Modelling Of Ammonia Nitrogen In River Using Artificial Intelligence Techniques

Chai, Voon Hao (2021) Modelling Of Ammonia Nitrogen In River Using Artificial Intelligence Techniques. Final Year Project, UTAR.

Preview

PDF Download (1366Kb) | Preview

Abstract

River with high water quality is essential for the survival of living organisms and human being. Ammonia Nitrogen is one of the water quality parameters or chemical pollutants that severely affect the water quality of rivers in Malaysia. Therefore, a precise estimation on ammonia nitrogen concentration in river is utmost important in various fields including agriculture, water resources and irrigation fields. The aim of this study is to predict the concentration of ammonia nitrogen in river using advance mathematical prediction models with the help of artificial intelligence (AI) techniques. The selected study area in this study is Langat River located in Selangor, Malaysia. Three AI models namely Back Propagation Neural Network (BPNN), Adaptive Neuro-Fuzzy Inference System (ANFIS) and Support Vector Machine (SVM) were developed, and their prediction performance were compared. Five water quality parameters were chosen as the input variables for the training and testing of the AI models. The five water quality parameters included Dissolved solids (DS), turbidity (T), total solids (TS), phosphate (PO4 3- ) and nitrate (NO3 - ) were obtained from Department of Irrigation and Drainage (DID) of Malaysia. The water quality parameters mentioned contain 77 dataset which was then utilized as input variables to train and test the AI models. 80% of the 77 dataset were used in the training process while the other 20% of the 77 dataset were used in the training process of the AI models. Min-max normalization was utilized to normalize the ammonia nitrogen concentration values to a range of 0 to 1 prior to the training process. The results generated by the AI models were interpreted and the performance of the AI models were evaluated by statistical analyses comprised of coefficient of determination (R2 ), mean squared error (MSE), root-meansquared error (RMSE), mean absolute error (MAE) and average percentage error. The performance of AI models was intra-compared among their own type and the AI model with the best performance was selected from each developed AI model type. The three best BPNN, ANFIS and SVM models were then compared among one another in term of prediction performance. The comparison result showed that BPNN model trained with log sigmoid function with 4 hidden neurons has the best performance compared to the ANFIS and vi SVM models. Therefore, this BPNN model is concluded to be the most suitable AI model to predict ammonia nitrogen concentration in Langat river. One of the recommendations for future research on this topic includes obtaining more dataset for AI model development. Another recommendation is to replace SVM with Support Vector Regression (SVR) as SVR is more effective in solving regression problem.

Actions (login required)