ESTIMATION OF TRAP EFFICIENCY OF RIHAND RESERVOIR USING ANN MODELLING

Abstract

Deposition of sediments in the reservoirs is a major critical issue in today’s scenario. The sediments flowing along river gets deposited in reservoir as velocity gets reduced in vicinity of reservoir. Practically it is very difficult to estimate the volume of sediments retained in a reservoir and the trap efficiency of reservoir. Trap efficiency of a reservoir is the percentage of sediments that are retained in a reservoir. There are many conventional methods that were developed to estimate the trap efficiency of reservoir. Some of the conventional methods that are used to estimate the trap efficiency of reservoir are Brown’s (1944), Churchill’s (1948), Brune’s (1953), Dendy’s (1974), Gill’s (1979), Heinemann (1981), etc. They all have given their empirical relations that can be used to estimate the trap efficiency of reservoir except Brown (1944). Brown has given the empirical equation relating the Ratio of capacity and watershed ratio. They all shows the relationship between capacity – inflow ratio and trap efficiency of reservoirs. Brune’s is the only method which is widely used all over world. In this present study Artificial Neural Network model was developed to estimate the trap efficiency of Rihand Dam located in Uttar Pradesh. Annual Rainfall, Age of Reservoir, Capacity and Inflow were taken as inputs. Using these inputs, the trap efficiency of reservoirs was estimated by ANN model. The model Developed has 5 and 10 hidden neurons for four and three inputs respectively. The data sets were trained using Levenberg Marquardt process. The MSE was 0.01 when the training was stopped. The average Trap Efficiency from ANN using four and three input parameter is 96.18% and 96.01% respectively. The outputs from ANN model was validated by the empirical equation developed by Jothiprakash and Vaibhav Garg. This equation has age of reservoir as a constraint in it unlike other equation. The average observed TE from Brune’s (1953) and Jothiprakash and Garg (2008) for Rihand Reservoir was found to be 96.11% and 96.03 respectively%. The results got validated and it shows that ANN model gives better results than conventional method. It is simple, easy and can solve complexity very easily. ANN model is not time consuming as well. ANN model gave best results with four inputs which were very close to Jothiprakash and Vaibhav Garg developed equation and other conventional method.