Application of Neuro-Genetic Algorithm to Determine Reservoir Response in Different Hydrologic Adversaries

The hydrologic adversaries like high magnitude storms, extreme dryness, aridity, more than normal demand for water etc. often cause a huge stress on the storage structures such as reservoirs and check dams. This stress implies a lot of adverse effects on the adjacent population. One of the major causes of floods and droughts were due to the mis-management of stored water during hydrologic adversaries. The present study tries to estimate the distribution of the surplus water in the case of hydrologic adversaries. In this regard, two years of daily discharge data of one of the reservoirs, Panchet, of the river Damodar was randomly selected and grouped into six categories based on their magnitude. Three neural models were built. One out of the three was selected due to better performance validating criteria. The behaviour of the inputs in the case of hydrologic abnormality was configured with respect to the available historical records and applied to the selected model. The output would give the magnitude of surplus in the case of the pre-configured hydrologic adversaries. According to the results, the Panchet reservoir could not mitigate the stress created due to the applied hydrologic adversaries. The study was conducted with a single reservoir and one major hydrologic pattern of the decade. A more detailed study with the help of this approach could further improve the model estimation.