ocean University, Shanghai 201306, China; 2.College of Fisheries and Life Science,Shanghai Ocean University, Shanghai 201306, China
Abstract: An ammonia nitrogen level forecasting model is developed based on the PCA-NARX neural network in which principal component variables extracted are used as exogenous inputs by principal component analysis (PCA) and the network structure was optimized in order to improve accuracy of ammonia nitrogen level forecasting and grasp the trend of the ammonia nitrogen levels accurately. The forecasting performance was conducted in a Chinese mitten handed crabEriocheirsinensis tank compared with NAR and NAR neural networks. Simulation results show that the proposed method has good nonlinear fitting ability and its root mean square error (RMSE) is lower than that in NAR and NARX in 24 h, consistent with NAR and NARX in 24 h and 48 h. Also, the comparison with other models indicates that PCA-NARX neural network has better nonlinear fitting ability and superior in forecasting dissolved oxygen level based on theRMSE in short term (48 h), and can be used to offer scientific guidance to control water quality.