A study of water quality modelling with the artificial neural network method in Surabaya river

Abstract

Decreasing river water quality due to the environmental conditions around the river can cause changes in the value of river benefits and endanger the environment. Thus, it becomes necessary to improve river water quality management by monitoring water quality. The purpose of this study is to model predictions of water quality conditions. The method is by analysing the parameters of water quality (BOD, COD, DO, pH, temperature) influenced by rainfall, catchment area, and land use. Data analysis was performed by the artificial neural network (ANN) method, using the Matlab R2014b software. In the process of network development, the most effective model was obtained with a variation of 75% training data with 5 hidden layers and a maximum epoch of 2000. From the results of the model, the relative errors (RE) for BOD, COD, DO, pH, and temperature were found to be 7.80%, 6.33%, 6.83%, 1.92% and 1.05% respectively, with an overall average of 4.79%. Because this RE value is quite small and less than 5%, it can be concluded that the artificial neural network method is quite effective for monitoring water quality conditions in the study area.