Publication: Prediction Of Water Quality Parameter By Applying Evolutionary Algorithm (EA) In Taiwan, China
Date
2020-02
Authors
Muhammad Shukri Bin Nor Azmi
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The research examined the ability of Artificial Nueral Network (ANN) to predict Nitrate
(NO3) in the Taiwan River. Regular monitoring of fluvial water quality criteria by the
Department of the Environment (DOE) for the span of 28 years from 1986 to 2014,
every month. The following four water quality systems have been selected in the
proposed Artificial Neural Network (ANN), namely ammonium (NH3), nitrogen
dioxide (NO2), dissolve oxygen (DO) and phosphate (PO4), for the proposed MLPNNN modelling. Two examples have been added, the first being the establishment of a
NO3 station for four input-based model and second model (Scen 2) Applications to
develop a NO3 prediction model based on prior station fourth inputs and NO3 expected
(upstream). The second scenario was to establish the NO3 prediction model. To order
to meet the verification requirements, the model needs tests when performance results
and the values observed are similar enough. Therefore, the verifications of MLP-NN on
the basis of field data gathering for the period 1986–2014 are presented to analyze the
performance of the model planned.
In order to identify the impact of input parameters on the model, the responsive
analysis was implemented on the model. Dissolving oxygen (DO) and phosphate (PO4)
have been found to be the most important inputs. On the other hand, the most important
parameter was ammonium ions, whilst the lowest contribution to the proposed modell
was Nitrogen dioxide (NO2). Three statistical indices were used in evaluating the
model's success: efficiency coefficient (CE), mid square error (MSE) and correlation
coefficient (CC). Scenario 1 indicates a comparatively low correlation in the test data
collection among observed and expected values. In addition, for the test set of 0.98, 0.96
and 0.97 for one station after the adoption of scenario 2 high correlation coefficients
were obtained between the observed and expected values. It seemed that Scenario 2's
findings were higher than Scenario 1, and all stations were improved significantly from
4 to 8 percent
Description
FYP Sem 2 2019/2020
Keywords
Water Quality Parameter , Water Quality Prediction , Evolutionary Algorithm