| Abstract | Reduction of appropriate sites for new reservoirs and the socio-environmental concerns have resulted in reduction of new water storage projects around the world. This slow augmentation of water storing capacities has increased the need of efficient and sustainable sediment evacuation from reservoirs. Many reservoirs around the world are being operated based on rule curves which are not providing sustainable operation of the reservoirs. These rule curves were developed using simulation techniques alone or using optimization models ignoring the sedimentation process while optimization. As such current reservoir simulation and optimization models fall short of incorporating the concept of sustainability, as the reservoir storage losses due to sedimentation are not considered. The present study develops a new reservoir simulation model with genetic algorithm (GA) based optimization capabilities and module to calculate sediment evacuation during simulation. The model developed in present study is called as Reservoir Optimization¬Simulation with Sediment Evacuation (ROSSE) Model. Sediments evacuated during each time step, in the model, are estimated either by the Tsinghua University equation for flushing or through coupling of the GSTARS3 model. Due to huge amount of computational time required by the GSTARS3 embedded in the GA, this study only presents the results using Tsinghua equation. The irrigation based operation policy, selected in the model, releases the full irrigation demands if the reservoir level is between upper and lower rule curves (normal zone). The released water for irrigation is used for power production and the sediment evacuation. No release quota is fixed specifically for sediment evacuation purpose or flushing. The optimization model in the present study is capable to optimize the rule curves both for single and multi objective criteria, minimization and maximization objectives, and with or without constraints. Hybrid GA, Elite GA, multiple ways of constraints handling, multiple selection operators (biased roulette wheel, tournament), multiple mutation operators (uniform, modified uniform), and multiple crossover operators (single, double, uniform) are distinguished features of the optimization model. The simulation and sediment evacuation modules are verified against the observed releases, water levels and reservoir storages, while the GA module is verified through De-Jong's test functions. The ROSSE model is applied to optimize the rule curves of the Tarbela Reservoir Pakistan. Ten daily inflows to the reservoir for a period of 1974 to 2003 are utilized to incorporate the stochastic effects of the inflows implicitly. The Tsinghua equation coefficient is calibrated using the observed sediment concentration data at various outlets of the dam for a period from 1984-2004. For low levels of the reservoir when such observed data is not available, the values recommended by Atkinson are utilized. Two applications of the ROSSE model are described in the dissertation. Firstly, optimal rule curves are obtained for maximization of net economic benefits from various components namely: water release for irrigation, power production, storage conservation by sediment evacuation and flood dis-benefits. Eight sets of the optimized rule curves are compared against the existing rule curves. In multi-objective optimization, the net benefits from irrigation release and sediment evacuation are always assigned first and second priorities. The simulation using optimized rule curves demonstrates an increase of net individual economic benefits in the range of 9% to 248% over the existing rule curves, while there is a small improvement of 8% in the total net economic benefits. The small increase in total net economic benefits is due to a small improvement in hydropower which is the main contributing factor. Shortage of irrigation supply is also reduced by 43% and sediment evacuation is improved up to 28% to enhance the reservoir sustainability. The results point out that the reservoir operation can be improved if the reservoir is drawn down twice in a year, once in February to March and next in June. Second application of the ROSSE model optimizes the rule curves for minimization of irrigation deficits per year, with constraints of production of current level of hydropower and current level of sediment evacuation. A reduction in irrigation deficits of about 24% is computed through simulation of optimized rule curves and existing rule curves. This reduction is attributed to storage conservation through sediment evacuation and optimized rule curves. A sensitivity analysis of nine different GA parameters is carried out to select the optimum values. It is found that the best results after 100 generations are obtained with population size of 200 strings, with probability of crossover of 0.70, and with mutation probability of 0.035. The study has shown the successful integration of sediment evacuation with simulation and optimization model. The study also concludes that there is plenty of scope for sustainable operation (through sediment evacuation) of the reservoir while there is a marginal scope of enhancement of economic benefit. It is also found that the use of the Tsinghua University flushing equation is quite satisfactory for computational intensive GA model and simulation based optimization of the rule curves. The developed methodology and the model can be used for optimization of rule curves of other reservoirs with sedimentation problems |