Inflow prediction and modelling for a seasonal flow reservoir

• Prediction of reservoir inflow for a partially gauged basin. • Neural network and mathematical model integration to obtain reliable results. • Capturing the variation in modelling by polynomial model. • The statistical correlation between the variables has been established for inflow prediction. Prediction of the inflows has been attempted to strengthen the water management system. The Inflows were predicted for the partially gauged Hemavathi river catchment. The various models which capture the correlation were used to predict flow for monsoon and post-monsoon period. These developed models were analyzed through reliability and suitability in prediction. These models were evaluated through statistical indicators. Absolute errors of the parameters were obtained to evaluate the best performance model. The best model (polynomial) within all the deterministic models was used to generate flows. Further, considering the generated flows the neural network model was trained and evaluated for Hemavathi reservoir, Karnataka India. The trained Back Propagation neural network with supervised learning was considered. The inflow model was validated from the statistical performances. The developed model showed good statistical performances, determining superior reliability, depicting higher Kling Gupta Efficiency of 0.927 and minimum absolute error of 0.939, proving to be an excellent model to rely upon.