18-09-2012, 12:11 PM
Surface Water Quality Prediction by Artificial Neural Network
Surface Water Quality.pptx (Size: 483.54 KB / Downloads: 90)
Introduction
Water quality modeling involves the prediction of water pollution of using mathematical simulation techniques.
The present study of Water Quality Prediction is based on the analysis of samples of water collected from Ganga River from various locations in Allahabad city, India.
The water quality parameters were analysed using statistical and soft computing techniques used to predict the Water Quality .
Mathematical models are generally complex & pose difficulty in implementation in real time systems.
Additionally, they fail to predict the future parameters from current & past measurements.
The ANN based model can reveal hidden relationships in the historical data, thus facilitating the prediction and forecasting of water quality.
Moreover, the soft computing techniques are flexible enough to accommodate additional constraints that may arise in the application.
LITERATURE REVIEW
Banerjee et. al. (2008) done assessment on the water quality characteristics of River Ganga at Kolkata Region, India using Water Quality Index and ANN simulation method.
M. Golabi et. al. (2006)used ANN in River Water quality Modelling of Karnoon river in Iran.
Hafizan Juahir et. al. (2005) compared ANN models with conventional models for predicting Water Quality Index (WQI).
Ali Najah et. al. (2001) predict the of Johor River Water Quality Parameters Using ANN.
Artificial Neural Network (ANN) Models
The most popular predictive model usually applied to model non-linear environmental relationship is the Artificial Neural Network (ANN) .
An ANN is composed of a large number of simple processing units, each interacting with others via excitatory or inhibitory connections.
Three different layers can be distinguished:
(i) An input layer
(ii) Hidden layer (one or more)
(iii) Output layer
In an ANN, one of main tasks is to determine the model input variables that significantly affect the output variable(s).
The choice of input variables for the present neural network modelling is based on a statistical correlation analysis of the field data, the prediction accuracy of water quality parameters, and the domain knowledge.
The ANN technique is flexible enough to accommodate additional constraints that may arise in the application.
Moreover, The ANN model can reveal hidden relationships in the historical data, thus facilitating the prediction and forecasting of water quality.
Validation of ANN Models
The best ANN Models developed from the historical data is evaluated with the data obtained during this year.
For every ANN Model, a corresponding regressive model is developed & is being compared.
It is found that ANN Models are better in predicting the water quality parameter than the conventional regressive models.
Conclusion
ANN Models are better tools to predict the water quality parameters rather than conventional statistical models.
D.O. ANN Model is the best model among the others because of higher value of Model Efficiency and lower value of RMSE, of predicted value predicted value.
Stronger the correlation between the parameters, better will be the ANN Model developed and consequently better prediction of water quality parameter is obtained.
The ANN models developed from the historical data is also better in predicting the present water quality data.