23-Thesis Report on Real Time Flood Forecasting Using Artificial Neural Network.

Abstract

Forecasting is the making predictions about the uncertainty of the future by using the historical data. Forecasting provide advance warning about any danger which is going to happen in future. Forecasting is broadly considered as a technique or a method for estimating or finding many future aspects of any operation. Planning for the future is a critical aspect. The main goal of forecasting is to give an accurate picture of future as possible as to. ANNs are a form of computing inspired by the functioning of the brain and nervous system. Recently, another class of black box models in the form of Artificial Neural Network (ANN) has been popularized in modeling real time problems wherein the non- linear relationship between the rainfall and runoff process is modeled. In this report, the application of Artificial Neural Network model (ANN) and a model combining the multiple layer regression (MLR) is investigated for modeling the real time flood prediction using rainfall-runoff data of Hamp River, Chattisgarh. . The rainfall in the catchment area Chirapani, Bodla, and Panadariya and the hourly discharge data is used to carry out this research work. The duration of data used is from 1981 to 2009. In this study the ANN and MLR model results are compared with each other. The ANN model performs better for real time flood prediction than that of MLR model during calibration and validation. In addition, the comparison of the scatter plots of ANN model is more precise than that of MLR. The result of all lead times of calibration and validation are compared. The RMSE of ANN model during calibration and validation for seven lead time were found to be 1.5881 and 1.0554, 1.8957 and 1.3638, 2.0828 and 1.5195, 2.2288 and 1.5826, 2.3039 and 1.6345, 2.3689 and 1.7120, 2.4547 and 1.7868 respectively, whereas for the MLR model, RMSE value during calibration and validation were 1.7244 and 1.1617, 2.0635 and 1.4664, 2.2671 and 1.6222, 2.3963 and 1.6742, 2.4826 and 1.7120, 2.5393 and 1.7745, 2.5935 and 1.860 respectively, and also the ANN model efficiency during calibration and validation were 0.8682 and 0.8946, 0.8122 and 0.8241, 0.7733 and 0.7817, 0.7405 and 0.7633, 0.7227 and 0.7476, 0.7068 and 0.7232, 0.6852 and 0.6986 respectively, whereas the MLR model efficiency during calibration and validation were